Introduction Hyperglycemia causes increased postoperative morbidity and mortality. Heterogeneous patient populations and differing definitions of hyperglycemia complicate broad applicability of past studies. We examined the effect of hyperglycemia on postoperative infections in patients with colorectal cancer resections.

Method A retrospective review of an IRB-approved prospectively maintained database was performed for all patients with a colorectal resection and anastomosis for malignancy at a single centre from 2008–2014. All analyses, confounding variables and outcomes were proposed a priori. Primary outcome was evidence of a surgical site infection (SSI) defined by the Centres for Disease Control criteria and analysed as a cumulative variable of superficial, deep and organ space infections. Main independent variable was post-operative day 1 (POD1) glucose level with hyperglycemia defined a priorias >180 mg/dL. Univariate analyses were performed using a Chi-squared test for categorical variables. Multivariable logistic regression was performed to adjust for confounding and identify the role of proposed variables on predicting the outcome of SSI.

Results 349 patients underwent colorectal resections for malignancy (mean age: 65.3 (+/-13.7) years). All patients received preoperative antibiotics. 20% had a stage 3 or 4 malignancy. 26 patients had POD1 hyperglycemia (mean 128.6 mg/dL +/- 33.6) with 70 (20%) identified with a SSI. Mean operative time was 190 ± 111 min. On univariate analysis, POD1 glucose (p = 0.312) nor POD1 hyperglycemia (p = 0.054) as continuous variables were not associated with the presence of SSI (p = 0.312). Active smokers and those who had an open resection had higher rates of SSI (p = 0.015, p < 0.001, respectively). A higher cancer stage (3 or 4) was significantly associated with SSI (p = 0.03). Multivariate logistic regression adjusting for known predictors (smoking status, surgical approach, diagnosis of diabetes, cancer stage and POD1 hyperglycemia) was performed with the dependent variable evidence of SSI (Hosmer-Lemeshow p = 0.45); operative time was excluded for concerns of co-linearity with surgical technique. Smoking (OR 4.6, 1.8–11.7; p = 0.002) and hyperglycemia (3.3, 1.2–9.1; p = 0.019) significant predictors of SSI while laparoscopy (n = 209; 60%) was protective against SSI (OR 0.21, 0.12–0.39; p < 0.001).

Conclusion Our study demonstrates that POD1 hyperglycemia is associated with increased rates of SSI when controlling for confounders. This reiterates the need for stringent postoperative glycemic control regardless of history of diabetes to further improve surgical outcomes.

Disclosure of interest None Declared.