External validation of predictive scores for diabetes remission after metabolic surgery

Purpose: Bariatric surgery has proven to be the most efficient treatment for obesity and type 2 diabetes mellitus (T2DM). Despite detailed qualification, desirable outcome after an intervention is not achieved by every patient. Various risk prediction models of diabetes remission after metabolic surgery have been established to facilitate the decision-making process. The purpose of the study is to validate the performance of available risk prediction scores for diabetes remission a year after surgical treatment and to determine the optimal model. Methods: A retrospective analysis comprised 252 patients who underwent Roux-en-Y gastric bypass (RYGB) or sleeve gastrectomy (SG) between 2009 and 2017 and completed 1-year follow-up. The literature review revealed 5 models, which were subsequently explored in our study. Each score relationship with diabetes remission was assessed using logistic regression. Discrimination was evaluated by area under the receiver operating characteristic (AUROC) curve, whereas calibration by the Hosmer–Lemeshow test and predicted versus observed remission ratio. Results: One year after surgery, 68.7% partial and 21.8% complete diabetes remission and 53.4% excessive weight loss were observed. DiaBetter demonstrated the best predictive performance (AUROC 0.81; 95% confidence interval (CI) 0.71–0.90; p-value > 0.05 in the Hosmer–Lemeshow test; predicted-to-observed ratio 1.09). The majority of models showed acceptable discrimination power. In calibration, only the DiaBetter score did not lose goodness-of-fit in all analyzed groups. Conclusion: The DiaBetter score seems to be the most appropriate tool to predict diabetes remission after metabolic surgery since it presents adequate accuracy and is convenient to use in clinical practice. There are no accurate models to predict T2DM remission in a patient with advanced diabetes.