Statistical models to predict type 2 diabetes remission after bariatric surgery

J Diabetes. 2014 Sep;6(5):472-7. doi: 10.1111/1753-0407.12127. Epub 2014 Feb 26.


Background: Type 2 diabetes (T2D) remission may be achieved after bariatric surgery (BS), but rates vary according to patients' baseline characteristics. The present study evaluates the relevance of several preoperative factors and develops statistical models to predict T2D remission 1 year after BS.

Methods: We retrospectively studied 141 patients (57.4% women), with a preoperative diagnosis of T2D, who underwent BS in a single center (2006-2011). Anthropometric and glucose metabolism parameters before surgery and at 1-year follow-up were recorded. Remission of T2D was defined according to consensus criteria: HbA1c <6%, fasting glucose (FG) <100 mg/dL, absence of pharmacologic treatment. The influence of several preoperative factors was explored and different statistical models to predict T2D remission were elaborated using logistic regression analysis.

Results: Three preoperative characteristics considered individually were identified as the most powerful predictors of T2D remission: C-peptide (R2 = 0.249; odds ratio [OR] 1.652, 95% confidence interval [CI] 1.181-2.309; P = 0.003), T2D duration (R2 = 0.197; OR 0.869, 95% CI 0.808-0.935; P < 0.001), and previous insulin therapy (R2 = 0.165; OR 4.670, 95% CI 2.257-9.665; P < 0.001). High C-peptide levels, a shorter duration of T2D, and the absence of insulin therapy favored remission. Different multivariate logistic regression models were designed. When considering sex, T2D duration, and insulin treatment, remission was correctly predicted in 72.4% of cases. The model that included age, FG and C-peptide levels resulted in 83.7% correct classifications. When sex, FG, C-peptide, insulin treatment, and percentage weight loss were considered, correct classification of T2D remission was achieved in 95.9% of cases.

Conclusion: Preoperative characteristics determine T2D remission rates after BS to different extents. The use of statistical models may help clinicians reliably predict T2D remission rates after BS.

Keywords: bariatric surgery; logistic regression analysis; metabolic surgery; prediction models; type 2 diabetes mellitus; 关键词:减肥手术,对数回归分析,代谢手术,预测模型,2型糖尿病.

Publication types

  • Research Support, Non-U.S. Gov't

MeSH terms

  • Adult
  • Aged
  • Bariatric Surgery*
  • Biomarkers / blood
  • Blood Glucose / metabolism
  • Diabetes Mellitus, Type 2 / blood
  • Diabetes Mellitus, Type 2 / diagnosis
  • Diabetes Mellitus, Type 2 / etiology
  • Diabetes Mellitus, Type 2 / surgery*
  • Female
  • Glycated Hemoglobin A / metabolism
  • Humans
  • Logistic Models
  • Male
  • Middle Aged
  • Models, Biological*
  • Models, Statistical*
  • Multivariate Analysis
  • Obesity / complications
  • Obesity / diagnosis
  • Obesity / surgery*
  • Odds Ratio
  • Remission Induction
  • Retrospective Studies
  • Risk Factors
  • Time Factors
  • Treatment Outcome
  • Weight Loss
  • Young Adult


  • Biomarkers
  • Blood Glucose
  • Glycated Hemoglobin A
  • hemoglobin A1c protein, human