Impact of Diabetes on Postoperative Outcomes Following
Colon Cancer Surgery
Nitasha Anand, MD1, Christopher A. Chong, MD2,3, Rachel Y. Chong, MD, PhD3,
and Geoffrey C. Nguyen, MD, PhD, FRCPC1,4
1Mount Sinai Hospital Division of Gastroenterology, University of Toronto Faculty of Medicine, Toronto, ON, Canada;2Queen’s University
Faculty of Health Sciences, Kingston, ON, Canada;3Department of Medicine, Lakeridge Health, Oshawa, ON, Canada;4Johns Hopkins
University School of Medicine, Baltimore, MD, USA.
BACKGROUND: Diabetes is the sixth most common
cause of death in the US and causes significant
postoperative mortality and morbidity.
OBJECTIVE: To characterize the impact of diabetes
among patients undergoing surgery for colorectal cancer.
DESIGN: This is is a retrospective cohort study.
PARTICIPANTS: Patients in the Nationwide Inpatient
Sample (NIS) who had undergone colorectal cancer
surgery between 1998 and 2005.
MEASUREMENTS: Using multivariate regression, we
determined the association of diabetes status with
postoperative mortality, postoperative complications,
and length of stay.
KEY RESULTS: An estimated 218,534 patients had
undergone surgery for colorectal cancer. We categorized
subjects by the presence of diabetes, the prevalence of
which was 15%. Crude postoperative in-hospital mortal-
ity was lower among diabetics compared to non-diabetics
(2.5% vs. 3.2%, P<0.0001). Adjusted mortality was 23%
lower in those with diabetes compared to non-diabetics
(aOR 0.77; 95% CI: 0.71–0.84). Diabetics also had lower
adjusted post-operative complications compared to non-
diabetics (aOR 0.82; 95% CI: 0.79–0.84). In uninsured
individuals and patients <50 years of age, there was no
protective association between diabetes and either in-
hospital mortality or postoperative complications.
CONCLUSIONS: Inpatients undergoingcolorectalcancer
surgery, those with diabetes had a 23% lower mortality
and fewer postoperative complications compared to non-
diabetics. The mechanisms underlying this unexpected
observation warrant further investigation.
KEY WORDS: diabetes; nationwide; colorectal cancer; hyperglycemia.
J Gen Intern Med 25(8):809–13
© Society of General Internal Medicine 2010
Diabetes affects 7% of the US population1and is the sixth
most common cause of death in the US. Diabetics have a 50%
chance of requiring surgery at some point in their lives,2and
the proportion of surgical patients with diabetes is close to
20%.3Diabetic patients often have microvascular and marcro-
vascular pathology4that influences their perioperative course5
and have a significantly higher risk of perioperative infection
and post operative cardiovascular morbidity and mortality.6-9
Hyperglycemia may be the underlying factor that mediates
worse outcomes in diabetics. The association between hyper-
glycemia and infection has been long recognized; studies
report diverse defects in neutrophil and monocyte dysfunction
such as adherence, chemotaxis and phagocytosis,10,11while
improvement in blood sugars helps restore granulocyte func-
tion.12Additionally, acute hyperglycemia may impair the
protective effects of cardiac ischemic preconditioning which
may lead to more extensive myocardial injury.13As well,
hyperglycemia appears to increase inflammatory markers and
oxidative stress resulting in endothelial cell dysfunction and
therefore promoting thrombosis.3
Diabetes may have a particularly negative impact on
individuals undergoing colorectal cancer surgery because of
the patients′ older age and the procedure’s inherent higher
risk. We have conducted this administrative claims study to
explore and generate further hypotheses regarding the poten-
tial impact of diabetes on mortality following colorectal surgery
for cancer, we conducted a nationwide, population-based
analysis using the Nationwide Inpatient Sample. We further
sought to distinguish whether the presence of the long-term
complications of diabetes further accentuated the disease’s
impact on post-surgical outcomes relative to uncomplicated
diabetes and no diabetes.
All data were extracted from the Nationwide Inpatient Sample
(NIS) between 1998 and 2005. The NIS is the largest all-payer
database of national hospital discharges, maintained as part of
the HealthcareCostand Utilization Project (HCUP)bythe Agency
for Healthcare Research and Quality (AHRQ). The NIS is a 20%
stratified sample of non-federal, acute-care hospitals in the
United States. Each record in the NIS includes a unique
identifier, demographic data (age, gender, and race), hospital
transfer status, admission type (emergent, urgent, or elective),
primary and secondary diagnoses (up to 15), primary and
Received August 27, 2009
Revised February 2, 2010
Accepted March 10, 2010
Published online March 30, 2010
secondary procedures (up to 15), expected primary and second-
ary insurance payers, total hospital charges, length of stay, and
size, teaching status). The NIS data concur with the National
Hospital Discharge Survey, supporting data reliability14.
We used Clinical Modification of the International Classifica-
tion of Diseases, 9th Revision (ICD-9-CM) procedural codes to
identify all patients in the NIS between 1998 and 2005 who met
the following criteria: (i) had a diagnosis of colorectal cancer
(ICD-9-CM codes: 153.0–154.8, 209.10–209.17, 230.3–230.6,
796.70–796.76, V10.05–V10.06); and (ii) who had undergone
partial colectomy (45.71–45.79), total colectomy (45.8), or
proctectomy (48.41–48.49, 48.5, 48.63–48.69) as identified by
ICD-9-CM procedural codes.
Predictor and Outcome Variables
ICD-9-CM diagnostic codes were used to identify patients with
uncomplicated diabetes (250.0-250.3, 250.8, 250.9) and diabe-
tes with complications (250.4–250.7), which included renal,
ophthalmological, neurological, or peripheral vascular manifes-
comorbidity conditions that are components of the validated
Charlson Index.15We modified the Charlson Index to exclude
diabetes as it was our main predictor of interest. In-hospital
mortality was the primary outcome. Secondary outcomes in-
cluded measures of post-operative complications and length of
stay. We used ICD-9-CM coding algorithms to identify postoper-
ative complications which were then further categorized as
wound infections, urinary, gastrointestinal (including acute liver
injury), pulmonary, cardiac, and intra-operative complications.
Data were analyzed using the Stata 10.0 SE software package
(Stata Corp LP, College Station, Texas). All analyses were stratified
by the presence of no diabetes, uncomplicated diabetes, and
complicated diabetes. These analyses took into account the
stratified two-stage cluster design using Stata’s SVY (survey data)
commands incorporating individual discharge-level weights. Two-
way χ2analyses and the Fischer’s exact test were performed to
compare categorical variables among different subgroups while
unpaired t-tests compared differences in means of continuous
variables. Multivariate logistic regression analysis was used to
determine the association between mortality and the presence of
uncomplicated and complicated diabetes, age, gender, primary
health insurance carrier, non-diabetic comorbidities, type of ad-
mission (elective vs. non-elective), median neighborhood income,
and teaching vs. non-teaching hospital status. We performed a
We used similar multivariate logistic and linear regression models
to determine the association between diabetes status and postop-
erative complications and length of stay, respectively.
The analysis of the Nationwide Inpatient Sample uses com-
pletely unidentified data with no risk of loss of confidentiality.
An initial expedited review by the Institutional Review Board of
the Johns Hopkins Medical Institutions deemed this study
exempt from further ethical review.
Baseline Demographic and Clinical
There were 218,534 patients who underwent surgery for
colorectal cancer between 1998 and 2005. Eighty five percent
of this population was non-diabetic, and the remainder (15%)
diabetic. The prevalence of uncomplicated and complicated
diabetes was 14% and 1%, respectively. Patients with diabetes
were older, more likely to be male and more likely to earn below
the national median than non-diabetics. The type of health
insurance did not differ among the four groups (Table 1).
Preoperative comorbidities are shown in Table 1. Patients
with diabetes, both complicated and uncomplicated, had a
higher rate of ischemic heart disease, heart failure, renal fail-
ure and cerebrovascular disease compared to non-diabetics.
All comorbid conditions occurred more frequently in the com-
plicated diabetes group compared to the uncomplicated dia-
Crude postoperative in-hospital mortality was 3.2% in those
without diabetes versus 2.5% in those with diabetes. After
adjustment for confounders, adjusted mortality was 23% lower
in those with diabetes compared to non-diabetics (aOR 0.77;
95% CI: 0.71–0.84), and this association did not differ for
elective vs. non–elective admissions.
When we stratified diabetic patients by the presence of
diabetic complications, the in-hospital mortality was 2.4% in
the uncomplicated diabetes group compared with 4.2% in the
complicated diabetes group. The adjusted mortality for un-
complicated diabetics was lower than that of non-diabetics
among both elective (aOR 0.71; 95% CI: 0.61-0.83) and non-
elective (aOR 0.75; 95% CI: 0.67–0.83) admissions. However,
among complicated diabetics, in-hospital mortality was two-
fold higher for elective admissions relative to non-diabetics
(aOR 2.14; 95% CI: 1.55–2.98). For non-elective admissions,
there was no difference in in-hospital mortality between non-
diabetics and those with complicated diabetes (aOR 0.91; 95%
The occurrence of any postoperative complication was lower
among diabetics compared to non-diabetics (28% vs. 31%, P<
0.0001). Rates of mechanical and infectious wounds, urinary,
pulmonary, gastrointestinal, and intraoperative complications
were lower in diabetics than in non-diabetics (Table 2). After
multivariate adjustment for confounders, diabetics had lower
odds of having at least one postoperative complication com-
Anand et al.: Diabetes and Colon Cancer Surgery
pared to non-diabetics (aOR 0.82; 95% CI: 0.79–0.84). In
secondary analysis that stratified complicated vs. uncompli-
cated diabetes, there was a difference in the rate of any
postoperative complication between uncomplicated diabetes
and non-diabetics (28% vs. 31%, respectively, P<0.0001),
while complicated diabetics complicated diabetes had
modestly higher rates of any postoperative complication
compared to non-diabetics (33.5% vs. 31.2%, P=0.02).
Additionally, uncomplicated diabetic subjects had a lower
rate of wound, urinary, pulmonary and gastrointestinal
complications compared to non-diabetics. In contrast,
patients with complicated diabetes had higher urinary,
pulmonary, gastrointestinal and cardiac complications than
non-diabetics (Table 2). After controlling for age, sex, elective
admission status, comorbidity, teaching hospital status, and
median neighbourhood income, those with uncomplicated
diabetes had lower adjusted post-operative complications
compared to non-diabetics (aOR 0.80; 95% CI: 0.78–0.83),
while subjects with complicated diabetes did not differ from
the latter with respect to any post-operative complication
(aOR 1.00; 95% CI: 0.91–1.16).
To address whether the mortality benefit associated with uncom-
plicated diabetes may have been partly due to unrecognized
comorbidity in the reference non-diabetic population, we con-
ducted a subgroup analysis of individuals younger than 50 years,
who are less likely to have comorbid conditions than older
patients. Patients who had uncomplicated diabetes but were
compared to non-diabetic counterparts after multivariate adjust-
complications, patients with uncomplicated diabetes who were
under 50 years had similar likelihood of postoperative complica-
tions as non-diabetics (aOR 1.02; 95% CI: 0.86–1.21).
Table 1. Baseline Demographics and Comorbidity
Mean age (yr) (SD)
Median income > National median
Charlson Index (SD)
Specific comorbid conditions
Ischemic heart disease
Congestive heart failure
Table 2. Postoperative Complications
Any postoperative complication
Anand et al.: Diabetes and Colon Cancer Surgery
To explore this possibility that diabetics may be more likely
referred to pre-operative internal medicine or anaesthesia clinics
and allowing for enhanced medical optimization and post-
operative follow-up, we conducted subgroup analysis with
uninsured patients who would be assumed to have less access
to peri-operative care. The reduced in-hospital mortality ob-
served in uncomplicated diabetics was absent among the
subgroup that was uninsured (aOR 0.99; 95% CI: 0.39–2.52).
Similarly, among uninsured patients, the likelihood of postoper-
ative complications was similar between non-diabetics and those
with uncomplicated diabetes (aOR 1.09; 95% CI: 0.86–1.39).
The average hospital length of stay was 9.9 days in both
diabetics and non-diabetics. However, after stratification for
diabetic complications, length of stay was modestly shorter in
uncomplicated diabetics compared to non-diabetics (9.7 vs.
9.9 days, P<0.01) while it was longer in complicated diabetics
than in either non-diabetics or uncomplicated diabetics (12.6
vs. 9.9 and 9.7 days, respectively, P<0.01). Compared to non-
diabetics, adjusted length of stay was 2% shorter for uncom-
plicated diabetics and 16% longer for complicated diabetics
after controlling for elective admission status, age, sex,
comorbidity, health insurance payer, teaching hospital status,
and median income.
Previous studies have shown that diabetics have a higher risk
of postoperative complications and mortality.6-9In our nation-
wide analysis of over two hundred thousand patients who
underwent surgery for colorectal cancer, patients with diabetes
had a lower mortality than those without diabetes. Ninety
three percent of the diabetic population had uncomplicated
diabetes. Patients with uncomplicated diabetes had up to a
25% reduction in adjusted mortality compared to those without
diabetes. Uncomplicated diabetics also had fewer postoperative
complications compared to non-diabetics. This protective effect
was not apparent in patients with complicated diabetes.
There are several factors that may contribute to our
unexpected findings. Firstly, we would expect that uncompli-
cated diabetics as a group were more likely to have adequate
glycemic control than those with complicated diabetes. Ran-
domized trials have demonstrated that in the critically ill,
aggressively trying to achieve normoglycemia does not appear
to confer any added benefit compared to less intensive
glycemic control, and may in fact cause greater harm15Thus,
mild hyperglycemia does not seem to lead to increased inpatient
complications. Possibly, uncomplicated diabetics are less likely
to sustain the more extreme levels of hyperglycemia necessary
to cause significant peri-operative morbidity.
However, this hypothesis does not account for why uncom-
plicated diabetics would have an improved outcome compared
to non-diabetics. We conjecture that patients with diabetes
may receive better peri-operative medical care than non-
diabetics. Diabetics may be more likely referred to pre-
operative internal medicine or anaesthesia clinics, allowing
for enhanced medical optimization and post-operative follow-
up. Our analysis of uninsured subjects demonstrated a loss of
the protective association between uncomplicated diabetes
and mortality and postoperative complications. Though this
subgroup analysis does not confirm our speculation, it does
confer credibility to the hypothesis. We should, however, note
that there is sparse evidence in the literature to support the
any clinically significant benefit of peri-operative medical
Another possible explanation for our paradoxical findings
may be that non-diabetics may have unrecognized and
therefore untreated diabetes or other comorbidities. One study
suggests that the true prevalence of hyperglycemia in hospital-
ized patients may be underestimated by as much as 40%.17-19
Patients with newly diagnosed hyperglycemia have a higher
mortality rate, longer length of stay, and increased rate of ICU
transfers.20If the protective benefit of uncomplicated diabetes
does arise from unrecognized comorbidity in the reference non-
diabetic population, then we would expect to see less of a
prominent effect among younger subjects who are less likely to
uncomplicated diabetics younger than 50 years relative to their
non-diabetic counterparts supports but does not confirm the
We also found that the increased mortality rate associated
with complicated diabetes relative to non-diabetics seemed to be
limited to elective admissions only. In non-elective admissions,
the urgent nature of the surgery may overshadow any benefit of
being non-diabetic. For example, in Goldman et al’s and Detsky
et al.’s preoperative cardiac risk assessment tools, diabetes does
not contribute to predicting increased risk in the face of more
significant issues such as emergent-nature surgery or surgical
site.21-23In Lee et al’s Revised Cardiac Risk Index, diabetes
needs to be sufficiently advanced to require insulin before being
considered an independent predictor on par with having a
“high-risk” surgical site.24
Another alternative explanation to our findings may be that
uncomplicated diabetic subjects may be self-selected to be
healthier than non-diabetics if they had relatively fewer comor-
bidities which may have resulted in fewer event rates among the
Index was higher and specific non-diabetes comorbidities such
as cardiovascular and renal disease were more common in the
uncomplicated diabetic group than non-diabetics.
Our current study has several limitations inherent to admin-
istrative data analyses. The NIS data set does not contain
personal identifiers that would allow linkage to medical records.
We are unable to classify the severity of diabetes in patients and
cannot assess what medications they were prescribed. Also,
mortality was classified as all cause mortality and we are unable
to break this down to specific causes. Furthermore, this is a
cross-sectional study, and we are unable to longitudinally follow
patients after discharge to assess long-term mortality and
morbidity associated with diabetes post operatively.
One of the main limitations is the use of the ICD-9 codes to
classify patients as having diabetes with and without complica-
tions. Studies assessing the accuracy and validity of ICD-9
codes for diabetes complications, quote up to 95% accuracy
rates.25,26However, because the NIS contains de-identified
subjects, it is not possible to validate diagnostic codes with the
medical record for this dataset. Because diabetic complications
were identified using administrative data, it is possible that
Anand et al.: Diabetes and Colon Cancer Surgery
complicated diabetes may have been misclassified as uncom-
plicated diabetes. We would expect this misclassification to
overestimate mortality and postoperative complications in the
uncomplicated diabetes group and therefore biasing the odds
ratio for mortality and complications relative to the non-diabetic
group toward the null. Nonetheless, acknowledging these
potential inaccuracies, we stratify analyses by the presence of
complications only in secondary analyses.
The main strength of this study is the large sample size which
offsets the possibility of a type I error induced by the multiple
subgroup comparisons performed in this study. Additionally,
this data set is a population-based representation of all
hospitalized patients in the United States and reflects all types
of hospital settings, insurance payers, and geographic regions,
enhancing the generalizability of our data and minimizing
referral bias associated with single-center studies from tertiary
Overall, our study illustrates that having uncomplicated
diabetes unexpectedly appears to reduce post-operative compli-
cations in colorectal cancer surgical patients. It is important to
note that, due to limitations of the administrative claims data,
this study is intended to explore and generate new hypotheses
in the field of peri-operative diabetic care. However, it would be
premature to implement changes in the preoperative and
inpatient management of diabetes based on these data alone.
Thus, future prospective studies are necessary to replicate these
results and establish these findings in other surgical groups.
Additionally, primary studies are warranted to investigate the
potential mechanisms underlying this seemingly paradoxical
association. Such research may help identify practices that
translate into meaningful reductions in surgical risk.
Acknowledgements: Dr. Nguyen had full access to all of the data
in the study and takes responsibility for the integrity of the data and
the accuracy of the data analysis.
Conflict of Interest: Dr. Nguyen has served as a consultant for
Schering Plough, Canada and Abbott Pharmaceutical, neither of
whom had any involvement in this study. The other authors report
no conflicts of interest.
Corresponding Author: Geoffrey C. Nguyen, MD, PhD, FRCPC;
Mount Sinai Hospital Division of Gastroenterology, University of
Toronto Faculty of Medicine, 600 University Ave., Ste. 433, Toronto,
ON M5G 1X5, Canada (e-mail: firstname.lastname@example.org).
1. National Estimates of Diabetes, Center for Disease Control and Preven-
tion. National Diabetes fact sheet:general information and national
estimates on diabetes in the United States, 2005. Available at: http://
2. Root HF. Preoperative medical care of the diabetic patient. Postgrad
3. Clement S, Braithwaite SS, Magee MF, Ahmann A, Smith EP, Schafer
RG, et al. Management of diabetes and hyperglycemia in hospitals.
Diabetes Care. 2004;27(2):553–91.
4. Haffner SM, Lehto S, Ronnemaa T, Pyorala K, Laakso M. Mortality
from coronary heart disease in subjects with type 2 diabetes and in
nondiabetic subjects with and without prior myocardial infarction. N
Engl J Med. 1998;339(4):229–34.
5. Coursin DB, Connery LE, Ketzler JT. Perioperative diabetic and
hyperglycemic management issues. Crit Care Med. 2004;32(4 Suppl):
6. Gil-Bona J, Sabate A, Pi A, Adroer R, Jaurrieta E. Mortality risk factors
in surgical patients in a tertiary hospital: a study of patient records in the
period 2004-2006. Cir Esp. 2009;85(4):229–37.
7. Lange CP, Ploeg AJ, Lardenoye JW, Breslau PJ. Patient-and proce-
dure-specific risk factors for postoperative complications in peripheral
vascular surgery. Qual Saf Health Care. 2009;18(2):131–6.
8. Chen SY, Zhou YB, Wang H, Li SK, Mao WZ, Wang HB. Risk factors of
intra-abdominal infection following gastrectomy in gastric cancer
patients. Zhonghua Wei Chang Wai Ke Za Zhi. 2009;12(2):137–40.
9. Malone DL, Genuit T, Tracy JK, Gannon C, Napolitano LM. Surgical
site infections: reanalysis of risk factors. J Surg Res. 2002;103(1):89–
10. Joshi N, Caputo GM, Weitekamp MR, Karchmer AW. Infections in
patients with diabetes mellitus. N Engl J Med. 1999;341(25):1906–
11. Mowat A, Baum J. Chemotaxis of polymorphonuclear leukocytes from
patients with diabetes mellitus. N Engl J Med. 1971;284(12):621–7.
12. Bagdade JD, Stewart M, Walters E. Impaired granulocyte adherence. A
reversible defect in host defense in patients with poorly controlled
diabetes. Diabetes. 1978;27(6):677–81.
13. Kersten JR, Schmeling TJ, Orth KG, Pagel PS, Warltier DC. Acute
hyperglycemia abolishes ischemic preconditioning in vivo. Am J Physiol.
1998;275(2 Pt 2):H721–5.
14. Whalen D, Houchens R, ELixhauser A. 2002 HCUP nationwide
Inpatient Sample (NIS) comparison report #2005-03 ed. Rockville, MD:
US. Agency for Healthcare Research and Quality 2005;1-89.
15. Griesdale DE, de Souza RJ, van Dam RM, Heyland DK, Cook DJ,
Malhotra A, et al. Intensive insulin therapy and mortality among
critically ill patients: a meta-analysis including NICE-SUGAR study
data. CMAJ. 2009;180(8):821–7.
16. Auerbach AD, Rasic MA, Sehgal N, Ide B, Stone B, Maselli J.
Opportunity missed: medical consultation, resource use, and quality of
care of patients undergoing major surgery. Arch Intern Med. 2007;167
17. Jencks SF. Accuracy in recorded diagnoses. JAMA. 1992;267(16):2238–
18. Levetan CS, Passaro M, Jablonski K, Kass M, Ratner RE. Unrecog-
nized diabetes among hospitalized patients. Diabetes Care. 1998;21
19. Umpierrez GE, Isaacs SD, Bazargan N, You X, Thaler LM, Kitabchi
AE. Hyperglycemia: an independent marker of in-hospital mortality in
patients with undiagnosed diabetes. J Clin Endocrinol Metab. 2002;87
20. van den Berghe G, Wouters P, Weekers F, Verwaest C, Bruyninckx F,
Schetz M, et al. Intensive insulin therapy in the critically ill patients. N
Engl J Med. 2001;345(19):1359–67.
21. Detsky AS, Abrams HB, Forbath N, Scott JG, Hilliard JR. Cardiac
assessment for patients undergoing noncardiac surgery. A multifactorial
clinical risk index. Arch Intern Med. 1986;146(11):2131–4.
22. Detsky AS, Abrams HB, McLaughlin JR, Drucker DJ, Sasson Z,
Johnston N, et al. Predicting cardiac complications in patients under-
going non-cardiac surgery. J Gen Intern Med. 1986;1(4):211–9.
23. Goldman L, Caldera DL, Nussbaum SR, Southwick FS, Krogstad D,
Murray B, et al. Multifactorial index of cardiac risk in noncardiac
surgical procedures. N Engl J Med. 1977;297(16):845–50.
24. Lee TH, Marcantonio ER, Mangione CM, Thomas EJ, Polanczyk CA,
Cook EF, et al. Derivation and prospective validation of a simple index
for prediction of cardiac risk of major noncardiac surgery. Circulation.
25. Kern EF, Maney M, Miller DR, Tseng CL, Tiwari A, et al. Failure of
ICD-9-CM codes to identify patients with comorbid chronic kidney
disease in diabetes. Healthy Serv Res. 2006;41(2):564–80.
26. Newton KM, Wagner EH, Ramsey SD, McCulloch D, Evans R, et al.
The use of automated data to identify complications and comorbidities of
diabetes: a validation study. J Clin Epidemiol. 1999;52(3):199–207.
Anand et al.: Diabetes and Colon Cancer Surgery