Response to Questions Posed by the Department of Defense

Regarding Microbiological Criteria as Indicators of Process

Control or Insanitary Conditions

Adopted 10JUN2015, Washington, DC

NATIONAL ADVISORY COMMITTEE ON MICROBIOLOGICAL

CRITERIA FOR FOODS

NACMCF Executive Secretariat, * U. S. Department of Agriculture, Food Safety and Inspection Service,
Office of Public Health Science,
Stop 3777, PP3, 9-210B, 1400 Independence Avenue S W, Washington, DC
20250-3700, USA

*Author for correspondence: NACMCF Executive Secretariat. Tel: 202-690-6537; Fax: 202-690-6364; Email:
james.rogers@fsis.usda.gov

Participating agencies include the U. S. Department of Agriculture, Food Safety and Inspection Service; U.S.
Department of Health and Human Services, Food and Drug Administration, and Centers for Disease Control and
Prevention; U.S. Department of Commerce, National Marine Fisheries Service; and U.S. Department of Defense,
Veterinary Service Activity. This article may be reproduced without prior permission.

Disclaimer: Mention of trade names or commercial products in this publication is solely for the
purpose of providing specific information and does not imply recommendation or endorsement by the
U. S. Department of Agriculture and other participating agencies.
 TABLE OF CONTENTS
Executive Summary
Recommendations
Introduction: Statement of Charge to NACMCF and the Rationale for the Approach
to Address the Charge
Specific Charge to the Committee
Public Health Focus
Committee’s Approach to Answering the Charge
Scope of Committee’s Work
General
Background: Department of Defense Procurement
Food Categories
Principles Used in Making the Process Flow Diagrams
Interpreting the Process Flow Diagrams
Intended Use of the Process Flow Diagrams
Manufacturing Processes and Opportunities for Loss of Process Controls
Measuring Insanitary Conditions
Sampling and Testing
Use of Statistical Sampling Plans in the Supply Chain
Finished Product Testing to Aid in the Management and Control of Suppliers
Process Control
Statistical Process Control Limits
Process Capability
SPC Monitoring via Microbiological Testing
Considerations for Finished-Product Testing
Sampling Plans for Screening and Auditing Suppliers
Surveillance at Point of Sale
Microbiological Limits and Criteria
Development of Limits and Criteria
Pathogens Important to Public Health
Indicators that Reflect Loss of Process Control or Insanitary Conditions
Comments on Microbiological Limits for Specific Food Categories
Routine and Non-routine Testing
Plan of Action if Limits are Exceeded
Commodity Specific Comments on Microbiological Limits
Other Indicators of Process Control and Sanitary Conditions
Glossary
Appendices
Bibliography
Response to Questions Posed by the Department of Defense Regarding Microbiological
Criteria as Indicators of Process Control or Insanitary Conditions

EXECUTIVE SUMMARY

The Department of Defense (DOD) purchases a grocery-store array of foods (hereafter to

include bottled water and packaged ice) throughout the world. DOD primarily uses the
assessment of a supplier’s food safety plan, including its HACCP system, to determine whether
a supplier is an acceptable supplier to meet its mission requirements. For these suppliers, DOD
can rely less on microbiological testing and more on process-oriented, risk-based preventive
controls that ensure the supplier’s manufacturing process is controlled and sanitary conditions
are maintained. However, some mission requirements include the need to purchase foods
where suppliers may not have fully developed food safety plans, including HACCP systems. In
these instances, DOD has a need for standardized sampling and testing programs that reflect
process control and assess sanitary manufacturing conditions. Such programs, defined herein,
would enable DOD to monitor suppliers from centralized locations, prioritize supplier audits, and
conduct cost-effective and meaningful verification testing.

To assist DOD with its ability to assess suppliers that do not have well-established food safety
plans, the NACMCF Committee (hereafter the Committee) has provided microbiological limits
for food categories that reflect process control and sanitary manufacturing conditions. These
limits are not microbiological criteria for finished products typically found in a product
specification, but are provided to help DOD assess process control and sanitary conditions in
those suppliers without evidence of a documented and functioning food safety plan. Combined
with process flow diagrams of manufacturing processes, the microbiological limits also provide
guidance to DOD auditors when assisting suppliers with corrective and preventive actions taken
when there is evidence of insanitary conditions and lack of process control. The processes for
statistical analyses of microbiological data for DOD and suppliers are provided to optimize the
use of the data in making decisions affecting process control and sanitation. These limits are
based on expert opinion, industry recommendations, and published finished-product
microbiological criteria from global sources.

RECOMMENDATIONS

• DOD should develop and implement a supplier expectations policy and program to address
supplier programs such as crisis management, environmental monitoring, sanitation
effectiveness monitoring, pest control, Good Manufacturing Practices (GMPs), Hazard
Analysis and Critical Control Point (HACCP) systems, preventive maintenance, the use of
statistical process control (SPC), and verification testing, as appropriate to the individual
operation.
• DOD should share the information contained herein with suppliers who do not have
documented and functioning food safety plans to begin the process of having them develop
SPC charts to demonstrate process control and sanitary conditions. These charts should be
based on microbiological limits provided in Appendix J. Suppliers also should examine
trends in the data from the supplier’s Environmental Monitoring Program (EMP) and

1
 sanitation effectiveness monitoring program. A timeline for development and use of these
charts should be set.
• DOD should provide a list of expert consultants that can assist suppliers with development
and implementation of the SPC charts, and EMP.
• DOD should develop purchasing specifications that include microbiological criteria for foods
purchased through the Worldwide Directory as well as for those foods purchased outside of
the Directory. These specifications should be set initially based on consultation with
industry experts, shared as draft specifications with the supplier community, and fully
adopted after feedback and data confirm that the specifications, and the microbiological
criteria imbedded therein, ensure safe and wholesome products, and are realistic and
practical.
• DOD should communicate microbiological standards, specifications and guidelines to all
suppliers and brokers.
• DOD should request that suppliers document their acceptance of the standards,
specifications and guidelines in manufacturing food for DOD.
• DOD should require that their suppliers, even if instructed through brokers, use the sampling
plan, specified limits, and analytical methods specified in the microbiological criteria (when
formally developed and implemented), and maintain documentation for audit purposes.
• DOD should require Certificates of Analysis and consider the use of Certificates of
Compliance with each shipment of product received to verify compliance with the specified
microbiological criteria (when formally developed and implemented).
• If there is a third-party intermediary that is involved in the food supply chain, the
intermediary should be required to receive, maintain and transfer the Certificate of Analysis
or Certificate of Compliance with the products.
• Whenever and wherever possible, meat, poultry and processed egg products should be
purchased from countries with United States Department of Agriculture (USDA)-equivalent
inspection programs, and from manufacturing establishments that meet the requirements of
the inspection system. When this is not possible, the manufacturing facility should meet the
requirements specified by USDA for production of meat, poultry and egg products. The
product specification for fresh (unfrozen) raw meat and poultry should include a maximum
time between slaughter and receipt by DOD.
• DOD should leverage the implementation of the Food Safety Modernization Act (FSMA)
legislation and regulations, requiring all suppliers that would be regulated by the Food and
Drug Administration (FDA) to meet statutory and regulatory requirements as mandated by
FSMA and corresponding regulatory rules.
• DOD should use an information technology solution that requires all suppliers to input key
data such as location, contacts, product identification, code dating and traceability program,
significant hazards, audit scores, regulatory actions (e.g., equivalent to recalls, market
withdrawals, non-compliance records), SPC data, and microbiological test data. DOD
should capture appropriate data in a standardized electronic spreadsheet or database.
• The risk of potential foodborne pathogens should be considered not only for fresh-cut and
frozen fruits and vegetables but also for whole or unprocessed fruits and vegetables.
• The risk of potential foodborne pathogens should be considered not only for processed nuts,
spices and herbs but also for unprocessed nuts, spices and herbs.
• DOD should develop procedures to collect appropriate meta-data associated with assay
results. Meta-data are data about the data, such as, methods, sample size, analytical unit,
point of sampling, and the reason the sample was collected.
• DOD should incorporate evaluation of sampling schemes and SPC into audit procedures for
those suppliers using the microbiological limits to assess process control and sanitary
conditions.

2
• DOD should consider enhancing diagnosis and reporting of foodborne illness, and
integrating this information amongst the Services, to help identify potential problems within
the supply chain.

INTRODUCTION: STATEMENT OF CHARGE TO NACMCF AND THE RATIONALE FOR

THE APPROACH TO THE CHARGE

DOD has specific action levels for various microbiological pathogens (e.g., Salmonella, Listeria
monocytogenes, Escherichia coli O157:H7, and Clostridium perfringens) and microbiological
toxins in certain raw and processed meat, poultry, egg products and other products, such as
fresh fruits and vegetables, procured globally for U.S. military personnel (U.S. Army Public
Health Command (USAPHC), Circular 40-1: Worldwide Directory of Sanitarily Approved Food
Establishments for Armed Forces Procurement, 2012; Appendix O, 2013 (27). Hereafter,
USAPHC Circular 40-1 is referred to as the Worldwide Directory. In addition, there are bacteria
that, when present in higher numbers, may indicate that processing conditions did not
adequately prevent bacterial growth or reduce bacterial contamination of the product. DOD has
encountered circumstances where the presence of potential pathogens or the numbers of non-
pathogenic indicator bacteria have generated concerns about the safety and/or wholesomeness
of products. DOD seeks updated microbiological limits to better evaluate process control and
insanitary 1 conditions at the point of production.

The Committee agreed with the need to establish microbiological limits to help assess process
control and sanitary conditions at DOD suppliers that do not have documented and functioning
food safety plans, including HACCP systems. In time, the testing by these suppliers, and to a
lesser extent by DOD, should assist these suppliers to develop functioning food safety plans
and enable the suppliers to meet the microbiological specifications established by DOD. DOD
also expressed interest in the use of criteria such as Staphylococcus aureus and Bacillus
cereus levels in ready-to-eat (RTE) products, mesophilic aerobic plate count (APC) in raw and
RTE products, and other possible indicators (e.g., generic E. coli, coliforms,
Enterobacteriaceae, enterococci and gas-forming anaerobes) for establishing that food was
manufactured with process controls and under sanitary conditions.

SPECIFIC CHARGE TO THE COMMITTEE

Because of the many questions regarding microbiological limits that might indicate poor process
control or insanitary conditions, the Committee was asked for its guidance to clarify the following
issues.

Describe processes and important considerations that could be used to develop a

microbiological criterion for a particular product (e.g., bagged leafy greens, dairy
products, grain-based products, raw ground beef, and RTE sliced luncheon meat) at
various points in the process that might indicate poor process control and/or insanitary

1
The terms insanitary and unsanitary are considered as one and the same in this document.
Insanitary is a word that has been used in regulatory language. In this document insanitary is
used as this term was provided in the charge to the NACMCF Committee.

3
 conditions. Describe how the processes and considerations could differ in other regions
of the world where processing conditions may make certain indicators or levels of
indicators more or less appropriate.

At the point of production, how many Staphylococcus aureus, Bacillus cereus, generic
Escherichia coli, coliforms, Enterobacteriaceae, enterococci and/or gas-forming
anaerobes in RTE finished products might indicate: a) a possible process control
problem or insanitary conditions, or b) potentially hazardous product unfit for
distribution? How might the levels and the applicability of these criteria vary between
different RTE products (e.g., processed meat, poultry, egg products, refrigerated
meat/poultry salads, and bagged leafy green salads)?

At the point of production, what level of mesophilic aerobic plate count in RTE finished
products and in non-intact raw meat and poultry products might indicate a possible
process control problem or insanitary conditions? How might these criteria vary between
different RTE products (e.g., processed meat, poultry and egg products, and refrigerated
meat/poultry salads)? How might these criteria vary between different non-intact raw
products (e.g., beef trimmings versus ground product)? How might these levels be
expected to change during the expected shelf-life of the product?

Are there other potential indicators (e.g., microbiological, biochemical or molecular

parameters) of process control that should be considered? If so, how might these apply
at various points in the process to major product categories (e.g., processed meat,
poultry and egg products, bagged leafy green salads and refrigerated meat/poultry
salads)?

Discuss various sampling plans (e.g., International Commission on Microbiological

Specifications for Food, ICMSF, 2- or 3-class plans) that may be applicable for the
various analytes and products identified in the questions above.

The Committee notes that the microbiological limits reflecting process control and sanitary
conditions requested by DOD should not be misinterpreted as microbiological criteria
(specifications and guidelines) for finished food products. It is important that persons reading
and using this document do not immediately transfer the limits provided herein to
microbiological criteria for foods. Over time, as suppliers without documented and functioning
food safety plans, including HACCP systems, use the microbiological limits to establish that
their processes are in control and that sanitary conditions exist during manufacturing, they can
complement this testing with their development of food safety plans that will demonstrate and
ensure that the products purchased by DOD meet the microbiological criteria for finished food
products. Once such documented and functioning food safety plans are audited by DOD and
found to be effective, testing using the microbiological limits provided herein will be secondary
and useful when there is evidence that there is a lack of process control or sanitary conditions
and investigative actions are undertaken to determine root causes.

PUBLIC HEALTH FOCUS

4
With the large number of personnel served by DOD, the wide variety of raw, RTE and fresh
foods procured, and the high number of countries, brokers and suppliers, the implications for
failures in the food safety systems are considerable. While insanitary conditions and process
failures can lead to higher numbers of indicator organisms (or classes of microorganisms such
as coliforms or aerobic bacteria detected by APC; hereafter “indicator organisms”), the greater
risks are failures leading to increased prevalence of pathogens in foods.

Verification testing by DOD, while limited in scope and absolute numbers of tests, should
provide feedback to suppliers to improve controls where necessary. DOD inspection and
auditing staff need to be equipped with tools to assist them in their evaluation of suppliers of a
wide array of products. One tool will be process flow diagrams that illustrate points in the
manufacturing process where loss of control or insanitary conditions can lead to introduction or
growth of microbial contamination.

COMMITTEE’S APPROACH TO ANSWERING THE CHARGE

The Committee leveraged the expertise of the Committee members, additional experts and
published literature and finished-product microbiological criteria to assist in developing
microbiological limits indicative of process control and sanitary conditions for food
manufacturing. The Committee prepared process flow diagrams to reflect the major food
categories purchased by DOD and used these diagrams to predict unit operations that would
lead to an increased prevalence of pathogens and levels of indicator organisms, or growth of
contaminants, based on loss of control or insanitary conditions. The diagrams also indicate
where in the process there are lethality steps.

SCOPE OF COMMITTEE’S WORK

The Committee focused on major food product categories to address the questions posed by
DOD. DOD purchases food products that include what one would find in a retail supermarket.
It was not in the scope of the Committee to recommend finished-product microbiological criteria
(i.e., product specifications and guidelines with levels of microorganisms describing acceptable,
marginally-acceptable and unacceptable products) for the vast array of products. In addition,
some food items purchased by DOD will no doubt fall outside of the major food categories
included by the Committee. DOD will need to work with food safety experts to address any
foods not covered in the major food categories.

The Committee recognized that a food safety program for DOD requires a farm to table
approach; but the charge did not ask for the Committee to address producer food safety
programs, supplier GMPs, broker responsibilities, management of the microbiological data,
information technology to optimize use of supplier testing and DOD verification testing, or food
service operations managed by DOD or their contractors. All of these components affect food
safety and quality of the food purchased and used by DOD and should be included in its
comprehensive food safety plan.

5
The Committee did not address the variability in food manufacturing around the world. The
Committee chose to recommend microbiological limits that reflect traditional processes that are
in control and running under sanitary conditions. The Committee did not address the
consequences for suppliers whose processes are deemed out-of-control or operating with
insanitary conditions. DOD will determine what steps it will take in the event a supplier is unable
to substantiate their process is in control or that sanitary conditions exist for manufacturing.
This report is intended to assist DOD in meeting mission requirements, particularly when
purchasing from suppliers without documented and functioning food safety plans, including
HACCP systems.

In addressing the charge, the Committee did not focus on establishing microbiological criteria as
part of purchasing specifications, which DOD does not currently use. The Committee does
discuss the use of microbiological limits for both assessment of process control and sanitary
conditions, and the use of the limits, when and where appropriate, as the initial step toward
developing microbiological criteria for lot acceptance.

The Committee did not address the programs and systems for delivering microbiological limits
to suppliers, ensuring suppliers implement testing against the limits, reviewing microbiological
data from suppliers, targeting of suppliers that do not test or do not meet the limits, collecting
and managing data on microbiological quality of the products produced for DOD, and for
selecting new suppliers or terminating existing suppliers.

GENERAL

While sampling and testing of food products are tools to verify compliance with preventive and
pre-requisite programs, process control and sanitary conditions, HACCP systems and
microbiological criteria, the results do not guarantee food safety. For all refrigerated and frozen
products, temperature monitoring should be done throughout storage and distribution channels,
as well as at receipt by DOD. Appropriate organoleptic and visual evaluation of the product and
the means of conveyance in which it was delivered should occur. Where possible, continuous
temperature recording documentation associated with the container delivering these products
should be reviewed before accepting the products.

For food products classified under the jurisdiction of FDA inspection, the facilities supplying
DOD should meet all applicable regulatory requirements, including those promulgated under the
authority of the FSMA with regard to preventive controls and product safety. Meat, poultry and
egg products that would be classified under the jurisdiction of the USDA Food Safety and
Inspection Service (FSIS) should meet the regulatory requirements defined by FSIS for the U.S.
and as equivalent for foreign suppliers.

BACKGROUND: DOD PROCUREMENT

DOD procures food products from all 50 states, U.S. territories, and over 60 countries. These
food products are made available to active duty and reserve service members and to retirees

6
and eligible family members who choose to purchase from on-post facilities. Clearly the ability to
safeguard these food products and ensure high quality is of paramount importance.

The DOD selection and approval process for new suppliers can take three months. In some
situations where foods are required more rapidly, expedited processes are used to approve
suppliers. All purchases of food for the military whether on bases, remote locations, ships, or
through commissaries or other commercial establishments, should occur using the Worldwide
Directory. Most of the purchases occur through the Defense Logistics Agency, but the Defense
Commissary Agency also purchases food products for grocery-type operations. Ship supply
officers will purchase food products for their ship. There are instances where procurement
occurs outside the Worldwide Directory, especially where fresh foods, including meat and
poultry, are purchased. In many instances, these non-standard situations are corrected when
detected; however, ship supply officers are granted more freedom in buying from unapproved
sources. It is noteworthy, and potentially problematic, that fresh fruits and vegetables are
currently exempt from requirements to purchase from approved suppliers.

Based on the food product and a DOD informal risk ranking, approved suppliers are scheduled
for DOD food protection audits on a quarterly, semi-annual, or annual basis. Food protection
audits encompass an establishment's total food safety and food protection systems and
programs. Those facilities receiving a passing score are then listed in the Worldwide Directory.
The audit scores are based on observations, with major and critical defects noted, and different
ramifications on the approval status for each type of finding. Audit documentation is reviewed
first at one of the 20 districts, then at one of the five regions, and finally at the Army Public
Health Command where new or continued approval is granted. If major or critical failures occur,
a corrective action request with a timeframe for completion is made of the supplier. Follow-up is
scheduled at a time reflective of the seriousness of the failure.

DOD evaluates the supplier’s food safety plan, including HACCP system, to help determine
whether the supplier can provide safe and wholesome food products. This evaluation also
includes a review of verification testing data that supports the efficacy of the supplier’s food
safety plan. In instances where a supplier is needed to meet mission requirements, but does
not have a documented and functioning food safety plan, DOD requires an alternative means to
assess the supplier’s processes and sanitary condition of the production environment.
Microbiological testing is one of the tools that help with this assessment. The microbiological
limits provided herein were requested by DOD to provide guidance on what tests are
appropriate for various foods and production processes, and what test results may be indicative
of process control and sanitary conditions.

Many food manufacturing facilities reference microbiological criteria from various entities or
have established their own criteria to monitor the safety and quality of raw or RTE components
used to manufacture finished products. The Codex Alimentarius defines a microbiological
criterion as consisting of the following components (30, 31):
• The purpose of the microbiological criterion (e.g., lot acceptance or process control);

7
 • The food, process or food safety control system to which the microbiological criterion
applies;
• The specified point in the food chain where the microbiological criterion applies;
• The microorganism(s) and the reason for its selection;
• The microbiological limits (e.g., m, M, or other action levels);
• A sampling plan defining the number of sample units to be taken (n), the size of the
analytical unit, and where appropriate, the acceptance number (c);
• Depending on its purpose, an indication of the statistical performance of the sampling
plan; and
• Analytical methods and their performance parameters.
DOD has established their own action levels (not-to-exceed limits) for finished products to assist
auditors in their evaluation of various processing systems and finished products. DOD
procurement requires that food products adhere to U.S. regulatory requirements; however, as
mentioned above, exceptions to this requirement may be granted under limited circumstances.

Laboratory analysis forms an integral part of the overall mission of protecting military personnel
and DOD beneficiary populations from foodborne and waterborne (hereafter foodborne will
include waterborne) illness. The DOD program allows for testing of food products and the
environments in which they are produced. Laboratory testing includes qualitative and
quantitative analyses for pathogenic and nonpathogenic bacteria, respectively, as well as
verifying other wholesomeness and quality parameters. Food testing equipment is located
within each DOD deployable veterinary detachment to provide presumptive (considered Level 1
testing by DOD) microbiological testing results, with the staff of each detachment responsible for
animal care, food protection, and review of area facilities that supply food. Testing by a food
manufacturing facility using an accredited laboratory (e.g., ISO 17025) is required for DOD
procurement. Currently, DOD uses microbiological test results in combination with audit findings
to determine the status of an establishment regarding initial and on-going approval, or whether
product that has been procured is safe and wholesome for military personnel.

Appropriate organoleptic evaluation of food products may be useful to assess quality. While
organoleptic examination has its value, it is inherently subjective and dependent upon sensory
capabilities that vary from analyst to analyst. Numbers of indicator bacteria such as APC might
be more effective for determining quality of products that may have been stored for a significant
period of time. However, fresh produce may have appropriate quality for use while also
containing substantial comparatively high concentrations of aerobic bacteria.

Food processors, including those who supply DOD with RTE multi-component products (e.g.,
meals, sandwiches), should be responsible for evaluating individual components (e.g.,
processed meats, cheese, poultry, egg products and spices) received at their establishments.
In many cases, these components may be included as ingredients in the final product without
further processing to inactivate biological hazards. The supplier establishments should perform
microbiological testing on these raw materials, require microbiological test results from the
secondary suppliers on a Certificate of Analysis, or require the listing of microbiological criteria
as elements of a Certificate of Conformance that accompanies the raw materials.

8
A variety of analytes (e.g., aerobic bacteria, E. coli, Enterobacteriaceae, coliforms, enterococci)
currently are monitored on a limited basis by DOD to suggest potential insanitary conditions or
poor process control. This report recommends that this testing should be done by suppliers
without documented and functioning food safety plans, including HACCP systems, using the
microbiological limits provided herein to demonstrate process control and sanitary conditions.
Currently, there is no consensus in the U.S. on acceptable microbiological limits for indicator
bacteria to indicate a process is in control. Such limits may vary by facility, process and food,
and may be best determined through the use of SPC as described herein.

FOOD CATEGORIES

Because of the vast array of food products purchased by DOD, categorization is complex. It is
beyond the scope of this document to list or cover all foods purchased by DOD. The major food
categories and the subcategories covered herein include:

Beverages
Bottled water
Ice, packaged
Juices and drinks, pasteurized, refrigerated
Shelf stable

Dairy
Butter, margarine
Cheese, hard
Cheese, soft, semi-soft, surface ripened
Cultured, pH<4.8
Cultured, pH>4.8 and < 5.4
Dried products (does not include dairy ingredients used to make infant formula)
Frozen desserts
Milk and milk products (fluid)
Processed cheese

Egg Products
Pasteurized, processed
Shell eggs, raw

Grain-based Products
RTE, baked items, refrigerated or temperature/time controlled for safety (TCS)
RTE, baked items, shelf stable or non-TCS
RTE, cereals
RTE, cold pressed bars
Non-RTE, Dry flour-based mixes
Non-RTE, Pasta, dried or refrigerated

Meals and Entrees

Non-RTE, ready-to-cook (RTC) meals, includes raw ingredients
RTE, deli salads, sandwiches, heat-eat meals, sushi
RTE, sous vide, cook and chill

9
Meat, Pork, Poultry Products
Non-RTE, beef and pork, raw, intact and non-intact
Non-RTE, poultry, raw
RTE, cooked, perishable
RTE, fermented, dried

Nuts and Nut Butters

RTE, not processed for lethality
RTE, processed for lethality

Produce
Fruits and vegetables, cut, frozen or refrigerated, minimally processed
Fruits and vegetables, whole
Mushrooms
Packaged salads and leafy greens
Vegetable sprouts

Seafood
Non-RTE, raw
RTE, fish, cold smoked
RTE, cooked or hot smoked
RTE, raw molluscan shellfish

Spices and Herbs, Coffee and Tea

PROCESS FLOW DIAGRAMS

The generic process flow diagrams for these food categories are included (Appendix A) to
identify for DOD auditors the steps in the manufacturing process where microbiological counts
could potentially increase with loss of process control or development of insanitary conditions.
In addition, the flow charts illustrate where there are lethality steps that reduce numbers of
indicator organisms and pathogens.

Principles Used in Making the Process Flow Diagrams

Steps for receiving and storing packaging materials were omitted to simplify the creation and
use of the process flow diagrams. It is expected that a DOD-approved food processing plant
would have appropriate control and documentation of these functions, either as part of product-
specific preventive controls or HACCP system, or as preventive and pre-requisite programs
such as Standard Operating Procedures (SOPs) for receiving and storage. It was recognized
that a finished food product could move through many storage and distribution facilities as part
of the supply chain. Moreover, it is possible that a finished product of one production system
could be an input for another production system. The final two steps were denoted “store
finished product” and “distribute finished product” to simplify the creation and use of the process
flow diagrams.

For several types of food, there are many different possible combinations of manufacturing
steps. Rather than try to show all multiple combinations and step sequences, the steps that

10
could be used in the relevant portion of the manufacturing process were listed collectively. For
example, in the process flow diagram for yogurt, the “add culture” step also includes the
information “(may be preceded by concentration)” and the “process” step also includes “filter,
heat, separate, concentrate, stir (optional)”. In the coffee process flow diagram the “process
raw coffee cherries” step lists the component steps of a wet method and a dry method to
process the coffee cherries. The Committee assumes that DOD personnel will be able to
recognize the specific steps observed at a food processing plant from among the general
manufacturing steps shown on the process flow diagrams.

Interpreting the Process Flow Diagrams

The name of a processing step may be followed by any of the following designations:
C, a step at which significant contamination may occur when adequate process controls are not
in place, G, a step in the process where growth of microorganisms can occur, K, a step where
there is a pathogen kill step, and S, a point where sampling and testing by the supplier are
recommended for verification or investigation.

The effectiveness of the expected process controls at preventing contamination may differ
considerably from step-to-step and product-to-product. For example, there would be a greater
likelihood of contamination during the harvesting of coffee cherries than during the packaging of
ground roasted coffee beans. Similarly, less contamination might be expected during yogurt
packaging than during the packaging of raw, non-RTE seafood.

Programs for minimizing contamination at the identified steps include Good Agricultural
Practices (GAPs), Sanitation Standard Operating Procedures (SSOPs), GMPs, SOPs for
specific steps, and purchasing specifications. Steps denoted as potential contamination points
may occur before or after a step causing significant reductions in the numbers of
microorganisms present in the food. For example, there may be a high level of concern about
L. monocytogenes contamination of RTE foods during the “package” step and this step will be
labeled with a “C.”

Intended Use of the Process Flow Diagrams

DOD personnel should use the process flow diagrams to review the general steps to
manufacture the food product under evaluation. From the process flow diagram, DOD
personnel should determine the step(s) at which sampling should be done by the supplier
without a documented and functioning food safety plan to demonstrate process control and
sanitary conditions. When microbiological or organoleptic analyses indicate that any supplier
may have shortcomings in process or sanitation controls, DOD personnel should use the
process flow diagram to determine steps at which contamination could occur or steps at which a
failure to achieve the expected destruction of bacteria may be occurring. It shall be important
that DOD consider that test results or organoleptic assessments for finished products at the
point of use (e.g., commissaries) may not reflect loss of process control or insanitary conditions
at the supplier since factors such as temperature control during storage and distribution can

11
affect microbiological results and organoleptic properties, and should be taken into account
when deriving conclusions about a supplier’s manufacturing processes.

MANUFACTURING PROCESSES AND OPPORTUNITIES FOR LOSS OF PROCESS

CONTROL

The designation of food categories and subcategories is based on criteria such as the food
description itself, type and extent of processing, RTE status, and chemical characteristics of the
food. For each subcategory a general process flow diagram depicts the manufacturing process
for the foods in that subcategory. If DOD investigates a process following the review of
verification test data or as part of an on-site audit, the process flow diagrams provide insights
into where in the manufacturing process the investigator or auditor could focus their attention.

Measuring Insanitary Conditions

The Committee believes that the best assessment of insanitary conditions is not necessarily
finished product testing, but is typically best achieved through strategic evaluation of the
production inputs, cleaning and sanitation practices and their efficacy, and the environmental
monitoring and sanitation effectiveness monitoring data generated by the supplier facility as part
of their preventive controls program.

SAMPLING AND TESTING

There are various reasons for sampling and testing by DOD itself. While relying primarily on
supplier testing, DOD may sample food products at locations such as distribution centers, field
locations or commissaries to determine the microbiological quality of the food product at a
particular point in the supply chain. The test results from analysis of these samples can provide
insights into supplier compliance with specified microbiological limits; although, as pointed out
above, the results would be affected by the warehousing, distribution and handling processes
and conditions in the supply chain from the time of manufacturing to the point of sampling. For
example, the results can provide indirect information regarding temperature control during
warehousing and its impact on the shelf life of the food product.

DOD also may take samples during supplier audits. If finished products are sampled, these
samples represent verification samples; the test results provide some indication of the ability of
the supplier to manufacture safe and wholesome food products and provide an incentive to
establish and maintain process control and sanitary conditions. The allocation of verification
testing resources should include consideration of the potential presence of biological, chemical
and physical hazards, type of food, supplier characteristics and where the supplier is located,
audit results, shelf life, the distribution system and likelihood of temperature abuse, as well as
the cost of sampling and testing. DOD has an informal risk ranking process that has been used
to define audit frequencies. A more systematic and analytical approach to risk ranking of foods
and suppliers by DOD considering the factors specified above would enhance controls over
food safety and quality, as well as resource allocation.

12
The DOD process of evaluating suppliers with documented and functioning food safety plans
should rely more on the documented evidence supporting effective food safety plans, including
verification testing results (EMP, sanitation effectiveness monitoring, and finished product
testing where appropriate) generated by the suppliers, with the DOD sampling and testing used
only for periodic verification. For those suppliers without documented and functioning food
safety plans, DOD should ensure the suppliers are conducting sufficient sampling and testing to
demonstrate process control and to establish that their manufacturing is occurring under
sanitary conditions, using the guidance provided in this report. When deemed necessary, more
finished product verification testing by the supplier and DOD may be appropriate for these
suppliers until they develop functioning food safety plans.

Use of Statistical Sampling Plans in the Supply Chain

Currently, DOD, through the USAPHC, maintains the Worldwide Directory but does not stipulate
purchase specifications, such as microbiological criteria including sampling plans,
microbiological limits, and reference methods for specific microorganism-commodity
combinations. This section addressing sampling plans is not intended to provide guidance to
DOD (or any other entity) for elaborating microbiological specifications for foods. Instead, the
aim is to provide some contextual and statistical background for DOD to consider when
evaluating food suppliers, their microbiological data, and the extent to which their manufacturing
process is in control.

Strategic microbiological testing of foods, as in-process samples or finished products, provides

useful information about microbiological quality, safety, sanitation, and the effectiveness and
extent of process control. While it is rarely possible to use microbiological testing of foods to
ensure safety and wholesomeness, it is possible to design strategic sampling schemes and
select appropriate analytes and assays that can aid in the management and control of suppliers.
Testing data can be used to help assess manufacturing and monitoring systems such as
HACCP and preventive control programs.

In some instances (e.g., immediate need by DOD for a supplier without a documented and
functioning food safety system, rapid development and implementation of HACCP systems and
preventive control programs by a supplier may not be possible in the short term. In such
instances, use of the microbiological limits provided in this report may be useful for suppliers
and DOD to evaluate the food safety and quality performance of the manufacturing process.
Furthermore, analysis of the data may help identify improvement opportunities. The Committee
recommends that a long-term goal be that all approved suppliers develop and implement
effective food safety plans, including HACCP systems, preventive control and prerequisite
programs. In doing so, suppliers and DOD can rely less on the use of the microbiological limits
described herein and finished-product testing, and more on data associated with the food safety
plan that demonstrate the manufacturing process is stable and capable, and sanitary conditions
are maintained continuously.

13
SPC methods are a powerful tool to evaluate process capability and monitor the extent of
control within a manufacturing process. In particular, SPC can be used to identify an out-of-
control process and consequently flag events warranting investigation for an assignable cause,
corrective action and potential preventive action. In this document, we focus on sampling
schemes that allow the use of SPC to assess process control and sanitary conditions,
particularly, but not exclusively, for suppliers without a documented and functioning food safety
plan. Some approaches described herein also may be suitable for a variety of other qualitatively
or quantitatively measurable observations such as those identifying chemical hazards or
physicochemical measurements; but control of these food process characteristics is beyond the
scope of this report.

Finished-Product Testing to Aid in the Management and Control of Suppliers

As mentioned previously, the microbiological limits provided in this report are not microbiological
criteria for finished products; although as data generated for SPC accumulate over time, they
may help define realistic finished-product criteria that reflect wholesomeness, safety, process
control and sanitary conditions. Finished-product testing does have a role for verification that
food is manufactured under sanitary conditions with processes that are under control.

As used herein, finished-products refer broadly to food products or ingredients that have
completed a manufacturing process by a supplier. It does not necessarily imply a RTE product.
For example, beef trim may be considered a finished product from the perspective of a
slaughter plant supplying trim to a customer (e.g., a producer of ground beef). Consequently, a
finished product of one process may be an input of another.

In order to ensure the integrity of its food supply, DOD should assess a supplier’s product as the
output of a process that should be under control and delivers wholesome and safe product. This
assessment is achieved through reviewing data supporting the supplier’s food safety plan,
supplier microbiological test data, surveillance of food products at receiving or in distribution,
monitoring of process control at the supplier, and supplier audits, among other activities. In
what follows, the elements of process control are reviewed, and guidelines are given for
statistically-based activities of surveillance and process control monitoring that help ensure
process control, sanitary conditions and high-quality finished products. It is important to
understand that assessing process control can take many forms including measurement and
documentation of critical processing parameters such as time, temperature and pressure,
documentation of employee compliance to personnel requirements, verification and monitoring
programs for SOPs and SSOPs, and evaluation of microbiological, chemical and physical
characteristics of food before, during and after processing.

Process Control

In simple terms as it relates to food manufacturing, storage and distribution systems, process
control can be defined as maintaining the output of a specific process within a desired range.
Control of a process (or management of a process in general) requires accomplishment of six
basic steps:

14
 1. The output of the process must be sampled and quantified on key attributes. Even
limited information (e.g., above or below target) can be used to establish control, if the
sampling rate is high enough. The higher the information content of the measurement
(e.g., enumeration vs. presence/absence), generally the lower the minimum required
sampling rate for control.
2. There must be predefined relevant process control performance limits and targets
traceable to the basic requirements for acceptable outputs (e.g., specifications) and the
history of the process.
3. The actual sample output results must be compared to the relevant process control
limits.
4. There must be a predetermined plan of action (POA, such as a corrective action plan)
based on the size and frequency of deviation from relevant limits. This POA should
include the conditions under which ‘take no action’ is the proper response to a deviation
from control limits. For example, a typical set of POA choices might be: take no action,
move to tightened inspection with increased sampling frequency or sample size, conduct
a pre-determined internal or external audit of the process that is typical for out-of-control
variability, or identify an assignable cause through root-cause analysis and take
corrective and preventive actions. The corrective actions specified must be validated to
ensure they do help to prevent future deviations.
5. The proper action must be decided upon based on the observed deviation.
6. The proper action must be promptly taken to adjust the process. Failure to be prompt is
equivalent to lowering sampling frequency and reduces the ability to control the process.

• Failure to execute any of these steps will obstruct control of the process.

Statistical Process Control Limits

A process is considered under statistical control when its output varies as expected within a
standard operating range (SOR) of variation (Appendix B). This refers to common cause
variation and represents the random variation inherent in a process. When a process becomes
out-of-control, its average shifts, variation increases beyond the SOR, or both. This loss of
control is typically is due to the introduction of a disturbance generated by an assignable cause.

SPC limits bracket the SOR, and indicate the boundary between controlled and out-of-control
operations. The SPC limits may be supplemented by additional statistical rules, such as run
tests (i.e., a rule defining loss of control based on a run of sequential observations, such as
seven measurements over the center line).

SPC limits typically are determined tin one of three ways:

1. Theoretically, from careful scientific analysis of the underlying process;

2. Nonparametrically, from quantiles of the empirical distribution function (EDF), derived
from historical data; or
3. Parametrically, from quantiles of an assumed model distribution (e.g., lognormal) whose
parameters (e.g., mean and standard deviation) are estimated from historical data.

15
The first way is difficult to carry out successfully, particularly for microbiological data. The third
method is typical for non-microbiological applications. However, all three may be useful options
for establishing SPC limits in various settings.

There is a trade-off involved in the choice of the quantiles used to establish the SOR. If the
upper control limit (UCL) is too low (or the lower control limit, LCL, is too high), the
corresponding false alarm rate (FAR) will be too high, and will monopolize resources in
performing corrective actions and searching for assignable causes when actually the process is
under statistical control. For example, if the UCL is chosen at the 90th percentile, then 10% of
testing can be expected to result in false alarms. If the chosen percentile is too high, the FAR
will be too low, and the process may drift out of control too far before it is discovered, or the
sampling rate would need to be increased to counteract this effect. Similar arguments apply to
the LCL used, if any.

Typical quantiles used for the upper control limit in SPC are 95%, 99%, 99.7% or 99.9%.
Choice of the quantile is related to FAR, production lots defined in part by time (e.g., hours,
days, or months), and the amount of resources budgeted for dealing with exceptions. Absent
other information, a reasonable rule of thumb might be to use 95% or 99% limits if the sampling
rate is low (e.g., weekly), so there is no more than one or two expected false alarms per year of
production; otherwise it is conventional to use 99.7% or 99.9% limits.

It is important to note that there is a difference between a process being in statistical control and
meeting specifications. A process is considered under statistical control if it is stable over time
and the observed variation is due to common, chance causes inherent to the process (e.g.,
background noise due to normal variation in ambient temperature and humidity) and there is no
between-lot variation. A food manufacturing process being under statistical process control
does not imply its capability with respect to meeting microbiological specifications. The ideal
situation is when a process is both under statistical control and is capable of manufacturing
products that meet specifications. However, a process can be in statistical control and not
capable of satisfying specifications. For example, the process consistently generates
substandard product. Alternatively, a process can be out of statistical control but capable of
satisfying specifications. For example, the process is designed to be robust in regard to
deviations from the norm, such that it meets specifications despite high variability. Given
seasonal and other sources of variability beyond a supplier’s control, the latter situation may be
particularly relevant to food production processes.

Process Capability

Observations that fall within the SPC limits indicate the SOR of production at a facility that is
under control. They indicate the typical range of results on product (in-process or finished
product samples) produced when the process is under control. Specification limits are different
in that they indicate the range of results that indicate company or customer requirements.

The degree by which the SPC limits fall within the specification limits reflects the process
capability to meet specifications when the process is in control. If the process UCL exceeds the

16
upper specification limit (USL) or the LCL is less than the lower specification limit (LSL), a
fraction of the product produced under normal conditions will not meet the specification, even
though the process is in control.

Process capability is traditionally quantified by a Process Capability Index (Cp, Appendix B).
Typically a recommendation for a new process is Cp = 1.45, or for an established process Cp =
1.25. Equivalent nonparametric rules would be that the USL corresponds to the 99.999
percentile for a new process or the 99.99 percentile for an established process. In both
instances, the USL is higher than the UCL.

SPC Monitoring via Microbiological Testing

SPC monitoring is meant to verify that a supplier’s process of production is operating in

statistical control (or in terms of previous discussions, there is control of the production
process), and therefore is expected to meet microbiological limits where they have relevance in
relation to the process control limits. SPC monitoring requires testing at a frequency that makes
the data valuable for assessment of stability and capability.

Microbiological testing presents some unique features not present in other applications where
SPC is used. Unless a chemical or physical surrogate variable is used, microbiological testing
typically results in a discrete count, not a continuous result. The count may be 0 or 1 (i.e.,
presence/absence testing) or a plate count, or the result of a sequence of serial dilutions. A
zero count represents a concentration below the limit of quantification or detection (e.g., <10/ml
or negative in 325 g) for the particular method and test portion size involved.

Because of the discrete count nature of microbiological testing, test results are governed
typically by one or more of three distributions:

1. Low prevalence (presence/absence) modeled by the binomial or Poisson distribution;

2. Single dilution plate counts, modeled by the Poisson distribution; and
3. Multiple dilution or large plate counts, governed by the lognormal distribution.

Examples of control charts (that illustrate statistical analysis of microbiological test results)
based on DOD data are provided in Appendices C, D, E, F, G and H. In addition, other
distributions that characterize microbiological populations include the Poisson lognormal
distribution. This distribution is a generalization of the Poisson that assumes that the mean
concentration varies log-normally rather than remaining constant throughout the product.
Furthermore, the combination of low prevalence and a range of concentrations when the analyte
is detected results in a zero-inflated distribution that complicates analysis. Zero-inflated refers
to a higher frequency of zero counts than expected under a parametric distribution. For
example, if the microbiological counts in a product follow a simple Poisson distribution with a
mean concentration of 0.04 CFU/g, zero counts in 25 g portions are expected with a frequency
of 37%. If a higher frequency of zero counts is observed, the distribution may be a
heterogeneous mixture in which the microorganism is completely absent from some proportion

17
of the product and present and Poisson-distributed in the remainder. The result would be a
zero-inflated Poisson distribution.

Considerations for Finished-Product Testing

The microbiological limits provided in this report for DOD are useful to establish process control
and sanitary conditions. If suppliers or DOD test finished products, the results may be useful in
assessing the microbiological quality of the product. However, to determine finished-product
acceptability, additional samples may be required (n>1), a three-class plan may be more
appropriate, and microbiological criteria for a food category (and not provided in this report)
shall be required. Considerations for finished-product testing are discussed herein to provide
insights and guidance as the suppliers without a documented and functioning food safety plan
move from establishing process control and sanitary conditions using microbiological limits to
collaborating with DOD to implement microbiological criteria for product acceptance.

Determining the beginning and endpoint of a clearly defined product lot, and delineating it
microbiologically from other lots is critical. A product lot may be defined using a number of
criteria, such as:

• The food manufactured between defined activities (e.g., clean-up to clean-up);

• The food manufactured within a period of time (e.g., day, week, or month); or
• A defined quantity of manufactured food.

The process of defining lots involves thoughtful balancing of various (and sometimes
competing) factors such as sampling costs, the likelihood that a lot is rejected by a customer,
and the cost of lot rejection. The International Organization for Standardization (ISO) observes
that from the point of view of the cost of sampling inspection, there is an advantage in large lots,
provided the same frequency distribution is maintained as lot size increases (12). However,
there are a number of reasons for limiting the lot size including: large lots might result in
inclusion of widely varying quality (i.e., heterogeneity of assignable causes), storage and
handling might preclude the formation of large lots, and the economic consequences of rejecting
or recalling large lots might be unacceptably large. In process control, therefore, there are
tradeoffs between the increased resolution of frequent testing (e.g., every shift or daily) and the
costs of sampling and laboratory analysis. While general rules are available for lot size,
frequency of lot sampling, and number of samples per lot, a sampling scheme can be devised to
optimize control subject to cost constraints (14) .

Lot definition also has implications for SPC when used for assessing the acceptability of a lot.
For purposes of SPC, an important consideration is that a lot is produced under reasonably
constant conditions so that a lot is a homogeneous volume of contemporaneous production.
Statistically, a volume of production is considered homogenous relative to a given characteristic
(e.g., concentration of the microorganism) if the characteristic follows the same probability
distribution throughout the volume (e.g., lognormal with fixed mean µ and fixed standard
deviation σ). It does not mean that the characteristic is the same throughout the volume (2).

18
That is, the conditions result in a homogenous frequency distribution that may or may not
produce a spatially uniform distribution within a lot.

A homogenous distribution is often interpreted in food microbiology to indicate a homogenized

product with the same mean concentration throughout (i.e., a Poisson spatial distribution);
however, statistically a consistent or homogeneous frequency distribution can result in spatial
heterogeneity within a lot (ILSI-Europe, 2010). For example, if two days of production have the
same mean concentration (µ1 = µ2) but substantially different variability (σ1 ≠ σ2), then the two
production lots are not characterized by a homogenous (the same) frequency distribution. This
concept is important because assignable causes that might occur between lots ought to be
different from those that occur within lots. As such, an important aim of SPC methods is to
evaluate between-lot variance compared to within-lot variance.

Selection of the appropriate microorganisms when deploying SPC is critical. Typically the best
organisms are either a) those that are predictably present within the sample matrix at some
quantifiable concentration; or b) those that are neither exceptionally rare (i.e., approaching 0%
prevalence) nor ubiquitous (i.e., approaching 100% prevalence) when detected with qualitative
assays. In some instances, microorganisms present at low prevalence may be useful for SPC
(Appendices D and E).

Sampling Frequency

Product samples may be taken systematically based on units of production or by duration of

production, e.g., by shift, day, week, month or quarter. Indicators of process control are best
obtained by more frequent sampling. As a general rule, sampling frequency should be high
enough to detect the presence of expected assignable causes within the first 10% of their
persistence time. SPC cannot function for process control if the sampling frequency is less than
twice during the assignable cause persistence time. Cost is associated with sampling and
testing, so considerable economic force is exerted to drive the frequency to the minimum
possible rate. However, disruptions that cause a loss of process control often persist for only a
finite time, and not much is learned if they are either not detected when happening, or are
detected too late for corrective action.

Although DOD currently conducts some sampling and testing during screening, auditing, and
surveillance, to develop fully the use of SPC, suppliers would need to do sampling and testing at
a frequency described above. As such, the supplier needs to have access to a competent
laboratory, have the technical ability to collect the appropriate samples, have the financial
resources to pay for the program, and have the knowledge of SPC necessary to interpret and
use the data.

Even under ideal conditions, a large quantity of data may be required before stable, precise
estimates are obtained for process parameters (e.g., mean, variance, prevalence). Shewhart
(16) cautioned that assignable causes of variation are almost always present in the early stages
of process control and that a long data sequence (e.g., a total sample size not less than 1000)

19
may be required to demonstrate that a process is in statistical control . However, acquiring
additional data is subject to diminishing returns, and requiring a very long sequence of data may
not be economically or technically feasible under operational conditions (Appendix I). For
example, the only suppliers of perishable foodstuffs required to support DOD operations in
austere areas may be small facilities without long production histories. Also, attainment of
process control is often a gradual, stepwise process. Therefore, in practice, a pragmatic
compromise is often warranted. As a general rule, Shewhart suggested a data sequence of not
less than twenty five samples of size four (e.g., sampling 25 lots at 4 samples per lot for a total
of 100 samples) is the minimum requirement for concluding that a process is in a state of
statistical control (16) . Similarly, the ICMSF (ICMSF, 2011) recommends that a minimum of 30
lots should be examined; but cautions that it may be necessary to conduct an initial process
control study for longer periods or in phases.

Sampling Plans for Screening and Auditing Suppliers

Screening of New Suppliers

The first step in screening a new supplier is to have the supplier conduct a self-audit against
DOD supplier expectations (currently a pre-audit checklist). With the self-audit, or upon an
initial visit, DOD should request that the supplier provide microbiological data that demonstrates
that their production process is under control and occurs under sanitary conditions. The
supplier could be asked for verification data supporting its food safety plan, or for those
suppliers without a documented and functioning food safety plan, SPC charts that help to
demonstrate their level of control (although it is unlikely such suppliers will have these charts
and will need to be provided direction, such as that given in this report). If either type of supplier
does not have the information, DOD should consider whether the supplier is willing to begin the
process of demonstrating that their process is under control and is operating under sanitary
conditions by collecting verification data to support their food safety plan or by using the
microbiological limits provided herein to support that their process is in control and production is
occurring under sanitary conditions. Suppliers might be accepted under a probationary status.
During the probationary period, finished product testing may be required to assess the
acceptability of the supplier’s product.

For-cause Auditing (Directed Audits)

When a potential problem has been identified (e.g., failure to achieve a microbiological criterion,
prematurely spoiled product, or an outbreak of illnesses associated with consumption of a
product), sampling is frequently required to determine the extent and source of the problem. The
ICMSF (ICMSF, 2002) refers to investigational sampling, which includes sampling for this
objective. While the sampling conducted in the course of for-cause auditing would typically
require more extensive sampling than normal sampling, it differs from tightened inspection in
that there are no conventional sampling plans specifically designed for determining the extent of
a problem and identifying the underlying cause. The success of such sampling depends greatly
on knowledge of the process, product, and microorganism. The process flow diagrams

20
presented in Appendix A should be a useful resource for guiding the sampling conducted during
for-cause auditing.

Surveillance at Point of Sale

DOD performs intermittent point of sale surveillance of finished products at locations such as at
commissaries. The accumulated data are valuable for various purposes such as assessing not
only the suppliers’ products and processes, but also the potential for contamination or abuse
during transportation, and storage and handling practices throughout the supply chain and at
the commissaries themselves. Various sampling plans are appropriate for surveillance
purposes including that sampling and testing being performed currently by DOD. However,
improvements in standardization of sampling plans and associated meta-data (characterization
of the data and the methods used) are warranted.

MICROBIOLOGICAL LIMITS AND CRITERIA

Development of Limits and Criteria

The ICMSF describes the establishment and application of microbiological criteria in

considerable depth in two publications, Microorganisms in Foods 7 (ICMSF, 2002) and
Microorganisms in Foods 8, Use of Data for Assessing Process Control and Product
Acceptance (ICMSF, 2011). The details described in these references will not be repeated
here; however, the following discussion relates to how the development of criteria relates to the
specific charges posed by DOD.

ICMSF defines three types of microbiological criteria: standards, specifications, and guidelines.
Standards are mandatory criteria incorporated into a law or ordinance (normally pathogen
oriented). Specifications are part of a purchasing agreement between a buyer and a supplier of
a food and may be advisory or mandatory according to use. Guidelines are advisory criteria
used to inform food operators and others of the microbiological content that can be expected in
a food when best practices are applied (ICMSF, 2002).

Regardless of where food products are manufactured in the world, the finished-product
microbiological criteria indicating safe, wholesome products for DOD would be the same. This
presents challenges for DOD because manufacturers around the world do not have the same
facility design requirements and standards, processing equipment and technology, sampling
and testing programs, regulatory requirements, preventive and pre-requisite programs,
oversight and auditing, customer expectations and food safety culture. Further complicating the
development of microbiological criteria for finished products purchased by DOD is the large
number and variety of products and suppliers.

In contrast to establishing appropriate microbiological criteria, if there was interest or a need to

truly reflect how microorganisms are related to process capability for each manufactured
product, data would need to be captured over many lots of production at each manufacturing
site to determine what levels of organisms measured at various points of production reflect

21
sanitary and insanitary conditions or lack of process control. This requires a site-specific
assessment for each product individually to gain an accurate assessment of these data; this
resource-intensive effort is not commonly done at manufacturing locations. Setting uniform
microbiological limits for process control, while purposeful, may not accurately reflect individual
processes and products within that general category. Thus, the suggested microbiological limits
(Appendix J) described herein should be considered guidance to DOD representing a
provisional starting point for developing empirically based microbiological data and a basis for
discussion of DOD expectations with suppliers that do not have documented and functional food
safety plans.

Microbiological analyses and comparison of the test results to microbiological limits, for the
purpose defined herein, or finished product microbiological criteria, yet to be fully defined by
DOD for the products they purchase, may be used to verify that a supplier’s control programs
for controlling microbiological contamination are effectively designed and implemented. When
there is evidence that the supplier’s controls are poorly designed or implemented, it may be
prudent to increase the frequency of microbiological testing; this testing may include testing
against microbiological limits provided herein, finished product testing, environmental
monitoring, and sanitation effectiveness monitoring. It seems reasonable to expect that
appropriate food safety and quality programs are more likely under the following conditions:
• the food safety regulatory program in the supplier’s country has been deemed equivalent
to its U.S. counterpart,
• the supplier has developed, implemented, and documented appropriate preventive and
pre-requisite food safety programs such as ensuring a safe and properly plumbed water
supply, GAPs, GMPs, and SSOPs,
• the supplier has developed, implemented, and documented a process-oriented risk-
based preventive food safety plan, including a HACCP system, that substantially
complies with risk-based preventive controls regulations authorized by FSMA, and
• the supplier’s food safety system has achieved third-party certification against standards
fulfilling the requirements such as those specified in the Global Food Safety Initiative
Guidance Document.

Pathogens Important to Public Health

It is somewhat easier to establish microbiological limits, and specifications, for certain

pathogens because whenever there is a likelihood of pathogens being present, sampling and
testing plans can be designed to require the absence of the pathogen at a given stringency of
testing, i.e., quantitative values need not be established.

The Committee considered where pathogens are reasonably likely to occur for each category of
food. The pathogens may have resulted from process control failures (e.g., contaminated raw
materials and ingredients, inadequate processing conditions and insufficient interventions,
failures in pre-requisite programs and preventive programs) or insanitary conditions (e.g., failure
in cleaning and sanitation, inferior facility and equipment design, poor personal hygiene).
Combining these analyses with summaries on the causative agents of foodborne outbreaks
allowed the Committee to prepare the microbiological limits for pathogens for the major food
categories that may reflect loss of process control or insanitary conditions (3).

22
Indicators that Reflect Loss of Process Control or Insanitary Conditions

Indicator organisms typically used to reflect process control or insanitary conditions include
those familiar to food manufacturers, e.g., APC, coliforms, E. coli, Enterobacteriaceae, S.
aureus, pseudomonads, and yeasts and molds. The levels of indicator organisms which
indicate loss of process control or insanitary conditions during processing are dependent upon
factors such as the cleaning and sanitation procedures and products, the types of processes
used, the sanitary design of equipment and the facility, and the food being manufactured.

One of the more difficult microbiological limits to establish to reflect loss of process control or
insanitary conditions is that for Gram-negative bacteria, whether coliforms, fecal coliforms,
Enterobacteriaceae or E. coli. Kornacki and others (13) provide an historical evaluation of these
criteria for foods and their utility based on current knowledge. None of these Gram-negative
bacteria accurately and consistently reflect fecal contamination of raw and processed foods nor
are they useful or reliable as index organisms predicting the presence of pathogens. These
criteria may be useful indicators of insanitary conditions and loss of process control; however
these uses are dependent upon many factors such as the type of food, the extent and type of
processing, the relationship between bacterial numbers and food quality, and the length of time
between production and sampling and testing. Kornacki et al. also reviewed the testing
methods and the many variables that affect the accuracy and utility of the results. For these
reasons, whichever indicator microorganisms are used, they are generally considered
guidelines for use. Based on this current review, in general, the indicator microorganisms of
most value would be Enterobacteriaceae, followed by E. coli, coliforms and fecal coliforms.

DOD is at a disadvantage without data from suppliers defining their normal cleaning and
sanitation practices, and their sanitation effectiveness monitoring program, as well as process
control data measured by manufacturers throughout their production runs. Setting arbitrary
quantitative limits for indicator organisms for a category of food products is guidance at best and
may or may not be reflective of insanitary conditions or lack of process control. For this reason,
the microbiological limits provided herein to DOD should be considered guidelines and a starting
point for suppliers and DOD to evaluate the process controls and sanitary conditions under
which the products were manufactured. The process flow diagrams indicating where bacterial
numbers may increase during manufacturing provide some guidance to DOD on questions to
ask of suppliers regarding where samples are taken, or process control measurements made,
during processing and what corrective actions might be taken based on the results of such
sampling and testing.

Comments on Microbiological Limits for Specific Food Categories

One of the limitations of microbiological limits as indicators of process control or insanitary

conditions is the balance of statistical validity with practicality (Appendices K, L and M).
Microbiological limits and sampling schemes are often dictated by common practice and are not
based on statistical design. The guidance below is based on review of the available literature,
expert opinion, and industry practice. Consequently, the limits discussed below should be

23
considered provisional starting points toward more formally designed microbiological limits for
process control that are updated and revised over time as additional data are acquired.

The tables (Appendix J) presented in this document are intended to provide guidance on
microbiological limits, proposed primarily for use by DOD for suppliers without documented and
functioning food safety plans, that reflect effective process controls and sanitary conditions used
to produce food products using good quality ingredients, validated pathogen intervention
strategies and lethality steps, GMPs and GAPs. Microbiological populations in raw commodities
are expected to be higher and more diverse than those in foods produced using a validated
lethality process. The limits identified are on a “per gram” or “per ml” basis and typically assume
a 25 g analytical unit unless otherwise described.

The microbiological limits are intended to help identify when a process is not in control so the
manufacturer can investigate causes and implement corrective actions. The limits reported for
indicator organism testing are not lot acceptance criteria. In some cases, the action to be taken
after exceeding the limit may be to increase sampling to determine the source of contamination
or to test for pathogens or other indicators of insanitary conditions. In cases where any
microorganism or class of indicator organisms exceed regulatory limits, then the lot should be
evaluated appropriately, and typically destroyed or diverted for reconditioning if appropriate. As
an example, the FDA Dairy Compliance Policy Guide 527.300 (23) considers cheese made with
pasteurized milk to be adulterated if the cheese contains 104 CFU/g S. aureus or B. cereus or
100 CFU/g E. coli; these lots should be rejected and additional investigation conducted. If
enterotoxins produced by S. aureus or B. cereus are detected, the product also should be
destroyed.

Enrichments (such as for pathogens in environmental sponge samples) may be performed on

composite samples. However, with compositing, if samples are pulled from multiple locations or
over the course of producing several lots of finished products, a positive result for the
enrichment would implicate all locations and the lots manufactured during the sampling period.
In contrast, enumeration data should be generated from a single sample analytical unit; pooling
samples might dilute unacceptable or marginal populations with samples having low populations
and thereby provide misleading results.

Assaying for APC to assess process control and sanitary conditions may be relevant for some
RTE foods but not others. APC values used to assess process control and sanitary conditions
during production should be low in RTE foods in which all components of the food have
received a lethality step (e.g., pasteurization, cooking, roasting). When RTE foods contain
some components that have received a lethality step, but then were further handled (e.g.,
sliced, assembled or mixed) before preparation of the final food product, APC levels would be
expected to be moderately higher. In contrast, using APC to assess process control and
sanitary conditions during the production of foods such as fresh fruits and vegetables,
fermented or cultured foods and foods incorporating these, has little value as these foods would
have an inherently high APC because of the normal microbiota present.

24
The presence of E. coli in RTE foods is undesirable because it represents poor hygienic
(insanitary) conditions or inadequate heat treatment (lack of process control). Thus, E. coli
should not be detected in RTE foods; generally, when microbiological specifications are
established, a microbiological limit of <10/g or <3 MPN/g (the limit of detection of usual test
methods) is typical for this microorganism. Levels exceeding 100/g are typically interpreted as a
level of contamination that may be associated with the introduction of pathogens or conditions
that allowed pathogen survival.

The Committee concurs with the common practices for environmental monitoring, i.e., testing for
Listeria spp. in wet, RTE-food processing environments, particularly for foods that support
growth of Listeria, and for Salmonella in dry, RTE-food processing environments. Salmonella
monitoring in warm, wet, RTE-food processing environments also may be appropriate
depending upon the product and facility. If product contact surfaces (Zone 1) are tested,
finished product should be held until results are confirmed negative; if testing demonstrates that
the product contact surfaces are positive for the pathogen, investigational testing in finished
product and corrective action is indicated. As of 2014, the U.S. maintains a standard of non-
detectable L. monocytogenes for all RTE food products. Other countries may allow up to 100
CFU/g for L. monocytogenes in RTE foods that do not support growth (e.g., frozen foods, those
with pH <4.4, water activity (aw) < 0.92, or pH < 5 and aw < 0.94) (5, 6).

All dairy food categories listed below are presumed to be made with pasteurized milk to
eliminate common vegetative bacterial pathogens. Therefore, the presence of any pathogens
when testing for process control or sanitary conditions represents post-process contamination.
Salmonella, E. coli O157:H7 and L. monocytogenes are considered adulterants in RTE dairy
products. In the U.S., these dairy products are either regulated under the PMO Pasteurized
Milk Ordinance (24) or microbiological standards are identified in the Dairy Compliance
Guidelines (23). Other resources for microbiological specifications and guidelines include the
Compendium of Methods for the Microbiological Examination of Foods (Milk and Milk Products
(1) and Standard Methods for the Examination of Dairy Products (29) . Alkaline phosphatase
level in pasteurized fluid bovine milk is limited to less than 2.0 micrograms phenol equivalent per
gram in one or more subsamples whereas cheeses may have higher limits. Actionable limits for
S. aureus and B. cereus are set to 104 CFU/g whereas limits for E. coli or coliforms are product
specific.

The general recommendation for DOD procurement of any beef, pork or poultry product,
whether raw or RTE, is to identify an establishment in the country which is authorized to ship
that product to the U.S. and procure product from that establishment. This will ensure the
establishment meets current FSIS performance standards and/or regulatory requirements. If
such an establishment cannot be identified, the testing recommended in Appendix J may be
used to determine the level of process control and sanitary conditions for establishments not
currently authorized to ship the product to the U.S.

Microbiological testing of finished products that receive a lethality step, such as baking or
cooking, may not be a good indicator of improper storage temperatures and hold times (process

25
controls) of ingredients or blends before the lethality step (such as extended runs between clean
up). Certain ingredients or foods may support microbiological growth and production of heat
stable toxins, such as those produced by S. aureus or B. cereus. Thermal treatments may
inactivate the vegetative cells in the final product but the toxin may remain. As a result, the
process must have validated microbiological control steps throughout the production to minimize
the risk of toxin being present in the finished product.

Routine and Non-routine Testing

In setting the microbiological limits to be used by suppliers that do not have documented and
functioning food safety plans, including HACCP systems, the Committee defined the
recommended testing frequency as routine and non-routine. Specific time intervals cannot be
set for each indicator organism, class of indicator organisms, pathogen, environmental
monitoring, or in some instances, chemical hazard (e.g., mycotoxin). The frequency of routine
and non-routine testing will be dependent upon numerous factors such as the production
process, the product being produced, the sanitary design of the facility and the equipment used
at the facility, the historical data generated by the supplier, the organism or class of indicator
organisms, and the investigative reason for testing. General guidance on the definition of these
frequencies is as follows.

Routine testing is defined as testing done at pre-determined intervals at sufficient frequency to

establish process control or sanitary conditions. The sampling interval may be on a physical lot
basis (e.g., 2,000 lb. combos for ground beef) or temporal basis (e.g., per shift, daily, weekly,
monthly). Non-routine testing can be investigational, for verification, validation, surveillance, or
for qualifying suppliers. Non-routine testing is less frequent and can be based on time intervals
(e.g., weekly, monthly, quarterly) or based on other indicators of lack of process control or
insanitary conditions. For example, if routine testing shows that samples of a pasteurized egg
product exceed limits for E. coli, testing for Salmonella may be appropriate. If routine testing of
a RTE food that can support growth of L. monocytogenes indicate contamination of the food
with Listeria spp., additional testing for L. monocytogenes may be appropriate. When a supplier
is manufacturing multiple- component foods (e.g., frozen desserts with inclusions, deli salads,
sandwiches, entrees), routine or investigational sampling and testing may be focused on those
components with the highest microbiological risk.

Plan of Action if Limits are Exceeded

The microbiological limits provided in Appendix J are useful to assess process control and
insanitary conditions. The action taken by a supplier if indicator organisms in samples taken at
the supplier location exceed the specified limits should be to investigate the cause of the high
counts, implement corrective and preventive actions, and reevaluate the effectiveness of the
actions after implementation. In the cases of a pathogen detected when there has been no
additional lethality step, an evaluation of the finished product associated with the sample tested
should occur to determine if the product should be rejected or, if appropriate, reworked or
diverted for processing that will inactivate the pathogen. Products contaminated with heat-
stable toxins typically will be destroyed as reconditioning likely will not eliminate the hazard.

26
If levels of indicator bacteria in samples assayed during distribution or at the point of sale
exceed the limits provided in Appendix J, a more thorough investigation should be taken by
DOD and the supplier to identify the cause of the higher counts. The investigation should note if
the food was at the end of the marked shelf-life, is considered perishable, if the packaging was
intact, and if the chill-chain was maintained during storage and distribution. Growth of spoilage
microbes is expected to occur during extended storage of perishable items. The higher counts
may have resulted from normal growth of spoilage microorganisms or temperature abuse rather
than the lack of process control or sanitary conditions during manufacture.

Commodity Specific Comments on Microbiological Limits

Beverages – Bottled water (artisan, mineral, purified, sparkling, spring) – Appendix A, Flow
Diagram A.1, Appendix J, Table J.1

The Committee recommends routine coliform testing for bottled water and ice to assess process
control and sanitary conditions. In countries where additional microbiological regulations apply,
testing for those organisms may be done periodically. A 2013 WHO Draft Report on regulations
and standards for drinking water quality recommends routine testing for E. coli or thermotolerant
coliforms to provide evidence that these microorganisms are undetectable in a 100-ml sample
(WHO, 2013). Other indicators also were reviewed in the WHO Draft Report and the following
recommendations were made. The presence of total coliforms immediately after treatment
indicates inadequate treatment. C. perfringens (undetectable in 100 ml) can be used an
indicator of the effectiveness of filtration process to eliminate enteric viruses or protozoan
oocysts (WHO, 2013). Enterococci (undetectable in 100 ml) may survive longer than E coli and
can be used as an indicator instead of E. coli. Total heterotrophic bacteria (limit of 100 CFU/ml
at 22 or 20 CFU/ml at 37°C) can be used for operational monitoring of treatment and
disinfection and assessing cleanliness of the distribution system. Pseudomonas aeruginosa,
parasites and enteric viruses were not considered in the WHO report; although they may be
required by individual country regulations.

Beverages – Ice, packaged – Appendix A, Flow Diagram A.2, Appendix J, Table J.2

Microbiological testing and limits will be similar to those for bottled water. In countries where
additional microbiological regulations apply, periodic testing for the organisms listed in those
regulations is appropriate.

Beverages – Juices and drinks, pasteurized, refrigerated – Appendix A, Flow Diagram A.3,
Appendix J, Table J.3

The Committee recommends routine coliform testing for process control purposes. Fruit juices
in the U.S. are subject to FDA regulations mandating HACCP and achievement of lethality
against pathogens of significance (E. coli O157:H7, Salmonella spp.); thus, periodic testing for
pathogens may be indicated (28) . This category also includes low acid drinks such as bottled
coffees, teas, and vegetable juices. For low-acid juices and drinks, the food safety plan should
address the control of pathogenic sporeformers, such as C. botulinum. For products that

27
support the growth of pathogenic sporeformers and where cold-chain management cannot be
guaranteed, alternative safety measures could be the inclusion of ingredients that inhibit growth
(e.g., blending with acidic juice to reduce pH) or alternative processing such as ultra-high
temperature processing to destroy spores. High levels of patulin can be produced in decaying or
moldy apples, and thermal processing does not destroy the mycotoxin. Therefore, apple juice
products should be tested for patulin (21) (U. S. Department of Health and Human Services,
2005)

Beverages – Shelf stable – Appendix A, Flow Diagram A.4, Appendix J, Table J.4

Process control of shelf-stable (commercially sterile) beverages is dependent upon control of

formulation and verification and monitoring of CCPs rather than routine microbiological testing.
If inspection observes indications of spoilage such as bulging containers, pH changes, and off-
odors then further investigation should be done by DOD and the supplier. Methods for
investigating failures in processing for commercial sterility are given in the Compendium of
Methods for the Microbiological Examination of Foods (4). Shelf-stable apple juice products
should be tested for patulin for the reasons described above for refrigerated juices (21) (U. S.
Department of Health and Human Services, 2005).

Dairy – Butter, margarine – Appendix A, Flow Diagram A.5, Appendix J, Table J.5

Although whipped butter held under unrefrigerated conditions has been associated with
outbreaks of S. aureus intoxication, the low moisture and high salt content, or lactic acid levels
of many of these products, generally preclude microbiological growth. However, routine
monitoring of sanitation and process control using indicators such as coliforms should be done.
Products containing added seasonings, herbs, or spices may have additional testing
requirements as the inclusion of unsafe adjunct ingredients has been linked to foodborne
illness. Testing for S. aureus, Enterobacteriaceae, and yeast and molds is useful under special
circumstances, such as the investigation of out-of-specification results. Due to listeriosis
outbreaks linked to contaminated butter, routine environmental testing of Zone 2 and 3 surfaces
for Listeria spp. should be done. Although not routinely tested, if Zone 1 environmental samples
are found to be positive for Listeria spp., investigational testing of finished product should be
undertaken.

Dairy – Cheese (hard) – Appendix A, Flow Diagram A.6, Appendix J, Table J.6

Although reported cases of foodborne illness have been linked to foods in this category,
microbiological safety issues in hard cheeses made with pasteurized milk and active starter
cultures are extremely rare. The presence of active cultures in these products makes the use of
routine microbiological testing for APC impractical as a tool for evaluation of process controls
and sanitary conditions. In contrast, routine testing for coliforms as an indication of sanitary
conditions should be conducted. Testing for S. aureus or E. coli is useful under special
circumstances such as validation, verification and investigation when production has occurred
without adequate process control. Finally, routine environmental testing of the food production

28
environment for the presence of Listeria spp. is recommended as a verification step for
sanitation programs.

Dairy – Cheese (soft, semi-soft, surface ripened) – Appendix A, Flow Diagram A.7, Appendix J,
Table J.7

This category represents a broad range of cheeses. Routine environmental monitoring for
Listeria spp. in the environment and coliforms in finished product should occur for all products in
this category. For products in this category which support the growth of L. monocytogenes and
have been implicated in illness such as soft cheeses with high pH values, in-plant monitoring for
this pathogen may be appropriate (15) . Testing for S. aureus and E. coli may be used when
processing or insanitary conditions indicate a potential increased microbiological risk.

Dairy – Cultured, pH<4.8 – Appendix A, Flow Diagrams A.8a and 8b, Appendix J, Table J.8

Rapid acidification and low final pH of these products precludes growth of bacterial pathogens.
The presence of active cultures in cultured dairy products make the use of most routine
microbiological testing impractical as a tool for evaluation of process controls and sanitary
conditions. Routine testing by suppliers for coliforms is recommended to assure compliance
with pertinent U.S. regulations and guidance (24). Non-routine testing for S. aureus is advisable
under limited conditions such as evaluating the impact of a slow fermentation processes. Mold
and yeast testing may be applicable when producing cultured products without mold inhibitors or
when products contain inclusions such as fruit puree that are known to carry spores. Finally,
routine environmental testing of the food production environment for the presence of Listeria
spp. is recommended as a verification step for sanitation programs.

Dairy – Cultured, pH>4.8 and < 5.4 – Appendix A, Flow Diagram A.9, Appendix J, Table J.9

The active starter culture and acid content present in these fermented products reduces the
growth rate of bacterial pathogens; but because the pH is higher than the aforementioned
cultured products with pH <4.8, prevention of post-pasteurization contamination is more critical.
The presence of active cultures in these products makes the use of most routine microbiological
testing impractical as a tool for evaluation of process controls or insanitary conditions.
However, routine testing by suppliers for coliforms is recommended to assure compliance with
pertinent US regulations and guidance (24) and routine environmental testing of the food
production environment for the presence of Listeria spp. is recommended as a verification step
for sanitation programs. Although typically not done, if Zone 1 environmental samples are
positive for Listeria spp., finished product testing for L. monocytogenes should occur. Testing for
S. aureus, psychrotrophic microorganisms, yeasts, and molds is useful under the special
circumstances described above for Dairy – Cultured, pH<4.8, when investigating results
exceeding microbiological limits, or during validation and verification efforts.

Dairy – Dried products (does not include dairy ingredients used to make infant formula) –
Appendix A, Flow Diagram A.10, Appendix J, Table J.10

29
The low moisture content of dried dairy product precludes microbiological growth. However,
routine monitoring of sanitation using coliforms and APC should occur. Furthermore, routine
testing for Salmonella by suppliers should occur as these products have been implicated in
cases of salmonellosis. Non-routine testing for S. aureus and B. cereus should be done under
special circumstances such as during investigation of possible mishandling prior to drying,
validation or verification efforts, or an investigation done in response to results indicative of
process failures or insanitary conditions.

Dairy – Frozen desserts, Appendix A, Flow Diagram A.11, Appendix J, Table J.11

Dairy ingredients used in a dessert mix are pasteurized and will have low microbiological
counts; frozen storage will control microbiological growth. Routine testing for coliforms by
suppliers should occur to establish process control and monitor sanitation. Although APC can
be used to monitor process control, inclusions, such as nuts, cookie dough and fruits, may result
in higher populations than the base mix. Periodic testing for Salmonella may be indicated under
special circumstances such as when lack of process control is suspected, the supplier is using
inclusions which have been previously associated with outbreaks, or during validation or
verification efforts.

Dairy – Milk and milk products (fluid) – Appendix A, Flow Diagram A.12, Appendix J, Table J.12

Fluid milk in the U.S. is produced under the PMO (24) which provides microbiological limits;
when done, such as when there is a pasteurization issue, alkaline phosphatase must be <2.0
micrograms phenol equivalent per gram as an indicator of adequate pasteurization. Routine
testing of APC and coliforms by suppliers should occur to ensure regulatory compliance, to help
establish process control, and to assist with evaluating sanitary conditions. Routine
environmental monitoring of Zone 2 and 3 surfaces for Listeria spp. is recommended.

Dairy – Processed Cheese – Appendix A, Flow Diagram A.13, Appendix J, Table J.13

This product is manufactured by heating cheese with water, emulsifier and other ingredients to
kill vegetative pathogens; molten cheese may then be hot-filled into loaves or blocks and chilled
and cut into individual slices for use; these cheeses are intended to be stored refrigerated.
Shelf-stable hot-filled cheese spreads or cheese sauces must be formulated for safety to inhibit
Clostridium botulinum. Cooling process cheese on casting belts or chill rolls may involve a
relatively high degree of environmental exposure of the product. The presence of non-
sporeforming microorganisms is indicative of post-process environmental contamination. Low
levels of such contamination are inevitable in these cases. Consequently, process cheese
producing facilities need to have robust environmental sampling and control plans for Listeria
spp. and Salmonella spp. Formulae with low levels of salt in the moisture phase could
potentially allow growth of enterotoxin producing Staphylococcus spp., principally S. aureus;
likely originating from human contact. The presence of generic E. coli on process cheese is
reflective of production in an insanitary environment.

Egg Products – Pasteurized, processed – Appendix A, Flow Diagram A.14, Appendix J, Table
J.14

30
Pasteurized egg products and pasteurized shell eggs receive a lethality treatment during
processing and may be used in dishes which are uncooked or lightly cooked. These products
may be recontaminated during packaging, handling and storage. These products should be
tested by suppliers routinely for S. aureus, coliforms, APC and Salmonella to verify process
control. Periodically, suppliers may test these products for B. cereus and Enterobacteriaceae.
Routine environmental testing for Listeria spp. and Salmonella is useful to evaluate sanitary
conditions. If samples exceed the microbiological limits, further investigation and correction
action should occur. Environmental monitoring of Zone 2 and 3 surfaces for Listeria spp. is
recommended; if Listeria spp. are found, it may lead to testing of Zone 1 surfaces for Listeria
spp. Finished product testing should occur for L. monocytogenes if Listeria spp. are detected
on Zone 1 surfaces (indicative of insanitary conditions) or suspected illnesses are reported.

Egg Products – Shell eggs, raw – Appendix A, Flow Diagram A.15, Appendix J, Table J.15

Raw shell eggs are not pasteurized and are not intended for consumption without an additional
lethality step, such as thorough cooking. Regulations in the U.S. require that high-volume
producers (>50,000 laying hens) test for Salmonella serotype Enteritidis to verify non-detection
of this pathogen in the shell eggs (22). High-volume producers supplying shell eggs to DOD
should test for S. Enteritidis. For other producers, the Committee recommends only periodic or
investigational testing of raw shell eggs and no microbiological limits are provided. Testing for
E. coli, coliforms or Enterobacteriaceae by suppliers may be useful to assess sanitary
conditions or establish process control.

Grain-based Products – RTE, baked items, refrigerated or temperature/time controlled for safety
(TCS) – Appendix A, Flow Diagram A.16, Appendix J, Table J.16

These products are prepared with a lethality step to eliminate pathogens; but the potential of
recontamination during handling and the pH-aw range (that can support microbiological growth
during extended out-of-refrigeration storage) warrants microbiological testing. Routine
monitoring of coliforms by suppliers should assess insanitary conditions (including post-process
contamination). APC testing should not be conducted if the products include ingredients which
are prepared using starter cultures (e.g., cheese, salami).

Grain-based Products – RTE, baked items, shelf stable or non-TCS – Appendix A, Flow
Diagram A.17, Appendix J, Table J.17

When manufacturing these products, the dough or batter goes through a baking step which
provides lethality against pathogens and pathogen growth is unlikely during storage due to
reduced water activity. While routine microbiological testing by suppliers generally is
unnecessary, environmental monitoring and in-process sample testing may be appropriate
under special circumstances that may increase the microbiological risk (e.g., excessive water
due to condensate or roof leaks) or when ingredients are added after the lethality step (e.g.,
dusting of bread surface with flour).

31
Grain-based Products – RTE, cereals – Appendix A, Flow Diagram A.18, Appendix J, Table
J.18

RTE cereals are made from grains that go through a lethality step sufficient to eliminate
pathogens of concern. Mycotoxin surveillance testing should be completed on incoming grains
to ensure the grains meet the individual country’s regulations. These RTE grain-based products
do not support the growth of microorganisms due to the very low aw. Routine microbiological
testing of finished product by suppliers is not recommended; but routine environmental testing
for Salmonella is useful to assess sanitary conditions. Non-routine testing for coliforms,
Enterobacteriaceae, APC and Salmonella by suppliers is appropriate for verification purposes,
qualifying lines, or when events occur during processing that may increase the microbiological
risk (e.g., excessive water due to condensate or roof leaks). If vitamin-containing or other such
solutions are sprayed atop cereals after heat-processing, and depending on the source and
processing of these solutions, sampling and testing of these solutions may be a useful measure
of process control.

Grain-based Products – RTE, cold pressed bars – Appendix A, Flow Diagram A.19, Appendix J,
Table J.19

Cold-pressed bars are made from cooked grains, carbohydrate-based binders, and inclusions
such as fruit, nuts and chocolate. Verification of the microbiological quality of ingredients used
in the cold-pressed bar formula is important since the bars will not receive a validated lethality
step during manufacturing. Recommendations for finished product and environmental testing by
suppliers are the same as for RTE cereals above.

Grain-based Products – Non-RTE, dry, flour-based mixes – Appendix A, Flow Diagram A.20,
Appendix J, Table J.20

These Non-RTE grain-based products harbor a complex and extensive microbiota and routine
microbiological testing by suppliers does not provide useful data to indicate process control and
sanitation (17). Flour is a minimally-processed commodity that is ground and sifted without any
lethality step. These products should receive a lethality step to eliminate pathogens before
consumption.

Grain-based Products – Non-RTE, pasta, dried or refrigerated – Appendix A, Flow Diagram

A.21, Appendix J, Table J.21

Pasta is produced by combining flour and water and sometimes other minor ingredients. The
microbiological profile may be similar to that of flour and routine testing by suppliers is not
particularly useful. However, the manufacturing process must be controlled to minimize
proliferation of naturally occurring microbiota after the introduction of moisture. Non-routine
testing of in-process samples by suppliers may be useful in special circumstances (e.g.,
evaluation of potential growth and enterotoxin production by S. aureus during extended down
time prior to drying or refrigeration). Although most of these products are intended to be cooked
by consumers before consumption, some varieties, such as instant noodles, may be prepared
with limited heating. Cooking of refrigerated pasta filled with meat or cheese may be sufficient to

32
cook the outer pasta, but not sufficient to provide a validated lethality step in the product interior.
Verification testing of raw materials (to support the Certificate of Analysis) and periodic testing of
product by suppliers for Salmonella may be appropriate; and environmental testing for Listeria
spp. or Salmonella should occur to verify sanitary conditions.

Meals and Entrees – Non-RTE, Ready-To-Cook (RTC) meals, includes raw ingredients –
Appendix A, Flow Diagram A.22, Appendix J, Table J.22

This category includes a wide range of multi-component (some raw), frozen or refrigerated food
products which are expected to be cooked by the consumer or food service operation. Routine
testing of these meals is not recommended; however manufacturers should be aware of the
following points. Suppliers should assess the pathogens and indicator organisms associated
with their products and sample and test if there is a reason to do so. Some of these meals and
entrees may be improperly prepared by the consumer using conventional or microwave ovens
and not undergo a validated lethality step. Pathogens of concern may vary depending on the
specific food. For example, meals prepared with cooked rice may pose a greater risk for B.
cereus; E. coli O157:H7 may be of concern for foods including raw, non-intact beef, and poultry
products may contain Salmonella. Histamine testing may be appropriate when scombroid
species are present.

Meals and Entrees – RTE, deli salads, sandwiches, heat-eat meals, sushi – Appendix A, Flow
Diagram A.23, Appendix J, Table J.23

This category includes a wide range of multi-component, short shelf-life, refrigerated food
products. They are expected to have diverse microbiological populations depending on the
ingredients used, may include ingredients which are raw, such as fresh produce, and are
frequently subjected to multiple handling steps which can introduce contamination. Routine
testing by suppliers of in-process or finished products for E. coli and environmental testing for
Listeria spp. and in some instances, Salmonella spp., should occur to assess process control
and sanitary conditions. As with the non-RTE, RTC meals, other non-routine testing of indicator
organisms and pathogens may be appropriate depending on the ingredients used and the type
of finished product. Although not routinely done, if Listeria spp. is found in Zone 1 environmental
samples, investigational testing for L. monocytogenes may be indicated.

Meals and Entrees – RTE sous vide, cook and chill – Appendix A, Flow Diagram A.24,
Appendix J, Table J.24

Sous vide products are prepared with raw or partially cooked foods, which are vacuum
packaged in an impermeable bag, cooked in the bag, rapidly chilled, and refrigerated with time-
temperature combinations that inhibit pathogen growth. If the cook process does not provide at
least a validated 6-log10 reduction of non-proteolytic C. botulinum spores (7), validation data
should be provided by the supplier to demonstrate that the process eliminates vegetative
pathogens. Because of the lack of inhibitory barriers in typical sous-vide products and the
concern for potential outgrowth of botulinum spores, strict adherence to refrigerated storage
after treatment is extremely important. If a validated cook step is used and verified, no routine
testing is recommended. In the absence of a validated cook process, testing for vegetative

33
microorganisms should be done by the supplier on post-cook samples to verify the thermal
process. Testing for E. coli can serve as a verification of thermal processing; periodic testing of
coliforms, Enterobacteriaceae and APC are useful for verification purposes. If cooling deviates
from prescribed requirements such as those given in USDA Appendix B (18) , testing for C.
perfringens may be useful as a part of the supporting documentation for safety. Routine testing
for C. perfringens typically is not done.

Meat, Pork, Poultry Products –Non-RTE, beef and pork, raw, intact and non-intact – Appendix
A, Flow Diagram A.25, Appendix J, Table J.25

These products include both intact (e.g., non-tenderized steaks, chops) and non-intact (e.g.,
whole muscle destined for ground product, trim, ground product, needle-tenderized steaks) raw
beef and pork products. Under normal operating conditions, no routine testing is recommended.
When it is necessary to meet a regulatory or customer requirement to confirm production is
occurring with process control and sanitary conditions, suppliers should test for E. coli (typical
for the U.S.) or Enterobacteriaceae (typical for the European Union). Those manufacturers
supplying DOD with non-intact product should request that their suppliers (secondary suppliers)
provide a Certificate of Analysis demonstrating that the raw materials have tested negative for
E. coli O157:H7 and other STEC, if appropriate. Suppliers to DOD also may test for Salmonella
to meet regulatory requirements or to provide evidence that they are meeting performance
standards that indicate production has occurred under sanitary conditions; this testing may
typically be done only for ground products.

Meat, Pork, Poultry Products –Non-RTE, poultry, raw – Appendix A, Flow Diagram A.26,
Appendix J, Table J.26

These products include both intact (e.g., non-injected whole birds, non-injected parts) and non-
intact (e.g., injected or “enhanced” or vacuum-tumbled poultry parts, ground poultry) raw poultry
products. Under normal operating conditions, no routine testing is recommended. Production of
these foods should include appropriate process controls to reduce pathogens to acceptable
levels and to prevent pathogen growth. When it is necessary to meet a regulatory or customer
requirement to confirm production is occurring with process controls and sanitary conditions, or
under specific circumstances when an investigation is underway, suppliers may test for
Salmonella and Campylobacter to verify process control and that pathogens are being reduced
to acceptable levels. In this case, testing should be performed on the relevant product type
such as raw poultry parts if they are the product type purchased. Testing for indicator
organisms or classes of organisms such as generic E. coli, coliforms, Enterobacteriaceae, or
APC, could provide additional information regarding maintenance of process control and
sanitary conditions.

Meat, Pork, Poultry Products – RTE, cooked, perishable – Appendix A, Flow Diagram A.27,
Appendix J, Table J.27

This group includes a spectrum of cooked beef, pork and poultry products which require strict
refrigeration for shelf life and safety (e.g., deli meats, hot dogs). While process control is often
monitored through routine testing of E. coli, potential contamination of L. monocytogenes is a

34
major concern and should be addressed by the supplier through routine environmental
monitoring of Zone 2 and 3 surfaces for Listeria spp. Although not routinely tested, if Zone 1
environmental samples are positive, finished product testing for L. monocytogenes may be
indicated. Non-routine testing of coliforms or Enterobacteriaceae, APC, Salmonella, and C.
perfringens may be useful for additional verification of sanitary conditions, adequate cooling, or
as periodic verification of process control.

Meat, Pork, and Poultry Products – RTE, fermented, dried – Appendix A, Flow Diagram A.28,
Appendix J, Table J.28

These products (e.g., jerky, dried pepperoni, meat sticks) are characterized by having
chemical/physical characteristics (e.g., aw and pH) that ensure the products will not spoil or
become unsafe when stored out of refrigeration throughout the manufacturer’s specified shelf-
life. However, it is essential that production of these foods include appropriate process steps to
reduce pathogens to acceptable levels and prevent growth of pathogens or the formation of
their toxins (e.g., cooking jerky with adequate humidity to prevent surface drying, active
fermentation to inhibit growth of S. aureus, and a lethality step to eliminate low-infectious dose
pathogens such as Salmonella and E. coli O157:H7) (8). Suppliers should use E. coli for
routine monitoring; coliforms and Enterobacteriaceae may be appropriate for verification
monitoring. Testing of products for bacteria, such as Salmonella, E. coli O157:H7 and S.
aureus may be appropriate when process controls are suspect, e.g., failed fermentation or
extended drying times.

Nuts and Nut Butters –RTE, not processed for lethality – Appendix A, Flow Diagram A.29,
Appendix J, Table J.29

Raw nuts (not processed for lethality) may be contaminated with microbiota from orchards, the
ground, or equipment and personnel during harvesting, shipping, processing, and handling.
Because consumption of raw nuts has been associated with illness, suppliers should test in-
process samples and finished products routinely for Salmonella and implement an environment
testing program that includes testing for Salmonella. For certain nuts (e.g., peanuts, pistachios,
Brazil nuts), routine testing for aflatoxin B1 should be done. Non-routine testing for E. coli and
aflatoxin B1 (for those not tested routinely for aflatoxin B1) may be done to assess sanitary
storage and production, and the quality of the raw nuts.

Nuts and Nut Butters – RTE, processed for lethality – Appendix A, Flow Diagram A.30,
Appendix J, Table J.30

In this category, peanuts and tree nuts are processed for lethality (e.g., by dry roasting, oil
roasting, or steam processing). Because nuts and nut butters have been associated with illness,
routine environmental testing, testing in-process samples, and finished product testing for
Salmonella should be done. For certain nuts (e.g., peanuts, pistachios, Brazil nuts), routine
testing for aflatoxin B1 should be done. Non-routine testing for E. coli and aflatoxin B1 (for
those not tested routinely) may be conducted to help assess sanitary storage and production,
and the quality of the raw nuts used in manufacturing.

35
Produce –Fruits and vegetables, cut, frozen or refrigerated, minimally processed – Appendix A,
Flow Diagram A.31, Appendix J, Table J. 31

Further processing of fresh fruits and vegetables may increase or decrease microbiological
populations depending on GMPs, sanitary design of equipment, washing, blanching, or the use
of antimicrobials. Routine testing by suppliers of product for E. coli and the environment for
Listeria spp. should be done to assess process control and sanitary conditions. Periodic testing
by suppliers of in-process or finished products for Salmonella or E. coli O157:H7 (or other
appropriate STEC) may be pertinent depending on the commodity, geographic location and use
of GAPs.

Produce –Fruits and vegetables, whole – Appendix A, Flow Diagram A.32, Appendix J, Table
J.32

Fruits and vegetables are expected to have microbiota associated with them. Whole fruits and
vegetables may be washed before introduction to commerce, but undergo no other lethality
step. Environmental testing in the packing house for Listeria spp. and Salmonella should be
done by the supplier to assess sanitary conditions, with the frequency dependent upon factors
such as the commodity, geographic location and use of GAPs. Although not listed in Table J.32
nor routinely done, the DOD may consider testing (by the supplier or DOD) for Cyclospora
cavetanensis, Cryptosporidium parvum, enteric viruses, or Shigella spp. as appropriate when
there is knowledge or suspicion high risk farming and handling practices (e.g., where evidence
of previous contamination exists, water contamination is likely, or contaminated fertilizer is
used).

Produce –Mushrooms – Appendix A, Flow Diagram A.33, Appendix J, Table J.33

Mushrooms are generally commercially produced indoors on composted substrate. They are
grown, harvested, sorted, graded, and packaged, and may or may not be sliced. No routine
testing of product is typically conducted because populations of indigenous microbiota likely will
be high. Routine monitoring and testing of the environment s by suppliers for Listeria spp. may
be deemed appropriate by DOD to assess sanitary conditions and process control. Such
testing would depend on factors such as the type of compost used, the water used, the
harvesting techniques, the storage and handling conditions, and the intended end use.

Produce – Packaged salads and leafy greens – Appendix A, Flow Diagram A.34, Appendix J,
Table J.34

Salad greens are expected to have microbiota that can originate from numerous sources such
as irrigation water, insects, birds, animals, and post-harvest handling and processing. When
salad greens are washed, particularly when a chemical such as chlorine is added to the wash
water, some microorganisms can be physically washed off; however, the washing process also
can contribute to cross contamination. Packaged salads and leafy greens generally have a
limited shelf life. Suppliers can use testing for E. coli to assess process control and sanitary
conditions. Environmental testing for Listeria spp. in processing facilities should be conducted
to monitor sanitary conditions.

36
Produce – Vegetable sprouts – Appendix A, Flow Diagram A.35, Appendix J, Table J.35

These are sprouted vegetable seeds before true leaves emerge that may be consumed raw or
cooked. Routine testing of in-process and finished products by suppliers for E. coli should be
done as an indicator of process control and sanitary production. Appropriate testing of spent
irrigation water for Salmonella and E. coli O157:H7 should be conducted to assess potential
product contamination. Routine environmental monitoring for Listeria spp. also should occur to
assess sanitary conditions.

Seafood – Raw – Appendix A, Flow Diagrams A.36a-e, Appendix J, Table J.36

Routine microbiological testing of in-process and finished products by suppliers is not

recommended for raw (fresh or frozen) finfish or raw crustaceans for either quality or safety.
Non-routine testing of in-process and finished products for coliforms and Salmonella may be
done to verify proper sanitation and process control. A visual inspection for parasites is
recommended if the product is intended for raw consumption. Alternatively, the supplier may
verify that freezing treatments are applied to destroy certain parasites. For scombroid species,
testing of finished product for histamine is recommended.

Seafood – RTE, fish, cold smoked – Appendix A, Flow Diagram A.37, Appendix J, Table J.37

Suppliers should conduct routine environmental testing for Listeria spp. to demonstrate that
production is occurring under sanitary conditions. The supplier also should test in-process and
finished products periodically for L. monocytogenes and Salmonella to demonstrate that the
product is produced under sanitary conditions. The pH of pickled herring should be verified
periodically. Scombroid species may contain histamine and products made from these species
should be tested to verify that proper temperature control was maintained.

Seafood –RTE, cooked or hot smoked – Appendix A, Flow Diagram A.38, Appendix J, Table
J.38

The supplier should apply a validated process that results in at least a 6-log10 reduction of L.
monocytogenes. When such a validated process is used, routine sampling of in-process and
finished product for S. aureus and the environment for Listeria spp. should occur to verify that
controls are in place to prevent recontamination. If required to further demonstrate that
production is occurring under process control and sanitary conditions, the supplier could also
test in-process and finished products for coliforms, APC, Salmonella and L. monocytogenes. If
it is apparent that there is a potential for recontamination through mechanical or manual
handling, testing finished products for Salmonella and L. monocytogenes should be done
routinely. Scombroid species may contain histamine if temperature abused and fish
decompose; finished products should be tested for histamine per FDA’s guidance documents
(26).

Seafood – RTE, raw molluscan shellfish – Appendix A, Flow Diagram A.39, Appendix J, Table
J.39

37
Suppliers must demonstrate traceability that establishes that the product was harvested from
approved waters in the U.S. or in countries (Canada, Mexico, New Zealand, South Korea) that
have a Memorandum of Understanding with the U.S. Under these conditions, no routine
microbiological testing of products is necessary by the supplier. Where the supplier is unable to
prove the status of the harvest waters, or where contamination is suspected, the DOD should
not accept the product. Non-routine in-process and finished product testing by suppliers on
RTE, raw molluscan shellfish from approved waters to demonstrate process control and sanitary
conditions may include analyses for APC, fecal coliforms, and Vibrio paraheamolyticus (or other
Vibrio spp. if warranted). In addition, Vibrio control plans as outlined in the National Shellfish
Sanitation Program (25) may be required if conditions warrant.

Spices and Herbs, Coffee and Tea – Appendix A, Flow Diagrams A.40.a, A.40.b, and A.40.c,
Appendix J, Table 40

Harvested spices are expected to have a varied microbiota associated with them, including
spore-forming bacteria and fungi. Also, when a dehydration process is performed outdoors
there is the potential to acquire additional contamination. Suppliers should test in-process and
finished products routinely for APC and Salmonella to assess process controls and sanitary
conditions. The suppliers also should routinely test the environment for Salmonella. Non-
routine testing of finished products by suppliers, when deemed necessary, to assess process
control and sanitary conditions may include testing for B. cereus (or other toxigenic Bacillus
spp.), E. coli, coliforms, mold and yeasts, and E. coli O157:H7 (or other STEC as appropriate).

OTHER INDICATORS OF PROCESS CONTROL AND SANITARY CONDITIONS

There are microbiological by-products, enzymes, products of decomposition (including those

detected through visual observation), and other analytes that may reflect lack of process control
or insanitary conditions. The following are examples of some of these indicators.

• Histamine in scombroid fish at high levels indicates possible temperature abuse, lack of
sanitary conditions, and decomposition of these fish.
• The presence of non-microbiological alkaline phosphatase in milk is an indication that the
milk has been inadequately pasteurized. Under these conditions microbiological pathogens
endemic to raw milk may survive and result in milk-borne illness.
• Peroxidase testing is used to indicate that blanching of fresh vegetables has been adequate.
Typical blanching temperatures (195 – 205°F for 3 minutes) would likely be sufficient to
provide a lethality step eliminating vegetative pathogens.
• The presence of aflatoxin or other mycotoxins is indicative of significant growth of molds.
The presence of aflatoxin or other mycotoxins may render the food unacceptable for human
consumption or for use in further food processing.
• Gas formation causing swollen product containers would be indicative of spoilage and
potential pathogen growth. Similarly, slime formation, visible mold growth, discoloration and
product leakage from a container would be indicative of spoilage or potential growth of
pathogens. Changes in product viscosity may be indicative of microbiological proteolysis or

38
 starch hydrolysis; such activity may be the result of post-processing contamination and
temperature abuse, or under processing.
• Peroxide values and concentrations of free fatty acids in nuts exceeding tolerance limits
would be indicative of poor storage conditions, extended age or temperature abuse. In such
situations, these changes would not indicate microbiological spoilage or growth, but
oxidation that impacts quality.
• When free fatty acid concentrations in milk exceed tolerances, this is indicative of hydrolytic
rancidity associated with poor raw material control and potential post-process
contamination.
• Any signs of pests or pest infestation indicate contaminated packaging materials, poor
storage conditions within a plant or distribution center, pest contamination within a transport
container or at the location of sampling. These products should be considered
compromised and unacceptable.
• Development of acidity (measured by pH or titration) is critical to the safe production of
many fermented products such as cheeses, and fermented sausages. Fermentation of
these products by harmless starter organisms retards or prevents the growth of pathogenic
bacteria like E. coli, Salmonella and L. monocytogenes. However, in other products acid
development is undesirable, e.g., flat sour defect in canned food resulting from undesirable
microbiological growth. Undesirable fermentation can result in expression of purge in RTE
meat products.

GLOSSARY

Acronym
Term Definition
/Symbol
Indicates the maximum number of non-conforming analytical
Acceptance units (two-class sampling plans) or marginally acceptable
C
number analytical units (three-class sampling plans) that can result
in lot acceptance.

The enumeration of colony forming units of mesophilic

Aerobic plate
APC aerobic and facultative anaerobic organisms on an
count
appropriate non-selective medium.
Target for assay detection, isolation or quantification, e.g.,
Analyte
Salmonella.

The relevant quantity – mass, volume or area – of the food

product that is being tested in each analytical unit. The
Analytical
analytical portion is less than or equal to the sample unit
portion
amount. For example, a 1 ml analytical portion of diluted
homogenate may be analyzed from a 25 g sample unit.

39
 A single unit of food, from which a predetermined analytical
portion is removed and tested for microorganisms. All or
Analytical unit part of the sample unit may be used as the analytical unit, or
multiple sample units may be composited into a single
analytical unit for presence/absence testing.
Attributes sampling plans are used when the measured
characteristics are qualitative or categorical. Microbial
Attributes
presence/absence data and quantitative concentration data
sampling plans
categorized into numerical ranges are classified as
attributes.

Bernoulli A Bernoulli process is a random process the result of which

process can only take one of two values, e.g., presence/absence.

The discrete probability distribution of the number of

Binomial "successes" in a sequence of n independent Bernoulli
distribution (yes/no) trials, each of which yields success with constant
probability (p)

Certificate of A document attesting to the quality and purity of a product

Analysis lot.
Certificate of A document issued by a competent authority that the
Conformance product meets required specifications.

Colony forming The number of single or clumped multiple cell aggregates

CFU
units giving rise to colonies recovered on a solid medium.

The probability of accepting a non-conforming lot. A false

Consumer's risk Β
negative or type II error.

The control limits delineate the expected extent of natural

Control limits, LCL and variability in the process. Conventionally defined as ±3
lower and upper UCL standard deviations about the mean, but can be adjusted
based on the desired false alarm rate.

The number of colony forming units recovered from an

Count
analytical portion
Criterion/criteria See microbiological criterion
Critical Control The point in food manufacturing at which effective control
CCP
Point can be exercised over a hazard.
Cumulative
Describes the probability that a random variable X will be
distribution CDF
found to have a value less than or equal to x: F(x) = P(X≤x).
function
Department of
DOD United States Department of Defense
Defense

40
 The prevalence that the sample is designed to detect with a
Design specified probability. May or may not be the assumed
prevalence prevalence of an attribute in a population from which
samples are drawn.

Empirical
The cumulative distribution function associated with the
cumulative
ECDF empirical (observed) measure of a sample. The non-
distribution
parametric estimator of the CDF.
function

Empirical
distribution EDF Synonymous with empirical cumulative distribution function
function

A program wherein equipment and facility sites are tested

Environmental
routinely for non-pathogens or pathogens to determine the
monitoring EMP
extent to which these microorganisms are present and could
program
likely contaminate food products manufactured in the facility.

The probability distribution that describes the time between

Exponential events in a Poisson process, i.e., a process in which events
distribution occur continuously and independently at a constant average
rate.

Exponentially
A curve smoothing technique applied to time series data
weighted EWMA
that exponentially down weights older observations.
moving average

The expected rate of false positives, e.g., indicating a loss of

False alarm rate FAR process control when the process actually remains under
control
A control chart used to monitor very low prevalence
G-chart contamination. Tracks the interval (number of samples)
between positives.
Good
Those hygienic practices described in the Code of Federal
Manufacturing GMP
Regulations, e.g., 21CFR 110.
Practices
Advisory criteria used to inform food operators and others of
Guidelines the microbiological content expected in a food when best
practices are applied.
High-event A production period when the observed prevalence likely
period exceeds the expected or design prevalence

41
 Statistically, a volume of production is considered
homogenous relative to a given characteristic (e.g.,
concentration of the microorganism) if the characteristic
Homogeneous
follows the same probability distribution throughout the
(statistical)
volume (e.g., lognormal with fixed mean µ and fixed
standard deviation σ). In contrast to a homogeneous
(uniform) spatial distribution.

Individuals Chart
Control chart for individual measurements
(i-chart)

In-process Refers to sampling of food products or ingredients that have

samples not completed a manufacturing process by a supplier

This word is used synonymously with unsanitary in this

document. It refers to conditions where lack of appropriate
Insanitary
hygienic conditions has resulted in unsatisfactory
microbiological contamination.

Lognormal A continuous probability distribution of a random variable

distribution whose logarithm is normally distributed.

A predefined quantity of food product, produced under

similar, or uniform, conditions so that the units in the lot are
Lot similar in their microbiological status. In lot acceptance
sampling, the quantity of food product represented by the
samples.
Mean time
between MTBP The average number of samples between positives
positives

The specification of a microbiological criterion includes the

selected microorganism(s); the microbiological limits; the
Microbiological sampling plan defining the number of sample units to be
criterion taken (n), the size of the analytical unit, and where
appropriate, the acceptance number (c); and the analytical
methods.

Microbiological limits are those levels above which might be

Microbiological indicative of loss of process control or insanitary conditions
limit and may lead to further investigation with corrective or
preventive actions.
Microbiological
limit for
Delimits acceptable and marginally acceptable
marginally m
concentrations. Used in 3-class sampling plans
acceptable
concentration

42
Microbiological
Marks the limit beyond which the level of contamination is
limit for
M hazardous or unacceptable Used in 2- and 3-class sampling
unacceptable
plans
concentration

The probability distribution of a random variable whose

Mixture
values can be interpreted as being derived from multiple
distribution
underlying probability distributions

Most probable An estimated quantitative concentration measurement

MPN
number developed using serial dilutions and detection methods.

When the target organism is not detected in the analytical

Negative unit, then the analytical unit is commonly referred to as
"negative."
Makes no assumptions about the probability distribution of
Nonparametric
the random variable
Non-routine testing can be investigational, for verification,
validation, surveillance, or for qualifying suppliers. Non-
Non-routine
routine testing is less frequent and can be based on time
testing
intervals (e.g., weekly, monthly, quarterly) or based on other
indicators of lack of process control or insanitary conditions.

A continuous probability distribution that is symmetric about

Normal
the mean (μ), with approximately 95% of values lying within
distribution
± 2 standard deviations (2σ) of the mean.

Operating
Describes the probability of accepting a lot as a function of
characteristic
lot quality
curve

Assumes that the data have come from a

Parametric
theoretical probability distribution defined by its parameters

A process control chart that monitors the proportion of non-

P-Chart conforming analytical units observed in a sample of size n,
applicable for moderate prevalence levels.
Pre-determined plan of action, such as corrective action
Plan of action POA
plan

Describes the probability of a given number of events

Poisson
occurring in a fixed interval of time and/or space if the
distribution
events occur independently with a constant average rate

43
 When the target organism is detected in the analytical unit,
Positive
then the analytical unit is commonly referred to as "positive."

The proportion of analytical units that contain the target

microorganism. The observed prevalence depends on the
Prevalence
analytical unit size and needs to be referenced to an
analytical unit size, i.e., prevalence of positives in X grams

Process
Cp The ability of a process to meet specification limits.
capability
Maintaining the output of a specific process (e.g., food
Process control manufacturing, storage and distribution system) within a
desired range.
The probability of rejecting a conforming lot. A false positive
Producer's risk Α
or type I error.
The value associated with a percentile of the cumulative
Quantile distribution function. If p(X≤A) = B, A is the quantile value
and B is the percentile of the CDF.
Range Chart used to monitor process variability for
R-Chart
continuous numerical data.
Routine testing is defined as testing done at pre-determined
intervals at sufficient frequency to establish process control
or sanitary conditions. The sampling interval may be on a
Routine physical lot basis (e.g., 2,000 lb. combos for ground beef) or
temporal basis (e.g., per shift, daily, weekly, monthly). The
frequency of testing should be determined based on
potential risks and performance of the system.
Ready-to-eat Food that is in a form that may be safely eaten without
RTE
food additional preparation to achieve food safety
A subset of units from the lot or production process,
Sample
selected in some predetermined manner.

Sample size n The number of samples units drawn to collect a sample

A single unit of food of a predetermined sample unit amount

(mass, volume, or area). All or part of the sample unit may
Sample unit be used as the analytical unit, or multiple sample units may
be composited into a single analytical unit for
presence/absence testing.
Defines the number of sample units to be taken (n), the size
Sampling plan of the analytical unit, and where appropriate, the acceptance
number (c).

44
 Specification
LSL and
limits, lower and Boundaries that define acceptable product
USL
upper
Specifications are part of a purchasing agreement between
Specifications a buyer and a supplier of a food and may be advisory or
mandatory according to use.

A process is considered under statistical control when its

Standard output varies as expected within a standard operating range
SOR
operating range (SOR) of variation. This refers to common cause variation
and represents the random variation inherent in a process.

Standards are mandatory criteria incorporated into a law or

Standards
ordinance (normally pathogen oriented)
A process is considered under statistical control if it is stable
over time and the observed variation is due to common,
Statistical chance causes inherent to the process and there is no
control between-lot variation. Statistical control means only that the
process output is predictable and is distinct from the
capability of a process to meet specifications.
Statistical A formal approach that uses statistical methods to monitor
SPC
Process Control and control a process.
A food that requires time/temperature control for safety to
limit pathogenic microorganism growth or toxin formation.
Temperature/tim
For a further description of TCS foods, refer to FDA 2013
e control for TCS
Food Code at
safety
http://www.fda.gov/downloads/Food/GuidanceRegulation/Re
tailFoodProtection/FoodCode/UCM374510.pdf
A single manufacturing or supply chain step, e.g., blanching
Unit operations
vegetables, slicing meat, loading a trailer.

This word is used synonymously with insanitary in this

document. It refers to conditions where lack of appropriate
Unsanitary hygienic conditions has resulted in unsatisfactory
microbiological contamination not conducive to or promoting
health; dirty or unhygienic.

The body of scientific evidence that demonstrates a process

Validation or procedure is effective in producing the outcome for which
it was intended

Variables sampling plans are used when the measured

Variables
characteristics are expressed on a continuous numerical
sampling plans
scale, e.g., concentration data.

Those activities, other than monitoring, that establish the

Verification validity of a food safety plan and that the food safety system
is operating according to the plan.

45
 A measurement between 0.00 and 1.00 defining the amount
of moisture available for microbiological or chemical activity.
Water activity aw
Deionized water has an aw of 1.00 under standard
conditions. Microbes are not known to grow below aw 0.60.
Worldwide Worldwide Directory of Sanitarily Approved Food
Directory Establishments for Armed Forces Procurement, 2012

REFERENCES

1. Bradley, R. L., K. S. Houck, and M. Smukowski. 2013. Milk and milk products. p. 1-16.
In, S. Doores, Y. Salfinger and M. L. Tortorello (ed.), Compendium of Methods for
Microbiological Examination of Foods, 5th Ed. American Public Health Association. doi:
10.21005/MBEF.0222.052
2. Codex (Codex Alimentarius Commission). 2004. General Guidelines on Sampling, CAC/GL
50-2004. www.codexalimentarius.org/download/standards/10141/CXG_050e.pdf Accessed
February 5, 2015.
3. Dey, M., J. A. Mayo, D. Saville, C. Wolyniak, and K. C. Klontz. 2013. Recalls of foods due
to microbiological contamination classified by the U.S. Food and Drug Administration, fiscal
years 2003 through 2011. Journal of food protection 76: 932-938.
4. Elliott, P. H., and A. Kataoka. 2013. Canned Foods - Test for Commercial Sterility. In, S.
Doores, Y. Salfinger and M. L. Tortorello (ed.), Compendium of Methods for
Microbiological Examination of Foods, 5th Ed. American Public Health Association,
Washington, D.C. doi: 10.21005/MBEF.0222.052
5. Food Standards Australia New Zealand. 2012. Standard 1.6.1 Microbiological Limits for
Food - F2012C00862 http://www.comlaw.gov.au/Details/F2012C00862 Accessed February
5, 2015
6. Food Standards Australia New Zealand. 2014. Criteria for Listeria monocytogenes in ready-
to-eat foods. http://www.foodstandards.gov.au/code/microbiollimits/Pages/Criteria-for-
Listeria-monocytogenes-in-ready-to-eat-foods.aspx Accessed February 5, 2015.
7. Hyytia-Trees, E., E. Skytta, M. Mokkila, A. Kinnunen, M. Lindstrom, L. La Hteenma Ki, R.
Ahvenainen, and H. Kornkeala. 2000. Safety evaluation of sous vide-processed products
with respect to nonproteolytic Clostridium botulinum by use of challenge studies and
predictive microbiological models. Appl. Environ. Microbiol. 66: 223-229.
8. Ingham, S. 2008. Critical Limit Summary: Validating Shelf-Stability and Prevention of
Staphylococcus aureus Growth on RTE Meat Products.
http://www.meathaccp.wisc.edu/validation/assets/CL_Staph_JFP_68.pdf Accessed
September 19 2014.
9. ILIS -Europe, . 2010. Impact of Microbial Distributions on Food Safety.
http://www.ilsi.org/Europe/Publications/Microbial%20Distribution%202010.pdf Accessed
February 5, 2015

46
10. International Commission for the Microbiological Specifications for Foods (ICMSF). 2011.
Microorganisms in Foods 8: Use of Data for Assessing Process Control and Product
Acceptance. Springer, New York.
11. International Commission for the Microbiological Specifications of Foods (ICMSF). 2002.
Microorganisms in Foods 7: Microbiological Testing in Food Safety Management. Springer,
New York.
12. International Organization for Standardization. 2007. Guidance on the selection and usage of
acceptance sampling systems for inspection of discrete items in lots — Part 1: Acceptance
sampling. ISO, Geneva.
13. Kornacki, J. L.., J. B. Gurtler, and B. A. Stawick. 2013. Enterobacteriaceae, Coliforms, and
Escherichia coli as Quality and Safety Indicators. . In, S. Doores, Y. Salfinger and M. L.
Tortorello (ed.), Compendium of Methods for Microbiological Examination of Foods, 5th Ed.
American Public Health Association. Washington, D.C. doi: 10.21005/MBEF.0222.052
14. Powell, M. R. 2014. Optimal Food Safety Sampling Under a Budget Constraint. Risk
Analysis 34: 93-100. doi: 10.1111/risa.12054
15. Ryser, E. T., and E. H. Marth. 2007. Listeria, Listeriosis, and Food Safety. 3rd ed. CRC
Press, Taylor and Francis Group, Boca Raton, FL.

16. Shewhart, W. A. 1986. Statistical Method from the Viewpoint of Quality Control. Dover
Publications, Inc. New York
17. Sperber, W. H., and North American Millers' Association Microbiology Working Group.
2007. Role of microbiological guidelines in the production and commercial use of milled
cereal grains: a practical approach for the 21st century. Journal of food protection 70: 1041-
1053.
18. U.S. Department of Agriculture. 1999. Appendix B. Compliance Guidelines for Cooling
Heat-Treated Meat and Poultry Products (Stabilization). Food Safety Inspection Service.
Washington D. C. http://www.fsis.usda.gov/OPPDE/rdad/FRPubs/95-033F/95-
033F_Appendix%20B.htm. Accessed February 5, 2015.
19. U.S. Department of Agriculture. 2005. Risk Assessment of the Impact of Lethality Standards
on Salmonellosis from Ready-to-Eat Meat and Poultry Products. Food Safety Inspection
Service. http://www.fsis.usda.gov/wps/wcm/connect/210d3d19-96bc-4b1e-8829-
325e25e63e0b/Salm_RTE_Risk_Assess_Sep2005.pdf?MOD=AJPERES Accessed February
5, 2015.
20. U. S. Department of Agriculture. 2014. FSIS Compliance Guideline for Meat and Poultry
Jerky Produced by Small and Very Small Establishments.
http://www.fsis.usda.gov/wps/wcm/connect/5fd4a01d-a381-4134-8b91-
99617e56a90a/Compliance-Guideline-Jerky-2014.pdf?MOD=AJPERES Accessed February
5, 2015.
21. U. S. Department of Health and Human Services.2005. Apple Juice, Apple Juice
Concentrates, and Apple Juice Products - Adulteration with Patulin CPG Sec.510.150.
http://www.fda.gov/ICECI/ComplianceManuals/CompliancePolicyGuidanceManual/ucm074
427.htm Accessed February 5, 2015.

22. U. S. Department of Health and Human Services. 2009. Prevention of Salmonella Enteritidis
in Shell Eggs During Production, Storage, and Transportation: Final Rule. 21 CFR Parts 16
and 118, Federal Register 74(130).

47
23. U. S. Department of Health and Human Services. 2010. Compliance Policy Guide Sec.
527.300 Dairy Products - Microbial Contaminants and Alkaline Phosphatase Activity.
http://www.fda.gov/downloads/ICECI/ComplianceManuals/CompliancePolicyGuidanceMan
ual/UCM238465.pdf Accessed February 5, 2015.
24. U. S. Department of Health and Human Services. 2011. Grade “A” Pasteurized Milk
Ordinance, Including Provisions from the Grade “A” Condensed and Dry Milk Products and
Condensed and Dry Whey--Supplement I to the Grade “A” Pasteurized Milk Ordinance.
http://www.fda.gov/downloads/Food/GuidanceRegulation/UCM291757.pdf Accessed
February 5, 2015.
25. U. S. Department of Health and Human Services. 2013. National Shellfish Sanitation
Program (NSSP).
http://www.fda.gov/Food/GuidanceRegulation/FederalStateFoodPrograms/ucm2006754.htm
Accessed February 5, 2015.
26. U. S. Department of Health and Human Services. 2014. Fish and Fishery Products Hazards
and Controls Guidance - Fourth Edition.
http://www.fda.gov/food/guidanceregulation/guidancedocumentsregulatoryinformation/seafo
od/ucm2018426.htm. Accessed February 5, 2015.
27. U.S. Department of Defense. 2013. Department of Defense Food Safety and Quality
Assurance Laboratory Action Levels. USAPHC Circular 40-1, Appendix O. In, Worldwide
Directory of Sanitarily Approved Food Establishments for Armed Forces Procurement, 2012.
28. U.S. Department of Health and Human Services, Food and Drug Administration. 2004.
Guidance for Industry Juice HACCP Hazards and Controls Guidance.
http://www.fda.gov/Food/GuidanceRegulation/GuidanceDocumentsRegulatoryInformation/J
uice/ucm072557.htm Accessed February 5, 2015.
29. Wehr, H. M., and J. F. Frank. 2004. Standard Methods for the Examination of Dairy
Products, 17th ed. American Public Health Association, Washington, DC.
30. World Health Organization. 2013. Water Sanitation Health – Drinking Water Quality.
http://www.who.int/entity/water_sanitation_health/draft_regscan_may_2014.pdf Accessed
August 14 2014.
31. World Health Organization, and Food and Agricultural Organization. 2013. Proposed Draft
Principles and Guidelines for the Establishment and Application of Microbiological Criteria
Related to Foods (at steps 5/8) Appendix III (Pages 43 to 49). Joint FAO/WHO Food
Standards Programme CODEX Alimentarius Commission Thrity-sixth Session Rome, Italy,
1-5 July 2013 and Report of the Forty-Fourth Sessionof the CODEX Committee on Food
Hygiene: 27-56.

48
Appendix A. Schematic Flow Diagrams of Production of Various Food Categories
and Bottled Water

These generic process flow charts are intended to provide DOD auditors with potential
steps in the manufacturing process where microbiological counts could increase with
loss of process control or development of insanitary conditions. In addition, the flow
charts illustrate where there are lethality steps that reduce numbers of indicator
organisms and pathogens if present.

Steps for receiving and storing packaging materials were omitted to simplify the creation
– and use – of the process flow diagrams. It is expected that a DOD-approved food
processing plant would have appropriate control and documentation of these functions,
either as part of product-specific HACCP plans, or as preventive and pre-requisite
programs such as Standard Operating Procedures for receiving and storage. It was
recognized that a finished food product could move through many storage and
distribution facilities as part of the supply chain. The final two steps were denoted “store
finished product” and “distribute finished product” to simplify the creation and use of the
process flow diagrams.

The intent was to include those steps relevant of the manufacturing process relevant to
microbial aspects of food rather than to include all possible aspects or combinations of
receipt, processing, storage, and distribution of production. The Committee assumes
that DOD personnel will be able to recognize the specific steps observed at a food
processing plant from among the general manufacturing steps shown on the process
flow diagrams.

Steps may be followed by any of the following designations:

• C, a step at which significant contamination may occur when adequate process
controls are not in place;
• G, a step in the process where growth of microorganisms can occur;
• K, a step where there is a pathogen kill step; and
• S, a point where sampling and testing by the supplier is recommended for
verification or investigation.

Programs for minimizing contamination at the identified steps include Good Agricultural
Practices, Sanitation Standard Operating Procedures, Good Manufacturing Practices,
and purchasing specifications. DOD personnel should use the process flow diagrams to
review the general steps to manufacture the food product under evaluation. From the
process flow diagram, DOD personnel should determine the step(s) at which verification
sampling should be done by the supplier. When analysis of verification samples
indicates that the supplier may have shortcomings in process or sanitation control, DOD
personnel should use the process flow diagram to determine steps at which
contamination might occur or steps at which a failure to achieve the expected
destruction of bacteria may be occurring.

1
 Flow Diagram A.1. BEVERAGES – BOTTLED
WATER (ARTESIAN, MINERAL, PURIFIED,
SPARKLING AND SPRING WATER)

Treat source water as

appropriate (S; K or
otherwise remove microbes)

Package (C, S)

Store finished product

Distribute finished product

2
Flow Diagram A.2. BEVERAGES – ICE, PACKAGED

Treat source water as

appropriate (S; K or
otherwise remove microbes)

Freeze

Store

Crush ice (optional; C)

Package (C, S)

Store finished product

Distribute finished product

3
 Flow Diagram A.3. BEVERAGES – JUICES AND DRINKS, PASTEURIZED,
REFRIGERATED
Harvest fruits or vegetables (C)

Transport

Storage (optional, G)

Wash

Treat source water as Extract juice (clarify, optional)

appropriate

Thermally concentrate (K)

Add ingredients for drinks
(optional; C)
Store in freezer

Blend water with drink Thaw

ingredients, or reconstitute
concentrate into 100% single-
strength juice
Pasteurize (K)

Cool/Chill

Package (C, S)

Store finished product (G)

Distribute finished product

4
 Flow Diagram A.4. BEVERAGES – SHELF STABLE

Receive and store Treat source water as

ingredients appropriate

Mix, blend, deaerate, filter

Treat for lethality (optional; K)

Options: Pasteurize, UHT,
HTST, Aseptic

Chill Hot fill container

Carbonate (optional) Chill

Package

Store finished product

Distribute finished product

5
 Flow Diagram A.5. DAIRY – BUTTER, MARGARINES

Butter Margarine

Receive and store Receive and store

ingredients (S)

Separate Cream Blend ingredients

Pasteurize (K) Emulsify (C)

Churn (C) Cool

Cool Package (C,S)

Work / add salt Store finished product

Form Distribute finished product

Package (C, S)

Store finished product (G possible in some nonstandard types)

6
Flow Diagram A.6. DAIRY – CHEESE (HARD)

Receive and store raw milk and other ingredients (G, S)

Filter, separate cream, standardize fat content (C)

Add ingredients (optional, C)

Pasteurize or apply sub-pasteurization heat treatment (K)

and homogenize (C)

Add culture and/or rennet or acid, annatto (C)/monitor pH

Process: form and cut curd, “cook” curd, drain whey,

cheddaring and similar steps (C, G)

Brine or add salt (C)

Press (C)

Ripen (optional; C; K- hard; G- soft, surface-ripened)

Subdivide (optional) and package (C, S)

Store finished product (G)

Distribute finished product

7
Flow Diagram A.7. DAIRY – CHEESE (SOFT, SEMI-SOFT AND SURFACE-
RIPENED)

Receive and store raw milk and other ingredients (G, S)

Filter, separate cream, standardize fat content (C)

Add ingredients (optional, C)

Pasteurize or apply sub-pasteurization heat treatment (K)

and homogenize (C)

Add culture and/or rennet or acid, annatto (C)

Process: form and cut curd, “cook” curd, drain whey,

cheddaring and similar steps (C, G)

Brine or add salt (C)

Press (C)

Ripen (optional; C; G- soft, surface-ripened)

Subdivide (optional) and package (C, S)

Store finished product (G)

Distribute finished product

8
 Flow Diagram A.8.a. DAIRY PRODUCTS Cultured pH<4.8 (Example – Yogurt)

Receive and store raw milk and other ingredients (G, S)

Filter, separate cream, standardize fat content (C)

Add ingredients (optional, C)

Pasteurize (K) and homogenize (C)

Add culture (may be preceded by concentration; C)

Ferment (may be packaged before fermentation; C, G)

Process: filter, heat, separate, concentrate, stir (optional; C)

Cool

Add fruits and other ingredients (optional; C, S for ingredients)

Package (C, S)

Store finished product

Distribute finished product

9
Flow Diagram A.8.b. DAIRY – CULTURED, pH<4.8 (Example – Sour Cream
Buttermilk, etc.)

Receive and store raw milk and other ingredients (G, S)

Filter, separate cream, standardize fat content (C)

Add ingredients (optional, C)

Pasteurize (K) and homogenize (C)

Add culture

Ferment (C, G; may be packaged before fermentation)

Cut curd and agitate (optional; C)

Cool

Package if not done previously (C, S)

Store finished product (S for product fermented in package)

Distribute finished product

10
Flow Diagram A.9. DAIRY – CULTURED, pH>4.8 AND <5.4 (Example – Cottage
Cheese)

Receive and store raw milk and other ingredients (G, S)

Filter, separate cream, standardize fat content (C)

Add ingredients (optional, C)

Pasteurize (K) and homogenize (C)

Add culture and rennet

Form, cut, “cook” curd (C, G)

Wash curds, drain whey, cool (C)

Add salt and dressing (milk/cream; C)

Package (C, S)

Store finished product

Distribute finished product

11
 Flow Diagram A.10. DAIRY – DRIED PRODUCTS
(does not include dairy ingredients used to make infant formula)

Receive and store milk or whey (G, S)

Filter, separate cream standardize fat content

(optional; C)

Pasteurize (K) and homogenize (optional; C)

Process by one or more of these steps: evaporate,

concentrate, pre-crystallize, remove lactose, spray-
dry, fluid-bed dry and cool, pneumatically transport
and cool (C)

Package (C, S)

Store finished product

Distribute finished product

12
 Flow Diagram A.11. DAIRY – FROZEN DESSERTS

Receive and store

ingredients (S optional)

Process: optional steps include measure (C); blend

(C), homogenize (C), pasteurize (K), cook (K),
assemble (C), build (C)

Freeze

Package (C, S)

Hard-freeze (optional)

Store finished product

Distribute finished product

13
 Flow Diagram A.12. DAIRY – MILK AND MILK PRODUCTS (Fluid)

Receive and store raw milk and other ingredients (G, S)

Filter, separate cream, standardize fat content (C)

Add vitamins (optional; recommended before homogenization; C)

Add other ingredients (optional, C)

Pasteurize (K) and homogenize (C)

Package (C, S)

Store finished product (G)

Distribute finished product

14
 Flow Diagram A.13. DAIRY – PROCESS CHEESE

Receive and store cheese

and other ingredients

Grind cheese (C)

Mix cheese and other

ingredients (C)

Cook(K)

Pack cold-pack cheese (C,

S)

Cast, slice, cool (C)

Package (C, S)

Store finished product

Distribute finished product

15
Flow Diagram A.14. EGG PRODUCTS – PASTEURIZED, PROCESSED

Receive and store eggs and

other ingredients (G)

Wash Eggs (C)

Crack eggs (C)

Separate yolk/white (C)

Blend yolk/white (C)

Add sugar/salt (optional, C)

Pasteurize/cook (K)

Cool

Package (S, C)

Store finished product

Distribute finished product

16
Flow Diagram A.15. EGG PRODUCTS – SHELL EGGS RAW

Receive and store eggs (G)

Wash and sanitize (C)

Candle

Check visually and grade

In-shell pasteurization
(optional; K)

Package finished eggs (S)

Store finished product (G)

Distribute finished product

17
Flow Diagram A.16. GRAIN BASED PRODUCTS – BAKED ITEMS, RTE,
REFRIGERATED OR TCS

Receive and store ingredients

Mix ingredients

Form dough

Proof

Bake (K)

Cool (C)

Add optional ingredients (C)

Slice (optional; C)

Package (S, C)

Store finished product (G)

Distribute finished product

18
Flow Diagram A.17. GRAIN BASED PRODUCTS – BAKED ITEMS, RTE, SHELF
STABLE, NON-TCS

Receive and store ingredients

Mix ingredients

Form dough

Proof

Bake (K)

Cool (C)

Add optional ingredients (C)

Slice (optional; C)

Package (S, C)

Store finished product

Distribute finished product

19
Flow Diagram A.18. GRAIN BASED PRODUCTS – RTE, CEREALS

Receive and store ingredients

Mix bulk and minor ingredients

Cook (K)

Extrude

Puff/toast (K)

Enrobe – Vitamins/coatings (C)

Dry (C)

Package (S)

Store finished product

Distribute finished product

20
Flow Diagram A.19. GRAIN BASED PRODUCTS – RTE, COLD PRESSED BARS

Receive and store ingredients

Mix ingredients (C)

Press/form (C)

Enrobe (optional; C)

Cool (C)

Package (S)

Store finished product

Distribute finished product

21
Flow Diagram A.20. GRAIN BASED PRODUCTS – NON-RTE, DRY FLOUR BASED
MIXES

Receive and store

ingredients

Blend ingredients (C)

Package (S)

Store finished product

Distribute finished product

22
 Flow Diagram A.21. GRAIN BASED PRODUCTS – NON RTE, PASTA, DRIED OR
REFRIGERATED

Receive and store ingredients

Mix ingredients (C)

Form dough

Extrude

Dried Pasta Refrigerated

Dry (C) Cook (K)

Cool (S, C, G)

Dry/dewater (C)

Package (S, C) Package (S)

Store finished product Store finished product (G)

Distribute finished product Distribute finished product

23
Flow Diagram A.22. MEALS AND ENTRÉES – NON-RTE, READY TO COOK (RTC)
MEALS, INCLUDES RAW INGREDIENTS

Receive and store

ingredients

Subdivide, chop, trim,

formulate, mix, assemble,
pre- or par-cook, cool
ingredients (C)

Package (S)

Cool/freeze (G, C)

Store finished product (G if

refrigerated)

Distribute finished product

24
 Flow Diagram A.23 MEALS AND ENTRÉES – RTE, DELI SALADS, SANDWICHES
HEAT-EAT MEALS, SUSHI

Receive and store Receive and store non-

ready-to-eat (RTE) ready-to-eat (NRTE)
ingredients ingredients

Subdivide, chop, trim, Subdivide, chop, trim,

formulate, assemble, formulate, assemble,
(optional C, G) (optional C, G)

Cook (or other lethality step; K)

Cool (S, G)

Subdivide, chop, trim,

formulate, assemble, re-cook
(optional, C, G)

Further cool (optional, G)

Package (C, S)

Store finished product (G)

Distribute finished product

25
Flow Diagram A.24. MEALS AND ENTRÉES – SOUS VIDE, COOK AND CHILL

Receive and store

ingredients

Trim, cut, prepare

ingredients (C)

Package and vacuum-seal

Pasteurize/cook (K)

Cool (S, G)

Store finished product at no

warmer than 3.3°C (S, G)

Distribute finished product

26
Flow Diagram A.25. MEAT, PORK AND POULTRY PRODUCTS – NON-RTE, BEEF
AND PORK RAW, INTACT AND NON-INTACT

Slaughter/remove head and hock

Singe swine hide

(optional; C) Dehide (C)

Wash and spot-clean (optional; pathogen reduction may occur; C)

Eviscerate and spot-clean (optional; pathogen reduction may

occur; C)

Apply steam or wash with organic acid and/or hot water (optional –
step varies with species and country; pathogen reduction may
occur; C, S)

Cool (G)

Apply steam or wash with organic acid and/or hot water (optional –
step varies with species and country; pathogen reduction may
occur; C, S)

Cut/”fabricate” (G, C)

Grind (optional; G, C)

Package (S)

Store finished product Distribute finished product

(G)

27
Flow Diagram A.26. MEAT, PORK AND POULTRY PRODUCTS – NON-RTE,
POULTRY, RAW

Receive live birds

Hang and stun, kill,

scald and pick,
eviscerate (C)

Wash

Chill with air/water

(depending on
conditions, pathogen
reduction
or contamination may
occur; (C)

Process: cut-up,
debone or further
process (G, C)

Store finished product

Distribute finished
product

28
Flow Diagram A.27. MEAT, PORK AND POULTRY PRODUCTS – RTE, COOKED
PERISHABLE

Receive and store meat, poultry, and other

ingredients

Process: temper, grind, cut/trim/portion/debone, cure,

mix/inject/rub/tumble, chop/emulsify, stuff (optional, C, G)

Cook (K)

Cool (S, C, G)

Package (G, C, S)

Store finished product

(G)

Distribute finished product

29
Flow Diagram A.28. MEAT, PORK AND POULTRY PRODUCTS – RTE
FERMENTED AND DRIED, DRIED

Receive and store meat, poultry, and other ingredients

Process: temper, weigh, combine ingredients;

form/shape products; rack/hang (C)

Ferment (optional; G)

Heat (K)

Cool (S, G)

Dry (may be part of previous Heat step;

additional pathogen reduction may occur)

Slice/cut; spray finished product with potassium

sorbate or other approved growth inhibitor
(optional; C)

Package (C, S)

Store finished product

Distribute finished product

30
Flow Diagram A.29. NUTS AND NUT BUTTERS – NUTS, RTE,
NOT PROCESSED FOR LETHALITY

Harvest (C) Optional: Dry

on orchard floor

Transport

Process (C) Options:

fumigate, hull, shell,
dehydrate, salt

Store processed nuts (C)

Sort, size, grade (C)

Package (C, S)

Store finished product

Distribute finished
product

31
 Flow Diagram A.30. NUTS AND NUT BUTTERS – RTE, PROCESSED FOR
LETHALITY

Receive untreated in-shell nuts Receive untreated shelled nuts

Remove debris, hull, sort, grade (C) Store

Store

Shell (C)

Remove debris, sort, grade (C)

Process for lethality (steam, PPO, dry- or oil-roast) (K)

Store roasted or treated nuts (may

be received from external supplier)

Grind roasted nuts (C, S)

Add/mix ingredients (optional; C)

Package (C, S)

Store finished product

Distribute finished product

32
 Flow Diagram A.31. PRODUCE – FRUITS AND VEGETABLES, CUT
FROZEN, OR REFRIGERATED, MINIMALLY PROCESSED

Harvest (C) Options: trim, core, cull, sort, pack

Transport

Pre-cool to remove field heat (optional, C)

Process options: inspect, sort, cull, trim, wash, de-water, shell, chop, cut,
slice, shred, grade, blend (C, G)

Wash and dewater (optional; C)

Blanch and cool (optional: C)

Package, may be preceded or followed by optional freezing (C, S)

Store finished product (G if refrigerated)

Distribute finished product

33
Flow Diagram A.32. PRODUCE – FRUITS AND VEGETABLES, WHOLE

Harvest (C) Options: trim, core, cull, sort, pack

(bulk or retail), inspect, grade

Transport

Pre-cool (optional, C)

Inspect, sort, cull, trim (optional, C)

Pre-cool (optional, C)

Wash (optional; C)

Inspect, grade (optional; C)

Package, if not already field packed (C, S)

Store finished product (may be refrigerated; G,

S-optional)

Distribute finished product

34
 Flow Diagram A.33. PRODUCE – MUSHROOMS – FRESH OR FROZEN,
WHOLE, SLICED, NOT CANNED OR MARINATED

Prepare compost substrate Prepare spawn (C, G)

Inoculate substrate with spawn

Incubate (G)

Harvest (C)

Trim and clean (C)

Sort and grade (C)

Package, may be preceded or

followed by optional freezing (C,S)

Store finished product (G)

Distribute finished product

35
Flow Diagram A.34. PRODUCE – PACKAGED SALADS AND LEAFY
GREENS

Harvest (C) Options: Trim, core,

cull, sort, pack

Transport

Pre-cool to remove field heat

(optional; C)

Process options: inspect, sort, cull, trim, wash (multiple

steps), de-water, cut/slice/shred, blend (C, G)

Package (C, S)

Store finished product (G)

Distribute finished product

36
Flow Diagram A.35. PRODUCE – VEGETABLE SPROUTS

Receive and store seeds

Sanitize for pathogen reduction

and rinse seeds

Transfer to growing bins (C)

Incubate and irrigate (C, G, S –

spent irrigation water)

Process, dehull (optional),

wash and de-water (C, G)

Package (C, S)

Store finished product

(refrigeration; G)

Distribute finished product

37
 Flow Diagram A.36a. SEAFOOD – NON-RTE, RAW

Harvest (G)

Board and sort (C)

Ice and pack (chilled Off-load (unless processed on- Store live
storage) (C) ship; G)

Scale, head, eviscerate, filet,

candle, portion, freeze, glaze as
appropriate (G, C)
Pack, w/ or w/o
shucking

Weigh, pack, label (G, C)

Store under
refrigeration or on
ice (G)

Store frozen, under

refrigeration, or on ice
(G, S)

Distribute finished product

38
Flow Diagram A.36b. SEAFOOD –RAW
Wild Salmon Shashimi

Receive frozen salmon (C, S)

Frozen storage of salmon

Tempering and filleting of salmon (G)

Pack and Label

Finished Product
Refrigerated storage (C, S, G)

39
 Flow Diagram A.36c. SEAFOOD –RAW
Wild Salmon Sushi

Receive Dry Materials Receive Frozen Salmon (C,S)

Dry storage Frozen

Dry of seaweed storage of
storage of (Nori) salmon
rice

Cook rice Tempering salmon (G)

Rice Acidification (C, S)

Assemble nori, rice and fish into rolls

Cut rolls

Pack and Label

Finished Product
Refrigerated Storage
(C, S, G)

40
Flow Diagram A.36d. SEAFOOD –RAW
Ceviche

Fillet fish into ½ inch (1.3cm)

strips (C,S)

Refrigerate (C,G)

Add lemon or lime juice and

ingredients (e.g., seasoned
chopped tomato, diced
cucumber, chopped onion,
etc.), salt and pepper

Store and Refrigerate (C,S,G)

41
Flow Diagram A.36e. SEAFOOD –RAW

Pickled Herring Fillets

Receive
(G,S)

De-ice and rinse

Freeze

Thaw

Head, gut and fillet

Dry Salt

Rinse

Cure
(C,S)

Drain

Pack and Label

Finished Product
Refrigerated Storage
(C, G)

42
Flow Diagram A.37. SEAFOOD – RTE FISH, COLD
SMOKED

Receive, wash and store fish (G, S)

Fillet and skin (G, C)

Brine under refrigeration (G,C)

Dry fish (G, C)

Smoke (G, C)

Cool (S, C, G)

Package (C)

Store finished product (refrigerated or frozen G, S)

Distribute finished product

43
Flow Diagram A.38. SEAFOOD – RTE, FISH OR CRUSTACEAN, COOKED OR
HOT SMOKED

Receive and store seafood and

other ingredients

Wash (may be preceded by

thaw)

Store under refrigeration (G)

Cut/portion (C)

Brine (may be preceded by

rinse, G)

Rinse

Smoke/dry (K)

Cool (G, C, S)

Package (G, C, S)

Store finished product (under

refrigeration or frozen; G)

Distribute finished products

44
 Flow Diagram A.39. SEAFOOD – RTE, RAW MOLLUSCAN
SHELLFISH

Harvest from approved, or tested and accepted, waters

(S optional for water)

Cool and/or wash onboard boat or ashore (C, G)

Receive and cool at dock/processing plant (C, G)

Wash and store (C, G)

Re-pack/shuck (C, G)

Store finished product (refrigerated or frozen; S, G)

Distribute finished product

45
 Flow Diagram A.40a. SPICES AND HERBS

Harvest (C)

Dry (C)

Pack (bulk; C)

Distribute to processor (multiple

steps possible; C)

Clean, sort, screen, grade (C)

Treat for lethality (optional; K)

Clean, mill, sort, grade (C)

Mix with other ingredients or spices

(optional; C)

Package (S)

Treat for lethality (optional; K

Store finished product

Distribute finished product

46
 Flow Diagram A.40b. BEVERAGES – COFFEE

Harvest (C)

Process raw coffee cherries (C, G):

Wet method: remove skin and pulp and separate from bean, ferment bean to
remove parenchyma, rinse and dry, mill, polish, grade, size and sort
Dry method: sun dry, mill, polish, grade, size sort

Roast (K) Cool (C)

Grind (C) Package roasted

beans (C)

Package ground coffee (C)

Brew (K)

Remove/discard
extracted grounds

Freeze- or spray-dry
(C)

Store finished product

Package instant
coffee (C, S) Distribute finished product

47
 Flow Diagram A.40c. BEVERAGES –

Harvest sort, screen tea leaves (C)

Wither (oolong, black) Pan fire (green) Steam (white)

Process: options are roll,

shape, bruise, cut, oxidize (C)

Dry and sort leaves (C)

Treat source water as

appropriate (S)

Extract tea leaves

Clarify liquid tea (C)

Evaporate and Package (S)

concentrate liquid tea.
Add recovered aroma
(C)

Freeze-dry or spray-dry
(C)l Store finished product

Package instant tea (C, S)

Distribute finished
product

48
Appendix B: Statistical Process Control (SPC) Charts

Control charts are plots of process data collected over time used to determine if a
process is in statistical control. It is important to note that there is a difference
between a process being in statistical control and meeting specifications. A
process is considered under statistical control if it is stable over time and the
observed variation is due to common, chance causes inherent to the process
(e.g., background noise due to normal variation in ambient temperature and
humidity) and there is no between-lot variation. A process is considered out of
statistical control if shifts in the process central tendency (e.g., mean), variability,
or both result from uncommon sources associated with special or assignable
causes (e.g., equipment malfunction, a change in raw materials, or failure of a
laboratory procedure). A food process being under statistical process control
does not imply its capability with respect to meeting microbiological
specifications. The ideal situation is when a process is both under statistical
control and is capable of manufacturing products that meet specifications.
However, a process can be in statistical control and not capable of satisfying
specifications. For example, the process consistently generates substandard
product. Alternatively, a process can be out of statistical control but capable of
satisfying specifications. For example, the process is designed to be robust to
deviations from the norm, such that it meets specifications despite high
variability. Given seasonal and other sources of variability beyond a supplier’s
control, the latter situation may be particularly relevant to food production
processes.

SPC charts can be classified as control charts for variables (e.g., average and
range charts) or control charts for attributes (e.g., p charts). Microbiological food
safety characteristics can be classified as variables or attributes. Microbiological
concentration data expressed on a continuous numerical scale are classified as
variable data. Microbiological presence/absence data or concentration data
classified into numerical ranges (e.g., m < x ≤ M) are classified as attribute data.

Montgomery (2) cautions: “It is not possible to give an exact solution to the
problem of control chart design, unless the analyst has detailed information about
both the statistical characteristics of the control chart tests and the economic
factors that affect the problem. A complete solution would require knowledge of
the costs of investigating and possibly correcting the process in response to out-
of-control signals, and the costs associated with producing a product that does
not meet specifications. Given this kind of information, an economic decision
model could be constructed to allow economically optimum control.” However,
such detailed information is not generally available for even a small subset of
food production processes, and the available information is subject to
considerable uncertainty, variability, and disagreement (e.g., regarding consumer
health impacts). Therefore, this discussion is limited to some general guidelines
that will aid in SPC chart design rather than identifying optimal designs.

1
Control Charts for Variables

Control of variable characteristics requires managing both the central tendency

and variability. Measures of central tendency include the mean (µ) and median.
Measures of variability include the standard deviation (σ) and the range (R). The
𝑥𝑥̅ chart is used to monitor control of the process average. Process variability can
be monitored with a control chart for the standard deviation (s chart) or the range
(R chart). Due to its simplicity, the R chart is widely used.

Suppose that the microbiological concentration data (y) from a food process are
lognormally distributed such that the log-transformed data (x = log10(y)) are
normally distributed with mean µ (log10 cfu/g) and standard deviation σ (log10
cfu/g). Estimates of µ and σ are based on an initial process capability study
conducted when the process is considered under statistical control. Let k =
number of lots (subgroups) sampled and n = number of samples per lot
(subgroup). As a rule of thumb, Shewhart (3) suggested that “a sequence of not
less than twenty five samples of size four” is the minimum requirement for
concluding that a process is in a state of statistical control (e.g., 4 samples per lot
from 25 lots).

Let 𝑥𝑥̅1 , 𝑥𝑥̅2 , … , 𝑥𝑥̅𝑘𝑘 be the geometric (log10) sample means from each lot (subgroup).
The estimate of process average (µ) is the grand mean (𝑥𝑥̿ ):
∑𝑘𝑘𝑗𝑗=1 𝑥𝑥̅𝑗𝑗
𝑥𝑥̿ =
𝑘𝑘

The sample range (R) is the difference between the largest and smallest
observations within each lot (subgroup): 𝑅𝑅 = 𝑥𝑥𝑚𝑚𝑚𝑚𝑚𝑚 − 𝑥𝑥𝑚𝑚𝑚𝑚𝑚𝑚 . The average sample
range (𝑅𝑅� ) provides an estimate of the process standard deviation (σ):
∑𝑘𝑘𝑗𝑗=1 𝑅𝑅𝑗𝑗
�
𝑅𝑅 =
𝑘𝑘

𝑅𝑅�
𝜎𝜎� =
𝑑𝑑2

Where d2 is the expected mean of R/σ. Table B.1 provides calculated d2 values
for n=2-25.

2
Table B.1. Factors for Control Charts for Variables Assuming a Normal
Distribution
n d2 d3 A2 D3 D4
2 1.128 0.853 1.880 0.000 3.267
3 1.693 0.888 1.023 0.000 2.575
4 2.059 0.880 0.729 0.000 2.282
5 2.326 0.864 0.577 0.000 2.114
6 2.534 0.848 0.483 0.000 2.004
7 2.704 0.833 0.419 0.076 1.924
8 2.847 0.820 0.373 0.136 1.864
9 2.970 0.808 0.337 0.184 1.816
10 3.078 0.797 0.308 0.223 1.777
11 3.173 0.787 0.285 0.256 1.744
12 3.258 0.778 0.266 0.283 1.717
13 3.336 0.770 0.249 0.307 1.693
14 3.407 0.763 0.235 0.328 1.672
15 3.472 0.756 0.223 0.347 1.653
16 3.532 0.750 0.212 0.363 1.637
17 3.588 0.744 0.203 0.378 1.622
18 3.640 0.739 0.194 0.391 1.609
19 3.689 0.733 0.187 0.404 1.596
20 3.735 0.729 0.180 0.415 1.585
21 3.778 0.724 0.173 0.425 1.575
22 3.819 0.720 0.167 0.435 1.565
23 3.858 0.716 0.162 0.443 1.557
24 3.895 0.712 0.157 0.452 1.548
25 3.931 0.708 0.153 0.459 1.541

By convention, statistical process control limits are based on µ ± 3σ, where

99.7% of values sampled from a normal distribution lie within a “Six Sigma”
interval centered on the mean. Tabled values for factors for control charts for
variables are typically based on the 3σ convention.

The 𝑥𝑥̅ chart monitors between-lot variability in the process mean. The equations
for constructing 3σ upper and lower control limits on the 𝑥𝑥̅ chart are as follows:

𝑈𝑈𝑈𝑈𝑈𝑈𝑥𝑥̅ = 𝑥𝑥̿ + 𝐴𝐴2 𝑅𝑅�

𝐿𝐿𝐿𝐿𝐿𝐿𝑥𝑥̅ = 𝑥𝑥̿ − 𝐴𝐴2 𝑅𝑅�
3
where 𝐴𝐴2 = 𝑑𝑑 . Table B.1 provides A2 values for n = 2 to 25. When 3σ control
2 √𝑛𝑛
limits are used and the process is under statistical control, the probability of a
sample mean being outside the 𝑥𝑥̅ control limits simply due to random chance (the
false-alarm rate (FAR)) is 0.3%. Note that the control limits are based on within-
lot variability (σ) only (2). Thus the conventional limits treat any between-lot

3
variability (𝜎𝜎𝑥𝑥̅ ) as indicating a lack of control, rather than a source of variation that
may be intrinsic to the process. Achieving negligible between-lot variability may
not be feasible in some food production processes. Even for relatively small
sample sizes, the sampling distribution of the sample mean is approximately
normal even if the underlying data are not; although the limit of quantitation
presents a potential complication for microbiological data if the proportion of
negative results is large (1). Montgomery (2) addresses monitoring processes
with a high proportion of data such as those that fall outside the detection limit or
are too numerous to count.

The R chart monitors within-lot variability. The equations for constructing 3σ

upper and lower control limits on the R chart for process variability are as follows:

𝑈𝑈𝑈𝑈𝑈𝑈𝑅𝑅 = 𝐷𝐷4 𝑅𝑅�

𝐿𝐿𝐿𝐿𝐿𝐿𝑅𝑅 = 𝐷𝐷3 𝑅𝑅�

3𝑑𝑑3 3𝑑𝑑3 𝜎𝜎𝑅𝑅

where 𝐷𝐷4 = 1 + , 𝐷𝐷3 = max �0, 1 − �, and 𝑑𝑑3 = . Table B.1 provides d3,
𝑑𝑑2 𝑑𝑑2 𝜎𝜎
D3, and D4 values for n = 2 to 25. Even for normally distributed data, the
sampling distribution of R (with the standard deviation, 𝜎𝜎𝑅𝑅 ) is non-negative and
positively skewed. Therefore, the symmetric 3σ control limits for R are only
approximate, and the actual FAR depends on n and the underlying distribution. In
food safety, the concern would typically be about excessive variation (UCLR), but
insufficient variation (LCLR) may indicate a problem with sampling or analytical
procedures.

The data used to construct 𝑥𝑥̅ and R charts also provide information about
process capability. The two-tailed process capability index (Cp) is defined in
terms of the upper and lower specification limits (USL and LSL):

𝑈𝑈𝑈𝑈𝑈𝑈 − 𝐿𝐿𝐿𝐿𝐿𝐿
𝐶𝐶𝑝𝑝 =
6𝜎𝜎

Note, however, that the equation for Cp only considers process variability. It
implicitly assumes that the process is centered on a mean of (USL-LSL)/2.
Compare to an upper-tail 𝐶𝐶𝑝𝑝 :

Cp = (USL – μ) /(3 σ)

Further details about the statistical basis for control charts for variables are
available in standard texts (e.g., (2) ). On-line calculators for control charts for
variables are available from a variety of sources (e.g.,
http://www.sqconline.com/).

When establishing control limits based on an initial process capability study, it

may be reasonable to remove a few extreme sample values due to assignable
causes from the dataset to better represent common cause variation of a stable

4
process under statistical control. However, extreme values may simply be
random outliers, and identifying an assignable cause for each extreme value may
not be possible. Similarly, apparent patterns in small datasets (e.g., a sequence
of extreme values or trends) may be simply due to random variation. If the initial
data indicate that process variability is not in statistical control, then the control
limits on the 𝑥𝑥̅ chart may not be meaningful. Therefore, beginning the analysis
with the R chart can be useful. It is customary to treat the control limits obtained
in the initial phase as provisional and to update and revise the control limits over
time as additional information is acquired (Appendix I) and the process matures.

BIBLIOGRAPHY

1. International Commission for the Microbiological Specifications of Foods

(ICMSF). 2002. Microorganisms in Foods 7: Microbiological Testing in Food
Safety Management. Springer, New York.
2. Montgomery, D. C. 2005. Introduction to Quality Control. 5th ed. John Wiley &
Sons, Hoboken, NJ.
3. Shewhart, W. A. 1986. Statistical Method from the Viewpoint of Quality Control.
Dover Publications.

5
Appendix C: Process Control for Attributes (p charts)

There are a variety of process control charts for attributes. The p chart is widely
used; it charts the fraction of non-conforming analytical units over a sampling
sequence. The p chart is based on the binomial distribution, which assumes that
there are only two possible outcomes for each observation (conforming or non-
conforming), the proportion of non-conforming analytical units (p) is constant, the
samples are independent (e.g., defects do not cluster), and a fixed sample size
(n). The sample proportion non-conforming (𝑝𝑝̂ ) is the ratio of the number of non-
conforming analytical units (d) observed in a sample of size n:
𝑑𝑑
𝑝𝑝̂ = 𝑛𝑛 (eq. C.1)

For the binomial distribution, the mean and variance of the sampling distribution
of 𝑝𝑝̂ are:
𝜇𝜇𝑝𝑝� = 𝑝𝑝 (eq. C.2)
and
𝑝𝑝(1−𝑝𝑝)
𝜎𝜎𝑝𝑝2� = (eq. C.3)
𝑛𝑛
respectively.

Assuming a constant sample size, the estimated average proportion non-

conforming (𝑝𝑝̅ ) across lots (subgroups) is:
∑𝑚𝑚 𝑑𝑑
𝑖𝑖 ∑𝑚𝑚 𝑝𝑝�
𝑖𝑖
𝑝𝑝̅ = 𝑖𝑖=1 = 𝑖𝑖=1 (eq. C.4)
𝑚𝑚𝑚𝑚 𝑚𝑚
where m = number of lots (subgroups) sampled and n = number of samples per
lot (subgroup).

Conventionally, p chart control limits are based on a symmetric ± 3σ interval

using a normal approximation to the binomial (1):
𝑝𝑝̅ (1−𝑝𝑝̅ )
𝑈𝑈𝑈𝑈𝑈𝑈 = 𝑝𝑝̅ + 3� (eq. C.5)
𝑛𝑛
𝑝𝑝̅ (1−𝑝𝑝̅ )
𝐿𝐿𝐿𝐿𝐿𝐿 = 𝑝𝑝̅ − 3� (eq. C.6)
𝑛𝑛

Note that equations C.5 and C.6 assume that 𝑝𝑝̅ represents the desired target
value of the proportion non-conforming for the process. In food safety, concern
would normally focus on exceeding the upper control limit (UCL); however,
observations below the lower control limit (LCL) could indicate problems with
sampling and analytical procedures, or it could represent an opportunity on how
to improve process quality. On-line calculators are available for computing
conventional 3 sigma control limits for p charts (e.g.,
http://www.sqconline.com/control-chart-calculator-attributes-discrete-data).
It should be noted that the further the target value of p is from 0.5, the larger the
sample size required for the normal approximation to be reasonable. As a
general rule, the normal approximation is reasonable if np ≥ 5 and n(1-p) ≥ 5. In
many food safety applications where the target value for p is substantially less

1
than 0.5, the sample size required for the normal approximation would be costly.
More generally, even if an exact binomial method is used to calculate control
limits, practical application of p charts is limited to cases where the target value
for p is not very small (Table C.1). The sample size should be large enough to
provide a reasonably high degree of confidence of observing at least one non-
conforming unit (1). For example, if the target value for p = 0.01 and n = 5, the
conventional upper control limit is 0.14 (eq. C.5). Consequently, observing a
single non-conforming unit in the sample (𝑝𝑝̂ = 1/5 = 0.2) would suggest a lack of
process control.

Table C.1. Sample size requirements for p chart control limits

p target n1 n2 n3
0.01 500 299 230
0.02 250 149 114
0.03 167 99 76
0.04 125 74 57
0.05 100 59 45
0.10 50 29 22
0.20 25 14 11
0.30 17 9 7
0.40 13 6 5
0.50 10 5 4
n1 = minimum sample size required for normal approximation
n2 = sample size required for 95% confidence of observing at least one non-conforming unit
n3 = sample size required for 90% confidence of observing at least one non-conforming unit

Standard texts provide additional details on constructing and interpreting p charts

(e.g., (1)).

BIBLIOGRAPHY

1. Montgomery, D. C. 2005. Introduction to Quality Control. 5th ed. John Wiley &
Sons, Hoboken, NJ.

2
Appendix D. High-Event Period Process Control

A high-event period may be defined as a production period when the observed

prevalence likely exceeds the expected or design prevalence. Here prevalence
refers to an attribute – either presence/absence or concentration in a range (e.g.,
CFU /g > M). The application of numerical criteria for identifying a high-event
period is intended for cases where the prevalence is impracticably low for p-
charts.

Suppose that the prevalence (p) is constant such that the number of positive test
results (x) out of n independent samples follows a binomial distribution. Then we
can determine combinations of x and n that are unlikely to occur by chance if the
true prevalence is no more than the design prevalence. A sampling period would
proceed until the testing results indicate a high-event period, or non-conformance
with the design prevalence. After appropriate action is taken in response to the
non-conformance, a new sampling period begins.

Tables D.1 and D.2 present the limits of conforming sample results for a false
alarm rate (FAR) of 5% and 1%, respectively. From Table D.1, for example, if the
number of positive test results observed is x = 4 out of n < 198, then there is less
than a 5% chance of observing the data if the true prevalence is 1%. A 5% FAR
might be appropriate for cases of low sampling frequency because a long
sampling period may elapse before a producer receives an indication that the
process is out of control.

Table D.1. High-Event Period Criteria for 5% False Alarm Rate

Design prevalence (p)

Positive 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.05
Test
Results
(x) Samples (n) for given design prevalence
1 71 35 24 18 14 12 10 9 8 7
2 164 82 55 41 33 27 23 21 18 16
3 274 137 91 69 55 46 39 35 31 28
4 395 198 132 99 79 66 57 50 44 40
5 523 262 175 131 105 88 75 66 59 53
6 658 329 220 165 132 110 95 83 74 67
7 797 399 266 200 160 134 115 101 90 81
8 940 471 314 236 189 158 135 119 106 95
9 1086 544 363 273 218 182 156 137 122 110
10 1235 618 413 310 248 207 178 156 139 125

Similarly from Table D.2, if the number of positive test results observed is x = 4
out of n < 129, then there is less than a 1% chance of observing the data if the

1
true prevalence is 1%. A 1% FAR might be appropriate for cases of high
sampling frequency because a sampling period of limited duration would elapse
before a producer receives an indication that the process is out of control.

Table D.2. High-Event Period Criteria for 1% False Alarm Rate

Design prevalence (p)

Acceptance 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.05
Number (c) Samples (n) for given design prevalence
1 30 15 10 7 6 5 4 4 3 3
2 88 44 29 22 18 15 13 11 10 9
3 165 83 56 42 34 28 24 21 19 17
4 257 129 86 65 52 44 37 33 29 27
5 359 180 120 90 73 61 52 46 41 37
6 468 235 157 118 95 79 68 60 53 48
7 583 292 195 147 118 99 85 74 66 60
8 704 353 236 177 142 119 102 90 80 72
9 829 415 278 209 167 140 120 105 94 85
10 957 480 320 241 193 161 139 122 108 98

2
Appendix E. Control Charts for Very Low Prevalence

Here ‘very low prevalence’ is taken to mean that less than 2% of the samples
taken are found positive for the analyte. The testing is typically limited to
presence/absence methods, and the finding of a positive is sufficient to implicate
the underlying lot.

The statistic of interest for this scenario is the prevalence proportion of positive
results, or, equivalently, the ‘mean time between positives’ (‘MTBP’).

E. coli O157:H7 Testing of Ground Beef

Lots of ground beef are tested for E. coli O157:H7. While positives may result in
rejected lots, this testing can also be used for process control.

The observed prevalence depends on the analytical unit size. In order to be

meaningful, the prevalence needs to be referenced to an analytical unit size, i.e.,
a prevalence of positives in X gram samples (e.g., 325 g). Guidelines for this
example require that prevalence of positives should average no higher than 1 in
500 samples or 0.2%, or a MTBP of 500 samples or more (the LSL). Because
the prevalence is so low, the nonparametric method is probably not viable for this
scenario, as a long sampling history would be required to find even a 99th
percentile. For this example we therefore use a parametric approach.

The normal operations are modeled as a Poisson process, with the MTBP
following an exponential distribution, which has a standard deviation equal to the
mean. Hypothetical data for 325-gram samples from several years indicate a
process MTBP of 690 sampling units with a standard deviation of 730, not much
different from 690, supporting the use of an exponential model.

A parametric ‘g-Chart’ is shown in Fig. E.1, with an upper control limit calculated
as:

UCL = MTBP + 3 √[ MTBP (MTBP + 1) (eq. E.1)

= 2762

In the example, the LCL is zero. The factor ‘3’ is based on the 3 sigma
convention and corresponds to the 99.9th percentile (Alternatively, the exact
exponential distribution quantiles could be used).

Also plotted is the exponentially weighted moving average (EWMA) of the MTBP
data, with a smoothing constant of 0.1 and starting value EWMA0 = 690:

EWMAk+1 = EWMAk + 0.10 (MTBPk+1 – EWMAk) (eq. E.2)

1
The EWMA smooths the rough curve and is helpful in visualizing the drift of the
MTBP average.

The g-Chart helps define operational conditions that lead to relatively stable low
prevalence. However, it has the limitation that it is an individual data trend chart
and the LCL is absent. If the data line moves above the UCL, this suggests
maintenance of this state of operations would result in lower prevalence, and
should be investigated to see how this lower prevalence could be sustained.
Also, if the data line exhibits a strong non-random pattern, this suggests a
systematic cause, which should be investigated. Figure E.1. shows a saw-tooth
appearance (‘up’ followed by ‘down’), indicating negative autocorrelation.

Finally, we can supplement the control limits with a ‘runs test’, e.g., if a run of 11
or more MTBP data consecutively fall below the mean MTBP, then ‘abnormal’
operations are detected, and an assignable cause should be sought (for the
exponential distribution, the mean is the 63rd percentile, so 0.6311 = 0.6% FAR).
The longest run observed below the mean MTBP is 7 samples here, which falls
within expectations for normal operations. As an alternative, a ‘center-line’ at
0.693 MTBP could be added, for which results under statistical control are
equally likely to fall on either side. A run of 7 or more results on the either side of
this center line represents detection of an abnormal change in the process that
should be investigated for an assignable cause, i.e., for the exponential
distribution, the median = 0.693 x mean, so 0.57 = 0.8% FAR.

2
 3000

10/5/2011
2500
Interval Between Positives

3/22/2012
6/3/2013
2000

1500

1000 9/21/2012

500

0
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45
Sequence number of Positive found

Data LSL MTBP UCL EWMA

Figure E.1. g-Chart for hypothetical E. coli O157:H7 MTBP data.

3
Appendix F. Control Chart for Low Prevalence with Quantification

This scenario corresponds to prevalence (i.e., presence/absences followed by

quantification of positive samples, or above or below limit of quantification) in the
range 2% to 10% where samples provide quantitative estimates. These data can
be used to create two different control charts, the g-Chart showing time between
positive results, and an individuals chart with adjusted quantiles. As in Appendix
E, the g-Chart helps define operational conditions that lead to relatively stable
low prevalence. The individuals chart detects data that may indicate the
presence of assignable causes that can support further investigation. The
sampling can still be interpreted as the output of a Poisson process, and the
prevalence and MTBP estimated. However, it is also assumed that positive
results now occur often enough that they are routinely quantified.

The prevalence from history may be estimated as

p = (# positives) / (total # samples) (eq. F.1.)

and the MTBP as 1/p, or as the average of the between-positive sampling

intervals as before.

The g-Chart will still be useful for maintaining normal prevalence (MTBP).

If the observed counts (not concentration) in quantitation are below 100 and a
single dilution is used, the counts may be modeled as Poisson distributed. If the
observed counts are 10 or more typically, or multiple dilutions or a most probable
number (MPN) technique is used, the counts (or estimates in the case of MPN)
may be modeled as normally distributed after a logarithmic transformation.

Processes that are in control may be characterized by a constant expected

prevalence with incidental modest contamination. A Poisson distributed
contamination may arise from isolated contamination events such as those
caused by aerosolized particles. Log normally distributed contamination may
arise from splatters or surface-to-surface contact. Processes that are out of
control may result from changes in prevalence of contamination, or increased
counts when they occur.

Data are logarithmic-transform from the original concentration results:

y = log10(x + 0.3 d) (eq. F.2.)

where ‘x’ is the concentration estimated for the positive result, ‘d’ is the
concentration corresponding to a single count result, and ‘y’ is the logarithmic
metamer. For example, if the analytical portion is 1 ml, and there is a one
decimal dilution, then a single count would result in an estimated concentration of

1
10 CFU/ml. Single count measurements (i.e., =10 CFU/ml) would be
transformed as 1.11 = log10(10+0.3 x 10).

Example: Coliforms in Soft Cheese

Lots of soft cheese (e.g., Brie) are sampled and tested for coliform bacteria.
Specifications require each lot should not exceed 1,000 CFU/25g test portion.
History is comprised of 702 samples, of which 28 were positive, for a prevalence
(p) of 3.99%. The MTBP was 24.4 samples with a standard deviation of 27.5
samples, close to the MTBP, supporting the exponential distribution assumption.

Figure F.1. shows the g-Chart, with an UCL exception at sample #309, and a run
of 9 values below the MTBP line. The exception indicates better control is
possible in normal operations, and this should be explored. The run of 9 below
the MTBP line, although not an exception, is suggestive of a problem with control
in this range of samples.

Contamination levels of individual samples may be plotted with an UCL based on

an extreme quantile of normal operations. From the historical data,
nonparametric quantiles were calculated and shown in the second column of
Table F.1.

Table F.1. Quantile results from Soft Cheese history

Quantile Nonparametric Adjusted Normal Quantile

Probability quantile Quantile
Probability
99% 2.30 74.9% 2.40
99.5% 2.60 87.5% 2.77
99.9% 3.65 97.49% 3.40
99.95% 3.84 98.75% 3.62

Note: All quantiles are for log10 transformed positive data using eq.(F.2).

The nonparametric quantiles are derived from the EDF of the 702-sample data.
As a rule, these are very imprecise for probabilities greater than 701.5 / 702 =
99.9%.

Parametrically, we can represent the distribution as a mixture of a binomial

distribution for prevalence and a truncated normal distribution for positive
observations. The hurdle threshold for positive counts is T = log10 (10 + 3) = 1.11.
The observed average and standard deviation of the positive data for y were 1.87
and 0.783, resp. The estimated z-score from a standard normal distribution for
the threshold is 1.11, with an associated probability of only 0.03, indicating a non-
truncated distribution might be useful as a rough approximation.

2
In Table F.1, the ‘Adjusted Quantile Probability’ column is the desired quantile
probability ‘P’ (given in the first column of the table) adjusted for prevalence p.
The adjusted quantile probability P* is given by

P* = [P – (1 – p)] / p (eq. F.3.)

Finally the normal distribution quantiles are the quantiles associated with P*.

Note that the nonparametric and normal quantiles are in excellent agreement
here, supporting the lognormal assumption. The probability of obtaining less than
a single count was about 0.1%, indicating no problem with bias at the low end.

Figure F.2. shows the individuals’ chart (‘i-Chart’) with associated UCLs for some
soft cheese data. The SPC UCL line corresponds to the mean 1.87 plus 2.575
times the standard deviation of 0.783. This SPC UCL line represents the 99.5%
quantile (normal vs. abnormal division) of the lognormally distributed positive
result data, given that a positive result occurs. The Quantile UCL line represents
the nonparametric 99.5% quantile across all results, including those which
correspond to zero counts. The point at sample #12 exceeds both UCLs,
indicating the point is unusual for observation as a result, and also unusual from
the baseline normal distribution point of view. Sample #12 represents an
unexpected shift in operations. The sample #24 result exceeds the Quantile UCL
line, which means it is a rarity in sampling, but does not exceed the SPC UCL
line, indicating it is not that unexpected from a positive data distribution point of
view, so still represents normal operations (same distribution of positive results).
This difference in interpretation between the nonparametric and parametric
approaches shows another advantage (besides allowing the estimation of high
probability quantiles using small samples) of the latter.

3
 5.0

4.0
Concentration (log10)

3.0

2.0

1.0

0.0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28
Sample number

Mean Quantile UCL Data SPC UCL

Figure F.1. g-Chart for coliforms in soft cheese. Note UCL exception around
sample #309 and run of 9 values below the MTBP.

4
 140
Observed Time Between Positives
120

100

80

60

40

20

0
53 77 128 189 312 416 442 460 462 502 553 590 664 678
Sample number

MTBP UCL Data

Figure F.2. Individuals chart (i-Chart) for positive samples observed for coliforms
in Soft Cheese.

5
Appendix G. Control Chart for Moderate to High Prevalence with
Quantitation

This example corresponds to prevalence in the range 10% to 95%. The sampling
can be interpreted as the output of a Bernoulli process, and the prevalence
estimated and controlled. It is also assumed that positive results are routinely
quantitated.

Because of the moderate to high prevalence, rational subgroups (i.e., lots that
represent test units belonging to a homogeneous population with the same
constant population parameters) of samples may be combined to increase
normality and provide better tools for prevalence SPC. If grouping is to be done
by time period, equal sampling for each period is advised, and the sample size
large enough to achieve an expected 5 positive results or more, and similarly for
negative results. For example, if the mean prevalence is 20%, the sample size
should be at least 5 / 0.2 = 25. If the mean prevalence is 80%, the sample size
would also be 25.

SPC may be carried out by a ‘p-Chart’, where the control limits are given by

𝑝𝑝̅ (1−𝑝𝑝̅ )
𝑈𝑈𝑈𝑈𝑈𝑈 = 𝑝𝑝̅ + 3� (eq. G.1.)
𝑛𝑛
𝑝𝑝̅ (1−𝑝𝑝̅ )
𝐿𝐿𝐿𝐿𝐿𝐿 = 𝑝𝑝̅ − 3� (eq. G.2.)
𝑛𝑛

where ‘n’ is the rational subgroup sample size.

The SPC of the distribution of the positive result concentrations may be

performing using an individuals’ chart with control limits typically as

UCL = μ + 3 σ ( eq. G.3.)

LCL = μ - 3 σ (eq. G.4.)

The values for ρ, σ and μ need to be determined from history or a process

capability study.

Although there is fixed sample size during a time period (in this case, 40 samples
per quarter), the number of positive results is a random variable, so a standard
X-bar chart cannot be used.

1
Example: Aerobic Plate Counts in Ground Beef

Lots of ground beef in cold storage are sampled and tested for aerobic plate
counts (APC). Assume the microbiological guidelines require that the APC
should not exceed 10,000,000 CFU/g (i.e., 7.0 on a log10(CFU) scale).

Data consist of 455 samples, of which 393 were positive (86.4% prevalence).
The positive samples had average log10-transformed concentration of 5.19 with
standard deviation of 1.34.

Table G.1. Quantile results from ground beef data

Quantile Nonparametric Adjusted Normal Quantile

Probability Quantile
Probability
99% 8.00 98.84% 8.23
99.5% 8.07 99.42% 8.57
99.9% 8.28 99.88% 9.28
99.95% 8.31 99.94% 9.54
Note: All quantiles are for log10-transformed positive data.

The nonparametric quantiles from the expected distribution function (EDF) are
precise up to 99.5%. The agreement between nonparametric and normal-based
quantiles is good, but not perfect. The probability of getting less than one count
under the normal distribution model is 2.2%, a small error which should be
adjusted out by using a truncated normal distribution, but which we will ignore
here.

Given a prevalence of 86%, the minimum sample size required for the normal
approximation to be valid is 36. Based on this, sampling was carried out by
quarter of the year, with 40 samples taken randomly each quarter. For each
quarter, the proportion of the 40 samples with enumerative results >0 was
calculated via eq. F.1 (Appendix F).

Figure G.1. shows a p-Chart for hypothetical data. Note the exceptions at periods
#7 and #22, either of which would indicate a drop in prevalence of samples with
enumerative results >0.

Figure G.2. shows an i-Chart for the samples with enumerative results >0 with
control limits based on + 3 σ. The data are under control with no exceptions.

2
 1.00

0.95

0.90
Proportion Positive

0.85

0.80

0.75

0.70

0.65

0.60
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
Period

p Average LCL UCL

Figure G.1. p-Chart for aerobic plate count in ground beef. Note the exceptions
at periods #7 and #22 which imply a drop in the proportion of samples with
enumerative results >0 from that expected. There is also a general tendency to
fall under the average.

3
 11.0

9.0
Concentratrion (log10)

7.0

5.0

3.0

1.0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
Positive #

Data Average LCL UCL NP Quantile

Figure G.2. Individuals’ chart for aerobic plate count in ground beef.

4
Appendix H. Control Chart for Very High Prevalence with Quantitation

This example corresponds to prevalence in the range 95% to 100%. For this
case, the number of samples with enumerative results equal to 0 is low enough
to apply eq. F.2. (Appendix F).

Small rational subgroups (i.e., lots) are now possible, corresponding to shift, day,
week, month or quarter, within which operations are expected to be consistent.
The use of such subgroups allows control not only of the mean, but also the
spread of the process.

The X-bar chart has control limits calculated as follows:

UCL ave = μ + A 2 R ave (eq. H.1.)

LCL ave = μ – A 2 R ave (eq.

H.2.)

where the center line is μ (determined from historical data), n is the subgroup
sample size, A 2 is a numerical factor available from standard control chart tables
(Appendix B) and R ave is the average range from the same historical dataset.
The range R is the difference between the maximum and minimum values in a
subgroup.

Variation is controlled by a R Chart, with control limits at

UCL R = R ave + D 4 R ave (eq. H.3.)

LCL R = 0 for n < 6 (eq. H.4.)

where D 4 is obtained from the same table as A 2 . For most purposes, n should
be between 2 and 6, with 4 or 5 preferred. The Range Chart responds to
changes in within subgroup variation.

Table H.1. shows various scenarios for using X-bar-R- control charts for
microbial testing.

1
Table H.1. Average-Range statistical process control chart characteristics for
different sampling time frames

Average-Range Charts for Outgoing Product Sampling

Purpose Verify a general good state of control at vendor output or
customer receiving
Operation 1. Compare variation and level across periods to variation within
periods.
2. Generally, start with long time-scale plans (higher numbers),
with shorter time-scale plans used for tightened inspection or
troubleshooting
Notes 1. Requires quantitative measurement result.
2. Each sampling unit may itself be a composite of specimens
taken from the common time period, with a single combined test
result
n Sample size taken (fixed size)
I. Sample Within Production Shift, n > 2 (best used for tightened inspection or internal
control)
Operation Take sample of size 'n' from each production shift. Plot average
and range of each sample
Average chart Range chart
Purpose Are the sample averages Is the variation observed
consistent across shifts and consistent within shift?
days, free of trends and
disturbances?
Assignable causes 1. Personnel changes 1. New employees within shift.
between shifts/days. 2. New management within
2. Introduction of new raw shift.
material lots. 3. New equipment or
3. Management changes procedures within shift.
between shifts/days.
4. Staffing and volume issues
between shifts.
5. New equipment or
procedures between
shifts/days.
II. Sample Within Production Day, n > 2 (best used for tightened inspection or internal
control)
Operation Take sample of size 'n' from each production day, across shifts.
Plot average and range of each sample
Average chart Range chart
Purpose Are the sample averages Is the variation observed
consistent across days, free of consistent within and across
trends and disturbances? different days' production?
Assignable causes 1. Personnel changes 1. New employees within day.
between days. 2. New management within
2. Introduction of new raw day.
material lots. 3. New equipment or
3. Management changes procedures within day.
between days.
4. Staffing and volume issues
between days.
5. New equipment or
procedures between days.
III. Sample Within Production Week, n = 5, 6, 7 (normal inspection)

2
 Operation Take single unit from each production day during week, across
shifts. The collection of results is the sample. Plot average and
range of each sample.
Average chart Range chart
Purpose Average Chart PURPOSE: Range Chart PURPOSE:
Are the sample averages Is the variation observed
consistent across weeks, free consistent within and across
of trends and disturbances? different weeks' production?
Assignable causes ASSIGNABLE CAUSES: ASSIGNABLE CAUSES:
1. Personnel changes 1. New employees within
between weeks. week.
2. Introduction of new raw 2. New management within
material lots. week.
3. Management changes 3. New equipment or
between weeks. procedures within week.
4. Staffing and volume issues 4. Day of week volume or
between weeks. procedure effects.
5. New equipment or
procedures between days
IV. Sample Within Production 'Month', n = 4 (loosened inspection)
Operation Take single unit from each production week, across days and
shifts. The collection of results is the sample. Plot average and
range of each sample
Average chart Range chart
Purpose Are the sample averages Is the variation observed
consistent across months, free consistent within and across
of trends and disturbances? different months' production?
Assignable causes 1. Personnel changes 1. New employees within
between months. month.
2. Introduction of new vendors. 2. New management within
3. Management changes month.
between months. 3. New equipment or
4. Staffing and volume issues procedures within month.
between months 4. New equipment or
5. New equipment or procedures within month
procedures between months.
6. Seasonal changes in raw
materials and production
volume

3
Example: Aerobic Plate Count in Bagged Salad

Received lots of bagged salad mixed are tested for APC. Five samples are
taken per lot. Based on prior test data, the long-term average APC log 10
concentration is 5.19 with standard deviation of 1.34 and an average range of
3.12.

Figure H.1. shows the Average Chart for recent data. No abnormal behavior is
apparent.

Figure H.2. shows the Range Chart for the same data. Two exceptions at the
UCL are prominent.

8.00

7.00

6.00

5.00
Sample Averages

4.00

3.00

2.00 LCL Ave. UCL Target

1.00
0 10 20 30 40 50 60

Lot

Figure H.1. Average Chart for APC in bagged salad mix.

4
 9.00

8.00

7.00

6.00
Sample Ranges

5.00

4.00

3.00

2.00

1.00 LCL Ave. UCL Target

0.00
0 10 20 30 40 50 60
Lot

Figure H.2. Range Chart for APC in bagged salad mix. Note the two out-of-
control points.

5
Appendix I. Number of Samples and Statistical Uncertainty about Setting
Control Limits for X Bar Charts

There are no firm rules for how much data are needed to develop control charts.
As the number of lots used to develop a control chart increases, the uncertainty
about setting the control limits (too high or too low) decreases. However, there
are diminishing returns to using more data.

For example, Figure I.1 is based on hypothetical data collected at a frequency of

n = 5 samples per lot represented by a lognormal distribution with a geometric
mean = 3 log10 CFU/g (1,000 CFU/g) and a standard deviation = 1 log10 CFU/g.
The grand mean (G mean) is represented by the central solid line. The upper and
lower 3 σ control limits for sample means are represented by dashed lines (UCL,
and LCL, respectively). The uncertainties associated with the statistics due to
random sampling variability are represented by dotted lines (90% confidence
limits). Assuming the process is stable over time, as the number of lots
(subgroups) used to develop an average control chart increases, the uncertainty
about the control limits decreases. The result of using more lots (subgroups) to
develop control limits is increased confidence that the limits are not set too high
or too low relative to the intended design (3 sigma). Uncertainty about the
control limits decreases more slowly than the uncertainty about the mean. This is
due to greater random sampling error in measures of variability.

Figure I.1. Uncertainty about 3 Sigma Control Limits for Mean (µ = 3 log10
CFU/g, σ = 1 log10 CFU/g)

The uncertainty about the control limits depends on the number of samples per
lot (n) as well as the number of lots (subgroups) used to develop the limits.
Assuming a lognormal distribution with geometric mean = 3 log10 CFU/g (1,000
CFU/g) and a standard deviation = 1 log10 CFU/g, Figure I.2 compares the

1
relationship between the uncertainty about average chart control limits and the
number of lots used to develop the limits for n = 5 and n = 3 samples per lot (i.e.,
rational subgroup). Both cases show an initial rapid decrease in uncertainty in
control limits followed by diminishing returns from additional lots. However, the
control limit uncertainty for n = 3 samples per lot (subgroup) starts from a
substantially higher level relative to n = 5 samples per lot (subgroup).

Figure I.2. 90% Confidence Range of 3 Sigma Control Limits for Mean (µ = 3
log10 CFU/g, σ= 1 log10 CFU/g)

It should be remembered that all data included in the dataset used to compute
the control limits should be consistent with a tenable assumption of a time period
of unchanging conditions (e.g., the same season of raw materials, the same
production process, the same equipment).

2
Appendix J. Microbiological and chemical limits for foods that are useful to assess process control and insanitary conditions for United
States Department of Defense (DoD) use.

Introduction

The microbiological and chemical limits provided in the following tables are useful for suppliers and DOD to assess process control and sanitary
conditions associated with the production of various foods. The limits and sample size or procedures described in these tables are not regulatory
limits, although in certain cases, they may reference regulatory limits.

The food categories correspond to those listed in Appendix A where flow diagrams for manufacturing of the foods are provided. The
microbiological data for the various microorganisms or classes of microorganisms can be used to develop statistical process control (SPC) charts,
as well as to gain an understanding of the microbiological quality and safety associated with the various products.

The environmental monitoring program (EMP) data have less utility for development of SPC charts because of the potential high number of
monitoring sites, and thus, the longer time frame required for sufficient data for SPC charts. However, the EMP data have been correlated with
food-product contamination (Kornacki, 2014) and have usefulness for assessing process control, cleaning and sanitation practices, targeting
supplier and DOD resources, and for trending EMP data over time to assess continuous improvement.

Each table in Appendix J includes the microorganisms that are useful for assessing process control and sanitary conditions during production of
foods within the given food category. The microbiological limits for these microorganisms are provided, as well as recommended actions to be
taken if the limits are exceeded. In many instances, the actions include investigating to determine a root cause, developing and implementing
corrective and preventive actions, and conducting follow-up sampling and testing to determine if the corrective and preventive actions have been
effective. In all tables, where applicable, these actions are identified as “Investigate” and “Implement Corrective Actions” to simplify the actions
listed. The investigative and corrective action processes likely will be unique to each situation.

Samples of the food may be taken at numerous points throughout production; these are considered as in-process samples. Samples taken at the
end of the production line also may be tested, with results compared against the microbiological limits. In some cases, these finished product data
may be useful to assist in the development of finished product microbiological criteria; however, initially, these data should be used to assess
process control and sanitary conditions, and compared against the limits provided for each criterion. When samples are taken at the end of the
production line and tested, and results exceed the limits, the recommended action may be to reject the lot of food represented by the sample. This
will be especially true when the microorganism detected is a pathogen and the food will not receive further processing using a validated kill step.

The number of in-process, finished product, or environmental samples to take and test may not be given for all criteria in the tables. In general,
taking more samples is better; and larger numbers of samples taken for pathogens can increase the confidence of detecting pathogens present at
a low prevalence. Analytical unit weights for testing should be a minimum of 25 grams; for pathogen testing, the analytical unit (usually a
composite weight) in the table may specify a particular weight (e.g., 325 or 375 grams) and provide the weights for the individual samples
contributing to the composite sample (e.g., 15 X 25-gram samples to result in a 375-gram analytical unit). The body of the report and Appendix I
discuss how sample numbers affect the design of SPC charts.

Microbiological and chemical limits for foods that are useful to assess process control and insanitary conditions for DoD use.
Table J.1. Microbiological Limits for Bottled Water

Notes: The bottled water category includes bottled water described as Artesian, Mineral, Purified, Sparkling or Spring.

Criteria & EMP Microbiological Limit Recommended Action if Comments

a
Target Microorganism Routine Non-Routine Limit is Exceeded
Reject lot. Investigate; test for generic
E. coli. Notify local authorities if they See 21 CFR 165.110 (b)(2)(i)(A) for
Coliforms <10 in 100 ml
are involved in providing water applicable regulatory standards
treatment.
Reject lot. Investigate. Notify local
authorities if they are involved in
providing water treatment. If water
E. coli (generic) or Negative in 100 See 21 CFR 165.110 (b)(2)(i)(B) for
comes in contact with food, it is
thermotolerant coliforms ml applicable regulatory standards
recommended that the food be
destroyed or reprocessed to kill
vegetative cells.
Investigate. Notify local authorities if
they are involved in providing water
Not routinely tested; however, some
treatment. If water comes in contact
Enterococcus Negative in 250 ml countries (in the EU) test for
with food, it is recommended that the
Enterococcus in lieu of coliforms.
food be destroyed or reprocessed to kill
vegetative cells.
Investigate. Notify local authorities if Reject or divert for further
<100/ml @ 22°C
Heterotrophic plate count they are involved in providing water processing; investigate, implement
<20/ml @ 37°C
treatment. corrective action
Pseudomonas aeruginosa Negative in 250 ml Investigate. Notify local authorities if Not routinely tested, but may be
they are involved in providing water required by individual country’s
treatment. regulations
Unless there is a particular concern
Investigate. Notify local authorities if for parasites or viruses, testing
Parasites & Viruses Negative they are involved in providing water typically is not done; this will be
treatment. very situational and location-
dependent.

a
(Code of Federal Regulations, 2014c; European Communities, 2007; World Health Organization (WHO), 2013, 2014)

Microbiological and chemical limits for foods that are useful to assess process control and insanitary conditions for DoD use.
 Table J. 2. Microbiological Limits for Ice, Packaged

Notes: Testing based on target microorganisms for bottled water.

Criteria & EMP Microbiological Limit Recommended Action if Comments

b
Target Microorganism Routine Non-Routine Limit is Exceeded
<10 in 100ml Investigate. Test generic E. coli. Notify
local authorities if they are involved in
Coliforms
providing water treatment for water
becoming ice.
Investigate. Notify local authorities if
they are involved in providing water
treatment for water becoming ice. If the
ice is for direct consumption, it is
E. coli (generic) or Negative in 100
recommended that the ice not be used.
thermotolerant coliforms ml
If ice comes in contact with food, it is
recommended that the food be
destroyed or reprocessed to kill
vegetative cells.
Investigate. Notify local authorities if Not routinely tested, but may be
Pseudomonas aeruginosa Negative in 250 ml they are involved in providing water required by individual country’s
treatment. regulations
Unless there is a particular concern
Investigate. Notify local authorities if for parasites or viruses, testing
Parasites & Viruses Negative they are involved in providing water typically is not done; this will be
treatment for water becoming ice. very situational and location-
dependent.

b
(Code of Federal Regulations, 2014c; European Communities, 2007; World Health Organization (WHO), 2013, 2014)

Microbiological and chemical limits for foods that are useful to assess process control and insanitary conditions for DoD use.
Table J. 3. Microbiological and Chemical Limits for Juices and Drinks, Pasteurized and Refrigerated

Notes: Examples of these products are orange juice, carrot juice, and some tea beverages. These products are pasteurized but must be kept
refrigerated to prevent spoilage. Raw citrus juices sold in the U.S. will require additional testing (Subpart B, Juice HACCP regulations). Juices with a
pH>4.6 should address control of Clostridium botulinum.

Criteria & EMP Microbiological Limit Recommended Action if Comments

c
Target Microorganism Routine Non-Routine Limit is Exceeded
Coliforms <10/ml Investigate, implement corrective action
This limit is based on the FDA Juice
HACCP regulation requiring a 5-
Negative in 10 Divert for reprocessing, if appropriate,
E. coli (O157:H7 or other log10 reduction. Processors with
individual 25-g or reject. Investigate and implement
STEC) demonstrated control may not need
samples corrective action
to test for E. coli O157:H7 except
for periodic verification purposes.
Investigate, consider Zone 1 and
d Negative for
Listeria spp. (EMP) finished product testing, implement
Zone 2 or 3
corrective action
Different countries may have
The presence of patulin in apple juice
different regulatory requirements. A
e above the limit should lead to rejection
Patulin (in apple juice) 50 µg/kg lower limit of10 µg/kg should be
of the product. Investigate and
considered when apple juice
implement corrective action.
products are intended for infants.
375 g analytical unit composed of
15 X 25-g samples
This limit is based on the FDA Juice
Divert for reprocessing, if appropriate,
HACCP regulation requiring a 5-
Salmonella (product) Negative in 375 g or reject. Investigate and implement
log10 reduction. Processors with
corrective action
demonstrated control may not need
to test for Salmonella except for
periodic verification purposes.

c
(Code of Federal Regulations, 2014a; U. S. Department of Health and Human Services, 2004)
d
(R.B. Tompkin, 2002; R. B. Tompkin, Scott, Bernard, Sveum, & Gombas, 1999)
e
(U. S. Department of Health and Human Services, 2005e)

Microbiological and chemical limits for foods that are useful to assess process control and insanitary conditions for DoD use.
 Table J.4. Microbiological and Chemical Limits for Shelf-stable Beverages

Notes: Examples of these products are carbonated beverages, commercial sterility/ultra-high temperature/aseptic beverages, and some juice
drinks. Microbiological control is accomplished by one or more of the following: low pH, pasteurization (UHT), and carbonation

Criteria & EMP Microbiological Limit Recommended Action if Comments

Target Microorganism Routine Non-Routine Limit is Exceeded
Different countries may have
The presence of patulin in apple juice
different regulatory requirements. A
f above the limit should lead to rejection
Patulin (in apple juice) 50 µg/kg lower limit of10 µg/kg should be
of the product. Investigate and
considered when apple juice
implement corrective action.
products are intended for infants.
g No microbiological testing
Microbiological criteria (NA) NA NA NA
recommended for this product

There are no microbiological limits set for shelf-stable beverages as these products are considered commercially-sterile (i.e., stable at room
temperature under normal handling and storage conditions). Suppliers should be verifying the raw materials used in the formulation of these
products before the process providing commercial sterility. Shelf-stable liquid products should be examined by means other than routine
microbiological testing; if inspection finds bulging containers, pH changes, odors, etc., then further investigation is warranted.

f
(U. S. Department of Health and Human Services, 2005e)
g
(Elliott & Kataoka, 2013)

Microbiological and chemical limits for foods that are useful to assess process control and insanitary conditions for DoD use.
 Table J.5. Microbiological Limits for Dairy- Butter, margarine

Notes: Either formulated with sufficient salt or lactic acid (for unsalted butter) to prevent growth or refrigerated; products containing added
seasoning/herbs/spices may have additional requirements

Criteria & EMP Microbiological Limit Recommended Action if Comments

h
Target Microorganism Routine Non-Routine Limit is Exceeded
i
Coliforms 10/g Investigate, implement corrective action
Due to certain strains being able to
survive milk pasteurization,
i
Enterobacteriaceae 10/g Investigate, implement corrective action enterococci are not widely adopted
as indicators of process hygiene in
j
the dairy industry
Mold/Yeast 20/g Investigate, implement corrective action
Investigate, consider Zone 1 and
k Negative for
Listeria spp. (EMP) finished product testing, implement
Zone 2 or 3
corrective action
Investigate, implement corrective
action.
S. aureus 100/g 4
if >10 /g, reject lot due to potential for
enterotoxin production

h
(National Academies of Science, 2003)
i
Coliforms or Enterobacteriaceae are acceptable for routine testing. (Kornacki, Gurtler, & Stawick, 2013)
j
(Craven, Eyles, & Davey, 2003)
k
(R.B. Tompkin, 2002; R. B. Tompkin et al., 1999)

Microbiological and chemical limits for foods that are useful to assess process control and insanitary conditions for DoD use.
Table J.6. Microbiological Limits for Dairy-Cheese, Hard

Notes: Ex. Parmesan, Cheddar, aw<0.95 and pH<5.6. All cheeses are made with pasteurized milk.

Criteria & EMP Microbiological Limit Recommended Action if Comments

l
Target Microorganism Routine Non-Routine Limit is Exceeded
Coliforms <100/g Investigate, implement corrective action
Investigate, implement corrective
E. coli (generic) <10/g
action; if >100/g, reject lot
Investigate, consider Zone 1 and
m Negative for
Listeria spp. (EMP) finished product testing, implement
Zone 2 or 3
corrective action
May be in-process vat sample due
Listeria spp. (product) Negative in 25 g Reject lot to the aging process for natural
cheese
May be in-process vat sample due
Salmonella Negative in 375 g Reject lot to the aging process for natural
cheese.
Investigate, implement corrective
Test for toxin if slow acid
action.
S. aureus 100/g 4 development; if positive for toxin,
If >10 /g, reject lot due to potential for
destroy product
enterotoxin production

l
(U.S. Department of Health and Human Services, 2010)
m
(R.B. Tompkin, 2002; R. B. Tompkin et al., 1999)

Microbiological and chemical limits for foods that are useful to assess process control and insanitary conditions for DoD use.
Table J.7. Microbiological Limits for Dairy-Cheese, Soft, Semi-Soft, Surface-Ripened

Notes: Ex. Brie, Fresh Mozzarella, aw>0.95 and pH>5.4. All cheeses are made with pasteurized milk.

Criteria & EMP Microbiological Limit Recommended Action if Comments

n
Target Microorganism Routine Non-Routine Limit is Exceeded
Coliforms <100/g Investigate, implement corrective action
Investigate, implement corrective
E. coli (generic) <10/g
action; if >100/g, reject lot
Investigate, consider Zone 1 and
o Negative for
Listeria spp. (EMP) finished product testing, implement
Zone 2 or 3
corrective action
125-g analytical unit composed of 5
Listeria spp. (product) Negative in 125 g Reject lot
x 25-g samples
Salmonella Negative in 375 g Reject lot
Investigate, implement corrective
Test for toxin if slow acid
action.
S. aureus 100/g 4 development; if positive for toxin,
if >10 /g, reject lot due to potential for
destroy product.
enterotoxin production

n
(Australia New Zealand Food Authority, 2014; U.S. Department of Health and Human Services, 2010)
o
(R.B. Tompkin, 2002; R. B. Tompkin et al., 1999)

Microbiological and chemical limits for foods that are useful to assess process control and insanitary conditions for DoD use.
Table J.8. Microbiological Limits for Dairy-Cultured, pH <4.8

Notes: Ex. Sour cream, yogurt, buttermilk; active pH control required

Criteria & EMP Microbiological Limit Recommended Action if Comments

p
Target Microorganism Routine Non-Routine Limit is Exceeded
Investigate, implement corrective
Coliforms 10/g action; if >10/g and used for RTE foods,
reject lot
Investigate, consider Zone 1 and
q Negative for
Listeria spp. (EMP) finished product testing, implement
Zone 2 or 3
corrective action
The presence of mold and yeast
may be influenced by added
ingredients such as fruit purees and
other inclusions. This needs to be
Mold/Yeast 10/g Investigate, implement corrective action considered in assessing mold and
yeast populations as well as
whether any detectable molds and
yeast would grow in the product
during its shelf life.
Investigate, implement corrective
No testing recommended unless
3 action.
S. aureus 10 /g 4 fermentation does not reach pH
if >10 /g, reject lot due to potential for
<4.8 in <8 h
enterotoxin production

p
(U. S. Department of Health and Human Services, 2011j; U.S. Department of Health and Human Services, 2010)
q
(R.B. Tompkin, 2002; R. B. Tompkin et al., 1999)

Microbiological and chemical limits for foods that are useful to assess process control and insanitary conditions for DoD use.
Table J.9. Microbiological Limits for Dairy-Cultured, pH >4.8 and < 5.4

Notes: Ex. Cottage cheese, cream cheese, moisture >50%; active pH control required

Criteria & EMP Microbiological Limit Recommended Action if Comments

r
Target Microorganism Routine Non-Routine Limit is Exceeded
Investigate, implement corrective
action.
Coliforms 10/g
If >10 /g and regulated under PMO,
reject lot due to regulatory limit
Investigate, consider Zone 1 and
s Negative for
Listeria spp. (EMP) finished product testing, implement
Zone 2 or 3
corrective action
The presence of mold and yeast
may be influenced by added
ingredients such as fruit purees and
other inclusions. This needs to be
Mold/Yeast 10/g Investigate, implement corrective action considered in assessing mold and
yeast populations as well as
whether any detectable molds and
yeast would grow in the product
during its shelf life.
Investigate, implement corrective
No testing recommended unless
3 action.
S. aureus 10 /g 4 fermentation does not reach pH
If >10 /g, reject lot due to potential for
<4.8 in <8 h
enterotoxin production

r
(Bradley et al., 2013; U. S. Department of Health and Human Services, 2011j)
s
(R.B. Tompkin, 2002; R. B. Tompkin et al., 1999)

Microbiological and chemical limits for foods that are useful to assess process control and insanitary conditions for DoD use.
 Table J.10. Microbiological Limits for Dairy-Dried Products

Notes: Ex. NFDM, whey powder. This does not cover dried dairy ingredients used in infant formula; those requirements are more stringent.

Criteria & EMP Microbiological Limit Recommended Action if Comments

t
Target Microorganism Routine Non-Routine Limit is Exceeded
4
APC/SPC 1X10 /g Investigate, implement corrective action
Investigate, implement corrective
action.
100/g 4
if >10 /g, reject lot due to potential for
B. cereus enterotoxin production
u
Coliforms 10/g Investigate, implement corrective action
Enterobacteriaceae 10/g Investigate, implement corrective action
Investigate, consider Zone 1 and
Negative for Zone 2
finished product testing, implement
v or 3
Listeria spp. (EMP) corrective actions
Investigate, implement corrective
action.
100/g 2
if >10 /g, reject lot due to potential for
S. aureus enterotoxin production
Investigate, consider Zone 1 and
Negative for
finished product testing, implement
w Zone 2 or 3
Salmonella (EMP) corrective action
As an alternative sampling option to
collecting and compositing 15-25 g
Divert for reprocessing, if appropriate, samples (total 375 g), an auto
Negative. in 375
or reject. Investigate and implement sampler can be used to collect
g
corrective action small amounts of samples
throughout a production run for a
x
Salmonella total of 375g

t
(International Commission for the Microbiological Specifications for Foods (ICMSF), 2011; U. S. Department of Health and Human Services,
2011j; U.S. Department of Health and Human Services, 2010)
u
Coliforms or Enterobacteriaceae are acceptable for routine testing. (Kornacki et al., 2013)
v
(R.B. Tompkin, 2002; R. B. Tompkin et al., 1999)
w
(Chen et al., 2009; Grocery Manufacturers Association, 2009)
x
Recommend 1500 g per lot when high volumes of product are produced per lot (or production day).

Microbiological and chemical limits for foods that are useful to assess process control and insanitary conditions for DoD use.
 Table J.11. Microbiological Limits for Dairy-Frozen Desserts

Notes:

Criteria & EMP Microbiological Limit Recommended Action if Comments

y
Target Microorganism Routine Non-Routine Limit is Exceeded
Populations may be influenced by
Investigate, implement corrective
ingredients; product specific APC
APC See comments action.
limits need to be established based
on baseline testing
Investigate, implement corrective
Populations may be influenced by
Coliforms 100/g action.
ingredients
Reject lot; Investigate, implement 250 g analytical unit composed of
Salmonella (product) Negative in 250 g
corrective action. 10 x 25-g samples
Listeria spp. (EMP) Negative for Investigate, consider Zone 1 and
Zone 2 or 3 product testing, implement corrective
action

y
(International Commission for the Microbiological Specifications for Foods (ICMSF), 2011)

Microbiological and chemical limits for foods that are useful to assess process control and insanitary conditions for DoD use.
Table J.12. Microbiological Limits for Dairy-Milk and Milk Products (Fluid)

Notes: Ex. Fluid milk, cream; Pasteurized, refrigerated; alkaline phosphatase negative (less than 2.0 micrograms phenol equivalent per g)

Criteria & EMP Microbiological Limit Recommended Action if Comments

z
Target Microorganism Routine Non-Routine Limit is Exceeded
4
APC/SPC 2.0 x 10 /g Investigate, implement corrective action
Investigate, implement corrective
aa action.
Coliforms 10/g
if >10 /g and regulated under PMO,
reject lot due to regulatory limit
Investigate, implement corrective
Enterobacteriaceae 10/g
actions
Investigate, consider Zone 1 and
bb Negative for
Listeria spp. (EMP) product testing, implement corrective
Zone 2 or 3
action
Investigate, implement corrective
action. No testing recommended unless
S. aureus 100/g 4
if >10 /g, reject lot due to potential for temperature abuse is suspected
enterotoxin production

z
(Bradley et al., 2013; U. S. Department of Health and Human Services, 2011j)
aa
Coliforms or Enterobacteriaceae are acceptable for routine testing. (Kornacki et al., 2013)
bb
(R.B. Tompkin, 2002; R. B. Tompkin et al., 1999)

Microbiological and chemical limits for foods that are useful to assess process control and insanitary conditions for DoD use.
Table J.13. Microbiological Limits for Dairy- Processed Cheese

Notes: Manufactured by heating cheese with water, emulsifier and other ingredients to kill vegetative pathogens; molten cheese may then be hot-filled
into loaves or blocks and chilled or cut into individual slices for use; these cheeses are intended to be stored refrigerated. Shelf-stable hot-filled cheese
spreads or cheese sauces must be formulated for safety to inhibit Clostridium botulinum.
Criteria & EMP Microbiological Limit Recommended Action if Comments
Target Microorganism Routine Non-Routine Limit is Exceeded
Test for products that are not hot-
filled directly into final container.
APC limit may be adjusted subject
3 Investigate and implement corrective to control chart associated with
APC/SPC 10 /g
action. Statistical Process Control.
Populations are predominantly
sporeformers or heat-stable
spoilage microorganisms.
Test for products that are not hot-
Coliforms <10/g Investigate, implement corrective action
filled directly into final container
Investigate and implement corrective Test for products that are not hot-
E. coli (generic) <10/g
action filled directly into final container
Investigate, consider Zone 1 and
cc Negative for
Listeria spp. (EMP) finished product testing, implement
Zone 2 or 3
corrective action
Investigate, implement corrective
action. Test for products that are not hot-
S. aureus <100/g 4
if >10 /g, reject lot due to potential for filled directly into final container
enterotoxin production
Investigate, consider Zone 1 and Test EMP in areas where products
dd Negative for
Salmonella (EMP) finished product testing, implement are not hot-filled directly into final
Zone 2 or 3
corrective action container
Divert for reprocessing, if appropriate,
Test for products that are not hot-
Salmonella (product) Negative in 125 g or reject. Investigate and implement
filled directly into final container
corrective action

cc
(R.B. Tompkin, 2002; R. B. Tompkin et al., 1999)
dd
(Chen et al., 2009; Grocery Manufacturers Association, 2009)

Microbiological and chemical limits for foods that are useful to assess process control and insanitary conditions for DoD use.
Table J.14. Microbiological Limits for Egg Products-Pasteurized, Processed

Notes:

Criteria & EMP Microbiological Limit Recommended Action if Comments

ee
Target Microorganism Routine Non-Routine Limit is Exceeded
4
APC 10 /g Investigate, implement corrective action
Investigate, implement corrective
3 action.
B. cereus 10 /g 4
if >10 /g, reject lot due to potential for
enterotoxin production
Coliforms 10/g Investigate, implement corrective action
Enterobacteriaceae 100/g Investigate, implement corrective action
Investigate, consider Zone 1 and
ff Negative for
Listeria spp. (EMP) finished product testing, implement
Zone 2 or 3
corrective action
Divert for reprocessing, if appropriate,
Listeria spp. (product) Negative in 100 g or reject. Investigate and implement
corrective action
Investigate, implement corrective
action.
S. aureus 10/g 4
if >10 /g, reject lot due to potential for
enterotoxin production
Investigate, consider Zone 1 and
gg Negative for
Salmonella (EMP) finished product testing, implement
Zone 2 or 3
corrective action
Divert for reprocessing, if appropriate,
Salmonella (product) Negative in 100 g or reject. Investigate and implement
corrective action

ee
(Australia New Zealand Food Authority, 2014; European Commission, 2005; International Commission for the Microbiological Specifications for
Foods (ICMSF), 2011; U. S. Department of Agriculture Food Safety Inspection Service, 2014c)
ff
(R.B. Tompkin, 2002; R. B. Tompkin et al., 1999)
gg
(Chen et al., 2009; Grocery Manufacturers Association, 2009)

Microbiological and chemical limits for foods that are useful to assess process control and insanitary conditions for DoD use.
Table J.15. Microbiological Limits for Egg Products-Shell Eggs, Raw
Notes:

Criteria & EMP Microbiological Limit Recommended Action if Comments

hh
Target Microorganism Routine Non-Routine Limit is Exceeded
ii 3
Coliforms 10 /g Investigate, implement corrective action
ii 3
E. coli (generic) 10 /g Investigate, implement corrective action
4
Enterobacteriaceae 10 /g Investigate, implement corrective action
If environment is positive for Salmonella
Negative for Salmonella Enteritidis
Salmonella (EMP) Enteritidis, conduct egg sampling per
Zone 2 or 3 environmental testing
FDA Final Rule 2009

hh
(U. S. Department of Health and Human Services, 2009)
ii
Coliforms, E. coli or Enterobacteriaceae are acceptable for routine testing. (Kornacki et al., 2013)

Microbiological and chemical limits for foods that are useful to assess process control and insanitary conditions for DoD use.
 Table J.16. Microbiological Limits for Grain-based products-RTE, baked items, refrigerated or TCS

Notes: Examples: focaccia, custard or cream-filled pastries, pies. Qualifying information: APC counts may be high due to containing ingredients
prepared with starter culture

Criteria & EMP Microbiological Limit Recommended Action if Comments

Target Microorganism Routine Non-Routine Limit is Exceeded
Investigate, implement corrective
3
10 /g action.
B. cereus 4
if >10 /g, reject lot due to potential for
enterotoxin production
Investigate process and implement
Coliforms 100/g
corrective action
Investigate process and implement
E. coli (generic) 10/g
corrective action
Investigate, consider Zone 1 and Periodic finished product testing
jj Negative for
Listeria spp. (EMP) finished product testing, implement (test/hold) for products that support
Zone 2 or 3
corrective action growth of L. monocytogenes
Investigate, implement corrective
3 action.
S. aureus 10 /g 4
if >10 /g, reject lot due to potential for
enterotoxin production
kk Reject lot. Investigate and implement
Salmonella negative in 375 g
corrective action

jj
(R.B. Tompkin, 2002; R. B. Tompkin et al., 1999)
kk
(Andrews, Jacobson, & Hammack, 2014)

Microbiological and chemical limits for foods that are useful to assess process control and insanitary conditions for DoD use.
 Table J.17. Grain-based products-RTE, baked items, shelf stable, non-TCS

Notes: Examples: bread. If raw ingredients added after baking step, additional risks should be considered. ICMSF 8 does not recommend routine
testing.
:

Criteria & EMP Microbiological Limit Recommended Action if Comments

ll
Target Microorganism Routine Non-Routine Limit is Exceeded
4 Investigate process and implement Populations predominantly
APC 10 /g
corrective action sporeformers
mm Investigate process and implement
Coliforms 100/g
corrective action
Investigate process and implement
Enterobacteriaceae 100/g
corrective action
Salmonella testing is appropriate if
Reject lot; investigate process and
Salmonella (product) Negative in 125 g raw ingredients (e.g., nuts, raw
implement corrective action
flour) are added post-baking

ll
(International Commission for the Microbiological Specifications for Foods (ICMSF), 2011)
mm
Coliforms or Enterobacteriaceae are acceptable for routine testing. (Kornacki et al., 2013)

Microbiological and chemical limits for foods that are useful to assess process control and insanitary conditions for DoD use.
 Table J.18. Microbiological Limits for Grain Based Products, RTE, cereals

Notes: Examples: breakfast cereals. Grain based product undergoes a lethality step; mycotoxin surveillance testing completed on incoming
grains as pre-requisite program with limits based on individual country’s regulations

Criteria & EMP Microbiological Limit Recommended Action if Comments

Target Microorganism Routine Non-Routine Limit is Exceeded
4 Investigate process and implement
APC 5x10 /g
corrective action
nn Investigate process and implement
Coliforms 100/g
corrective action
Investigate process and implement
Enterobacteriaceae 100/g
corrective action
Investigate, consider Zone 1 and
oo Negative for
Salmonella (EMP) finished product testing, implement If Zone 1 positive, reject lot
Zone 2 or 3
corrective action
Salmonella Negative in 375 g Reject lot; investigate process and 375-g analytical composed of 15 x
implement corrective action 25-g samples

nn
Coliforms or Enterobacteriaceae are acceptable for routine testing. (Kornacki et al., 2013)
oo
(Chen et al., 2009; Grocery Manufacturers Association, 2009)

Microbiological and chemical limits for foods that are useful to assess process control and insanitary conditions for DoD use.
 Table J.19. Microbiological Limits for Grain Based Products – RTE, cold-pressed bars

Notes; Ex. granola bars; Qualifying information: ingredients will undergo mycotoxin surveillance testing as appropriate; shelf-stable, aw <0.85

Criteria & EMP Microbiological Limit Recommended Action if Comments

Target Microorganism Routine Non-Routine Limit is Exceeded
4 Investigate process and implement
APC 5x10 /g
corrective action
pp Investigate process and implement
Coliforms 100/g
corrective action
Investigate process and implement
Enterobacteriaceae 100/g
corrective action
Investigate, consider Zone 1 and
qq Negative for
Salmonella (EMP) finished product testing, implement
Zone 2 or 3
corrective action
Salmonella Negative in 375 g Reject lot; investigate process and 375-g analytical unit composed of
implement corrective action 15 x 25-g samples

pp
Either coliform or Enterobacteriaceae testing is appropriate. (Kornacki et al., 2013)
qq
(Chen et al., 2009; Grocery Manufacturers Association, 2009)

Microbiological and chemical limits for foods that are useful to assess process control and insanitary conditions for DoD use.
 Table J.20. Microbiological Limits for Grain Based Products non-RTE, dry, flour based mixes
rr
Notes: Flour can contain pathogens occasionally and should be subjected to a lethality step prior to consumption

Criteria & EMP Microbiological Limit Recommended Action if Comments

Target Microorganism Routine Non-Routine Limit is Exceeded
No microbiological testing
Not Applicable (NA) NA NA NA
recommended for this product

rr
(Sperber & North American Millers' Association Microbiology Working Group, 2007)

Microbiological and chemical limits for foods that are useful to assess process control and insanitary conditions for DoD use.
 Table J.21. Microbiological Limits for Grain Based Products – Non-RTE, pasta, dried or refrigerated

Criteria & EMP Microbiological Limit Recommended Action if Comments

Target Microorganism Routine Non-Routine Limit is Exceeded
High APC counts for unheated
6
APC 10 /g products made with raw flour are
not unexpected
Investigate process and implement Periodic testing recommended for
E. coli (generic) 100/g
corrective action refrigerated pasta
Investigate, consider Zone 1 and
ss Negative for Recommended in facilities
Listeria spp. (EMP) finished product testing, implement
Zone 2 or 3 manufacturing refrigerated pasta
corrective action
Investigate, implement corrective
3 action.
S. aureus 10 /g 4
if >10 /g, reject lot due to potential for
enterotoxin production
Investigate, consider Zone 1 and
tt Negative for Recommended in facilities
Salmonella (EMP) finished product testing, implement
Zone 2 or 3 manufacturing dried pasta
corrective action
Divert for reprocessing, if appropriate,
Salmonella Negative in 125 g or reject. Investigate and implement
corrective action

ss
(R.B. Tompkin, 2002; R. B. Tompkin et al., 1999)
tt
(Chen et al., 2009; Grocery Manufacturers Association, 2009)

Microbiological and chemical limits for foods that are useful to assess process control and insanitary conditions for DoD use.
Table J.22. Microbiological and Chemical Limits for Meals and Entrees—Non-RTE, ready-to-cook meals, includes raw ingredients
Notes: This category includes a wide variety of products and processes that will influence appropriate testing choices. Depending on their
ingredients, foods that are likely to be prepared by microwave heating or those that are not labeled with validated cooking instructions may require
more stringent testing (e.g. Raw poultry, beef or flour ingredients may necessitate routine, instead of non-routine, testing for Salmonella or E. coli
(O157:H7 or other STEC).

Criteria & EMP Microbiological Limit Recommended Action if Comments

Target Microorganism Routine Non-Routine Limit is Exceeded
Investigate, implement corrective
Test if food contains components
3 action.
B. cereus 10 /g 4 that are high risk for B. cereus, such
if >10 /g, reject lot due to potential for
as cooked rice
enterotoxin production
uu 3
Coliforms 10 /g Investigate, implement corrective action
Testing if raw, non-intact beef
Divert for reprocessing, if appropriate,
E. coli (O157:H7 or other component is present. Validated
Negative in 25g or reject. Investigate and implement
STEC) cooking instructions should be
corrective action
present on package.
4
Enterobacteriaceae 10 /g Investigate, implement corrective action
Investigate, implement corrective
3 action.
S. aureus 10 /g 4
if >10 /g, reject lot due to potential for
enterotoxin production
Investigate, consider Zone 1 and
vv Negative for Zone 2
Salmonella (EMP) finished product testing, implement
or 3
corrective action
Divert for reprocessing, if appropriate,
Salmonella (product) Negative in 25 g or reject. Investigate and implement
corrective action
Histamine testing appropriate only
when scombroid species are
ww Reject lot; Investigate, implement
Histamine 50 ppm in 250 g present; The FDA Hazards and
corrective action
Controls Guide lists a defect
action level of 50ppm
Table J.23. Microbiological and Chemical Limits for Meals and Entrees--RTE, deli salads, sandwiches, heat-eat meals, sushi

uu
Coliforms or Enterobacteriaceae are acceptable for routine testing. (Kornacki et al., 2013)
vv
(Chen et al., 2009; Grocery Manufacturers Association, 2009)
ww
(U. S. Department of Health and Human Services, 2011a)

Microbiological and chemical limits for foods that are useful to assess process control and insanitary conditions for DoD use.
Notes: Survey data indicates a wide range in microbial populations depending on specific food. Items may include ingredients that are raw.
Criteria & EMP Microbiological Limit Recommended Action if Comments
Target Microorganism Routine Non-Routine Limit is Exceeded
Investigate, implement corrective
Test if food contains components
3 action.
B. cereus 10 /g 4 that are high risk for B. cereus, such
If >10 /g, reject lot due to potential for
as cooked rice
enterotoxin production
xx 4
Coliforms 10 /g Investigate, implement corrective action
E. coli (generic) 100/g Investigate, implement corrective action
4
Enterobacteriaceae 10 /g Investigate, implement corrective action
Investigate, consider Zone 1 and
yy Negative for
Listeria spp. (EMP) finished product testing, implement
Zone 2 or 3
corrective action
Investigate, implement corrective
3 action.
S. aureus 10 /g 4
If >10 /g, reject lot due to potential for
enterotoxin production
Investigate, consider Zone 1 and
zz Negative for
Salmonella (EMP) finished product testing, implement
Zone 2 or 3
corrective action
Reject lot; Investigate, implement 375-g analytical unit composed of
Salmonella (product) Negative in 375 g
corrective action 15 x 25-g samples
Histamine testing appropriate only
when scombroid species are
aaa Reject lot; Investigate, implement
Histamine 50 ppm in 250 g present; The FDA Hazards and
corrective action
Controls Guide lists a defect action
level of 50ppm.

xx
Coliforms, E. coli, or Enterobacteriaceae are acceptable for routine testing. (Kornacki et al., 2013)
yy
(R.B. Tompkin, 2002; R. B. Tompkin et al., 1999)
zz
(Chen et al., 2009; Grocery Manufacturers Association, 2009)
aaa
(U. S. Department of Health and Human Services, 2011a)

Microbiological and chemical limits for foods that are useful to assess process control and insanitary conditions for DoD use.
Table J.24. Microbiological Limits for Meals and Entrees—RTE, sous-vide, cook and chill

Notes: These products receive a lethality treatment; presence of vegetative microbes represents post-process contamination. If not using validated
sous-vide process for a 6-log reduction of non-proteolytic Clostridium botulinum, testing of vegetative microorganisms is warranted.

Criteria & EMP Microbiological Limit Recommended Action if Comments

Target Microorganism Routine Non-Routine Limit is Exceeded
3
APC 10 /g Investigate, implement corrective action
Investigate, implement corrective
Test if food contains components
3 action.
B. cereus 10 /g 4 that are high risk for B. cereus, such
If >10 /g, reject lot due to potential for
as cooked rice
enterotoxin production
bbb
Coliforms 100/g Investigate, implement corrective action
Investigate, implement corrective
E. coli (generic) 10/g action; reject lot or divert for recooking if
appropriate
Enterobacteriaceae 100/g Investigate, implement corrective action
If greater than 500 CFU/g, indicator
Divert for reprocessing, if appropriate,
ccc of loss of process control or
Clostridium perfringens 500/g or reject. Investigate and implement
potential deviation from USDA
corrective action
cooling requirements

bbb
Coliforms or Enterobacteriaceae are acceptable for routine testing. (Kornacki et al., 2013)
ccc
(U. S. Department of Agriculture Food Safety Inspection Service, 1999; U.S. Department of Agriculture Food Safety Inspection Service, 1999)

Microbiological and chemical limits for foods that are useful to assess process control and insanitary conditions for DoD use.
Table J.25. Microbiological Limits for Meat—Beef and Pork, Non-RTE, raw (intact, non-intact)

Criteria & EMP Microbiological Limit Recommended Action if Comments

ddd
Target Microorganism Routine Non-Routine Limit is Exceeded
5
APC 10 /g Investigate, implement corrective action
3
Coliforms 10 /g Investigate, implement corrective action
E.coli (generic) 500/g Investigate, implement corrective action
Divert for lethality step, if appropriate, or Test non-intact beef (ground,
E. coli (O157:H7 and/or reject. Investigate and implement tenderized, enhanced) product and
eee Negative in 325 g
other STEC) corrective action intact product intended to become
non-intact
4
Enterobacteriaceae 10 /g Investigate, implement corrective action
See sampling for
fff USDA-FSIS Not used an accept/reject criterion;
Salmonella (product) Investigate, implement corrective action
Performance used for process control
Standards

ddd
(National Advisory Committee on Microbiological Criteria for Foods, 2013; U.S. Department of Agriculture Agricultural Marketing Service (AMS),
2015)
eee
(U. S. Department of Agriculture, 2011)
fff
(U. S. Department of Agriculture Food Safety Inspection Service, 2014e)

Microbiological and chemical limits for foods that are useful to assess process control and insanitary conditions for DoD use.
Table J.26. Microbiological Limits for Meat—Poultry, Non-RTE, raw (intact, non-intact)

Criteria & EMP Microbiological Limit Recommended Action if Comments

Target Microorganism Routine Non-Routine Limit is Exceeded
6
APC 10 /g Investigate, implement corrective action
ggg 3
Coliforms 10 /g Investigate, implement corrective action
3
E. coli (generic) 10 /g Investigate, implement corrective action
4
Enterobacteriaceae 10 /g Investigate, implement corrective action
See sampling for
hhh USDA-FSIS Not used an accept/reject criterion;
Salmonella (product) Investigate, implement corrective action
Performance used for process control
Standards

ggg
Coliforms, generic E. coli, Enterobacteriaceae are acceptable for testing. (Kornacki et al., 2013)
hhh
(U. S. Department of Agriculture Food Safety Inspection Service, 2014e); Campylobacter is also proposed In the USDA-FSIS Performance
Standards.

Microbiological and chemical limits for foods that are useful to assess process control and insanitary conditions for DoD use.
Table J.27. Microbiological Limits for Meat— RTE cooked, perishable

Notes: Ex. Includes beef, pork and poultry products, deli meats, frankfurters

Criteria & EMP Microbiological Limit Recommended Action if Comments

Target Microorganism Routine Non-Routine Limit is Exceeded
3
APC 10 /g Investigate, implement corrective action
iii
Coliforms 100/g Investigate, implement corrective action
E. coli (generic) 10/g Investigate, implement corrective action
Enterobacteriaceae 100/g Investigate, implement corrective action
Investigate, consider Zone 1 and
jjj Negative for
Listeria spp. (EMP) finished product testing, implement
Zone 2 or 3
corrective action
Divert for reprocessing, if appropriate,
kkk 125-g analytical unit composed of 5
Listeria spp. (product) Negative in 125 g or reject. Investigate and implement
x 25-g samples
corrective action
Salmonella Negative in 125 g Investigate, implement corrective action
If greater than 500 CFU/g, indicator
Divert for reprocessing, if appropriate,
lll of loss of process control or
Clostridium perfringens 500/g or reject. Investigate and implement
potential deviation from USDA
corrective action
cooling requirements

iii
Either coliforms or Enterobacteriaceae are acceptable for testing. (Kornacki et al., 2013)
jjj
(R.B. Tompkin, 2002; R. B. Tompkin et al., 1999)
kkk
(Australia New Zealand Food Authority, 2014; Canada Food Inspection Agency, 2012; U. S. Department of Agriculture Food Safety Inspection
Service, 2014a; U.S. Department of Health and Human Services, 2008)
lll
(U. S. Department of Agriculture Food Safety Inspection Service, 1999)

Microbiological and chemical limits for foods that are useful to assess process control and insanitary conditions for DoD use.
Table J.28. Microbiological Limits for Meat— RTE fermented, dried

Notes: Includes beef, pork and poultry products, Jerky, dried fermented sausage, dried acidified meat sticks; e.g. aw <0.85 or pH <5.3 and aw <0.92 for
vacuum-packaged meat sticks. Products should be manufactured with a validated kill step for E. coli O157:H7 (beef) or Salmonella (pork, poultry) as
appropriate for the given meat matrix.

Criteria & EMP Microbiological Limit Recommended Action if Comments

mmm
Target Microorganism Routine Non-Routine Limit is Exceeded
nnn
Coliforms 100/g Investigate, implement corrective action
E. coli (generic) 10/g Investigate, implement corrective action
Divert for reprocessing, if appropriate,
E. coli (O157:H7 or other 125-g analytical unit composed of 5
Negative in 125 g or reject. Investigate and implement
STEC) x 25-g samples
corrective action
Enterobacteriaceae 100/g Investigate, implement corrective action
Testing of in-process (uncooked)
Investigate, implement corrective
product may be appropriate if
3 action.
S. aureus 10 /g 4 fermentation does not meet USDA-
If >10 /g, reject lot due to potential for
accepted guidelines for
enterotoxin production ooo
temperature-hours
Divert for reprocessing, if appropriate,
125-g analytical unit composed of 5
Salmonella Negative in 125 g or reject. Investigate and implement
x 25-g samples
corrective action

mmm
(U. S. Department of Agriculture Food Safety Inspection Service, 2014a)
nnn
Either coliforms or Enterobacteriaceae are acceptable for testing. (Kornacki et al., 2013)
ooo
(American Meat Institute Foundation, 1997)

Microbiological and chemical limits for foods that are useful to assess process control and insanitary conditions for DoD use.
Table J.29. Microbiological and Chemical Limits for Nuts and Nut Butters – RTE, Not processed for lethality

Notes: Ex.: include peanuts, tree nuts (e.g., walnuts, almonds, pecans, pistachios, macadamia)

Criteria & EMP Microbiological Limit Recommended Action if Comments

Target Microorganism Routine Non-Routine Limit is Exceeded
If 2 of 10 samples are >0.36
E. coli (generic) 0.36 MPN/g Investigate, implement corrective action
MPN/g, the product is violative
Investigate root cause of positive
Negative for
ppp results, conduct vector sampling and
Salmonella EMP Zone 2 and 3
repeat sampling until confirm negative
surfaces
results
Divert for reprocessing, if appropriate,
qqq Negative in 2 X Two 375-g analytical units derived
Salmonella (product) or reject. Investigate and implement
375-g samples from 30 x 25-g samples
corrective action
This is routine testing for peanuts,
rrr pistachios & Brazil nuts, but non-
Toxins – Aflatoxin B1 20 ppb Investigate, implement corrective action
routine for other nut types or
situations

ppp
(Chen et al., 2009; Grocery Manufacturers Association, 2009)
qqq
(Andrews & Hammack, 2003; U.S. Food and Drug Administration, 2015)
rrr
(U. S. Department of Health and Human Services, 2005a, 2005c, 2005g, 2005i)

Microbiological and chemical limits for foods that are useful to assess process control and insanitary conditions for DoD use.
Table J.30. Microbiological and Chemical Limits for Nuts and Nut Butters – RTE, processed for lethality
Notes: Ex.: peanut butter, almond butter, roasted nuts

Criteria & EMP Microbiological Limit Recommended Action if Comments

Target Microorganism Routine Non-Routine Limit is Exceeded
Investigate root cause of positive
Negative for
sss results, conduct vector sampling and
Salmonella EMP Zone 2 and 3
repeat sampling until confirm negative
surfaces
results
Divert for reprocessing, if appropriate,
ttt Negative in 2 x Two 375 g analytical units derived
Salmonella (product) or reject. Investigate and implement
375-g samples from 30 x 25-g samples
corrective action
This is routine testing for peanuts,
uuu Reject lot; investigate, implement pistachios & Brazil nuts, but non-
Toxins – Aflatoxin B1 20 ppb
corrective action routine for other nut types or
situations

sss
(Chen et al., 2009; Grocery Manufacturers Association, 2009)
ttt
(Andrews & Hammack, 2003; U.S. Food and Drug Administration, 2015)
uuu
(U. S. Department of Health and Human Services, 2005a, 2005c, 2005g, 2005i)

Microbiological and chemical limits for foods that are useful to assess process control and insanitary conditions for DoD use.
Table J.31. Microbiological Limits for Produce—Fruits and Vegetables, Cut, Frozen or Refrigerated

Notes: minimally processed

Criteria & EMP Microbiological Limit Recommended Action if Comments

vvv
Target Microorganism Routine Non-Routine Limit is Exceeded
Consider improvements in production
E. coli (generic) 100/g
hygiene and selection of raw materials
Depending on commodity,
E. coli (O157:H7 or other Reject lot; investigate, implement
Negative geographical location and use of
STEC) corrective action
GAPs
www Negative zone 2 Consider zone 1 and finished product
Listeria spp. (EMP)
or 3 testing; implement corrective action
Depending on commodity,
Listeria spp. (finished Reject lot; investigate, implement geographical location and use of
Negative
product) corrective action GAPs; Sample size may vary; e.g.
25 g
Depending on commodity,
Reject lot; investigate, implement geographical location and use of
Salmonella (product) Negative
corrective action GAPs; Sample size may vary; e.g.
25 g

vvv
(European Commission, 2005; Health Canada, 2008)
www
(R.B. Tompkin, 2002; R. B. Tompkin et al., 1999)

Microbiological and chemical limits for foods that are useful to assess process control and insanitary conditions for DoD use.
Table J.32. Microbiological Limits for Produce—Fruits and Vegetables, Whole

Notes: Ex. products customarily consumed without cooking, tomatoes, cantaloupes, avocado, mangoes, apples, celery, carrots, berries, whole lettuce

Criteria & EMP Microbiological Limit Recommended Action if Comments

xxx
Target Microorganism Routine Non-Routine Limit is Exceeded
Depending on commodity,
E. coli (O157:H7 or other Reject lot; investigate, implement geographical location and use of
Negative
STEC) corrective action GAPs; Sample size may vary; e.g.
25 g
yyy Negative zone 2 Consider zone 1 and product testing;
Listeria spp. (EMP)
or 3 implement corrective action
zzz Negative zone 2 Consider zone 1 and finished product
Salmonella (EMP)
or 3 testing; implement corrective action
Depending on commodity,
Reject lot; investigate, implement geographical location and use of
Salmonella (product) Negative
corrective action GAPs; Sample size may vary; e.g.
25 g

xxx
(European Commission, 2005; Health Canada, 2008; International Commission for the Microbiological Specifications for Foods (ICMSF), 2011)
yyy
(R.B. Tompkin, 2002; R. B. Tompkin et al., 1999)
zzz
(Chen et al., 2009; Grocery Manufacturers Association, 2009)

Microbiological and chemical limits for foods that are useful to assess process control and insanitary conditions for DoD use.
Table J.33. Microbiological Limits for Produce—Produce, Mushrooms

Notes: fresh, whole, sliced, not canned or marinated

Criteria & EMP Microbiological Limit Recommended Action if Comments

aaaa
Target Microorganism Routine Non-Routine Limit is Exceeded
3
E. coli (generic) 10 /g Investigate, implement corrective action
Negative for 10 Reject or divert for further processing if
E. coli (O157:H7 or other
individual 25-g appropriate; investigate, implement No composite testing
STEC)
samples corrective action
4
Enterobacteriaceae 10 /g Investigate, implement corrective action
bbbb Negative zone 2 Consider zone 1 and finished product
Listeria spp. (EMP)
or 3 testing; implement corrective action
Investigate, implement corrective
3 action.
S. aureus 10 /g 4
If >10 /g, reject lot due to potential for
enterotoxin production
cccc Negative zone 2 or Consider zone 1 and finished product
Salmonella (EMP)
3 testing; implement corrective action
Negative for 2 x Reject or divert for further processing if
Two 375-g analytical units
Salmonella (product) 375-g composite appropriate; investigate, implement
composed of 30 x 25-g samples
samples corrective action

aaaa
(European Commission, 2005; Health Canada, 2008)
bbbb
(R.B. Tompkin, 2002; R. B. Tompkin et al., 1999)
cccc
(Chen et al., 2009; Grocery Manufacturers Association, 2009)

Microbiological and chemical limits for foods that are useful to assess process control and insanitary conditions for DoD use.
Table J.34. Microbiological Limits for Produce—Packaged Salads and Leafy Greens

Criteria & EMP Microbiological Limit Recommended Action if Comments

dddd
Target Microorganism Routine Non-Routine Limit is Exceeded
Consider improvements in production
E. coli (generic) 100/g
hygiene and selection of raw materials
E. coli (O157:H7 or other Reject lot; investigate, implement
Negative Sample size may vary e.g.25 g
STEC) corrective action
eeee Negative zone 2 Consider zone 1 and finished product
Listeria spp. (EMP)
or 3 testing; implement corrective action
Listeria spp. (finished Reject lot; investigate, implement
Negative Sample size may vary; e.g. 25 g
product) corrective action
Salmonella (finished Reject lot; investigate, implement
Negative Sample size may vary; e.g. 25 g
product) corrective action

dddd
(International Commission for the Microbiological Specifications for Foods (ICMSF), 2011)
eeee
(R.B. Tompkin, 2002; R. B. Tompkin et al., 1999)

Microbiological and chemical limits for foods that are useful to assess process control and insanitary conditions for DoD use.
Table J.35. Microbiological Limits for Produce— vegetable sprouts

Criteria & EMP Microbiological Limit Recommended Action if Comments

ffff
Target Microorganism Routine Non-Routine Limit is Exceeded
3
E. coli (generic) 10 /g Investigate, implement corrective action
E. coli (O157:H7 or other Negative in 2 50- Reject lot; investigate, implement
STEC) product gm analytical units corrective action
E. coli (O157:H7 or other Negative in 2 x
Reject lot; investigate, implement
STEC) (spent irrigation 100-g analytical
gggg corrective action
water) units
hhhh Negative zone 2 Consider zone 1 and finished product
Listeria spp. (EMP)
or 3 testing; implement corrective action
Negative in 2 x
iiii Reject lot; investigate, implement Each 250-g analytical unit
Listeria spp. (product) 250-gm analytical
corrective action composed of 5 x 50-g samples
units
Negative in 30 x
Reject lot; investigate, implement
Salmonella (product) 50-gm analytical
corrective action
units
Negative in 2 x
Salmonella (spent irrigation Reject lot; investigate, implement
cccc 375-gl analytical
water) corrective action
units

ffff
(Health Canada, 2006; U.S. Food and Drug Administration, 1999)
gggg
Sampling spent irrigation water; collect the total of 1-liter of spent irrigation water from various trays of growing sprouts. Two x 375-ml
subsamples are used for Salmonella detection and 2 x 100-ml subsamples are used for detection of E. coli O157:H7.
hhhh
(R.B. Tompkin, 2002; R. B. Tompkin et al., 1999)
iiii
(Australia New Zealand Food Authority, 2014; Hitchens & Jinneman, 2013)

Microbiological and chemical limits for foods that are useful to assess process control and insanitary conditions for DoD use.
Table J.36. Microbiological and Chemical Limits for Seafood, Raw

Notes: Ex.: fish, shrimp, crabs. Verification testing for histamine in scombroid species only. The FDA Hazards and Controls Guide lists a defect action
level of 50ppm. See Table J.37, J.38 and J. 39 if raw seafood may be used for applications without full cook such as for sushi or ceviche, additional
testing may be appropriate.

Criteria & EMP Microbiological Limit Recommended Action if Comments

jjjj
Target Microorganism Routine Non-Routine Limit is Exceeded
Routine testing is not recommended
3
Coliforms 10 /g Investigate, implement corrective action for a raw product that is not
intended to be consumed raw.
Consider sampling in facilities or
countries where growing, harvesting
or handling conditions result in
significant prevalence of Salmonella
Divert for reprocessing if appropriate or
in the product. Treat raw seafood as
Salmonella Negative reject; investigate, implement corrective
RTE food if it may be used for
action
applications without full cook such
as sushi or ceviche (See Table J.37
and J.38);Sample size may vary;
kkkk
e.g. 25 g.
Reject lot; investigate, implement Scombroid species only; see
Histamine 50 ppm llll
corrective action sampling details

jjjj
(Ahmed & Institute of Medicine (U.S.). Committee on Evaluation of the Safety of Fishery Products., 1991; Australia New Zealand Food Authority,
2001, 2014; International Commission for the Microbiological Specifications for Foods (ICMSF), 2011; U. S. Department of Health and Human
Services, 2011a)
kkkk
Sample size may vary depending on intended usage, e.g. for sushi or ceviche without full cook (Andrews & Hammack, 2003)
llll
(U. S. Department of Health and Human Services, 2011a)

Microbiological and chemical limits for foods that are useful to assess process control and insanitary conditions for DoD use.
Table J.37. Microbiological and Chemical Limits for Seafood--RTE, Fish, Cold Smoked

Notes: Verification testing for histamine in scombroid species only. The FDA Hazards and Controls Guide lists a defect action level of 50ppm

Criteria & EMP Microbiological Limit Recommended Action if Comments

mmmm
Target Microorganism Routine Non-Routine Limit is Exceeded
Routine testing for general status of
cleaning and disinfection can be
2 done by swab sampling and
APC (EMP) 10/cm Investigate, implement corrective action
determining the aerobic plate count.
Product contact surfaces should
2
contain less than 10 CFU/cm
Investigate, consider Zone 1 and
nnnn Negative for
Listeria spp. (EMP) finished product testing, implement
Zone 2 or 3
corrective action
Negative in 5 Divert for reprocessing, if appropriate,
Listeria monocytogenes
individual 25-g or reject. Investigate and implement
(product)
analytical units corrective action
Divert for reprocessing if appropriate or
Salmonella Negative reject; investigate, implement corrective Sample size may vary; e.g. 25 g
action
Reject lot; investigate, implement Scombroid species only; see
Histamine 50 ppm oooo
corrective action sampling details

mmmm
(International Commission for the Microbiological Specifications for Foods (ICMSF), 2011; U. S. Department of Health and Human Services,
2011a)
nnnn
(Scott et al., 2005; R.B. Tompkin, 2002; R. B. Tompkin et al., 1999)
oooo
(U. S. Department of Health and Human Services, 2011a)

Microbiological and chemical limits for foods that are useful to assess process control and insanitary conditions for DoD use.
Table J.38. Microbiological and Chemical Limits for Seafood-- RTE, cooked or hot smoked
Notes: Ex.: includes cooked and hot smoked products, cooked crabmeat, lobster meat, shrimp, crayfish, surimi, seafood salads, hot-smoked fish.
Histamine testing recommended for scombroid species only with Defect Action Level 50ppm
Criteria & EMP Microbiological Limit Recommended Action if Comments
Target Microorganism pppp Routine Non-Routine Limit is Exceeded
Investigate Source of Post Cook Handling
APC 105/g
and Storage Contamination
qqqq For cooked seafood other than
Coliforms 100/g Investigate; implement corrective action
shrimp and crabmeat
nnnn 3 For cooked crabmeat- fresh (Handled
Coliforms 5x10 /g Investigate; implement corrective action
after final cook)
nnnn 3 For cooked shrimp (Handled after
Coliforms 10 /g Investigate; implement corrective action
final cook)
Investigate, consider Zone 1 and finished
Negative for Zone
Listeria spp. (EMP) rrrr product testing, implement corrective
2 or 3
action
Negative in 5 Divert for reprocessing, if appropriate, or
Listeria monocytogenes
individual 25-g reject. Investigate and implement
(product)
analytical units corrective action
Investigate, implement corrective action. Routine testing is recommended
S. aureus 103/g If >104 /g, reject lot due to potential for especially for products that are
enterotoxin production handled after the final cook (kill) step

Validated cooking will destroy

Salmonella; if HACCP plans are in
Divert for reprocessing, if appropriate, or place to control cooking and
Salmonella Negative reject. Investigate and implement recontamination after cooking, there
corrective action should be no need for routine testing.
Testing could be done as a
verification or investigation. Sample
size may vary; e.g. 25 g
Reject lot; investigate, implement Scombroid species only; see
Histamine 50 ppm
corrective action sampling details ssss

pppp
(Australia New Zealand Food Authority, 2014; International Commission for the Microbiological Specifications for Foods (ICMSF), 1986; U. S.
Department of Health and Human Services, 2011a)
qqqq
(Buchanan, 1991)
rrrr
(Scott et al., 2005; R.B. Tompkin, 2002; R. B. Tompkin et al., 1999)
ssss
(U. S. Department of Health and Human Services, 2011a)

Microbiological and chemical limits for foods that are useful to assess process control and insanitary conditions for DoD use.
Table J.39. Microbiological Limits for Seafood--RTE, Raw Molluscan Shellfish

Notes: Ex.: molluscan shellfish such as oysters, clams, mussels, scallops intended to be eaten without a full cook. Shellfish must be from approved
harvest waters from countries with MOU (I.e., New Zealand, Mexico, Korea, Canada) with the United States. Shellfish from any other source should not
be accepted by DOD. Investigational testing for aquatic toxins. While V. vulnificus and parahaemolyticus may be a concern in RTE, raw molluscan
shellfish, no limits can be recommended at this time.

Criteria & EMP Microbiological Limit Recommended Action if Comments

tttt
Target Microorganism Routine Non-Routine Limit is Exceeded
Divert for reprocessing, if appropriate,
6
APC 1.5x10 /g or reject. Investigate and implement
corrective action
2 Investigate; implement corrective
Coliforms (fecal) 3.3x10 /100g
action; enumerate generic E. coli
2
E. coli (generic) 3.3x10 /100g Investigate; implement corrective action

tttt
(National Advisory Committee on Microbiological Criteria for Foods, 1992; U. S. Department of Health and Human Services, 2011a)
Each analytical unit is comprised of 10-12 individual shellfish composited into one unit. See text FDA Fish and Fishery Products Hazards and
Controls Guidance for sampling procedure.

Microbiological and chemical limits for foods that are useful to assess process control and insanitary conditions for DoD use.
 Table J.40. Microbiological Limits for Spices, Herbs, Coffee and Tea

Notes: defined as ready-to-eat

Criteria & EMP Microbiological Limit Recommended Action if Comments

uuuu
Target Microorganism Routine Non-Routine Limit is Exceeded
5
APC 10 /g Investigate
Investigate, implement corrective
4 action.
B. cereus 10 /g 4
if >10 /g, reject lot due to potential for
enterotoxin production
4
Coliforms 10 /g Investigate, implement corrective action
E. coli (generic) 10/g Investigate, implement corrective action
Divert for reprocessing, if appropriate,
E. coli (O157:H7 or other Number and size of samples will
Negative or reject. Investigate and implement
STEC) vary depending on product
corrective action
4
Mold/Yeast 10 /g Investigate, implement corrective action
vvvv Negative zone 2 Consider zone 1 and finished product
Salmonella (EMP)
or 3 testing; implement corrective action
Number and size of samples will
Divert for reprocessing, if appropriate,
vary depending on product; routine
Salmonella Negative or reject. Investigate and implement
testing of Salmonella in roasted
corrective action
coffee may not be applicable
5
Mesophilic sporeformer 10 /g Investigate, implement corrective action

uuuu
(Health Canada, 2008; Sagoo et al., 2009)
vvvv
(Chen et al., 2009; Grocery Manufacturers Association, 2009)

Microbiological and chemical limits for foods that are useful to assess process control and insanitary conditions for DoD use.
References:

1. Ahmed, F. E., & Institute of Medicine (U.S.). Committee on Evaluation of the Safety of Fishery Products. (1991).
Seafood safety Retrieved from http://www.nap.edu/catalog/1612.html
2. American Meat Institute Foundation. (1997). Good Manufacturing Practices for Fermented Dry & Semi-Dry
Sausage Products. Retrieved February 18, 2015, from
http://www.meathaccp.wisc.edu/Model_Haccp_Plans/assets/GMP%20Dry%20Sausage.pdf
3. Andrews, W. H., & Hammack, T. S. (2003). BAM: Chapter 1. Food Sampling/Preparation of Sample Homogenate.
In Bacteriological Analytical Manual, Edition 8, Revision A, 1998. Retrieved April 13, 2014, from
http://www.fda.gov/Food/FoodScienceResearch/LaboratoryMethods/ucm063335.htm
4. Andrews, W. H., Jacobson, A., & Hammack, T. (2014). BAM: Salmonella, Chapter 5 FDA, Bacteriological Analytical
Manual (Vol. 2014).
5. Australia New Zealand Food Authority. (2001). Guidelines for the microbiological examination of ready-to-eat
foods. Retrieved September 23, 2014, from
http://www.foodstandards.gov.au/publications/pages/guidelinesformicrobi1306.aspx
6. Australia New Zealand Food Authority. (2014). Criteria for Listeria monocytogenes – Microbiological Limits for
Foods.
7. Bradley, R. L., Houck, K. S., Smukowski, M., Bradley, R. L., Houck, K., & Smukowski, M. (2013). Milk and Milk
Products. In S. Doores, Y. Salfinger, & M. L. Tortorello (Eds.), Compendium of Methods for Microbiological
Examination of Foods (5th ed.): American Public Health Association.
8. Buchanan, R. L. (1991). Microbiological criteria for cooked, ready-to-eat shrimp and crabmeat. Journal of Food
Technology, 45(4), 157-160.
9. Canada Food Inspection Agency. (2012). Implementation of the 2011 Health Canada Policy on Listeria
monocytogenes in Ready-to-Eat Foods. Retrieved September 19, 2014, from
http://www.inspection.gc.ca/food/fish-and-seafood/product-inspection/fish-
listeria/eng/1338572158747/1338572678417
10. Chen, Y. H., Scott, V. N., Freier, T. A., Kuehm, J., Moorman, M., Meyer, J., . . . Banks, J. (2009). Control of
Salmonella in low-moisture foods II: Hygiene practices to minimize Salmonella contamination and growth. Food
Protection Trends, 29, 435-445.
11. Code of Federal Regulations. (2014a). 21 CFR 120. Hazard Analysis and Critical Control Point (HACCP) Systems.
Retrieved February 19, 2015, from
http://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfCFR/CFRSearch.cfm?CFRPart=120&showFR=1

Microbiological and chemical limits for foods that are useful to assess process control and insanitary conditions for DoD use.
 12. Code of Federal Regulations. (2014c). 21 CFR 165. Requirements for Specific Standardized Beverages.
Retrieved February 3, 2014, from http://www.ecfr.gov/cgi-
bin/retrieveECFR?gp=&SID=d26e287c93ddfdc1dcc06d9b896b5d5f&r=PART&n=21y2.0.1.1.38
13. Craven, H. M., Eyles, M. J., & Davey, J. A. (2003). Enteric indicator organisms in foods. In A. D. Hocking (Ed.),
Foodborne Microorganisms of Public Health Significance, 6th ed. Tempe, Australia: Australian Institute of Food
Science and Technology Inc.
14. Elliott, P. H., & Kataoka, A. (2013). Canned Foods - Test for Commercial Sterility. In S. Doores, Y. Salfinger, & M.
L. Tortorello (Eds.), Compendium of Methods for Microbiological Examination of Foods (5th ed.): American Public
Health Association.
15. European Commission. (2005). Microbiological Criteria for Foodstuff. Commission Regulation (EC) No. 2073/2005.
Retrieved September 23, 2014, from http://www.fsai.ie/uploadedFiles/Reg2073_2005(1).pdf
16. European Communities. (2007). Drinking Water Regulations. Retrieved September 23, 2014, from
http://www.irishstatutebook.ie/pdf/2007/en.si.2007.0278.pdf
17. Grocery Manufacturers Association. (2009). Control of Salmonella in low-moisture foods. Retrieved August 14,
2014, from http://www.gmaonline.org/downloads/technical-guidance-and-tools/SalmonellaControlGuidance.pdf
18. Health Canada. (2006). Health Products and Food Branch (HPFB) Guidance for Industry: Sample Collection and
Testing for Sprouts and Spent Irrigation Water (pp. 1-18).
19. Health Canada. (2008). Standards and Guidelines for Microbiological Safety of Food – An Interpretative Summary.
Retrieved September 2014, 2014, from http://www.hc-sc.gc.ca/fn-an/res-rech/analy-meth/microbio/volume1-
eng.php
20. Hitchens, A. D., & Jinneman, K. (2013). BAM: Chapter 10. Detection and Enumeration of Listeria monocytogenes,
Revised 2013 FDA, Bacteriological Analytical Manual 8th Edition, Revision A, 1998. . Retrieved April 13, 2014,
from http://www.fda.gov/food/foodscienceresearch/laboratorymethods/ucm071400.htm
21. International Commission for the Microbiological Specifications for Foods (ICMSF). (1986). Microorganisms in
Foods 2: Sampling for Microbiological Analysis: Principles and Specific Applications (2nd ed.). University of
Toronto, Canada: University Toronto Press.
22. International Commission for the Microbiological Specifications for Foods (ICMSF). (2011). Microorganisms in
Foods 8: Use of Data for Assessing Process Control and Product Acceptance. New York: Springer.
23. Kornacki, J. I. (2014). An Environmental Sampling Approach to Product Risk Assessment. Food Safety Magazine.
24. Kornacki, J. I., Gurtler, J. B., & Stawick, B. A. (2013). Enterobacteriaceae, Coliforms, and Escherichia coli as
Quality and Safety Indicators, Updated September 2013. In Y. S. S. Doores, M. L.Tortorello (Ed.), Compendium of
Methods for the Microbiological Examination of Foods (5th ed.).
http://ajph.aphapublications.org/doi/book/10.2105/MBEF.0222: American Public Health Association.

Microbiological and chemical limits for foods that are useful to assess process control and insanitary conditions for DoD use.
 25. National Academies of Science. (2003). Scientific Criteria and Performance Standards to Control Hazards in Dairy
Scientific Criteria to Ensure Safe Foods (pp. 225-247). Washington, D.C.: Institute of Medicine. National Research
Council of the National Academies.
26. National Advisory Committee on Microbiological Criteria for Foods. (1992). Microbiological criteria for raw
molluscan shellfish. J. Food Prot, 55(6), 463-480.
27. National Advisory Committee on Microbiological Criteria for Foods. (2013). Expedited response to the questions
posed by the USDA agricultural marketing service to support ground beef purchase for the federal food and
nutrition assistance programs. J. Food Prot, 76, 523-5237.
28. Sagoo, S. K., Little, C. L., Greenwood, M., Mithani, V., Grant, K. A., McLauchlin, J., . . . Threlfall, E. J. (2009).
Assessment of the microbiological safety of dried spices and herbs from production and retail premises in the
United Kingdom. Food Microbiol, 26(1), 39-43. doi: 10.1016/j.fm.2008.07.005
29. Scott, V. N., Wiedmann, M., Hicks, D., Collette, R., Jahncke, M. L., & Gall, K. (2005). Guidelines for Listeria testing
of environmental, raw product and finished product samples in smoked seafood processing facilities. Food
Protection Trends, 25(1), 23-34.
30. Sperber, W. H., & North American Millers' Association Microbiology Working Group. (2007). Role of microbiological
guidelines in the production and commercial use of milled cereal grains: a practical approach for the 21st century.
J. Food Prot, 70(4), 1041-1053.
31. Tompkin, R. B. (2002). Control of Listeria monocytogenes in the food-processing environment. J. Food Prot, 65(4),
709-725.
32. Tompkin, R. B., Scott, V. N., Bernard, D. T., Sveum, W. H., & Gombas, K. S. (1999). Guidelines to prevent post-
processing contamination from Listeria monocytogenes. Dairy, Food and Environmental Sanitation, 19(8), 551-562.
33. U. S. Department of Agriculture. (2011). Shiga-toxing Producing Escherichia coli in Certarin Raw Beef Products.
Federal Register, 76(182), 58157-58165.
34. U. S. Department of Agriculture Food Safety Inspection Service. (1999). Appendix B: Compliance Guidelines for
Cooling Heat-Treated Meat and Poultry Products (Stabilization). June 1999.
http://www.fsis.usda.gov/OPPDE/rdad/frpubs/95-033F/95-033F_Appendix%20B.htm. Retrieved December 12,
2012
35. U. S. Department of Agriculture Food Safety Inspection Service. (2014a). Compliance guidelines to control Listeria
monocytogenes in post-lethality exposed ready-to-eat meat and poultry products. Updated January 2014.
Retrieved August 14, 2014, from http://www.fsis.usda.gov/wps/wcm/connect/d3373299-50e6-47d6-a577-
e74a1e549fde/Controlling_LM_RTE_Guideline_0912?MOD=AJPERES
36. U. S. Department of Agriculture Food Safety Inspection Service. (2014c). FSIS Microbiological Testing Program for
Salmonella in Pasteurized Egg Products, 1995–2013. Retrieved August 14, 2014, from

Microbiological and chemical limits for foods that are useful to assess process control and insanitary conditions for DoD use.
 http://www.fsis.usda.gov/wps/portal/fsis/topics/data-collection-and-reports/microbiology/testing-program-for-rte-
meat-and-poultry-products/microbiological-testing-program-for-pasteurized-egg-products/pasteurized-egg-products
37. U. S. Department of Agriculture Food Safety Inspection Service. (2014e). Pathogen Reduction – Salmonella and
Campylobacter Performance Standards Verification Testing. Retrieved January 23, 2015, from
http://www.fsis.usda.gov/wps/wcm/connect/b0790997-2e74-48bf-9799-
85814bac9ceb/28_IM_PR_Sal_Campy.pdf?MOD=AJPERES
38. U. S. Department of Health and Human Services. (2004). Guidance for Industry Juice HACCP Hazards and
Controls Guidance.
39. U. S. Department of Health and Human Services. (2005a). Aflatoxin in Peanuts and Peanut Products. CPG Sec.
570.375.
40. U. S. Department of Health and Human Services. (2005c). Brazil Nuts - Adulteration with Aflatoxin. CPG Sec.
570.200.
41. U. S. Department of Health and Human Services. (2005e). Patulin in Apple Juice, Apple Juice Concentrates and
Apple Juice.
42. U. S. Department of Health and Human Services. (2005g). Pistachio Nuts - Aflatoxin Adulteration. CPG Sec.
570.500.
43. U. S. Department of Health and Human Services. (2005i). Tree Nuts - Adulteration with Filth, Involving the
Presence of the Organism Escherichia coli. CPG Sec. 570.450.
44. U. S. Department of Health and Human Services. (2009). Prevention of Salmonella Enteritidis in Shell Eggs During
Production, Storage, and Transportation: Final Rule. 21 CFR Parts 16 and 118, Federal Register 74(130).
45. U. S. Department of Health and Human Services. (2011a). Fish and Fishery Products Hazards and Controls
Guidance - Fourth Edition. Retrieved August 14, 2014, from
http://www.fda.gov/food/guidanceregulation/guidancedocumentsregulatoryinformation/seafood/ucm2018426.htm.
46. U. S. Department of Health and Human Services. (2011j). Grade “A” Pasteurized Milk Ordinance, Including
Provisions from the Grade “A” Condensed and Dry Milk Products and Condensed and Dry Whey--Supplement I to
the Grade “A” Pasteurized Milk Ordinance.
47. U.S. Department of Agriculture Agricultural Marketing Service (AMS). (2015). Microbiological Testing of AMS
Purchased Meat, Poultry and Egg Commodities. . Retrieved February 22, 2015, from
http://www.ams.usda.gov/AMSv1.0/ams.fetchTemplateData.do?template=TemplateA&navID=MicrobialTestingofCo
mmodities&rightNav1=MicrobialTestingofCommodities&topNav=&leftNav=&page=FPPMicroDataReports&resultTy
pe=&acct=lsstd
48. U.S. Department of Agriculture Food Safety Inspection Service. (1999). Performance standards for the production
of certain meat and poultry products. Retrieved October 18, 2014, from

Microbiological and chemical limits for foods that are useful to assess process control and insanitary conditions for DoD use.
 https://www.federalregister.gov/articles/1999/01/06/99-32/performance-standards-for-the-production-of-certain-
meat-and-poultry-products
49. U.S. Department of Health and Human Services. (2008). Guidance for Industry: Control of Listeria monocytogenes
in Refrigerated or Frozen Ready-To-Eat Foods; Draft Guidance. Retrieved January 22, 2015, from
http://www.fda.gov/Food/GuidanceRegulation/GuidanceDocumentsRegulatoryInformation/FoodProcessingHACCP/
ucm073110.htm
50. U.S. Department of Health and Human Services. (2010). Compliance Policy Guide Sec. 527.300 Dairy Products -
Microbial Contaminants and Alkaline Phosphatase Activity.
51. U.S. Food and Drug Administration. (1999). Guidance for industry: Sampling and microbial testing of spent
irrigation water during sprout production. Federal Register, 64, 57896-57902.
52. U.S. Food and Drug Administration. (2015). Investigations Operations Manual 2015, Ch. 4 Sampling. Retrieved
June 13, 2015, from http://www.fda.gov/downloads/ICECI/Inspections/IOM/UCM123507.pdf
53. World Health Organization (WHO). (2013). Report on regulations and standards for drinking water quality - draft 12.
Retrieved April 14, 2014, from http://www.who.int/water_sanitation_health/regulations_and_standards_for_drinking-
water_quality/en/
54. World Health Organization (WHO). (2014). Drinking- water quality. Retrieved September 19, 2014, from
http://www.who.int/water_sanitation_health/dwq/en/

Microbiological and chemical limits for foods that are useful to assess process control and insanitary conditions for DoD use.
Appendix K. Sampling for Lot Acceptance or Rejection

In lot acceptance sampling, a sample of n units is drawn from a lot, and a characteristic
of interest (e.g., presence of a pathogen or concentration of an indicator organism) of
the sample units is analyzed. The test portion or analytical unit may represent an entire
sample unit (e.g., enrichment of a 25-g sample), a composite of multiple sample units
(e.g., enrichment of sixty 6.25-g sample units into an approximate 375 g composite test
portion), or a portion or aliquot of a sample unit (e.g., enumeration of 1 ml of diluted
homogenate prepared from a 25-g sample unit). Based on the sample results and the
lot acceptance criteria, a lot disposition decision is made to either accept or reject the
lot.

For lot acceptance sampling to have direct impact, lots must vary with respect to the
analyte of interest. If the analyte is homogeneous among lots, sampling will accept
some lots and reject others by chance alone, and the accepted lots are no better than
the rejected lots. Lot acceptance sampling schemes may involve normal, tightened, or
reduced inspection levels. Lot acceptance sampling may be applied lot-by-lot, or skip-lot
sampling may be applied to less than 100% of lots. Conventional rules for switching
between inspection levels depend on the history of supplier conformance to
specifications and are available as guidelines for frequency and intensity of lot
acceptance sampling schemes (e.g., MIL-STD-1916 [available at
http://guidebook.dcma.mil/34/milstd1916(15).pdf], ISO 2859-1, ISO 2859-3, /ASQ Z1.4
[available at www.asq.org]).

Lot Acceptance Sampling for Attributes

Current microbiological lot acceptance sampling schemes are based on lot acceptance
sampling for attributes. Both microbial presence/absence data obtained from enriched
samples and quantitative concentration data divided into numerical ranges are classified
as attributes. Two-class sampling plans are applicable where product quality is divided
into two attribute classes. For sampling based on detection methods, the classes are
presence or absence. For sampling based on enumeration methods, the classes are
x≤m or x>m, where x is a measure of concentration (e.g., CFU/g, CFU/ml, CFU/cm2, as
appropriate), and m is the microbiological limit separating acceptable from unacceptable
concentrations. Three-class sampling plans are applicable where product quality is
measured by enumeration methods and results are divided into three attribute classes:
x≤m, m<x≤M, or x>M; where m is the limit for a marginally acceptable concentration
(note the change in meaning from the 2-class plan), and M is the limit for an
unacceptable concentration (i.e., similar in meaning to m in the 2-class plan). Two- and
three-class sampling plans also may be used for process control (3), but in the context
of lot acceptance sampling, a consequence of non-conformance with a microbiological
criterion is rejection of the lot represented by the samples.

1
Performance Characteristics of the Sampling Plans

The performance characteristics of microbiological lot acceptance sampling plans are

generally described in terms of the probability of acceptance (pa) for a given lot
composition under specified microbiological limits. The operating characteristic (OC)
curve plots the probability of accepting the lot versus a measure of its quality (e.g., the
proportion of analytical units exceeding a given limit). In practice, the probability of lot
acceptance also depends on the sampling procedures and analytical methods specified
under the microbiological criteria (e.g., test sensitivity and specificity, percent recovery
in enumeration, compositing). The statistical performance characteristics of a lot-
acceptance sampling plan reflect the different probabilities of rejecting lots of different
qualities. In addition to this direct, or curative, effect of a sampling plan, for a continuing
series of lots presented for inspection, lot acceptance sampling also may have an
indirect, or preventative, effect by exerting economic pressures on suppliers to prevent
or limit the frequency and severity of non-conformance through process control
measures (2, 6, 8) .

Two-Class Plans

For two-class sampling plans based on a maximum limit (m), the probability of lot
acceptance (pa) can be calculated directly from the proportion of non-conforming
analytical units (p). For presence/absence sampling plans, p refers to the proportion of
test-positive analytical units. For concentration-based sampling plans, p refers to the
proportion of analytical units with CFU/g ≥ m. A two-class sampling plan is defined by
the sample size n and the maximum number of non-conforming analytical units allowed
(acceptance number) c. In general, the probability of lot acceptance (pa) is the
probability that the number of non-conforming independent sample units in the sample
of n is less than or equal to c:

𝑖𝑖=𝑐𝑐 𝑛𝑛
𝑝𝑝𝑎𝑎 = ∑𝑖𝑖=0 𝐶𝐶𝑖𝑖 (𝑝𝑝)𝑖𝑖 (1 − 𝑝𝑝)𝑛𝑛−𝑖𝑖 (eq. K.1.)

𝑛𝑛!
where 𝐶𝐶𝑖𝑖𝑛𝑛 = 𝑖𝑖!(𝑛𝑛−𝑖𝑖)! (the binomial coefficient, or combination of n things taken i at a time),
p is the proportion of non-conforming analytical units, n is the number of samples drawn
from the lot, and c is the acceptance number.

Most two-class sampling plans for pathogens specify c = 0. In this case, pa simplifies to:

pa = (1-p)n (eq. K.2.)

For presence/absence sampling plans, it is important to note that the proportion of test
positives depends on the size of the analytical unit (test portion). If analytical units of
different sizes are examined from the same food lot, the proportion of test positives will
be lower for smaller analytical units. For example, for a given concentration, a 25-g
sample is less likely to contain at least one microorganism than a 100-g sample.
Therefore, prevalence needs to be referenced to the size of the analytical unit (e.g.,

2
prevalence in 25 grams). If the detection method is less than 100% sensitive, the
apparent prevalence (proportion of test positives) is less than the true prevalence
(actual proportion of positives). For concentration-based sampling plans, equations K.1.
and K.2. assume that measurement error is negligible (~100% recovery) and that false
positive results are very unlikely (~ 100% specificity). If measurement error is
substantial, the probability of lot acceptance may be affected by the microbial
distribution (not just the proportion of non-conforming sample units) because
measurement error can result in misclassification of sample units above or below the
limit (m).

Figure K.1. presents the operating characteristic curves for c = 0 two-class sampling
plans over a range of sample sizes (n).

Figure K.1. Operating Characteristic Curves for Two-Class Sampling Plans with c = 0.

Table K.1. summarizes the performance of an n = 5, c = 0 two-class sampling plan with

m = absence in 25 g with probability of acceptance (pa) = 0.95, 0.5, and 0.05.

Table K.1. Summary Performance of n = 5, c = 0 Two-Class Sampling Plan

n 5
c 0
m absence in 25 g
pa p
0.95 0.0102
0.50 0.1294
0.05 0.4507

Alternatively, the probability of acceptance for two-class sampling plans can be

calculated based on an assumed statistical distribution of the microbial concentration in

3
a lot. Assuming that the average concentration is log normally distributed and that the
number of CFU in an analytical unit varies randomly according to the Poisson
distribution (a Poisson-Lognormal distribution), Table K.2. summarizes the performance
of an n = 5, c = 0 two-class sampling plan with m = absence in 25 g. The values
(probability of acceptance) shown in Table K.2. assume a perfect detection method
(100% sensitivity and 100% specificity). The calculations can be performed using on-
line resources (http://www.icmsf.org/publications/sampling_plans.html;
http://www.fstools.org/sampling). For further details about the Poisson-Lognormal used
by these on-line calculators, see (7) .

Table K.2. Performance of n = 5, c = 0, m = absence in 25-g Two-Class Sampling Plan

N c M Probability stdev* = stdev = stdev = stdev = 1.2

of lot 0.25** 0.50 0.80 (log10
acceptance (log10 (log10 (log10 CFU/g)
CFU/g) CFU/g) CFU/g)
Geometric mean concentration (log10 cfu/g)***
5 0 absence 0.95 -3.46 -3.67 -4.08 -4.81
in 25 g 0.50 -2.32 -2.48 -2.74 -3.14
0.05 -1.64 -1.69 -1.74 -1.79
Note: Probability of acceptance (pa) values assume negligible measurement error.
*stdev = standard deviation
**In many applications, measurement error alone exceeds 0.25 log10 CFU/g. We include this standard
deviation value to help the reader understand the derivation of values for sampling plans commonly
presented in the literature (e.g. (3, 4))
***Technically, the geometric mean is the exponentiated mean of the logarithms of individual
concentrations. For example, if the mean of the log-transformed values = -2 log10 CFU/g, the geometric
mean = 0.01 CFU/g. However, because -2 log10 and 0.01 are equivalent, we adopt the common usage in
the food microbiology literature of “geometric mean” to refer to the mean of the log-transformed values to
differentiate it from the arithmetic mean on the original scale (CFU/g). For the lognormal distribution, the
geometric mean represents the median because the normal distribution is symmetric.

It should be noted that in terms of consumer health risk, the arithmetic mean (the
average, or expected value) concentration is more relevant than the geometric mean
(median) concentration (5) . Table K.3. presents the arithmetic mean concentration for
the 5 percent probability of acceptance distributions in Table K.2.

Table K.3. Arithmetic Mean for Lognormal Distributions

geometric standard arithmetic

mean deviation mean
(log10 CFU/g) (log10 CFU/g) (CFU/g)
-1.64 0.25 0.0270
-1.69 0.50 0.0396
-1.74 0.80 0.0993
-1.79 1.20 0.7377

4
As an example interpretation of Table K.3, if the arithmetic mean concentration is
0.0396 CFU/g, a 25-g serving would contain an average of 1 CFU. Note that there is no
direct correspondence between the probability of lot acceptance and the level of risk
indicated by the arithmetic mean concentration. This illustrates that evaluating the food
safety impact of sampling plans is not as straightforward as calculating their statistical
operating characteristics.

Three-Class Plans

Three-class sampling plans are based on a marginal limit (m) and a maximum limit (M).
A three-class sampling plan is defined by the sample size n and c the maximum number
of marginal analytical units allowed, or acceptance number. (The acceptance number
for analytical units exceeding M is zero.) For three-class sampling plans, the probability
of lot acceptance (pa) can be calculated directly from the proportion of marginally
acceptable (pm) and unacceptable (pd) analytical units (1):

𝑖𝑖=𝑐𝑐 𝑛𝑛
𝑝𝑝𝑎𝑎 = ∑𝑖𝑖=0 𝐶𝐶𝑖𝑖 (𝑝𝑝𝑚𝑚 )𝑖𝑖 (1 − 𝑝𝑝𝑑𝑑 − 𝑝𝑝𝑚𝑚 )𝑛𝑛−𝑖𝑖 (eq. K.3)

𝑛𝑛!
where 𝐶𝐶𝑖𝑖𝑛𝑛 = 𝑖𝑖!(𝑛𝑛−𝑖𝑖)! (the binomial coefficient), pm is the proportion of marginally
acceptable analytical units (with m < x ≤ M), pd is the proportion of analytical units with x
> M, n is the number of samples drawn from the lot, and c is the acceptable number of
marginal units.

Figure K.2 summarizes the operating characteristics of an n = 5, c = 2 three-class

sampling plan for probability of acceptance (pa) = 0.95, 0.5, and 0.05.

Figure K.2. Operating Characteristic Contours for Three-Class Plan: n = 5, c = 2

Table K.4 presents selected performance characteristics for an n = 5, c = 2 three-class

sampling plan.

5
Table K.4. Performance Characteristics for n = 5, c = 2 Three-Class Sampling Plan:
Probabilities of Acceptance (pa) for Lots Containing Indicated Proportions pd and pm

Proportion Proportion Marginal (pm)

unacceptable
(pd) 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45
0.90 <* 0.0000
0.85 < < 0.0000
0.80 < < < 0.0000
0.75 < < < < 0.0000
0.70 < < < < < 0.0000
0.65 0.0051 < < < < < 0.0000
0.60 0.0101 0.0092 0.0074 0.0051 < < < 0.0000
0.55 0.0182 0.0170 0.0146 0.0111 0.0073 < < < 0.0000
0.50 0.0310 0.0294 0.0262 0.0213 0.0156 0.0099 0.0051 < <
0.45 0.0500 0.0481 0.0438 0.0374 0.0294 0.0209 0.0129 0.0065 <
0.40 0.0774 0.0750 0.0697 0.0614 0.0508 0.0389 0.0270 0.0163 0.0080
0.35 0.1156 0.1127 0.1063 0.0959 0.0822 0.0663 0.0497 0.0338 0.0201
0.30 0.1675 0.1642 0.1564 0.1438 0.1267 0.1062 0.0840 0.0618 0.0414
0.25 0.2367 0.2327 0.2236 0.2084 0.1875 0.1620 0.1334 0.1039 0.0753
0.20 0.3270 0.3224 0.3117 0.2938 0.2687 0.2375 0.2018 0.1638 0.1258
0.15 0.4429 0.4377 0.4253 0.4044 0.3748 0.3373 0.2938 0.2463 0.1974
0.10 0.5896 0.5837 0.5695 0.5454 0.5108 0.4666 0.4143 0.3563 0.2952
0.05 0.7727 0.7661 0.7501 0.7225 0.6826 0.6310 0.5692 0.4995 0.4250
0.00 0.9988 0.9914 0.9734 0.9421 0.8965 0.8369 0.7648 0.6826 0.5931
pm
pd 0.50 0.55 0.60 0.65 0.70 0.75 0.80 0.85 0.90
0.50 0.0000
0.45 < 0.0000
0.40 < < 0.0000
0.35 0.0098 < < 0.0000
0.30 0.0243 0.0117 < < 0.0000
0.25 0.0498 0.0289 0.0137 < < 0.0000
0.20 0.0902 0.0590 0.0339 0.0160 0.0053 < 0.0000
0.15 0.1500 0.1064 0.0689 0.0393 0.0184 0.0060 < 0.0000
0.10 0.2342 0.1762 0.1239 0.0797 0.0451 0.0210 0.0068 < 0.0000
0.05 0.3488 0.2742 0.2046 0.1428 0.0912 0.0513 0.0237 0.0077 <
0.00 0.5000 0.4069 0.3174 0.2352 0.1631 0.1035 0.0579 0.0266 0.0086
Note: Probability of acceptance (pa) values calculated assuming negligible measurement error.
*pa < 0.005.

Note that some combinations of pm and pd presented in Figure K.2. and Table K.4. may
not be plausible for certain applications. For example, consider a three-class sampling

6
plan for mesophilic aerobic bacteria in dried milk with m = 104 CFU/g, and M = 106
CFU/g. Based on Table K.3, pa = 0.05 for the combination pm = 0.05 and pd = 0.45.
However, if the distribution is lognormal, this combination of pm and pd values would
imply a lot of dried milk with a geometric mean aerobic plate count of 4 log10 CFU/g
(10,000 CFU/g), a standard deviation of 16 log10 CFU/g, and a maximum concentration
in excess of 40 log10 CFU/g.

Alternatively, the probability of acceptance for three-class sampling plans can be

calculated based on an assumed statistical distribution of the microbial concentration in
a lot. In this case, eq. K.3. above is still used to calculate the probability of lot
acceptance, but the marginal and unacceptable proportions (pm and pd) are derived
from the assumed statistical distribution. Table K.5. summarizes the performance of n =
5, c = 2 three-class sampling plans based on assuming a lognormal distribution.

Table K.5. Performance of n = 5, c = 2 Three-Class Sampling Plans

n c m M Probability stdev* = stdev = stdev = stdev =

of lot 0.25 0.50 0.80 1.2
acceptance (log10 (log10 (log10 (log10
CFU /g) CFU /g) CFU /g) CFU /g)
Geometric mean concentration (log10 CFU
/g)
5 2 104 CFU 106 CFU 0.95 3.78 3.56 3.29 2.82
/g /g 0.50 4.00 4.00 3.99 3.89
(4 log10 (6 log10 0.05 4.22 4.44 4.68 4.90
CFU /g) CFU /g)
5 2 <3MPN/g 9.8 0.95 0.25 -0.19 -0.87 -1.79
MPN/g 0.50 0.47 0.33 0.05 -0.38
0.05 0.68 0.76 0.79 0.78
*stdev = standard deviation

As an example of Table K.5. calculations, consider the lognormal distribution with

geometric mean of 4.68 log10 CFU /g (47,863 CFU /g) and standard deviation of 0.8
log10 CFU /g (denoted by shaded cell of Table K.5.) For the first sampling plan, the
marginal limit (m) of 4 log10 CFU /g is the 20th percentile of the distribution. The
maximum limit (M) of 6 log10 CFU /g is the 95th percentile of the distribution. As a result,
pm = 0.95-0.20 = 0.75 and pd = 1-0.95 = 0.05. Looking up these values in Table 3 (or
inserting the values into eq. 3) results in probability of lot acceptance (pa) = 0.05. The
calculations can be performed using on-line resources
(http://www.icmsf.org/publications/sampling_plans.html;
http://www.fstools.org/sampling).

Impact of Lot Acceptance Sampling Plans

As noted above, the impact of lot acceptance sampling plans depends on variability
among lots. The Food and Agriculture Organization/World Health Organization jointly

7
supported development of a web-based analytical tool that can analyze the direct
impact of lot acceptance sampling plans based on user-specified distributions of
contamination between and within lots (http://www.fstools.org/sampling).

BIBLIOGRAPHY

1. Codex Alimentarius Commission. 2004. General Guidelines on Sampling.

2. Hill, I. D. 1960. The Economic Incentive Provided by Sampling Inspection. Journal of
the Royal Statistical Society. Series C (Applied Statistics) 9: 69-81.
3. International Commission for the Microbiological Specifications for Foods (ICMSF).
2011. Microorganisms in Foods 8: Use of Data for Assessing Process Control and
Product Acceptance. Springer, New York.
4. International Commission for the Microbiological Specifications of Foods (ICMSF).
2002. Microorganisms in Foods 7: Microbiological Testing in Food Safety Management.
Springer, New York.
5. International Life Sciences Institute-Europe (ILSI-Europe). 2010. Impact of Microbial
Distributions on Food Safety.
6. International Organization for Standardization. 2007. Guidance on the selection and
usage of acceptance sampling systems for inspection of discrete items in lots — Part 1:
Acceptance sampling. ISO, Geneva.
7. van Schothorst, M., M. H. Zwietering, T. Ross, R. L. Buchanan, and M. B. Cole. 2009.
Relating microbiological criteria to food safety objectives and performance objectives.
Food Control 20: 967-979.
8. Whittle, P. 1954. Optimum Preventative Sampling. Journal of the Operations Research
Society of America 2: 197-203.

8
Appendix L. Design of 2-class and 3-class Sampling Plans

Two- and three-class sampling plans are attribute sampling plans where
quantitative microbiological concentration data are divided into two or three
classes, respectively. The key parameters of these plans are: ‘n’, the sample
size; ‘cm’, the acceptance number for sample units which exceed m in
concentration; and ‘cM’, the acceptance number for sample units which exceed M
in concentration. The sample of n units is presumed to be a ‘rational subgroup’,
such as units chosen from the same lot or time period of production.

Some additional notation that will be useful in the discussion of 2- and 3-class
sampling plans are:

Pm: The percentile rank of m in normal production.

Qm = 1 - Pm
PM: The percentile rank of M in normal production.
QM = 1 - P M
ym = log10(m + 0.3 d), where ‘d’ is the dilution factor from counts to concentration
yM = log10(M + 0.3 d), where ‘d’ is the dilution factor from counts to concentration

2-Class Sampling Plans

In a ‘2-Class’ sampling plan, only the M quantile is used.

There are several approaches to designing a 2-class sampling plan.

Approach1: Given M and associated error fraction target, find n and cM

This approach is used primarily in designing sampling plans for acceptance

testing, and the focus is on the false negative fraction (‘consumer’s risk’ or ‘type II
error’ or ‘β’).

Here ‘M’ denotes a median concentration that, if exceeded, represents

‘unacceptable’ product which should be rejected with high reliability by the plan.

Subject matter expertise or specifications provides a value for M that should be

rejected with high reliability (‘Power’ = 1 – β), say 95% or 99% or 99.9%. The
parameters n and cM are chosen to achieve the power required. For this purpose,
M is presumed to be the median of the distribution of concentration for an
unacceptably contaminated lot of product.

The probability of rejecting this type of unacceptable lot is

1
 1–β = P[(# > M) > cM] (L.1.)

calculated from a binomial distribution with p = P[ X < M] = 0.50.

Some possible plans are:

Table L.1. 2-Class plans, given M as median of 'unacceptable' lots

Power Power
n c Target Actual
4 0 95.00% 93.75%
5 0 95.00% 96.88%
6 0 95.00% 98.44%
10 1 95.00% 98.93%
20 5 95.00% 97.93%
10 1 99.00% 98.93%
20 4 99.00% 99.41%
10 0 99.90% 99.90%
20 3 99.90% 99.87%

Note that this type of plan is based on specifications, so is not generally useful
for SPC.

Approach 2: Given PM and associated error fraction target, find n and cM

This approach is used in designing sampling plans for process control, and the
focus is on the false positive fraction (‘producer’s risk’ or ‘type I error’ or ‘α’ or
‘FAR’).

M here represents a specified high quantile of the concentration of ‘normal’

production. Its value is determined from its percentile rank PM either
nonparametrically or parametrically based on historical data. The parameters n
and cM are chosen to achieve the maximum false positive fraction α allowed,
typically 1% or 0.1%, depending upon sampling rate. Note that M is presumed to
be a high quantile of the distribution of concentration during ‘normal’ production
under statistical process control.

The probability of a randomly chosen lot from normal production failing is

α = P[(# > M) > cM] (eq. L.1.)

calculated from a binomial distribution with n trials and p = QM.

2
Some possible plans are:

Table L.2. 2-Class plans, given M and α

α Prob. α
n c Target <M Actual
1 0 5.00% 95.00% 5.00%
2 1 5.00% 95.00% 0.25%
3 1 5.00% 95.00% 0.72%
5 1 5.00% 95.00% 2.26%
10 2 5.00% 95.00% 1.15%
1 0 1.00% 99.00% 1.00%
2 1 1.00% 99.00% 0.01%
3 1 1.00% 99.00% 0.03%
5 1 1.00% 99.00% 0.10%
10 1 1.00% 99.00% 0.43%
1 0 0.10% 99.90% 0.10%
2 1 0.10% 99.90% 0.00%
3 1 0.10% 99.90% 0.00%
5 1 0.10% 99.90% 0.00%
10 1 0.10% 99.90% 0.00%

Approach 3: Given n, cM and error fraction target, find PM

This approach results in achievement of an exact false positive fraction α, and

again is used in designing sampling plans for process control with the focus on
the false positive fraction (‘producer’s risk’ or ‘type I error’ or ‘α’ or ‘FAR’).

M represents a high quantile of ‘normal’ production whose value is determined

from its percentile rank PM which solves the following equation

α = P[(# > M) > cM] (eq. L.2.)

calculated from a binomial distribution with n trials and p = QM.

3
Some possible plans are:

Table L.3. 2-Class plans, given n, c and α

α Prob. α
n c Target <= M Actual
1 0 5.00% 95.00% 5.00%
2 1 5.00% 77.65% 5.00%
3 1 5.00% 86.46% 5.00%
5 1 5.00% 92.36% 5.00%
10 2 5.00% 91.28% 5.00%
1 0 1.00% 99.00% 1.00%
2 1 1.00% 90.00% 1.00%
3 1 1.00% 94.10% 1.00%
5 1 1.00% 96.73% 1.00%
10 1 1.00% 98.45% 1.00%
1 0 0.10% 99.90% 0.10%
2 1 0.10% 96.80% 0.10%
3 1 0.10% 98.20% 0.10%
5 1 0.10% 99.00% 0.10%
10 1 0.10% 99.52% 0.10%

4
3-Class Sampling Plans

In a ‘3-Class’ sampling plan, both the m and M quantiles are used.

We will discuss again the same three approaches to designing a 3-class

sampling plan.

Approach 1: Given m, M and error fraction target, find n, cm and cM

This approach once again is used for acceptance testing, and the focus is on the
false negative fraction (‘consumer’s risk’ or ‘type II error’ or ‘β’).

Here ‘M’ again denotes the median concentration that, if exceeded, represents
‘unacceptable’ product that should be rejected with high reliability by the plan.
The quantile ‘m’ denotes the median concentration that separates ‘acceptable’
from ‘marginal’ product. Presumably ‘acceptable’ product should be rejected
infrequently.

Subject matter expertise or specifications provide a value for M that should be

rejected with high reliability (‘Power’ = 1 – β), say 95% or 99% or 99.9%. Such
expertise or specifications also provides the value for m that should be rejected
50% of the time (i.e., the ‘indifference’ level). The parameters n, cm and cM are
chosen to achieve the statistical power required.

The probability of rejecting unacceptable lots with median concentration M is

determined from eq.(L.1.) and the probability of rejecting indifferent lots with
median concentration m is

0.5 ~ P[(# > m) > cm] (eq. L.3.)

calculated from a binomial distribution with p = P[ X < m] = 0.50.

5
Some possible plans are:

Table L.4. 3-Class plans, given m, M as median concentrations

M M m m
m M Power Power Power Power
n c c Target Actual Target Actual
4 1 0 95.00% 93.75% 50.00% 68.75%
5 2 0 95.00% 96.88% 50.00% 50.00%
6 2 0 95.00% 98.44% 50.00% 65.63%
7 3 1 95.00% 93.75% 50.00% 50.00%
10 4 1 95.00% 98.93% 50.00% 62.30%
7 3 0 99.00% 99.22% 50.00% 50.00%
10 4 1 99.00% 98.93% 50.00% 62.30%
10 4 0 99.90% 99.90% 50.00% 62.30%

Approach 2: Given Pm, PM and error fraction target, find n, cm and cM

This approach is used primarily in designing sampling plans for process control,
and the focus is on the false positive fraction (‘producer’s risk’ or ‘type I error’ or
‘α’ or ‘FAR’).

M and m represent specified quantiles of the concentration of ‘normal’ production

under statistical control, with m < M. Their values are determined from their
percentile ranks Pm and PM either nonparametrically or parametrically based on
historical data. The parameters n, cm and cM are chosen to achieve the maximum
false positive fraction α allowed, typically 1% or 0.1%, depending upon sampling
rate.

The probability of a randomly chosen lot from normal production failing from both
rejection rules for the special case that cM = 0 is
𝑐𝑐
1-α = � �𝑛𝑛𝑘𝑘�𝑢𝑢𝑘𝑘 𝑝𝑝𝑛𝑛−𝑘𝑘 (eq. L.4.)
𝑘𝑘=0

where p = Pm and u = PM – Pm.

An additional design rule for 3-class plans is to make the rejection rules for m
and M roughly of equal impact in the determination of the total false positive
fraction. This can be achieved by determining PM for the 2-class plan with sample
size n and acceptance number cM = 0 and half the false positive fraction α, and
then choosing the parameters related to m. We assume therefore that PM has
been determined from

6
 α/2 = P[(# > M) > 0] (eq. L.5.)

calculated from a binomial distribution with n trials and p = QM.

Some possible plans are:

Table L.5. 3-Class sampling plan given m and M probabilities

Exact
α Probability Probability α
n c Target <m <M m+M
5 1 0.50% 98.50% 99.95% 0.454%
5 1 1.00% 97.50% 99.90% 1.047%
5 1 5.00% 94.50% 99.49% 4.720%
5 1 10.00% 91.50% 98.98% 9.650%
5 2 0.50% 93.50% 99.95% 0.493%
5 2 1.00% 91.50% 99.90% 1.019%
5 2 5.00% 85.00% 99.49% 4.894%
5 2 10.00% 80.00% 98.98% 9.919%
5 3 0.50% 85.00% 99.95% 0.470%
5 3 1.00% 81.00% 99.90% 1.041%
5 3 5.00% 71.00% 99.49% 5.028%
5 3 10.00% 64.50% 98.98% 10.045%

Approach 3: Given n, cm, cM and error fraction target, find Pm and PM

This approach results in achievement of an exact false positive fraction α, and

again is used in designing sampling plans for process control with the focus on
the false positive fraction (‘producer’s risk’ or ‘type I error’ or ‘α’ or ‘FAR’).

M and m again represent specified quantiles of the concentration of ‘normal’

production under statistical control, with m < M. Their values are determined from
their percentile ranks Pm and PM either nonparametrically or parametrically based
on historical data. The parameters n, cm and cM are specified, and Pm and PM
chosen to achieve the maximum false positive fraction α allowed, typically 1% or
0.1%, depending upon sampling rate.

7
As in the section titled ‘Given Pm, PM and error fraction target, find n, cm and cM’,
we assume cM = 0 to simplify the equations, and choose PM to satisfy eq.(L.5.).
The implicit eq.(L.4.) is then solved for Pm.

8
Some possible plans are:

Table L.6. Find m for 3-class plan from given α, M and σ

FAR Probability Probability

N c Target <M <m
5 1 0.200% 99.980% 98.966%
5 1 0.200% 99.980% 98.966%
5 1 0.200% 99.980% 98.966%
5 1 0.500% 99.950% 98.342%
5 1 0.500% 99.950% 98.342%
5 1 0.500% 99.950% 98.342%
5 1 1.000% 99.900% 97.608%
5 1 1.000% 99.900% 97.608%
5 1 1.000% 99.900% 97.608%
5 1 5.000% 99.495% 94.176%
5 1 5.000% 99.495% 94.176%
5 1 5.000% 99.495% 94.176%
5 2 0.200% 99.980% 95.237%
5 2 0.200% 99.980% 95.237%
5 2 0.200% 99.980% 95.237%
5 2 0.500% 99.950% 93.436%
5 2 0.500% 99.950% 93.436%
5 2 0.500% 99.950% 93.436%
5 2 1.000% 99.900% 91.610%
5 2 1.000% 99.900% 91.610%
5 2 1.000% 99.900% 91.610%
5 2 5.000% 99.495% 84.770%
5 2 5.000% 99.495% 84.770%
5 2 5.000% 99.495% 84.770%
5 3 0.200% 99.980% 87.778%
5 3 0.200% 99.980% 87.778%
5 3 0.200% 99.980% 87.778%
5 3 0.500% 99.950% 84.500%
5 3 0.500% 99.950% 84.500%
5 3 0.500% 99.950% 84.500%
5 3 1.000% 99.900% 81.392%
5 3 1.000% 99.900% 81.392%
5 3 1.000% 99.900% 81.392%
5 3 5.000% 99.495% 71.088%
5 3 5.000% 99.495% 71.088%
5 3 5.000% 99.495% 71.088%

9
Approach 4: Finding m and M from Pm and PM

The concentrations m and M may be determined nonparametrically as quantiles

of an observed empirical cumulative distribution function (ECDF) at probabilities
Pm and PM. Acceptable accuracy requires the number of underlying observations
N > 2 / QM. E.g., for PM = 99.9%, QM = 1 – PM = 0.001 and the requirement is that
N > 2000. Lack of available data typically limits this approach to PM = 99% or
less.

Extension to higher quantiles is possible by fitting available data to a parametric

model, such as a normal distribution for the log10-transformed concentrations.
Suppose this has been done, and the prevalence of zero results is P0, and the
mean and standard deviation of the log10-transformed concentrations are
estimated at ‘m’ and ‘s’.

If the entire mixture ECDF (including zero results) is the desired basis for
quantiles, compute adjusted quantiles for Pm and PM as

Pm’ = 1 – (1 – Pm) / (1 – P0) (eq. L.6.)

PM’ = 1 – (1 – PM) / (1 – P0) (eq. L.7.)

Find ym and yM as the quantiles of the normal distribution with mean m and
standard deviation s corresponding to Pm’ and PM’. Then compute m and M as
the antilogs of ym and yM. Note that observed zero results are still used in
applying the rejection rules.

10
Appendix M. Implied False Alarm Rate in a 3-class Sampling Plan Given m,
M, n, c, and σ.

In addition to designing 3-class sampling plans to achieve a desired false alarm

rate (FAR), we can also investigate the FAR implied by existing plans. We begin
by assuming that the unacceptable concentration limit M is chosen according to
some criterion and then evaluate the impact of different levels of process
variability (σ) on the FAR under existing 3 class sampling plans.

M percentile criterion: First, assume that M is chosen to be an extreme

percentile in the right hand tail of the distribution, such as the 99.5th percentile.
This means that M is chosen such that pd = 0.5% of individual samples have an
unacceptable concentration. For a given sampling plan, the FAR attributable to M
(FARM) = 1-(1-pd)n. This is simply the complement of eq. K.2. (Appendix K). For
an n = 5 sampling plan, FARM = 2.5%. For a given sampling plan, the overall
FAR is given by the complement of eq. K.3. (Appendix K). The overall FAR can
be partitioned into portions attributable to M and m: FAR = FARM + FARm.
Assuming a lognormal distribution, given pd and σlog10 values, we can calculate
pm from existing sampling plans based on the ratio of the limits (M/m). Given a
fixed M percentile, the implied µlog10 and percentile of the specified marginal limit
value (m) will vary depending on the process variability (σ). This has important
consequences for the FAR. Table M.1 presents the FARs implied by existing n =
5, c = 2 three-class sampling plans over a range of σlog10 values.

Table M.1. False alarm rates implied by existing 3-class sampling plans (M
percentile)

M
n c log(M/m) σlog10 FAR(%) FARM(%) FARm(%)
percentile
5 2 1 0.25 7.7 99.6 2.5 97.1
0.50 71.8 15.8 2.5 13.3
0.80 90.8 3.1 2.5 0.6
1.20 95.9 2.5 2.5 0.0
5 2 2 0.25 0.0 100.0 2.5 97.5
0.50 7.7 99.6 2.5 97.1
0.80 53.0 45.2 2.5 42.8
1.20 81.8 6.6 2.5 4.1

FARM criterion: Second, assume that M is chosen to achieve a desired FARM for
a given sampling plan, such as the conventional 0.3%. For a given sampling
plan, pd = 1-(1-FARM)1/n. For an n = 5 sampling plan, pd = 0.05% (M = 99.95th
percentile). Based on the same approach used to generate Table M.1, Table M.2
presents implied FAR.

1
Table M.2. False alarm rates implied by existing 3-class sampling plans (FARM)

m
σlog10 FAR(%) FARM(%) FARm(%)
n c log(M/m) percentile

5 2 1 0.25 23.2% 91.4% 0.3% 91.2%

0.50 89.8% 1.2% 0.3% 0.9%
0.80 97.8% 0.3% 0.3% 0.0%
1.20 99.3% 0.3% 0.3% 0.0%

5 2 2 0.25 0.0% 100.0% 0.3% 99.7%

0.50 23.2% 91.4% 0.3% 91.2%
0.80 77.9% 7.8% 0.3% 7.5%
1.20 94.5% 0.4% 0.3% 0.1%

The take away message from Tables M.1 and M.2 is that designing 3-class
sampling plans based on n, c, m, and M without considering process variability
(σ) can result in highly inconsistent false alarm rates. On the other hand, the
analysis in this section assumes that limits are based on reliable data from an
ideal process that is under statistical control and capable of meeting
microbiological specifications. If the m and M limit values are chosen instead on
the basis of strong observed associations with certain concentrations being
exceeded (e.g., measurements exceeding M are associated with observed
pasteurization process failures, measurements exceeding m and associated with
significantly reduced shelf life), then frequent occurrences where the
microbiological criteria were exceeded may indicate unsatisfactory process
quality. Under such circumstances, the events would be mischaracterized as
false alarms because the exceedances indicate a lack of process capability to
meet microbiological specifications.