In Reply: Dr Dowdy points out that a substantial proportion of the person-time in our study occurred among former smokers. As individual former smokers cannot change their status to never-smokers, he questions whether our presentation of PARs is meaningful. Although individual former smokers cannot become never-smokers, it is conceivable to eliminate cigarette smoking at the population level. The first use of the PAR was by Levin in 1953 for the purpose of estimating the burden of lung cancer caused by ever smoking, 1 and this approach has been used many times subsequently, including the US Surgeon General's Report on the Health Consequences of Smoking. 2 We suggested 3 possible explanations for higher HRs in our study than in previous studies, including changing cigarette composition, increasing detection bias for bladder cancer in smokers, and chance. Dowdy suggests a fourth possible explanation, whereby unmeasured or poorly measured confounders, perhaps associated with socioeconomic status, may be more prominent in current smokers in the present day than in the 1980s. In our study cohort, as in the reference cited by Dowdy, 3 current smoking was associated with lower educational attainment (18% of participants with less than a college education were current smokers compared with 11% of participants with a college education), whereas the proportion of former smokers was 52% in both groups. The proportion of current smokers also varied by race/ethnicity, with 15% non-Hispanic white individuals, 20% non-Hispanic black individuals, 11% Hispanic individuals, and 10% Asian/Pacific Islander/Native American individuals being current smokers. For former smoking, the corresponding proportions were 52%, 43%, 47%, and 41%, respectively.In our models, we adjusted for education and race/ ethnicity and in this way indirectly for environmental exposures associated with these factors. Although models were adjusted for education and race/ethnicity (current vs never smoking, HR, 4.06; 95% CI, 3.66-4.50), risk estimates were quite similar in models adjusted for only age and sex (HR, 4.17; 95% CI, 3.77-4.62). Nevertheless, as in all observational studies, unmeasured or poorly measured confounders may have affected our risk estimates. Future studies are needed to replicate our findings in other populations and to directly evaluate the role of cigarette composition in bladder cancer etiology.