Background The discrepancy between genotype and expressed phenotype of the polymorphic N‐acetyltransferase (NAT2) has been suggested by separate genotypic and phenotypic studies in populations with human immunodeficiency virus (HIV). Only one study has examined both genotype and phenotype in the same population, and no discrepancies were observed. Methods In a cross‐sectional study, 105 HIV‐positive patients and patients with acquired immunodeficiency syndrome (AIDS) were phenotyped for NAT2 activity with use of caffeine as an in vivo probe; 50 of these patients were also genotyped by restriction mapping and allele‐specific amplification. In a longitudinal study, 23 patients were phenotyped at least twice during the 2‐year study. Results The distribution of the NAT2 phenotype among the 105 patients was unimodal and skewed toward slow acetylators as opposed to the bimodal distribution observed in healthy white populations. The genotype distribution was 26:24 slow:fast. There were 18 discrepancies between genotype and phenotype: 12 slow acetylators with fast genotypes and six fast acetylators with slow genotypes. No drug‐related effects on NAT2 activity were apparent, but the role of disease progression was evident. Among the slow acetylators whose genotype was fast, the incidence of AIDS was higher (six of 12) than that among the fast acetylators whose genotype was fast (two of 14). Among patients phenotyped more than once (mean time between samples, 10.4 months) changes in phenotype from fast to slow were associated with progression of HIV infection. Conclusions Disease progression in HIV infection and AIDS may alter expression of the NAT2 gene. The genotype and the phenotype are not interchangeable measurements. In the HIV population, to know the genotype is useful only if the phenotype is also known and vice versa. Clinical Pharmacology & Therapeutics (1997) 62, 261–271; doi: