Review, Performance Comparison, and Validation of Models Predicting Type 2 Diabetes Remission After Bariatric Surgery in a Western European Population

Purpose: The majority of patients with type 2 diabetes (T2DM) achieve remission after bariatric surgery. Several models are available to preoperatively predict T2DM remission. This study compares the performance of these models in a Western population one year after surgery and explores their predictive value in comparison to a model specifically designed for our study population.

Materials and methods: Prediction models were retrieved using a literature search. Patients were retrospectively selected from a database of the Antwerp University Hospital. Performance of the models was assessed by determining the area under the receiver operating characteristic curve (AUROC), the accuracy, and the goodness of fit, and by comparing them to a custom-made logistic model.

Results: The probability of T2DM remission was calculated using 11 predictive scoring models and 8 regression models in a cohort of 250 patients. Complete T2DM remission occurred in 64.0% of patients. The IMS score (AUROC = 0.912; accuracy = 83.6%), DiaBetter score (0.907; 82.0%), and Ad-DiaRem score (0.903; 82.8%) best predicted T2DM remission and closely approached the performance of the custom-constructed model (0.917; 84.0%). The model by Ioffe et al. (0.630; 69.2%), Umemura et al. (0.692; 71.4%), and the ABCD score (0.757; 72.8%) were the least accurate.

Conclusion: Most T2DM remission models reliably predicted one-year T2DM remission, with limited inter-model differences. The accuracy of most models approached that of the custom-constructed model, indicating a high predictive capability and performance in our patient cohort. To date, most models are only validated to estimate T2DM remission one year after surgery and they do not predict long-term remission.