Cancer studies suggest that the null polymorphisms of glutathione S-transferase M1 or T1 (GSTM1/GSTT1) may affect the ability to detoxify or activate chemicals in cigarette smoke. The potential modification of the association between smoking and coronary heart disease (CHD) by GSTM1 and GSTT1 has not been studied in humans. A case-cohort study was conducted to test the hypotheses that specific genotypes of GSTM1 or GSTT1 affect susceptibility to smoking-related CHD. CHD cases (n=400) accrued during 1987-1993 and a cohort-representative sample (n=924) were selected from a biracial cohort of 15792 middle-aged men and women in four US communities. A significantly higher frequency of GSTM1-0 and a lower frequency of GSTT1-0 were found in whites (GSTM1-0=47.1%, GSTT1-0=16.4%) than in African-Americans (AAs) (GSTM1-0=17.5%, GSTT1-0=25.9%). A smoking-GSTM1-0 interaction for the risk of CHD was statistically significant on an additive scale, with ever-smokers with GSTM1-0 at a approximately 1.5-fold higher risk relative to ever-smokers with GSTM1-1 and a approximately 2-fold higher risk relative to never-smokers with GSTM1-0, after adjustment for other CHD risk factors. The interaction between having smoked >/=20 pack-years and GSTT1-1 was statistically significant on both multiplicative and additive scales. The risk of CHD given both GSTT1-1 and >/=20 pack-years of smoking was approximately three times greater than the risk given exposure to >/=20 pack-years of smoking alone, and approximately four times greater than the risk given exposure to GSTT1-1 alone. The modification of the smoking-CHD association by GSTM1 or GSTT1 suggests that chemicals in cigarette smoke that are substrates for glutathione S-transferases may be involved in the etiology of CHD.