BACKGROUND Many Japanese believe that low-yield cigarettes are less hazardous than regular cigarettes, and many smokers consume low-yield cigarettes to reduce their risks from smoking. We evaluate the association between actual nicotine intake and brand nicotine yield, and the influence of nicotine dependence on this association. METHODS The study subjects included 458 Japanese male smokers, aged 51.2 +/- 9.9 years, who participated in health check-ups in a hospital in 1998 and 2000. Each subject filled out a self-administered smoking questionnaire and the score of each on the Fagerström Test for Nicotine Dependence was calculated. Urinary cotinine concentration was measured at the time of participation. RESULTS The geometric mean of urinary cotinine concentration was 535 ng/mgCr for those who smoked brands with the lowest nicotine (0.1 mg on the package), compared with 1010 ng/mgCr for those who smoked brands with the highest (0.9-2.4 mg, weighted mean of 1.1 mg). Thus, despite the 11-fold ratio of nicotine yield on the packages, the ratio of urinary cotinine level was less than twofold. Both nicotine yield on the package and nicotine dependence significantly increased urinary cotinine concentration, and the negative interaction between them almost attained statistical significance. Cotinine concentration in heavily dependent smokers was consistently high regardless of the nicotine yield of brands. CONCLUSIONS The nicotine yield of cigarettes measured by machine-smoking does not reliably predict the exposure of smokers. Smokers consuming low-yield nicotine cigarettes did not reduce actual intake of nicotine to the level that might be expected, especially for those heavily dependent on nicotine. Current labeling practices are misleading for the two-third of smokers who are moderately or highly dependent on nicotine.