In reply

We reported the linkage of DRD2 haplotypes to heroin addiction in 799 Chinese and 663 German individuals.¹ In both populations, linkage was driven by a large difference between cases and controls in frequencies of haplotypes that are uncommon. This led Curtis and Gurling to propose that the case-control differences are artifacts generated by the inability of MLOCUS (Jeffrey Long, PhD, University of Michigan, Ann Arbor), an expectation-maximization program, to accurately infer haplotypes. Genotyping error and population stratification are issues not raised in their letter but were potential problems of concern to us. The genotyping error rate was 0.02 in Chinese and 0.01 in German individuals, and there is no reason to think that error would cause cases and controls to differentiate. Stratification was not apparent with a 32-locus–single nucleotide polymorphism panel but a larger, more informative panel is now being applied.