Table 3

Independent predictors of late relapse of diabetes after bariatric surgery in multivariable analyses

Predictors	Main model^*			Second model^*
Predictors	OR	95% CI	P	OR	95% CI	P
Preoperative number of diabetes medications	2.04	(1.54, 2.71)	<0.001	2.13	(1.58, 2.87)	<0.001
Preoperative duration of diabetes (years)	1.08	(1.02, 1.14)	0.005	1.10	(1.03, 1.16)	0.002
Type of surgery (SG vs. RYGB)	2.21	(1.22, 4.01)	0.009	—	—	—
Weight loss at 1 year (%)	—	—	—	0.92	(0.89, 0.95)	<0.001
Late weight regain^#	—	—	—	3.69	(2.18, 6.24)	<0.001
	Main model^*			Second model^*
Multiple logistic regression equation^{^}	L = −2.4917 + (0.0789 × preoperative duration of diabetes) + (0.714 × preoperative number of diabetes medications) + (0.7929 × [1, for SG, or 0, for RYGB])			L = −0.729 + (0.0919 × preoperative duration of diabetes) + (0.7565 × preoperative number of diabetes medications) − (0.0877 × short-term weight loss) + (1.3059 × [1, for late weight regain, or 0, for no weight regain])

↵* Main model to be used before surgery to predict late diabetes relapse based on preoperative characteristics; second model to be used 1 year after bariatric surgery to predict late relapse based on preoperative characteristics and weight changes.
↵# Categorical variable defined as >5% weight regain from the weight recorded 1 year after bariatric surgery.
↵^ Estimated probability of late relapse of diabetes after initial remission for a given patient (100%) = EXP [L]/(1 + EXP [L]). The notation EXP is equivalent to ex, where “e” is the base of the ln (2.718).