Table 3

Prediction of diabetes remission: logistic models and ROC analyses

OverallMedical*Surgical overall C (n = 415)GOGD
A (n = 727)B (n = 312)Without T. of S. C1With T. of S.** C2D (n = 318)E (n = 97)
VariableβP valueβP valueβP valueβP valueβP valueβP value
Logistic regression after stepwise elimination
 Intercept3.71<0.0011.8950.1030.0890.9422.83<0.0013.27<0.0012.838<0.001
 Age−0.0420.005
 Fasting glycemia (mmol/L)−0.1180.004−0.3060.051−0.1450.004−0.1410.011−0.1860.006
 Diabetes duration−0.192<0.001−0.9100.013−0.210<0.001−0.220<0.001−0.1970.004−0.273<0.001
 BMI0.0590.036
 T. of S.2.180<0.001
 Type of therapy (level 1)−0.819<0.001−2.2470.042−1.22<0.001−1.364<0.001
 Type of therapy (level 2)−0.2820.4730.8320.582−1.130.054−1.7830.03
Univariate ROC analysis
FGH
Cutoff (Sens., Spec.)AUC (95% CI)Cutoff (Sens., Spec.)AUC (95% CI)Cutoff (Sens., Spec.)AUC (95% CI)
ROC analysis
 BMI≥44 (32%, 83%)0.58 (0.52–0.64)
 Fasting glycemia (mmol/L)≤7.65 (59.1%, 73.1%)0.69 (0.63–0.75)≤7.6 (62.6%, 72.6%)0.71 (0.65–0.77)
 Diabetes duration≤1.5 (65.9%, 73.8%)0.74 (0.68–0.79)≤1.5 (77.9%, 70.6%)0.78 (0.72–0.84)≤7.5 (92.4%, 55%)0.75 (0.61–0.90)
  • Sens., sensitivity; Spec., specificity; T. of S., type of surgery (GO vs. GD).

  • *A further model including percent weight change was tested; results were β = −0.41 (P = 0.009) for basal glycemia, β = −0.53 (P < 0.001) for diabetes duration, and β = −0.09 (P = 0.005).

  • **When type of surgery was included in the model, BMI was no longer significant.