Preoperative glucose in surgical oncology patient is not associated with postoperative outcomes after adjustment for frailty

BACKGROUND: Observational studies have shown associations between even small elevations in preoperative glucose and poorer outcomes, including increased length of stay (LOS) and higher mortality. This has led to calls for aggressive glycemic control in the preoperative period, including delay of treatment until glucose is reduced. However, it is not known whether there is a direct causal effect of blood glucose or whether adverse outcomes result from overall poorer health in patients with higher glucose. METHODS: Analysis was performed using a retrospective database of patients aged 65 and older who underwent cancer surgery. The last measured preoperative glucose was the exposure variable. The primary outcome was extended LOS (>4 days). Secondary outcomes included mortality, acute kidney injury (AKI), major postoperative complications during the admission period, and readmission within 30 days. The primary analysis was a logistic regression with prespecified covariates: age, sex, surgical service, and the Memorial Sloan Kettering-Frailty Index. In an exploratory analysis, lasso regression was used to select covariates from a list of 4160 candidate variables. RESULTS: This study included 3796 patients with a median preoperative glucose of 104 mg/dL (interquartile range: 93-125). Higher preoperative glucose was univariately associated with increased odds of LOS > 4 days (odds ratio [OR]: 1.45, 95% confidence interval [CI]: 1.22-1.73), with similar results for AKI, readmission, and mortality. Adjustment for confounders eliminated these associations for LOS (OR: 0.97 [95% CI: 0.80-1.18]) and attenuated all other glucose-outcome associations. Lasso regression gave comparable results to the primary analysis. Using the upper bound of the respective 95% confidence interval, we estimated that, at best, successful reduction of elevated preoperative glucose would reduce the risk of LOS > 4 days, 30-day major complication, and 30-day mortality by 4%, 0.5%, and 1.3%, respectively. CONCLUSIONS: Poor outcomes following cancer surgery in older adults with elevated glucose are most likely related to poorer overall health in these patients rather than a direct causal effect of glucose. Aggressive glycemic management in the preoperative period has very limited potential benefits and is therefore unwarranted.