Human acetylation phenotypes were determined with caffeine (137X) as the test substance, improved by measuring urinary caffeine metabolites with a previously described HPLC method. Caffeine, 5-acetylamino-6-formylamino-3-methyluracil (AFMU), 1-methylxanthine (IX), 1-methyluric acid (IU), 1,7-dimethylxanthine (17X), and 1,7-dimethyluric acid (17U) were quantified. This study tested the hypothesis, suggested by previous studies, that the acetylation polymorphism is strongly influenced by a major gene. Phenotypes were assessed by using four urinary caffeine metabolite ratios: AFMU/1X, AFMU/[1X + 1U + 17U], AFMU/[AFMU + 1X + 1U], and AFMU/[1X + 1U + 17X + 17U] in a population included 281 nuclear family members who were healthy volunteer subjects. Each urinary ratio revealed strong familial aggregation with correlations between parents and offspring varying from 0.340 to 0.486 as a function of the ratio considered, and between sibs from 0.410 to 0.512 whereas correlations among spouses were not significant, excluding an effect of environmental factors. Segregation analyses were conducted upon these four ratios testing a series of specific models of inheritance and gave evidence for single locus control of N-acetyltransferase (NAT) activity, with Mendelian codominant transmission using the AFMU/1X, AFMU/[1X + 1U + 17U], and AFMU/[1X + 1U + 17X + 17U] ratios. The slow allelic frequencies were 0.739, 0.753, and 0.724, respectively, and the phenotypic concordance was 90 to 92% with the AFMU/1X ratio. The familial aggregation observed in using the AFMU/[AFMU + 1X + 1U] ratio was consistent with a recessive transmission for the allele controlling the homozygous slow phenotype. This last ratio is not convenient to differentiate rapid heterozygous and homozygous phenotype. Considering the number of misclassified subjects, this study would be completed by genotyping the same families as demonstrated by some authors who predicted 97.5% of acetylation phenotype when using PCR-based DNA amplification test.