Accuracy of prediction models for long-term type 2 diabetes remission after gastric bypass

Abstract

Aim: To evaluate the accuracy of DiaBetter, DiaRem, Ad-DiaRem and 5y-Ad-DiaRem scores' at predicting T2D remission 10 or more years after surgery.

Methods: Patients with obesity and T2D (n = 126) submitted to RYGB with 10 or more years of follow-up. It was a unicentric trial. Pre-operative anthropometric and clinical data was retrieved to calculate DiaRem, DiaBetter, Ad-DiaRem and 5y-Ad-DiaRem scores, while a hospital visit was conducted to assess current diabetes status. The area under the receiver operating characteristic (AUROC) curve was calculated as estimate of the scores' accuracy to predict long-term T2D remission.

Results: Among the entire cohort (n = 126), 70 subjects (55.6%) achieved and maintained T2D remission 10 or more years after RYGB. The 5y-Ad-DiaRem score was the one that depicted the highest discriminative power (AUROC = 0.838) to predict long-term T2D remission when compared to DiaBetter (AUROC = 0.735), DiaRem (AUROC = 0.721) and Ad-DiaRem (AUROC = 0.720).

Conclusion: The score with highest accuracy to predict long-term T2D remission after RYGB surgery was the 5y-Ad-DiaRem. Yet, the available scores accuracy to predict T2D remission in the long term is still suboptimal, highlighting the unmet need for a better scoring system.

Keywords: Prediction models; Roux-en-Y gastric bypass; Type 2 diabetes; Type 2 diabetes remission.