A previous case control study by Vandenbergh et al. [1997: Am J Med Genet 74:439-442] showed an association between the high activity catechol O-methyltransferase (COMT) polymorphism and polysubstance abuse in a group of North American subjects. In the current study we confirm these results by genotyping 38 Israeli heroin addicts and both parents using a robust family-based haplotype relative risk (HRR) strategy. There is an excess of the val COMT allele (likelihood ratio = 4.48, P = 0.03) and a trend for an excess of the val/val COMT genotype (likelihood ratio = 4.97, P = 0.08, 2 df) in the heroin addicts compared to the HRR control group. We also genotyped an additional 101 nonrelated heroin addicts and 126 control subjects using a case control design and found no significant difference in COMT val allele frequency (25.4% vs. 29.7%, likelihood ratio = 1.04, P = 0.31). A significant difference is observed in COMT allele frequency among the three principal Israeli ethnic groups (Ashkenazi Jewish, non-Ashkenazi Jewish, and Palestinian Arab) in a large group of control subjects we have so far examined (chi-square = 7.9, P = 0.019, df = 2, n = 1,422 alleles) suggesting that population stratification is responsible for our failure to observe an excess of the COMT val allele when using the case-control design.