PUBD Training
PUBD Training
PUBD Training
(PUBD)
October/November, 2013
PUBD - Overview
Introduction
How to recognize PUBD
Study Documents that identify PUBD
SOPs, MANs and FORMs
Processes
Roles
Potentially Unblinding events
Escalation and Notification
Next Steps
Consequences
PUBD Introduction
In recent months there have been several potentially
unblinding events, including:
PUBD files in unrestricted folders (folders not properly restricted
or files placed on standard folders)
Potentially unblinding information in email to blinded staff (LDM)
Scrambled dataset with potentially unblinding information
Potentially Unblinding Data (PUBD) is any information or data that is collected during
the conduct of an Amgen clinical trial and prior to initial lock of the database which
may:
Indicate or infer treatment assignment
Create bias in the conduct of a study
Impact the integrity of the clinical trial or data
MAN-001619 Accessing and Distributing Restricted Data for Amgen Clinical Trials**
External Data
Transfer to external group
Loads to CTDB
Transfer to CTSHARED for Non-CTDB or for unblinded group review
Any GSO DM TA and Systems (including FSPs) should notify the following oversight
roles immediately when a potential unblinding risk is identified:
Studys Amgen PL, FSP Ops Lead, Amgen Ops Lead
refer to the DM Systems ECS Portal for FSP and Amgen Ops Lead contacts
Amgen Ops lead, FSP Ops Lead and Amgen PL will engage the reporter to establish
initial risk assessment confirming the reported risk is not a false alarm
If it is not a false alarm, Amgen PL will notify the following distribution list immediately
and within the same business day of identified incident with summary details and
provide daily updates until the issue is closed
Summary details to be included
Distribution List
1.
2.
3.
4.
5.
6.
10
Restricted folder should not include the following unix groups: cdmonc, cdmearlydev, cdmneph, cdmbone,
cdmgenmed, cdminflam
Request restriction of folder via Amgen FAST business owners** if not correctly restricted
For PUB data/outputs on CTSHARED,
Amgen Ops Lead requests a Varonis report for impacted file access audit history listed by user IDs, file
access action type, and timestamps within the established exposure timeframe via Amgen FAST Systems
Owner (Min Lee)
Amgen PL/Amgen Ops Lead perform unblinding exposure analysis using the Varonis report
EDA to set BUI setting to delete if missing and load empty file. This will remove data and audit trail. Load
latest header data. (inactivate if missing would not remove audit trail).
LDM to request the site to blank out the PUD data value from EDC and provide appropriate reason for
change
Ops lead to escalate to the CSI and EDC Ops for a full assessment of exposure for the trial
Per
confirmation
from Compliance,
For Internal
Use
Only. Amgen Confidential.
Always verify the correct folder under the correct study prior to placing PUBD.
Never send any datasets via email. Always place on ctshared and send the ctshared
file link to the requester
Always review email body to ensure there in no PUB data before forwarding or
responding
12
Additional Information on
Processes
Any testing should use non PUBD data (either dummy or nonblinded production data)
Confirm output location is accurate and the folder is restricted appropriately by manually
reviewing the folder properties
PUBD Checks not to be added to automated scheduled until after the first PROD data is
received
After the first PROD data is received, the EDA will place the data in the correct location and
advise study team data transferred per current process
Study team will request SAS EC team to execute the SAS EC for the PUBD data for the first
time
Prior to first execution, programmer will confirm output directory in program one additional time
Programmer will execute PUBD checks and confirm that the output is as expected and in the
proper location
Programmer will add PUBD check to the automated schedule and the checks will run
automatically on future transfers
14
External Data
SOW/Charter/etc
Vendor document specification should include blinded information
Covance DM will contact FSP EDA Ops Lead if any new SOWs drafts contain
blinded information for review.
Example is for repeat or unscheduled visits as PUBD check run by EDA cannot
discriminate by certain unscheduled or repeat visits, it is best to have all
unscheduled/repeats as blinded.
15
External Data
EDA to execute pubdchk for all header data prior to load into CTDB and non-ctdb
data files
Pubdchk output will need to be reviewed by EDA and FSP Ops lead to verify
PUBD
Do not load data if any PUB data found in header file
Do not place non-ctdb data files if any PUB data found in header file
Loading of PUBD data to CTDB only after signed unblinding form (i.e. FORM008236) is received and confirmation from LDM to load PUBD data. This is typically
within 3 days from Initial Lock
If folder does not exist, request the Amgen FAST Business Owners (Mickey Cheng or Mehul Shukla) to
create the folder with appropriate restrictions.
16
17
18
19
20
21
22
23
24
25
would be changed to
a916_20070561-PUBDVAL-COVANCEIN_DBL_UNIT-CHEMLAB_3FULL-2011AUG05124025.txt
26
27
28
29
30
31
If the PubD specs are not provided with the request, ask
for it.
PubD specs are created by Biostatistics/GSP. Ideally,
they should reside in EDM Teams
32
Result Files
/cdm/amgXYZ/Protocol#/lab/bsm
/cdm/amgXYZ/Protocol#/lab/covance
/cdm/amgXYZ/Protocol#/lab/mrl
33
/
ctshared/amgXYZ/indication/Protocol#/cdm/sas/
edc/data/ssd
Non-CTDB Result Files
/cdm/amgXYZ/Protocol#/lab/bsm
/
ctshared/amgXYZ/indication/Protocol#/cdm/sas/
edc/data/ssd_blinded_qlab
For Internal Use Only. Amgen Confidential.
34
Scrambled data
/ctshared/amgXYZ/indication/Protocol#/cdm/sas/edc/data/ssd_disguised
35
36
37
38
Scrambling: Process
Rave
Extract Relevant RAVE data using CDMS2SAS
E.g. LAB, AB, PK etc.
This will extract the data under ../data/ssd
Check and confirm that ssd_blinded and ssd_da both are
restricted.
Non-CTDB
39
Process (contd.)
If the program are not already set up, set it up. This can
be done by copying various programs from other study
and modifying it according to the needs
Programs reside under /programs/data_disguised at
study level
40
Scrambling: Steps
Run MakeDA_Rave.sas macro to create disguised assignments.
This program needs to run the first every time there is a new
request for scrambling
It creates disguiseassignments.sas7bdat SAS Dataset under
ssd_da
Three variables: _pt_, Randomization Date, Disguised Assignments
Since this contains actual patient ids with corresponding disguised
assignments, one needs to be very careful with this. Accidental
leakage of the file can cause potentially unblinding
41
What this does is Reads the Result Files, merges with Header data and
than creates the unblinded file with blinded data that is stored under
ssd_blinded which again is a restricted directory
Outputs from these are e.g. lab_u, ab_u, pk_u etc. Specs on naming the
files will be provided in the PubD specs
42
What this does is Reads the Result Files, merges with Header data
and then creates the unblinded file with blinded data that is stored
under ssd_blinded which again is a restricted directory
Outputs from these are e.g. db_ab_1_u, db_pk_1_u, db_lab_1_u etc.
Specs on naming the files will be provided in the PubD specs
43
44
45
46
47
48
49
50
Use Copy and Paste, but forget to update the code per specs
In order to calculate the SI values, the analyte unit and conversion factor (from reported unit to SI unit) must
be set up in Rave Lab Admin module for the analyte. In addition, the range (specifically the high range) must
be assigned to the analyte for the lab vendor used for the study.
Remediation:
If conversion factor is missing, LDM to confirm with LDR if there is an available conversion factor. If so,
the conversion factor is to be approved and added to the analyte in the Rave Lab Admin Module. If no
conversion factor is available, SI value cannot be provided in the scrambled dataset only reported
values
If only the range is missing, LDM to confirm with LDR if textbook ranges are available for use in the
study. If available, LDM to request to update ranges for the study. If not available, LDM to request dummy
ranges to be used (i.e. 0-999). Note, the dummy range selected should be such that none of the results
are falsely flagged as abnormally low or high.
51
Move all unblinded/PubD data from ssd_blinded directory to the Month_year folder
created under it.
E.g. Move _u files from /ctshared/amgXYZ/indication/Protocol#/cdm/sas/edc/data/ssd_blinded
to /ctshared/amgXYZ/indication/Protocol#/cdm/sas/edc/data/ssd_blinded/Month_Year
Move all Scrambled Data from ssd_disguised to the Month_year folder created under
it.
E.g. Move _d files from /ctshared/amgXYZ/indication/Protocol#/cdm/sas/edc/data/ssd_disguised
to /ctshared/amgXYZ/indication/Protocol#/cdm/sas/edc/data/ssd_disguised/Month_Year
NOTE: The SLA to process PubD is 3 Business Days. In case of New studies that need this set up, it may
take a day or two additional.
52
53
54
55
56
57