Oral Presentation ANZOS Annual Scientific Meeting 2021

Predicting diabetes resolution after metabolic - external validation of predictive scores. (#77)

Izabela Karpińska 1 , Joanna Choma 1 , Michał Wysocki 2 , Alicja Dudek 1 , Piotr Małczak 1 , Magdalena Szopa 3 , Michał Pędziwiatr 1 , Piotr Major 1
  1. 2nd Department of General Surgery , Jagiellonian University Medical College, Kraków, Poland
  2. Department of General Surgery and Surgical Oncology, Ludwik Rydygier Memorial Hospital, Kraków, Poland
  3. Department of Metabolic Diseases, Jagiellonian University Medical College, Kraków, Poland

Background:

Bariatric surgery is the most efficient treatment of obesity and type 2 diabetes mellitus (T2DM). Despite detailed qualification, not every patient achieve T2DM remission after intervention. Recently, new scores: Individualised Metabolic Surgery (IMS), DiaRem, Ad-DiaRem, DiaBetter and Robert’s score have been developed to predict diabetes remission after bariatric surgery.

Objectives:

The aim of the study was to validate and compare the performance of different models as the predictors of diabetes remission 1 year after surgical treatment.

Methods:

The retrospective analysis included patients with T2DM who underwent Roux-en-Y gastric bypass (RYGB) or sleeve gastrectomy (SG) and completed 1-year follow-up. Each score relationship with diabetes remission was assessed using logistic regression. Discrimination was evaluated by area under the receiver operating characteristic curve (AUROC) whereas calibration by Hosmer–Lemeshow test.

Results:

Out of 252 patients 150 (59.5%) were women whereas 102 (40.5%) were men with median age 48 years. 46.83% of patients underwent SG whereas 53.17% had RYGB. The T2DM remission rate reached 90.5%. Median of preoperative A1c was 6.75% and preoperative BMI was 45.39 kg/m2, both decreased to 5.8% and 33.09 kg/m2 respectively after 1 year. %EWL amounted to 53.4%.
All IMS, DiaRem, Ad-DiaRem, DiaBetter and Robert’s scores were predictive of diabetes remission in a logistic regression analysis (OR 0.97, p<0.0001; OR 0.83; p<0.0001; OR 0.80, p=0.0001; OR 0.51, p<0.0001; OR  1.93, p=0.0031, respectively). The majority of models showed acceptable discrimination power. Robert’s score had poor discrimination with AUROC=0.67 (p<0.0001) whereas DiaBetter presented excellent discrimination with AUROC 0.81 (p<0.0001). Most of scores except IMS did not lose their goodness of fit.

Conclusion:

All developed scores can be used in preoperative assessment of patients before bariatric surgery. Since DiaBetter score seem to be more accurate than others scores in predicting metabolic outcomes it is more likely to be implemented into day-to-day practice.