Defining and Using Preoperative Predictors of Diabetic Remission Following Bariatric Surgery

Background: Diabetes remission is defined as the return of glycemic control in the absence of medication or insulin use after bariatric surgery. We sought to identify and assess the clinical utility of a predictive model for remission of type 2 diabetes mellitus in a population seeking bariatric surgery. Method: A retrospective cohort design was applied to presurgical data on patients referred for Roux‐en‐Y gastric bypass (RYGB) or vertical sleeve gastrectomy (VSG). The model developed from logistic regression was compared with a published model through receiver operating characteristic analyses. Results: At 12 months postoperatively, 59.7% of the cohort was remitted, with no differences between RYGB and VSG. Logistic regression analyses yielded a model in which 4 preoperative variables reliably predicted remission. A Hosmer‐Lemeshow goodness‐of‐fit test result of 0.204 indicated good fit of the developed prediction model to our outcome data. The predictive accuracy of this prediction model was compared with a published model, and an associated variation with diabetes years was substituted for age in our patient population. Our model was the most accurate. Conclusions: Using these predictors, healthcare providers may be able to better counsel patients who are living with diabetes and considering bariatric surgery on the likelihood of achieving remission from the intervention. This refined prediction model requires further testing in a larger sample to evaluate its external validity.