ORIGINAL SCIENTIFIC ARTICLES
Percent Body Fat and Prediction of Surgical
Site Infection
Emily Waisbren, BS, Heather Rosen, MD, MPH, Angela M Bader, MD, MPH, Stuart R Lipsitz, ScD,
Selwyn O Rogers, Jr, MD, MPH, FACS, Elof Eriksson, MD, PhD, FACS
Obesity is a risk factor for surgical site infection (SSI) after elective surgery. Body mass index
(BMI) is commonly used to define obesity (BMI ⱖ30 kg/m2), but percent body fat (%BF)
(obesity is ⬎25%BF [men]; ⬎31%BF [women]) might better predict SSI risk because BMI
might not reflect body composition.
STUDY DESIGN: This prospective study included 591 elective surgical patients 18 to 64 years of age from
September 2008 through February 2009. Height and weight were measured for BMI. %BF was
calculated by bioelectrical impedance analysis. Preoperative, operative, and 30-day postoperative data were captured through interviews and chart review. Our primary, predetermined
outcomes measurement was SSI as defined by the Center for Disease Control and Prevention.
RESULTS:
Mean %BF and BMI were 34⫾10 and 29⫾8, respectively. Four-hundred and nine (69%)
patients were obese by %BF; 225 (38%) were obese by BMI. SSI developed in 71 (12%)
patients. With BMI defining obesity, SSI incidence was 12.3% in nonobese and 11.6% in obese
patients (p ⫽ 0.8); Using %BF, SSI occurred in 5.0% of nonobese and 15.2% of obese patients
(p ⬍ 0.001). In univariate analyses, significant predictors of SSI were %BF (p ⫽ 0.005), obesity
by %BF (p ⬍ 0.001), smoking (p ⫽ 0.002), National Nosocomial Infections Surveillance score
(p ⬍ 0.001), postoperative hyperglycemia (p ⫽ 0.03), and anemia (p ⫽ 0.02). In multivariable
analysis, obese patients by %BF had a 5-fold higher risk for SSI than nonobese patients (odds
ratio ⫽ 5.3; 95% CI, 1.2⫺23.1; p ⫽ 0.03). Linear regression was used to show that there is a
positive, nonlinear relationship between %BF and BMI.
CONCLUSIONS: Obesity, defined by %BF, is associated with a 5-fold increased SSI risk. This risk increases as
%BF increases. %BF is a more sensitive and precise measurement of SSI risk than BMI.
Additional studies are required to better understand this relationship. (J Am Coll Surg 2010;
210:381–389. © 2010 by the American College of Surgeons)
BACKGROUND:
deaths, and add between $4.5 and $5.7 billion per year to
health care costs.2,3
The majority of studies on obesity and its risk factors
define obesity using body mass index (BMI). In the existing
body of literature, obesity, regardless of the definition used,
is thought to increase the risk of SSI, possibly because of
technical difficulty, altered wound physiology, and impaired immune response.4-6 Because 40% of the elective
surgical population is overweight or obese, SSIs that might
be a result of an obesity-related risk factor deserve additional scrutiny.7 However, the term obesity is often overused
or misused. The Oxford English Dictionary defines obesity
as “very fat,” and the World Health Organization defines it
as a “condition with excess body fat to the extent that health
and well-being are adversely affected.”8
Instead of these colloquial definitions, quantitative tools
can be employed to provide an objective measure of obesity.
The most common is BMI, which uses an index of weight
and height to classify patients into 1 of 3 categories: healthy
Surgical site infection (SSI) is defined by the Centers for
Disease Control and Prevention as an infection occurring
at or near the surgical incision within 30 days of a procedure.1 SSIs occur in approximately 2 million patients each
year in the United States, are associated with about 90,000
Disclosure Information: Nothing to disclose.
Presented at the American College of Surgeons 95th Annual Clinical Congress, Chicago, IL, October 2009.
Received November 6, 2009; Revised December 21, 2009; Accepted January
5, 2010.
From the Plastic Surgery Division (Waisbren, Eriksson), Center for Surgery
and Public Health (Rosen, Lipsitz, Rogers), Anesthesia and Center for Preoperative Evaluation (Bader), and Division of Trauma, Burn, and Surgical
Critical Care (Rogers), Brigham and Women’s Hospital, Boston, MA; and
Department of Plastic Surgery, Children’s Hospital Boston, Boston, MA
(Rosen).
Correspondence address: Elof Eriksson, MD, PhD, Plastic Surgery Division,
Brigham and Women’s Hospital, 45 Francis St, Boston, MA 02115. Email:
eeriksson@partners.org
© 2010 by the American College of Surgeons
Published by Elsevier Inc.
381
ISSN 1072-7515/10/$36.00
doi:10.1016/j.jamcollsurg.2010.01.004
382
Waisbren et al
Body Fat and Surgical Site Infection Risk
Abbreviations and Acronyms
BIA
%BF
BMI
SSI
⫽
⫽
⫽
⫽
bioelectrical impedance analysis
percent body fat
body mass index
surgical site infection
weight, overweight, or obese. According to the World
Health Organization8 and for the purposes of this study,
obesity is defined as a BMI ⱖ30. However, BMI has
known limitations, such as failure to distinguish fat mass
from lean mass.9 The ability of BMI to correctly identify
obesity in a population is widely debated.10-14
This study aims to examine a different quantitative tool,
percent body fat (%BF), which is measured by bioelectrical
impedance analysis (BIA). This measure can be obtained
easily and inexpensively by using a standing scale with
metal plates onto which each foot is placed. After entering
variables such as patient age; height; and weight, %BF,
along with a variety of additional variables (eg, total body
water, total body fat, fat mass), is calculated within seconds.
This information is then stored in a computer spreadsheet
with a patient identifier. The implementation of BIA to
measure %BF is easy, inexpensive, and accessible, and it is
foreseeable that this device will be implemented in clinics
everywhere.15,16 The American Council on Exercise defines
obesity as men with ⬎25%BF and women with ⬎31%BF.
The ability of BMI to identify “fatness” or obesity is based
on the assumption that body mass correctly estimates body
fat.17 However, in some situations, BMI fails to identify
“fatness.” Some individuals, eg, body builders, are overweight for their height but not necessarily “overfat,” and
others have a normal BMI but their body fat exceeds what
is healthy. %BF can better measure obesity than BMI.18
In this report, we describe a prospective cohort study
that uses data to analyze the relationship between %BF and
SSI incidence to determine whether %BF is a more accurate predictor than BMI. Our aims were as follows: to
define the relationship between obesity and SSI using the
quantitative definitions for obesity, including %BF and
BMI; to determine the predictive value of %BF for development of SSI; and to characterize the relationship between %BF and BMI.
METHODS
Patient population
After approval from our Institutional Review Board, we
culled the daily registry of surgical procedures at the Weiner
Center for Preoperative Evaluation at Brigham and Women’s Hospital for patients who met eligibility criteria. Pa-
J Am Coll Surg
tients were eligible for inclusion in the study if they were
between the ages of 18 and 64 years and planning to undergo elective surgery in 1 of 7 categories, ie, general, orthopaedic, cardiothoracic, gynecologic, plastic surgery,
otolaryngology, or urology. This particular age group was
chosen in an attempt to be consistent with prior studies of
using BIA to measure %BF, the majority of which focuses
on this mid-range age group, and because we aimed to
study as homogenous a group as possible.15,16,18,19 Trauma
patients, patients undergoing emergency procedures, and
immune-compromised patients (eg, HIV/AIDS infection,
organ transplant recipients, and patients receiving corticosteroids or chemotherapy) were excluded. The final cohort
was a 591-patient representative sample of the patients seen
at the Center from September 2008 through February
2009. This group of patients was selected randomly from
the clinic schedule before their appointment.
Measurement of adiposity and end points
Preoperative data were collected by 1 investigator (EW)
using the following procedure: medical record review to
determine inclusion/exclusion criteria; administration of a
questionnaire through personal interview; and focused
physical examination, which included measurement of
height and weight (for calculation of BMI) and measurement of %BF by BIA (model TBF310; Tanita). Use of
body-fat estimation by BIA has been validated; the correlation between BIA and dual-energy x-ray absorptiometry
was 0.85 for men and 0.88 for women. The correlation
between fat-free mass by BIA and hydrodensitometry in
healthy adults under age 50 has been reported to be 0.98,
with a standard error of estimation of 2.29 kg.18,19 The cost
of the %BF measuring device was under $2,000 and was
very easy to use, similar to a traditional standing scale. It
involves the patient standing on the metal foot pads with
bare feet and the measurement is printed out and transferred to a computer spreadsheet within seconds. Perioperative variables were collected from the anesthesia record
and operative note, and postoperative outcomes variables
were collected from the medical record (including laboratory values) and follow-up telephone questionnaire.
The primary surgical outcomes of interest included SSI,
as defined by Centers for Disease Control and Prevention
criteria.1 Secondary outcomes included other wound complications, such as wound dehiscence; fat and skin necrosis;
seroma; hematoma; or other perioperative non⫺woundrelated outcomes, such as mortality; myocardial infarction;
arrhythmia; pulmonary embolism, deep-vein thrombosis,
cerebral vascular accident, acute renal failure, urinary tract
infection; pneumonia; sepsis; bacteremia; and Clostridium
difficile infection.
Vol. 210, No. 4, April 2010
Waisbren et al
Body Fat and Surgical Site Infection Risk
383
Table 1. Univariate Analysis for Development of Surgical Site Infection
Variable
%BF, mean ⫾ SD
Obese (%BF), n (%)
Current smoker, n (%)
Pedal edema, n (%)
History of cancer, n (%)
Diabetes, n (%)
Peripheral vascular disease, n (%)
Duration of operation, h, mean ⫾ SD
Postoperative length of stay, d, mean ⫾ SD
Drain placement, n (%)
Wound classification, n (%)
Clean
Clean-contaminated
Contaminated
Dirty
ASA classification, n (%)
I
II
III
IV
NNIS score, n (%)
0
1
2
3
Postoperative glucose,* mg/dL, mean ⫾ SD
Postoperative hyperglycemia, n (%)†
Postoperative hyperglyemia in nondiabetics, n (%)
Postoperative anemia‡,§, n (%)
History of SSI, n (%)
Participants with
postoperative SSI (nⴝ71)
Participants without
postoperative SSI (n ⴝ 520)
37.4 ⫾ 8.8
62 (87.3)
22 (31)
18 (25.4)
38 (53.5)
12 (16.9)
6 (8.5)
3.7 ⫾ 2.9
4.0 ⫾ 4.5
28 (39.4)
33.7 ⫾ 10.5
347 (66.7)
83 (16)
73 (14)
200 (38.5)
49 (9.4)
17 (3.3)
2.7 ⫾ 1.7
2.0 ⫾ 2.8
132 (25.4)
31 (43.7)
36 (50.7)
2 (2.8)
2 (2.8)
303 (58.3)
200 (38.5)
13 (2.5)
4 (0.8)
4 (5.6)
39 (54.9)
28 (39.4)
0 (0)
52 (10)
354 (68.1)
109 (21.0)
5 (1.0)
18 (25.4)
28 (39.4)
24 (33.8)
1 (1.4)
161.6 ⫾ 57.8
38 (84.4)
29 (82.9)
42 (59.2)
5 (7.0)
250 (48.1)
210 (40.4)
55 (10.6)
5 (1.0)
142.2 ⫾ 41.5
170 (68.8)
132 (63.8)
233 (44.8)
7 (1.4)
p Value
0.005
⬍0.001
0.002
0.01
0.02
0.05
0.03
⬍0.001
⬍0.001
0.01
0.04
0.01
⬍0.001
0.007
0.03
0.03
0.02
0.001
All variables were included in the stepwise multivariate analysis.
*Postoperative glucose recorded in 298 patients.
†
Fasting glucose ⬎120 mg/dL.
‡
Hematocrit ⬍36 women, ⬍41 men.
§
Postoperative hematocrit recorded in 286 patients.
ASA, American Society of Anesthesiologists; %BF, percent body fat; NNIS, National Nosocomial Infections Surveillance; SSI, surgical site infection.
Statistical analysis
We consulted with the National Surgical Quality Improvement Program at our institution. Their data were used to
guide our sample size calculations. With at least 500 total
patients and at least one-third of them nonobese (determined by BMI or %BF), this study is designed to have 80%
power to detect an odds ratio of at least 2.3 between SSI
and obesity using multivariable logistic regression with a
type I error of 5%; this calculation assumed a coefficient of
determination of 0.3 between the dichotomous variable
obesity and all possible confounders.20
Patient characteristics and study results were summarized using proportions, means with standard deviations,
and medians with interquartile ranges. Wilcoxon rank sum
tests and t-tests (where appropriate) were used for comparing
continuous variables between 2 groups. For dichotomous and
categorical variables, chi-square and Fisher’s exact tests (where
appropriate) were used to determine differences in proportions between 2 groups.The main a priori confounders, which
were identified through a literature search before data collection, are given in Table 1. To determine if these possible confounders are related to the SSI, the Wilcoxon rank sum test
was used for comparing continuous confounders between patients with and without SSI; similarly, Fisher’s exact tests were
used to determine differences in proportions between patients
with and without SSI.
384
Waisbren et al
Body Fat and Surgical Site Infection Risk
Multivariable logistic regression was used to determine if
obesity (defined either by BMI or %BF) was an important
predictor of SSI, while controlling for potential confounders. Forward selection was used to identify potential confounders. When added to the logistic regression model in
the forward selection, any potential confounder was kept in
the multivariable model if it led to a ⱖ10% change in the
odds ratio estimate between outcomes and obesity. Once a
confounder entered the model in the forward selection, it
was not dropped out afterward. Diabetes mellitus was kept
in the multivariable model because all authors agreed that it
was most likely the strongest confounder. Linear regression
was used to characterize the relationship between %BF and
BMI. We report odds ratios, p values, and 95% confidence
intervals from the logistic regressions. After finding the
best-fitting logistic regression model, adjusted probabilities
of SSI (adjusted for significant confounders) were calculated and plotted as a function of obesity.
Statistical analyses were performed using Stata/IC (version 10, Stata Corp). All p values were 2-sided.
RESULTS
Clinical characteristics
Of the 591 recruited participants, 12 were excluded from
follow-up because of cancellation or deferral of surgery.
The postoperative telephone questionnaire was completed
by 586 of 591 patients (99%). Preoperative characteristics
and operative variables for the final cohort are shown in
Table 2. A total of 68% of participants were defined as
“relatively healthy” at baseline, based on the American Society of Anesthesiologists score of 2. Most surgical wounds
were classified as “clean” or “clean-contaminated.” Of the
surgical specialties included, the majority of participants
underwent general (26%, n ⫽ 153) and orthopaedic surgery (18%, n ⫽ 107). One-hundred and nine participants
(18%) underwent laparoscopic procedures, 28 (4.7%) underwent bariatric surgery, and permanent implants were
placed in 165 (28%) of the patients. Pre- and postoperative
laboratory data were available for a subgroup of patients.
Analysis of serum laboratory values showed no statistically significant difference in preoperative glucose levels in
SSI and non-SSI patients (p ⫽ 0.15). However, those in
whom SSI developed were more commonly hyperglycemic
(glucose level ⬎120 mg/dL) after the operation than those
in whom SSI did not develop (84% of SSI patients versus
69% of non-SSI patients; p ⫽ 0.03). Nondiabetics in
whom SSI developed had a higher incidence of hyperglycemia than those in whom SSD did not develop (83%
versus 64%; p ⫽ 0.03). Of 286 patients for whom a postoperative hematocrit was recorded, those in whom SSI developed had a higher incidence of postoperative anemia
J Am Coll Surg
(women with hematocrit ⬍36 and men with hematocrit
⬍41) than non-SSI patients (59% versus 45%; p ⫽ 0.02).
Nonobese patients by %BF were younger than obese
patients by %BF, but the ages were the same when obese
versus nonobese patients by BMI. Additionally, there
were more obese women (%BF) than the nonobese
group (%BF), but there was no difference in gender
between obese and nonobese patients when obesity was
measured using BMI.
SSI and postoperative complications
In the 30-day postoperative follow-up period, complications were experienced by 127 (25%) participants, among
whom SSI was identified in 71 (12%). Eighty percent of
SSIs were classified as superficial incisional infections. SSI
was most common after oncologic (23% of oncologic procedures resulted in SSI) and cardiothoracic (17% of cardiothoracic procedures resulted in SSI) operations. Gynecologic
surgery had the lowest rate of SSI (8% of gynecologic procedures resulted in SSI). There was no statistically significant
difference in the incidence of SSI between men and women (p
⫽ 0.36). Wound-related complications, including dehiscence, necrosis, and seroma or hematoma formation occurred
in 80 (13.5%) participants and 57 (9.6%) experienced non–
wound-related complications. Wound intervention was required in 22 (3.7%) participants, and the incidence of hospital
readmission was 5%.
SSI and obesity
There was a positive, nonlinear relationship between %BF
and BMI (BMI and BMI2 were both significant at p ⬍
0.001 in a linear regression model with %BF as one of the
outcomes). %BF identified more patients as obese (69% of
the cohort) than did BMI (38% of the cohort). Patients in
whom SSI developed had a significantly higher %BF than
those in whom SSI did not develop (37 ⫾ 9 versus 34 ⫾
11; p ⫽ 0.005). More patients in the SSI group than in the
non-SSI group were obese by %BF (87% versus 67%; p ⬍
0.001). There was no statistically significant difference in
the percentage of obese patients as defined by BMI in those
with and without SSI (p ⫽ 0.79). Additionally, there was
no significant difference in the incidence of SSI between
the obese and nonobese BMI groups (12% of patients versus 12.1%; p ⫽ 0.79). When patients are grouped according to discrete ranges of BMI, the relationship between SSI
and BMI is not nearly as linear or predictable as it is when
patients are grouped according to %BF (Fig. 1A, 1B).
With %BF to define obesity, patients who were obese
had a higher occurrence of SSI than nonobese patients
(15% of obese patients versus 5% nonobese patients; p ⬍
0.001). Differentiation in SSI incidence between the obese
and nonobese groups was attenuated when men were ana-
Vol. 210, No. 4, April 2010
Waisbren et al
Body Fat and Surgical Site Infection Risk
385
Table 2. Demographic and Clinical Characteristics of Obese and Nonobese Patients
Characteristic
All participants
(nⴝ591)
Obese (%BF)
(nⴝ409)
49.7 ⫾ 10.1
50.5 ⫾ 9.9
Nonobese (%BF)
(nⴝ182)
p Value*
Obese (BMI)
(nⴝ225)
Nonobese (BMI)
(nⴝ366)
p Value†
0.89
Preoperative demographic and clinical
characteristics
48 ⫾ 10.3
0.006
49.8 ⫾ 9.6
49.7 ⫾ 10.4
Women, n (%)
Age, y, mean ⫾ SD
396 (67)
288 (70)
108 (59)
0.008
148 (66)
248 (68)
0.62
History of cancer, n (%)
238 (40.3)
168 (41)
70 (39)
0.55
73 (32)
165 (45)
0.002
Antihypertensives, n (%)
187 (31.6)
159 (39)
28 (15)
⬍0.001
103 (46)
84 (23)
⬍0.001
Current smoker, n (%)
105 (17.8)
70 (17)
35 (19)
0.53
40 (18)
65 (18)
1.0
Alcohol (⬎6⫻/wk), n (%)
95 (16.1)
67 (16)
28 (15)
0.76
29 (13)
66 (18)
0.1
COPD, n (%)
97 (16.4)
72 (18)
25 (14)
0.24
44 (20)
53 (15)
0.11
Pedal edema, n (%)
91 (15.4)
73 (18)
18 (10)
0.01
53 (24)
38 (10)
⬍0.001
Recent surgery, n (%)
72 (12.2)
50 (12)
22 (12)
0.96
32 (14)
40 (11)
0.24
Remote infection, n (%)
64 (10.8)
45 (11)
19 (10)
0.84
27 (12)
37 (10)
0.47
History of SSI, n (%)
12 (2)
10 (2)
2 (1)
0.36
5 (2)
7 (2)
0.8
⬍0.001
Diabetes, n (%)
61 (10.3)
53 (13)
8 (4)
0.002
41 (18)
20 (6)
History of myocardial infarction, n (%)
14 (2.4)
12 (3)
2 (1)
0.25
9 (4)
5 (1)
0.04
Angina, n (%)
42 (7.1)
29 (7)
13 (7)
0.98
20 (9)
22 (6)
0.19
Peripheral vascular disease, n (%)
23 (3.9)
15 (4)
8 (4)
0.67
8 (4)
15 (4)
0.74
Cerebrovascular accident, n (%)
19 (3.2)
17 (4)
2 (1)
0.07
10 (4)
9 (3)
0.18
192 (32.5)
133 (33)
59 (32)
0.98
75 (33)
117 (32)
0.73
Duration of operation, (mean ⫾ SD, h)
2.8 ⫾ 1.9
2.9 ⫾ 2.1
2.4 ⫾ 1.4
0.003
2.9 ⫾ 1.6
2.8 ⫾ 2.1
0.51
Perioperative antibiotics, n (%)
569 (97.1)
395 (98)
174 (96)
0.15
217 (97)
352 (97)
Recent weight loss, n (%)
Operative variables
Wound classification, n (%)
0.01
Clean
334 (56.5)
238 (58)
96 (53)
129 (57)
205 (56)
Clean-contaminated
236 (39.9)
160 (39)
76 (42)
89 (40)
147 (40)
15 (2.5)
7 (2)
8 (4)
5 (2)
10 (3)
6 (1)
4 (1)
2 (1)
2 (1)
4 (1)
Contaminated
Dirty
ASA classification, n (%)
0.01
0.001
I
56 (9.5)
28 (7)
28 (15)
10 (4)
46 (13)
II
393 (66.5)
277 (68)
116 (64)
148 (66)
245 (67)
III
137 (23.2)
100 (25)
37 (20)
64 (30)
73 (20)
IV
5 (0.9)
4 (1)
1 (1)
3 (1)
2 (1)
NNIS score, n (%)
0.02
0.008
0
268 (45.4)
179 (44)
89 (49)
92 (41)
176 (48)
1
238 (40.3)
164 (40)
74 (41)
89 (40)
149 (41)
2
79 (13.4)
64 (16)
15 (8)
43 (19)
36 (10)
2 (0.5)
4 (2)
5 (2)
7 (2)
3
6 (1)
0.81
0.001
Implant recipient, n (%)
165 (27.9)
121 (30)
44 (24)
0.18
77 (34)
88 (24)
Drain placement, n (%)
160 (27.1)
116 (28)
44 (24)
0.31
64 (28)
96 (26)
0.007
0.57
Catheter placement, n (%)
332 (56.2)
239 (58)
93 (51)
0.1
139 (62)
193 (53)
0.03
Postoperative length of stay, d, mean ⫾ SD
2.2 ⫾ 3.1
2.4 ⫾ 3.2
1.9 ⫾ 2.9
0.09
2.2 ⫾ 2.8
2.8 ⫾ 2.1
0.77
Blood loss, mL, mean ⫾ SD
138 ⫾ 281
157 ⫾ 317
96 ⫾ 168
0.02
148 ⫾ 251
133 ⫾ 298
0.55
*p Value for obese versus nonobese (%BF).
†
p Value for obese versus nonobese (BMI).
SSI, surgical site infection; Obesity, defined by body-mass index (BMI ⬎30 kg/m2) or % body fat (males ⬎25 %BF, females ⬎31 %BF); Current smoker, cigarette use
within 1 year; Recent surgery, surgical procedure within 1 mo of current operation; Angina, within 1 y of current operation; Weight loss, ⬍5 lbs within 6 mo; ASA Score,
American Society of Anesthesiology preoperative assessment score: 1. Healthy, 2. Mild systemic disease, 3. Severe systemic disease, 4. Severe systemic disease that is a constant
threat to life; NNIS, National Nosocomial Infections Surveillance System, risk index developed to predict a surgical patients’ risk of acquiring a SSI, ranges from 0-3,
variables included are ASA score ⬎3, and operation classified as “contaminated (3)” or “dirty (4)”, and an operation lasting over T hours (T depends on operative procedure
being preformed); Wound classification clean, operative procedure does not enter into a normally colonized viscus or lumen of the body; clean-contaminated, operative
procedure enters into a colonized viscus or cavity of the body, but under elective and controlled circumstances; contaminated, gross contamination is present at the surgical
site in the absence of obvious infection; dirty, active infection is already present.
386
Waisbren et al
Body Fat and Surgical Site Infection Risk
J Am Coll Surg
Figure 1. Risk of surgical site infection (SSI) developing (A) measured by body mass
index in men and women. There was a nonlinear relationship between BMI and SSI risk
for the whole group. (B) SSI measured by percent body fat in men and women, revealing
a linear, predictable relationship for percent body fat.
lyzed separately from women (13.2% versus 5.4%; p ⫽
0.09). When patients were grouped in discrete ranges of
body fat, there was a positive linear relationship between
%BF and incidence of SSI. When this analysis was stratified by gender, the positive linear relationship was maintained for women, but attenuated for men.
Univariate predictors of SSI are shown in Table 1. Both
continuous %BF and obesity as defined by %BF were significantly different between the SSI and non-SSI groups
(for %BF continuous: p ⫽ 0.005; for obese by %BF: p
ⱕ0.001). All components of National Nosocomial Infections Surveillance score (which includes duration of operation, wound classification, and American Society of Anesthesiologists classification) were significantly different as
well (duration of operation: p ⬍ 0.001; wound classification: p ⫽ 0.04; ASA score: p ⫽ 0.01). A higher percentage
of patients had a history of previous SSI in the current SSI
group (7% versus 1%, p ⫽ 0.001). There was no difference
in the receipt of perioperative antibiotics between the
groups.
After adjustment for smoking, National Nosocomial Infections Surveillance score, and diabetes in a multivariable
analysis, obese patients as defined by %BF were ⬎5 times
more likely to have an SSI than nonobese patients (odds
ratio ⫽ 5.3; 95% CI, 1.2⫺23.0; p ⫽ 0.03). In a separate
model using BMI to define obesity, obesity was not an
independent predictor of SSI (odds ratio ⫽ 1.3; 95% CI,
0.6⫺2.7; p ⫽ 0.45), see Table 3 for results.
DISCUSSION
The aims of this study were to define the relationship between obesity and SSI, determine the predictive capabilities of %BF for development of SSI, and identify correla-
Vol. 210, No. 4, April 2010
Waisbren et al
Body Fat and Surgical Site Infection Risk
387
Table 3. Results of Multiple Logistic Regression Analysis for Surgical Site Infection
Characteristic
Nonobese by percent body fat (reference)
Obese by percent body fat
Nonobese by BMI (reference)
Obese by BMI, kg/m2
Cigarette smoking
NNIS score 0 (reference)
NNIS score 1
NNIS score 2
NNIS score 3
Diabetes
Odds ratio
NA
NA
1
1.3
2.3
1
1.8
5.8
1.9
1.3
BMI model
95% CI
NA
NA
NA
0.6⫺2.7
1.3⫺4.1
NA
0.9⫺3.4
2.8⫺12.0
0.2⫺18.5
0.6⫺2.8
p Value
Odds ratio
% Body fat model
95% Cl
p Value
NA
NA
NA
0.45
0.006
NA
0.07
⬍0.001
0.58
0.52
1
5.3
NA
NA
2.2
1
1.8
5.2
3.7
1
NA
1.2⫺23.1
NA
NA
1.2⫺4.0
NA
0.95⫺3.4
2.6⫺10.7
0.36⫺38.9
0.47⫺2.13
NA
0.03
NA
NA
0.008
NA
0.07
⬍0.001
0.27
1
BMI, body mass index; NNIS, National Nosocomial Infections Surveillance System.
tions between %BF and BMI. We found that obese
patients as defined by %BF were ⬎5 times more likely to
have SSI develop after surgery than nonobese patients. In
contrast, BMI-defined obesity was not an independent risk
factor for SSI. Although the risk of SSI developing increases
with an increase in %BF in a predictable pattern, there is no
corresponding consistent relationship when using BMI.
Given these results, we conclude that %BF is better than
BMI in determining the association between obesity and
risk of occurrence of postoperative SSI.
Many studies have shown that obesity by BMI is associated with an increased risk for SSI development.7,21-23
However, another body of literature supports our results
that obesity measured by BMI is not a risk factor for
SSI.24,25 Our results agree with the latter, most likely a
reflection of our large sample size. No studies to date have
shown the strength of the association when using %BF.
Regardless of the BMI debate, our results show a very
strong relationship between SSI and obesity measured by
%BF. Obesity can increase the risk of SSI by several proposed mechanisms, such as technical difficulty during surgery, tissue trauma, and tension on the healing wound. The
wound might also have decreased circulation and decreased
oxygenation, which have been shown to be risk factors for
SSI.26 An impaired immune response can also result from
the interactions of adipocytes with the immune system.27-32
Percent body fat allows us to accurately categorize patients in a way that reflects the risk of SSI developing. Even
though BMI is the one of the most widely used measures of
obesity, our data question how successfully this classification scheme identifies individuals at increased risk because
of obesity. In our study, patients who are obese per BMI do
not necessarily have increased morbidity. With a more consistent relationship between %BF and SSI risk, measuring
%BF allows us to understand the risks of SSI developing
associated with each increment of %BF (Fig. 1B). Using
%BF in the clinical setting helps physicians and nurses
identify patients who are in the high-risk category. Our
finding that obese patients by %BF are ⬎5 times more
likely to have SSI develop than nonobese patients should be
taken seriously. It can be beneficial to counsel patients on
the merits of preoperative weight loss as a preventive
measure.
Our study identified a nonlinear relationship between
BMI and %BF, suggesting that BMI does not adequately
define obesity. Other studies have also found a quadratic
relationship between body fat and BMI,10,11,18 which might
explain why the incidence of SSI was much higher when
using %BF than BMI. The quadratic relationship suggests
that BMI does not accurately reflect “adiposity” or “fatness”
for many patients. In a large study using 2,032 adult patients from the Framingham study, Roubenoff and colleagues found that BMI was a poor predictor of fatness in
women (R2 ⫽ 0.55) and men (R2 ⫽ 0.38) and imprecise
(standard error of estimate ⫽ 5 percentage points), and
there was a quadratic relationship between %BF and
BMI.18 Frankenfield and colleagues10 showed that measurement of body fat is a more appropriate way to assess
obesity in people with a BMI ⬍30. They found that 30%
of men and 46% of women with BMI ⬍30 had obesity
levels of body fat and were misclassified by BMI.10 In the
clinical setting, %BF is a far more accurate and precise
measure of fatness than BMI.
Interestingly, nonobese patients by %BF were younger
than obese patients by %BF, but the ages were the same
when obese versus nonobese patients by BMI (Table 2).
Perhaps this is because %BF is sensitive to differences in
body composition that change subtly with age (ie, with age
muscle mass is replaced by fat mass33), which might not be
detected by BMI because it calculates only height and
weight, which tend to be consistent as we age.
Similarly, we found that there were more obese women
(%BF) than the nonobese group (%BF), but there was no
difference in gender between obese and nonobese patients
388
Waisbren et al
Body Fat and Surgical Site Infection Risk
when obesity was measured using BMI (Table 2). Our
study group did not have equal numbers of men and
women, so a prospective study looking specifically at this
would be interesting. However, we hypothesize that
women have different body composition than men and
perhaps the sensitivity of %BF allows the subtle differences
to be detected. To further investigate this relationship, we
altered Figure 1 to include only men and only women, but
we did not find a substantial difference from the Figure 1
shown in this publication.
Limitations of our study are as follows: We collected data
using both medical records and oral questionnaires for each
participant. A telephone interview was conducted that
asked questions useful in identifying SSI and other woundrelated complications (eg, Did you observe purulent discharge from the wound? Did erythema/edema/pain develop around the wound?). These data were used to make
sure that we did not miss any SSIs or complications that
might not have been documented in medical records. This
questionnaire was conducted in the postoperative period,
introducing recall bias as one limitation. Use of questionnaires to obtain details of the medical history can be complicated by bias inherent in self-reported data, although
this bias should be nondifferential between the groups of
patients. Because patients were recruited from a tertiary
care hospital that serves as a referral center in a major city,
they might not be representative of the general population,
as they were typically older than 50 years or had substantial
comorbidities. Although obesity by %BF was found to be
an independent predictor of SSI by both bivariate and multivariable analysis, it is a complex disease and it is possible
that %BF is a surrogate for other factors that are the true
cause of the increased risk of SSI. One limitation of our
measurement technique is that BIA values are affected by
numerous variables, including body position, hydration
status, consumption of food and beverages, ambient air
and skin temperature, recent physical activity, and conductance of the examining table. Although we were unable to
control for all of these, we requested that all patients were
fasting at the time of measurement. We also measured %BF
in the morning to be consistent. Additionally, we measured
body fat 3 times in each participant and took the mean of
this score to reduce intrameasurement variability.
In conclusion, obesity, as defined by %BF, is an independent predictor of SSI. As %BF increases, the risk of SSI
developing also increases. Additionally, we conclude that a
positive, nonlinear relationship exists between %BF and
BMI. Based on these findings in the preoperative setting,
%BF is a more accurate predictor of SSI than BMI.
The 5-fold greater SSI incidence in %BF obese than
BMI obese patients is convincing evidence that %BF
J Am Coll Surg
should be used to assess preoperative SSI risk. %BF was
consistently found to be a more sensitive marker for obesity
risks and a better way to detect obesity than BMI. Yet, even
with %BF’s potential use, prospective studies are needed to
pinpoint the proper obesity category cutoffs for men and
women.
Data within this study should be used for several groups,
such as obese patients considering an elective surgical procedure and surgeons assessing risk and providing informed
consent. There might be a benefit to counseling patients,
when possible and clinically appropriate, to consider preoperative weight loss as a preventive measure. Although
BMI is used almost universally in the clinical setting, it
might not be a useful or necessary measurement. Rather, to
calculate the risk posed by obesity, it is better to measure
%BF, which is easy, quick, and inexpensive and yields more
useful information than BMI. Additional studies should be
undertaken to determine if our results are consistent with
other adverse outcomes in addition to SSI.
Author Contributions
Study conception and design: Waisbren, Eriksson
Acquisition of data: Waisbren
Analysis and interpretation of data: Waisbren, Rosen, Bader,
Lipsitz, Rogers, Eriksson
Drafting of manuscript: Waisbren, Eriksson
Critical revision: Waisbren, Rosen, Bader, Rogers, Lipsitz,
Eriksson
REFERENCES
1. Mangram AJ, Horan TC, Pearson ML, et al. Guideline for prevention of surgical site infection, 1999. Hospital Infection Control Practices Advisory Committee. Infect Control Hosp Epidemiol 1999;20:250–278; quiz 79⫺80.
2. Haley RW, Culver DH, White JW, et al. The nationwide nosocomial infection rate. A new need for vital statistics. Am J Epidemiol 1985;121:159–167.
3. Stone PW, Larson E, Kawar LN. A systematic audit of economic
evidence linking nosocomial infections and infection control
interventions: 1990⫺2000. Am J Infect Control 2002;30:145–
152.
4. Anaya DA, Dellinger EP. The obese surgical patient: a susceptible host for infection. Surg Inf 2006;7:473–480.
5. Choban PS, Flancbaum L. The impact of obesity on surgical
outcomes: a review. J Am Coll Surg 1997;185:593–603.
6. Flancbaum L, Choban PS. Surgical implications of obesity. Ann
Rev Med 1998;49:215–234.
7. Choban PS, Heckler R, Burge JC, Flancbaum L. Increased incidence of nosocomial infections in obese surgical patients. Am
Surg 1995;61:1001–1005.
8. World Health Organization. Obesity: Preventing and Managing
the Global Epidemic. Report on a WHO Consultation on Obesity. Geneva: World Health Organization, 305 June, 1997:
WHO/NUT/NCD/98.1; 1998.
Vol. 210, No. 4, April 2010
9. Willett WC, Dietz WH, Colditz GA. Guidelines for healthy
weight. N Engl J Med 1999;341:427–434.
10. Frankenfield DC, Rowe WA, Cooney RN, et al. Limits of body
mass index to detect obesity and predict body composition.
Nutrition 2001;17:26–30.
11. Gallagher D, Heymsfield SB, Heo M, et al. Healthy percentage
body fat ranges: an approach for developing guidelines based on
body mass index. Am J Clin Nutr 2000;72:694–701.
12. Micozzi MS, Albanes D, Jones DY, Chumlea WC. Correlations
of body mass indices with weight, stature, and body composition in men and women in NHANES I and II. Am J Clin Nutr
1986;44:725–731.
13. Ravussin E, Swinburn BA. Pathophysiology of obesity. Lancet
1992;340:404–408.
14. Smalley KJ, Knerr AN, Kendrick ZV, et al. Reassessment of
body mass indices. Am J Clin Nutr 1990;52:405–408.
15. Kyle UG, Bosaeus I, De Lorenzo AD, et al. Bioelectrical impedance analysis—part I: review of principles and methods. Clin
Nutr 2004;23:1226–1243.
16. Kyle UG, Bosaeus I, De Lorenzo AD, et al. Bioelectrical impedance analysis-part II: utilization in clinical practice. Clin Nutr
2004;23:1430–1453.
17. Bray GA. Health hazards of obesity. Endocrinol Metab Clin
North Am 1996;25:907–919.
18. Roubenoff R, Dallal GE, Wilson PW. Predicting body fatness:
the body mass index vs estimation by bioelectrical impedance.
Am J Public Health 1995;85:726–768.
19. Lukaski HC, Bolonchuk WW, Hall CB, Siders WA. Validation
of tetrapolar bioelectrical impedance method to assess human
body composition. J Appl Physiol 1986;60:1327–1332.
20. Hsieh FY, Bloch DA, Larsen MD. A simple method of sample
size calculation for linear and logistic regression. Stat Med 1998;
17:1623–1634.
21. Canturk Z, Canturk NZ, Cetinarslan B, et al. Nosocomial infections and obesity in surgical patients. Obes Res 2003;11:
769–775.
Waisbren et al
Body Fat and Surgical Site Infection Risk
389
22. Dindo D, Muller MK, Weber M, Clavien PA. Obesity in general
elective surgery. Lancet 2003;361:2032–2035.
23. Merkow RP, Bilimoria KY, McCarter MD, Bentrem DJ. Effect
of body mass index on short-term outcomes after colectomy for
cancer. J Am Coll Surg 2009;208:53–61.
24. Mullen JT, Davenport DL, Hutter MM, et al. Impact of body
mass index on perioperative outcomes in patients undergoing
major intra-abdominal cancer surgery. Ann Surg Oncol 2008;
15:2164–2172.
25. Tuech JJ, Regenet N, Hennekinne S, et al. Laparoscopic colectomy for sigmoid diverticulitis in obese and nonobese patients: a
prospective comparative study. Surg Endosc 2001;15:1427–
1430.
26. Hopf HW, Hunt TK, West JM, et al. Wound tissue oxygen
tension predicts the risk of wound infection in surgical patients.
Arch Surg 1997;132:997–1004; discussion 1005.
27. Fantuzzi G. Adipose tissue, adipokines, and inflammation. J
Allergy Clin Immunol 2005;115:911–919; quiz 20.
28. Hartmann M, Jonsson K, Zederfeldt B. Effect of tissue perfusion and oxygenation on accumulation of collagen in healing wounds. Randomized study in patients after major abdominal operations. Eur J Clin Nutr Acta Chir 1992;158:
521–526.
29. Lamas O, Marti A, Martinez JA. Obesity and immunocompetence. Eur J Clin Nutr 2002;56[Suppl 3]:S42⫺S45.
30. Nieman DC, Henson DA, Nehlsen-Cannarella SL, et al. Influence of obesity on immune function. J Am Diet Assoc 1999;99:
294–299.
31. Pond CM. Adipose tissue and the immune system. Prostaglandins Leukot Essent Fatty Acids 2005;73:17–30.
32. Stallone DD. The influence of obesity and its treatment on the
immune system. Nutr Rev 1994;52:37–50.
33. Elmadfa I, Meyer AL. Body composition, changing physiological functions and nutrient requirements of the elderly. Ann Nutr
Metab 2008;52:2–5.