Performance and Improvement of the DiaRem Score in Diabetes Remission Prediction - A Study with Diverse Procedure Types

Abstract Objective Test the validity of the DiaRem score as a predictor of T2D remission at one-year post-surgery in patients who underwent LABG, LSG, RYGB, and BPD/DS, as well as to develop and test a novel model that uses DiaRem to predict T2D remission by procedure type (LABG, LSG, RYGB, and BPD/DS). Background The DiaRem score has proven to be a reliable predictor tool for diabetes remission in RYGB, as well as LAGB and LSG. To our knowledge there are no studies that analyze DiaRem in patients undergoing BPD/DS. Methods A retrospective review of institutional records identified patients who underwent primary bariatric procedures (LAGB, LSG, RYGB, and BPD/DS) between 01/01/2000-04/10/2017, had a diagnosis of diabetes and had complete preoperative and one-year postoperative follow-up information. A univariable logistic regression model was fit to assess the association between DiaRem score and diabetes remission. A multivariable logistic regression model was created, including procedure type and other preoperative characteristics. The area under the Receiver operating curve (AUCROC) curve was calculated to analyze the performance of both models for the entire cohort as well as a BPD/DS only subgroup. Results A total study cohort of 602 was obtained. The majority of patients underwent RYGB (456; 75.7%), followed by LSG (84; 14.0%), BPD/DS (35; 5.8%) and LAGB (27; 4.5%). The multivariable regression model with RYGB as the reference procedure, showed that BPD/DS results in higher odds of T2D remission at one-year post-surgery (adjusted odds ratio (OR) 3.29, 95% CI 1.27, 8.51), while LSG (adjusted OR 0.52, 95% CI 0.29, 0.93) and LAGB resulted in lower odds (adjusted OR 0.23, 95% CI 0.09, 0.60). The univariable DiaRem model and the novel model were determined to be moderately strong in classifying diabetes remission in the entire cohort (AUROC: 0.79, 95% CI: 0.75, 0.83 and 0.82, 95% CI: 0.79, 0.85, respectively) as well as in the BPD/DS sub-group (AUROC: 0.85, 95% CI: 0.70, 0.99 and 0.84, 95% CI 0.69, 0.99, respectively). Conclusion Our study shows that the DiaRem score is a reliable tool to predict diabetes remission, amongst a wide variety of different procedures as well as specifically those receiving BPD/DS. Our novel model, which takes into account procedure type, not only shows that BPD/DS patients have the highest odds of diabetes remission than other procedures, but also that this model performs significantly better at predicting diabetes remission than DiaRem alone.