Some aromatic amines are considered to be putative bladder carcinogens. Hemoglobin (Hb) adducts of 3-aminobiphenyl (3-ABP) and 4-aminobiphenyl (4-ABP) have been used as biomarkers of exposure to aromatic amines from cigarette smoke. One of the goals of this study was to determine intra- and inter-individual variability in 3-ABP and 4-ABP Hb adducts and to explore the predictability of ABP Hb adduct levels based on caffeine phenotyping. The study was conducted in adult smokers (S, n = 65) and non-smokers (NS, n = 65). The subjects were phenotyped for CYP1A2 and NAT2 using urinary caffeine metabolites. Blood samples were collected twice within 6 weeks and adducts measured by GC/MS. The levels of 4-ABP Hb adducts were significantly (p < 0.0001) greater in S (34.5 ± 21.06 pg/g Hb) compared to NS (6.3 ± 3.02 pg/g Hb). The levels of 3-ABP Hb adducts were below the limit of quantification (BLOQ) in most (82%) of the NS and about 10-fold lower in S (3.6 ± 3.29 pg/g Hb) compared to 4-ABP Hb adducts. No differences were observed in the adduct levels between weeks 1 and 6 in the smokers, suggesting that a single sample would be adequate to monitor cigarette smoke exposure. The regression model developed with CYP1A2, NAT2 phenotype and number of cigarettes smoked (NCIG) accounted for 47% of the variability in 3-ABP adducts, whereas 32% variability in 4-ABP adducts was accounted by CYP1A2 and NCIG. The ratio of 4-ABP Hb adducts in adult S:NS was ∼ 5:1, whereas 3-ABP Hb adducts levels were BLOQ in some S, exhibited large interindividual variability (∼ 91% compared to 57% for 4-ABP Hb) and poor dose response relationship. Therefore, 4-ABP Hb adduct levels may be a more useful biomarker of aminobiphenyl exposure from cigarette smoke.