Table 2. Subgroup analyses of studies included in a meta-analysis of GSTM1 null genotype and CHD risk.

Null versus present	Studies	Cases/controls	OR (95% CI)	Heterogeneity		Model	P for Begg's test	P for Egger's test
				I2	PH
Total studies	37	16684/36510	1.38 (1.15, 1.69)	93.8%	<0.001	Random	0.097	0.499
Total studies (adjustment for heterogeneity1)	22	5341/8322	1.17 (1.05, 1.31)	40.2%	0.027	Random	0.236	0.424
Smoker	14	2249/1300	1.64 (1.12, 2.40)	82.2%	<0.001	Random	0.189	0.387
Non-smoker	11	1962/2195	1.26 (0.70, 2.27)	94.2%	<0.001	Random	0.755	0.043
Ethnicity
Caucasians	23	12268/31531	1.34 (1.04, 1.72)	95.7%	<0.001	Random	0.045	0.605
Caucasians (adjustment for heterogeneity2)	13	2980/4021	1.18 (1.07, 1.31)	18.1%	0.261	Fixed	1.00	0.763
Asians	13	3547/3406	1.40 (1.11, 1.77)	78.6%	<0.001	Random	0.583	0.903
Asians (adjustment for heterogeneity3)	9	2001/2094	1.60 (1.32, 1.95)	44.2%	0.073	Random	0.348	0.557
Source of controls
HB	12	4302/3023	1.47 (0.86, 2.51)	96.10%	<0.001	Random	0.244	0.238
PB	25	12382/33487	1.33 (1.11, 1.58)	91.20%	<0.001	Random	0.199	0.418
Sample size
<600	20	3628/2973	1.32 (1.09, 1.61)	68.5%	<0.001	Random	0.041	0.016
>600	17	13056/33537	1.40 (1.07, 1.84)	96.9%	<0.001	Random	0.650	0.472

Abbreviations: HB, hospital based; PB, population based.

P_H: P-value based on Q test for between-study heterogeneity.

¹Adjustment for heterogeneity was performed by excluding 15 studies as the outliers and the possible major source of heterogeneity.

²Adjustment for heterogeneity was performed by excluding 10 studies as the outliers and the possible major source of heterogeneity.

³Adjustment for heterogeneity was performed by excluding 4 studies as the outliers and the possible major source of heterogeneity.