Table 7.

Mediation effect of insulin on the association between BMI and intermediate hyperglycemia adhering to the principles of Baron and Kenny and by using the standard equation modeling analysis, adjusted for age and sex.

Baron and Kenny mediation analysis
	Model 1			Model 2			Model 3
	Intermediate hyperglycemia (HbA1c > 5.6– < 6.4%)			Serum insulin			Intermediate hyperglycemia (HbA1c > 5.6– < 6.4%)
	OR	P -value	95% CI	OR	P -value	95% CI	OR	P -value	95% CI
Obese vs. non-obese	3.39	0.008*	1.30,8.30	5.40	<0.001*	3.20,9.09	2.20	0.09	0.80,5.80
Serum insulin	··	··	··	··	··	··	3.60	0.03*	1.10,11.70
Standard equation modeling analysis
	Estimate (bootstrap SE)	95% CI	P -value
Direct effect	−0.00045 (0.0006)	−0.0016, 0.00077	0.47
Indirect effect	0.0008 (0.00035)	0.000094, 0.0014	0.02*
Total effect	0.00033 (0.0004)	−0.00047, 0.0011	0.41
Ratio of the mediation effect (indirect to total effect)	0.0008/0.00033 = 2.4
Ratio of indirect to direct effect	0.0008/−0.00045 = 1.7
Ratio of total to direct effect	0.00033/-0.00045 = 0.73

Serum insulin were stratified by 50th percentile cutoff (median).

The standard equation modeling calculated the confidence intervals and standard errors were from 5,000 random samplings with replacement by the bootstrapping method.

OR, Odds Ratio, CI, 95% Confidence Interval, SE, Standard Error.

^*P-value <0.05. ··, N/A.