Table 12. Univariate and multivariate meta-regression analysis of eligible longitudinal studies for odds ratio of type 2 diabetes mellitus for 1 increment in risk alleles in relation to diabetes-associated single nucleotide polymorphism (SNP)^a according to characteristics of study design^a.

Variable	Univariatec N = 21			Multivariate 1 N = 20b			Multivariate 2 N = 20b
	coefficient	SE	P	coefficient	SE	P	coefficient	SE	P
Number of SNPs ≥10 (Yes/No)	−0.206	0.066	0.005	−0.182	0.083	0.049	−0.186	0.082	0.04
Western country (Yes/No)	−0.067	0.078	0.40	−0.035	0.091	0.7	d
White-dominant (Yes/No)	−0.003	0.039	0.94	d			0.002	0.04	0.97
Mean age ≥50 years (Yes/No)b	0.002	0.034	0.95	0.01	0.04	0.82	0.017	0.037	0.67
Women-dominant	−0.045	0.033	0.19	−0.052	0.042	0.24	−0.049	0.042	0.27
BMI ≥25 kg/m2 (Yes/No)	−0.025	0.036	0.50	0.01	0.048	0.84	0.000	0.043	1.00
Adjustment for BMI	0.002	0.038	0.96	−0.011	0.04	0.79	−0.007	0.042	0.88
Using OGTT to diagnose DM	0.008	0.032	0.80	−0.006	0.035	0.87	−0.004	0.035	0.91
R-squared					49.1%			47.5%
F-test					F(7,12) = 1.39	P = 0.29		F(7,12) = 1.37	P = 0.30

BMI, body mass index; DM, diabetes mellitus; OGTT, oral glucose tolerance test; SE, standard error.

^aLogarithm of odds ratio for diabetes mellitus was a dependent variable, and each study characteristic was entered as an explanatory variable.

^bN = 20 because one study¹⁷ did not present data on mean age or BMI.

^cFundamentally, the results were consistent with the stratified analysis.

^dWestern country and Asian ethnicity was not entered simultaneously because of collinearlity.