Thomas Jefferson University
Jefferson Digital Commons
Department of Surgery Faculty Papers &
Department of Surgery
Impact of Obesity on Perioperative Morbidity and
Mortality Following Pancreaticoduodenectomy
Timothy K. Williams
Thomas Jefferson University
Ernest L. Rosato
Thomas Jefferson University
Eugune P. Kennedy
Thomas Jefferson University
Karen A. Chojnacki
Thomas Jefferson University
Thomas Jefferson University
See next page for additional authors
This Article is brought to you for free and open access by the Jefferson Digital Commons. The Jefferson Digital Commons is a service of Thomas
Jefferson University'sAcademic & Instructional Support & Resources Department (AISR). The Commons is a showcase for Jefferson books and
journals, peer-reviewed scholarly publications, unique historical collections from the University archives, and teaching tools. The Jefferson Digital
Commons allows researchers and interested readers anywhere in the world to learn about and keep up to date with Jefferson scholarship. This article
has been accepted for inclusion in Department of Surgery Faculty Papers & Presentations by an authorized administrator of the Jefferson Digital
Commons. For more information, please contact: JeffersonDigitalCommons@jefferson.edu.
Williams, Timothy K.; Rosato, Ernest L.; Kennedy, Eugune P.; Chojnacki, Karen A.; Andrel,
Jocelyn; Hyslop, Terry; Doria, Cataldo; Sauter, Patricia K.; Bloom, Jordan; Yeo, Charles J.; and
Berger, Adam C., "Impact of Obesity on Perioperative Morbidity and Mortality Following
Pancreaticoduodenectomy" (2009).Department of Surgery Faculty Papers & Presentations.Paper 15.
Timothy K. Williams, Ernest L. Rosato, Eugune P. Kennedy, Karen A. Chojnacki, Jocelyn Andrel, Terry
Hyslop, Cataldo Doria, Patricia K. Sauter, Jordan Bloom, Charles J. Yeo, and Adam C. Berger
This article is available at Jefferson Digital Commons:http://jdc.jefferson.edu/surgeryfp/15
As submitted to Journal of the American College of
Surgeons and later accepted and published in Journal of
the American College of Surgeons Volume 208, Issue 2,
February 2009, Pages 210-217
Impact of Obesity on Perioperative Morbidity and Mortality Following
Timothy K. Williams, MD*, Ernest L. Rosato, MD, FACS*, Eugene P. Kennedy, MD, FACS*,
Karen A. Chojnacki, MD, FACS*, Jocelyn Andrel, MSPH†, Terry Hyslop, PhD†, Cataldo Doria,
MD, PhD, FACS*, Patricia K. Sauter, CRNP*, Jordan Bloom, BS*, Charles J. Yeo, MD,
FACS*, and Adam C. Berger, MD, FACS*
From the *Department of Surgery, Thomas Jefferson University; Jefferson Pancreas, Biliary and
Related Disease Center, †Division of Biostatistics, Kimmel Cancer Center, Philadelphia, PA.
Correspondence: Adam C. Berger, MD; 1100 Walnut Street; MOB, Suite 500; Philadelphia, PA
19107; (215) 955-1622; fax—(215) 923-8222; email—email@example.com
Key Words—obesity, pancreaticoduodenectomy, complications, blood loss
Running Title—Obesity increases serious complications in patients undergoing
Presented in part at the 2007 meeting of the Pancreas Club (Washington, DC) and the 2008 3rd
Academic Surgical Congress (Huntington Beach, CA).
Background: Obesity has been implicated as a risk factor for perioperative and postoperative
complications. The aim of this study was determine the impact of obesity on morbidity and
mortality in patients undergoing pancreaticoduodenectomy (PD).
Study Design: Between January 2000 and July 2007, 262 patients underwent PD at Thomas
Jefferson University Hospital (TJUH), of whom 240 had complete data, including body mass
index (BMI) for analysis. Data on BMI, preoperative parameters, operative details, and post-
operative course were collected. Patients were categorized as obese (BMI >30 kg/m2),
overweight (25≤BMI<30), or normal weight (BMI<25). Complications were graded according
to previous published scales. Other endpoints included length of postoperative hospital stay,
blood loss, and operative duration. Analyses were performed using univariate and multivariable
Results: There were 103 (42.9%) normal weight, 71 (29.6%) overweight and 66 (27.5%) obese
patients. There were 5 perioperative deaths (2.1%) with no differences across BMI categories. A
significant difference in median operative duration and blood loss between obese and normal
weight patients was identified (439vs. 362.5minutes, p= 0.0004; 650 vs. 500 ml, p=0.0139).
Furthermore, median length of stay was marginally significantly longer for by BMI (9.5 vs. 8
days, p=0.095). While there were no significant differences in superficial wound infections,
obese patients did have an increased rate of serious complications compared to normal weight
patients (24.2% vs. 13.6%, respectively; p=0.10).
Conclusions: Obese patients undergoing PD have a significantly increased blood loss and longer
operative time, but do not have a significantly increased length of postoperative hospital stay or
rate of serious complications. These findings should be considered when assessing patients for
operation and when counseling patients regarding operative risk, but do not preclude obese
individuals from undergoing definitive pancreatic surgery.
Pancreaticoduodenectomy is the surgical standard for several disease states. Although
largely performed for periampullary malignancies, it is often performed for a variety of benign
pathologies as well.1 Despite advances in the technique of pancreaticoduodenectomy over the
years, it remains a procedure that carries significant morbidity, even when performed at high
volume centers.2 Accordingly, numerous studies have been performed that evaluate preoperative
risk factors for perioperative and postoperative morbidity and mortality in these patients.
Advanced age, as well as certain co-morbid conditions (i.e. diabetes and coronary artery disease)
and serologic factors (i.e. elevated blood urea nitrogen and low serum albumin) have been
demonstrated to be independent risk factors for the incidence of postoperative complications
following pancreaticoduodenectomy.3-5 However, few studies to date have evaluated the impact
of obesity on postoperative complications following pancreaticoduodenectomy.2
The National Institutes of Health (NIH) defines obesity as a body mass index (BMI) of
≥30 kg/m2. Normal weight individuals are considered to have a BMI of <25 kg/m2, while
overweight is defined as having a BMI ≥25 and <30 kg/m2. At epidemic proportions in our
society today, obesity affects approximately 30 percent of adults in the USA. By the year 2015,
that value is projected to reach 40 percent.6 Obesity is linked to many illnesses such as diabetes,
osteoarthritis, hypertension and others. It has also been shown to be a risk factor for the
development of several malignancies, including pancreatic cancer.7 Obesity has also been
implicated as a risk factor for the development of postoperative complications following a
variety of operations.8-11 The purpose of this study is to evaluate the impact of obesity on the
occurrence of perioperative and postoperative morbidity and mortality in patients undergoing
pancreaticoduodenectomy at a high volume tertiary care hospital.
This is a retrospective analysis of an Institutional Review Board approved prospectively
maintained database for all patients undergoing pancreatic surgery at the Thomas Jefferson
University Hospital (TJUH). All cases included in this study were resected at our institution
between January 2000 and July 2007. During this time period, 262 pancreaticoduodenectomies
were performed. Of these cases, 240 had data on BMI at the time of surgery and were selected
for analysis. Twenty patients were excluded due to incomplete data collection, and two cases
performed for penetrating abdominal trauma were also excluded from analysis.
The prospective database used in this study was populated from patient charts, an
electronic medical record and anesthesia records. Demographic and preoperative clinical data
utilized for this study included age, sex, height, weight, serum albumin level, the presence or
absence of a preoperative biliary stent, and the American Society of Anesthesiologists (ASA)
score. Body mass index was calculated using the following formula: BMI=
weight(kg)/height(m)2. For this analysis, patients were divided into three groups based on BMI:
normal weight (BMI <25), overweight (25 ≤ BMI < 30), or obese (≥30). This BMI classification
system is endorsed by the World Health Organization and the National Institutes of Health, and
is the most widely accepted means of stratifying individuals based on weight.
A total of thirteen surgeons performed Whipple operations during this time period;
however, the majority of cases (217) were performed by four surgeons. Surgical technique was
based largely on surgeon preference. The two main techniques used were the
pancreaticoduodenectomy (PD) with antrectomy (classic type), and the pylorus preserving
pancreaticoduodenectomy (PPPD). Over this time period, there were a total of 129 (54%)
pylorus preserving and 103 (43%) classic type PD. Eight patients (3%) underwent other types of
pancreatic resections, mainly completion or total pancreatectomies. Over the study period, there
was a transition from classic Whipple operations to pylorus-preserving PD. Furthermore, the
volume of operations performed increased substantially over the last 2 years of the study period.
Sixty-six patients were enrolled in a randomized clinical trial which began in July 2006 at TJUH
and randomized patients to a duct-to-mucosa or invagination method of pancreaticojejunostomy
(NCT00359320). Postoperative care was again by surgeon preference. However, in October
2005, we initiated a critical pathway for patients undergoing PD—the results of this have been
described previously.12 For this study, 76 patients were treated before the pathway and 164 were
treated after its implementation.
Perioperative data collected for analysis included intraoperative blood loss, length of
operation and specimen pathology. Postoperative clinical endpoints included length of
postoperative hospital stay and complications. Postoperative complications were categorized
based on a modified version of the surgical complication classification system of Clavien et al
(Table 1).13,14 This system has been established as a means of objectively classifying
postoperative complications. Grade 1 and 2 complications were considered minor, whereas
grades 3-5 were considered severe. While some patients did experience more than one
complication, only the highest grade complication was included in the comparative analysis
across BMI groups. Delayed gastric emptying (DGE) was determined by whether the patient had
a nasogastric tube for more than 10 days after surgery, could not proceed to a regular diet within
10 days, and had vomiting for more than 3 consecutive days after the fifth postoperative day.
Pancreatic fistula was defined using the definition of the International Study Group.
Patients with normal weight preoperatively (BMI < 25) were defined as the control group
for this analysis. BMI was categorized according to usual convention, <25, 25-29.9, and
≥30k/m2. Complications were categorized according to severity, as 0, 1-2 and 3-5. A number of
pre- and post- operative continuous variables had skewed, non-normal distributions.
Comparisons between BMI categories for these variables were made using the Kruskal-Wallis
test, and if significant, followed by pairwise Bonferroni-corrected Wilcoxon tests. Pre- peri- and
post-operative categorical variables were analyzed using univariate logistic regression (for
dichotomous variables) or proportional odds modeling (for polytomous variables). Post-hoc
Bonferroni-corrected pairwise Wald chi-square tests were completed if univariate results were
significant. Pairwise test results for univariate analyses are reported in the text only.
The proportional odds model, a form of logistic regression that establishes separate
starting points (intercepts) with single odds ratios estimated for each level of outcome across
covariates, was used to build multivariable models of association with endpoints estimated blood
loss (EBL), operative duration, and length of stay. EBL was categorized as <400, 400-600, 601-
900, or >900ml for these analyses,, surgical duration was categorized according to quartiles
observed in the population, namely ≤329 minutes, >329-399, >399-472, or >472 minutes, length
of stay was categorized as ≤ 6, 7-9, 10-14, and 15 or more days, and ASA was categorized as 1-2
and 3-4. Each of these models initially included the following: sex, BMI, ASA, albumin, type of
resection, age, and presence of complications. For each outcome, multivariable modeling
proceeded in a backwards stepwise manner, retaining those effects where p<0.05. In the case of
multi-level covariates, if an individual level of the covariate was significant at p<0.05, the
variable was also retained.
Of the 240 patients, 103 (42.9%) were normal weight, 71 (29.6%) were overweight, and
66 (27.5%) were obese. There were no significant differences between the BMI groups with
regard to age, sex, ASA class, or preoperative albumin (Table 2). The American Society of
Anesthesiologists (ASA) class was used as an indication of co-morbid illness for this study. For
the subgroups based on BMI, 66% of normal weight, 60.9% of overweight and 72.3% of obese
individuals were ASA class III. Most of the remaining patients were ASA class II (Table 2).
Although there was an increased number of ASA class III obese patients, this did not reach
statistical significance. There was a statistically higher prevalence of preoperative biliary stents
in the obese and overweight groups compared to normal weight group (p = 0.0003, and P = 0.02,
All surgical specimens were analyzed by a pathologist skilled in evaluating pancreatic
pathology. Most operations (74%) were performed for cancer and nearly half were for
pancreatic cancer (n=117; 49%). Other malignant diagnoses included ampullary and duodenal
adenocarcinomas, cholangiocarcinoma, duodenal GIST (n=2), neuroendocrine tumors, and
metastatic lesions (n=2). The most common benign lesions for which PD was performed was
intraductal papillary mucinous neoplasm (IPMN, n=23). Resections were also performed for
mucinous or serous cystadenoma, autoimmune pancreatitis, and adenoma (Table 3).
Intraoperative Estimated Blood Loss
There were significant differences in blood loss in the 3 groups of patients. The median
EBL for normal weight patients was 500 ml vs. 650 ml in both overweight and obese patients,
respectively. On univariate analysis (Table 4), blood loss was globally associated with BMI
(p=0.02). Using the Bonferroni-corrected alpha of 0.025, overweight values were marginally and
obese values significantly increased compared to normal weight (p=0.06, p=0.04, respectively).
By multivariable analysis, obesity proved to be an independent risk factor for increased
intraoperative blood loss (OR 2.00, p=0.02), however being overweight did not quite reach
statistical significance (OR=1.74, p=0.05). Interestingly, multivariable analysis identified male
sex as an independent risk factor for increased intraoperative blood loss (OR 3.88, p<0.0001)
Duration of Surgery
The lengths of the procedures were analyzed utilizing the BMI categories described
above. Normal weight patients had the lowest median operative time (363 minutes). In
comparison, overweight and obese patients had increased operative duration in the univariate
model, with a medians of 413 and 439 minutes (Bonferroni-adjusted p = 0.0736 and p=0.0008,
respectively) (Table 4). Again, the multivariable model supported obesity, but not overweight as
a significant risk factor for increased surgery duration (OR=2.39, p=0.005) (Table 5). As was
seen for intraoperative blood loss, male sex was identified as an independent risk factor for
increased duration of surgery (OR=2.55, p=0.0002).
Length of Stay
Postoperative length of hospital stay was chosen as an endpoint, as it serves as a measure
of resource utilization. In this analysis, normal weight individuals had a median postoperative
length of stay of 8 days. Overweight individuals did not have a significantly different median
LOS in comparison to normal weight patients (Table 4). Obese individuals did stay longer
postoperatively compared with normal weight patients but this was not significant (9.5 days,
p=0.06). Based on the multivariable analysis, obesity proved to be an independent risk factor for
increased length of stay (OR=1.9, p=0.04) (Table 5). Also seen on the multivariable analysis is
that the type of resection is an independent risk factor for the increased length of stay (OR =
15.2, p<0.001) (Table 5). This is partially reflected by a doubling of the occurrence of delayed
gastric emptying (7% vs. 3%) in patients undergoing classic PD.
A total of 120 (50%) patients experienced one or more complications postoperatively.
There were 29 (12.1%) grade I, 50 (20.8%) grade II, 25 (10.4%) grade III, 11 (4.6%) grade IV
and 5 (2.1%) grade V complications (Table 6). On univariate analysis, there was no significant
difference in the complication rate by BMI, although the individual comparison between obese
and normal individuals was marginally significant (56.3% vs. 42.7%, p = 0.08) (Table 4). The
prevalence of serious complications (grades 3-5) across BMI groups was: normal weight
(13.6%), overweight (15.5%), and obese (24.2%). On univariate analysis, this increase in serious
complications between normal weight and obese patients was not statistically significant
(p=0.13), neither was this trend statistically significant in the multivariable analysis. The
mortality rates were similar across the groups. The complication rate was higher amongst
patients undergoing classic PD compared to PPPD (54% vs. 49%); however, this was not
statistically significant. Additionally, after October 2005, when the volume of PD significantly
increased, the complication rate also decreased from 56% to 49%. Again this was not a
significant change. It does partially reflect the change from predominantly classic PD and PPPD
and the increase in volume at our institution as well as the implementation of a critical pathway.
In regards to specific complications, the risk of local wound infections was not
statistically significantly different across BMI categories (6%, 4%, and 5%). Additionally, the
occurrence of pancreatic fistula was similar in all BMI categories (4%, 6%, and 6%,
respectively). There was no difference in pancreatic fistula rate in classic PD versus PPPD.
However, there was a higher rate of intra-abdominal collections requiring either an interventional
or open drainage procedure for obese patients, as compared to normal weight patients (7% vs.
14%, respectively, p=0.05).
Our series implicates obesity (defined as BMI >30) as an independent predictor for
several perioperative and postoperative complications following pancreaticoduodenectomy (PD).
These include increased operative blood loss, increased operative duration and increased length
of postoperative hospital stay. Furthermore, obesity appears to be a risk factor for an increased
overall complication rate. More specifically, it serves as a risk factor for intra-abdominal
collections requiring a drainage procedure.
Intraoperative Blood Loss and Duration of Surgery
Our series demonstrates a significantly increased intraoperative blood loss in obese
patients compared to normal weight individuals. While this finding has not been previously
described in patients undergoing PD, it has been noted for a variety of other surgical procedures.
In a recent series, increased blood loss was noted for obese patients undergoing open
gastrectomy.15 Other studies have reached similar conclusions in patients undergoing
laparoscopic hysterectomy and transhiatal esophagectomy.16,17 An increase in the occurrence of
perioperative complications in patients with increased blood loss is supported by many studies in
the literature. For example, in a study of over 500 laparoscopic nephrectomies, Turna et al
demonstrated that increased blood loss was an independent and significant predictor of
postoperative complications.18 Additionally, other studies have also demonstrated increased
complication rates in those undergoing spine surgeries when blood loss was greater.19
Interestingly, in a study of 367 patients undergoing PD, House et al found that there was not a
significant difference in blood loss for patients with BMI ≥ 30 kg/m2 compared to those with
BMI <30 kg/m2.
There is also a common association between perioperative blood transfusion and
increased complications, with transfusion usually being a surrogate for increased blood loss in
the operating room. The impact of blood transfusion on patients undergoing cardiac surgery has
been clearly established as an independent risk factor for postoperative morbidity and mortality,
as well as increased length of postoperative hospital stay and increased incidence of low-output
heart failure.20-22 In patients undergoing liver resection, perioperative blood transfusion is a
prognostic factor for postoperative complications in general.22 In a study of 100 patients
undergoing hepatobiliary resections for cancer, the complication rate was 94% in those receiving
transfusion vs. 52% for those who were not transfused.23
Additionally, we report a significantly longer operative duration for obese patients.
While intuitive, this conclusion has not been described for patients undergoing PD. This finding
has been reported for patients undergoing gynecologic, and general surgical/oncologic
procedures however.15,16,24 The clinical significance of this finding is less well studied, but in
one study of 136 patients undergoing liver surgery, the complication rate was 10% for operations
of 2 hours or less and 44% for operations greater than 3.5 hours.22 In a study of abdominal
wound dehiscence after laparotomy using the Veteran’s Health Administration NSQIP database,
Webster et al found that operative time greater than 2.5 hours was a significant contributing
factor for dehiscence (albeit not nearly as important as many other factors).25 Obviously patients
undergoing PD will have operative times greater than two and a half hours, but increased
operative time is associated with increased complications.
Length of Postoperative Hospital Stay
13 Download full-text
In this analysis, obese patients stayed in the hospital significantly longer than normal
weight patients. This endpoint was not evaluated in a recent series by House et al examining
preoperative predictors (including obesity) of postoperative complications in patients undergoing
PD.2 Several studies involving non-pancreatic surgery have evaluated this endpoint but failed to
show that obesity correlates with increased length of stay.15-17,24,26,27
Although not deemed a postoperative complication directly, increased postoperative
length of stay frequently reflects the occurrence of a postoperative complication, and may serve
as a surrogate for postoperative morbidity. Our multivariable analysis shows that postoperative
complications do correlate with increased length of stay. This conclusion was supported by the
recent study of House et al.2 Since our analysis did show an increased risk of serious
complications in obese compared to normal weight patients, it seems intuitive that obese patients
would have an increased length of stay. The fact that the above mentioned studies failed to show
an increased length of stay in obese patients may therefore be explained by the fact that these
studies did not show an increased rate of complications in obese patients.15,16,24,26,27
Previous work from our institution has demonstrated that the implementation of a critical
pathway resulted in a decreased length of stay and a significant decrease in hospital costs.12 In
this study, the median length of stay decreased from 13 days to 7 days with a decrease in hospital
charges by approximately $120,000.12 Interestingly, when we re-examined the length of stay
data for patients operated on after the initiation of the clinical pathway, the median length of stay
was 6, 7, and 7 days respectively for normal weight, overweight, and obese patients. Therefore,
it seems that when patients are maintained on a strict clinical pathway, large differences in
postoperative length of stay may not be seen for obese and overweight patients.