Freedman ND, Silverman DT, Hollenbeck AR, Schatzkin A, Abnet CC. Association Between Smoking and Risk of Bladder Cancer Among Men and Women. JAMA. 2011;306(7):737-745. doi:10.1001/jama.2011.1142
Author Affiliations: Division of Cancer Epidemiology and Genetics, National Cancer Institute, National Institutes of Health, Department of Health and Human Services, Rockville, Maryland (Drs Freedman, Silverman, Schatzkin, and Abnet); and AARP, Washington, DC (Dr Hollenbeck).
†Died January 20, 2011.
Context Previous studies indicate that the population attributable risk (PAR) of bladder cancer for tobacco smoking is 50% to 65% in men and 20% to 30% in women and that current cigarette smoking triples bladder cancer risk relative to never smoking. During the last 30 years, incidence rates have remained stable in the United States in men (123.8 per 100 000 person-years to 142.2 per 100 000 person-years) and women (32.5 per 100 000 person-years to 33.2 per 100 000 person-years); however, changing smoking prevalence and cigarette composition warrant revisiting risk estimates for smoking and bladder cancer.
Objective To evaluate the association between tobacco smoking and bladder cancer.
Design, Setting, and Participants Men (n = 281 394) and women (n = 186 134) of the National Institutes of Health-AARP (NIH-AARP) Diet and Health Study cohort completed a lifestyle questionnaire and were followed up between October 25, 1995, and December 31, 2006. Previous prospective cohort studies of smoking and incident bladder cancer were identified by systematic review and relative risks were estimated from fixed-effects models with heterogeneity assessed by the I2 statistic.
Main Outcome Measures Hazard ratios (HRs), PARs, and number needed to harm (NNH).
Results During 4 518 941 person-years of follow-up, incident bladder cancer occurred in 3896 men (144.0 per 100 000 person-years) and 627 women (34.5 per 100 000 person-years). Former smokers (119.8 per 100 000 person-years; HR, 2.22; 95% confidence interval [CI], 2.03-2.44; NNH, 1250) and current smokers (177.3 per 100 000 person-years; HR, 4.06; 95% CI, 3.66-4.50; NNH, 727) had higher risks of bladder cancer than never smokers (39.8 per 100 000 person-years). In contrast, the summary risk estimate for current smoking in 7 previous studies (initiated between 1963 and 1987) was 2.94 (95% CI, 2.45-3.54; I2 = 0.0%). The PAR for ever smoking in our study was 0.50 (95% CI, 0.45-0.54) in men and 0.52 (95% CI, 0.45-0.59) in women.
Conclusion Compared with a pooled estimate of US data from cohorts initiated between 1963 and 1987, relative risks for smoking in the more recent NIH-AARP Diet and Health Study cohort were higher, with PARs for women comparable with those for men.
More than 350 000 individuals are diagnosed with incident bladder cancer per year worldwide,1 including more than 70 000 per year in the United States.2 In data from Surveillance, Epidemiology, and End Results Program, incidence rates in white individuals aged 50 years or more have remained stable during the past 30 years (1976-2006), from 123.8 per 100 000 person-years to 142.2 per 100 000 person-years in men and from 32.5 per 100 000 person-years to 33.2 per 100 000 person-years in women; similar patterns are observed in other ethnic and racial groups.3
Tobacco smoking is the best established risk factor for bladder cancer in both men and women.4,5 Although rates of bladder cancer have remained stable during the past 30 years, prevalence of cigarette smoking in the United States has substantially decreased during the same period.6
Typically, risk estimates for current smokers have been approximately 3 in previous studies.5,7 However, the composition of cigarettes has changed during the past 50 years, leading to a reduction in tar and nicotine concentrations in cigarette smoke,8 but also to an apparent increase in the concentration of specific carcinogens, including β-napthylamine, a known bladder carcinogen, and tobacco-specific nitrosamines.9 Concurrent with these changes in the constituents of cigarette smoke, epidemiological studies have observed higher relative risks associated with cigarette smoking for lung cancer.10 A recent article from the New England Bladder Cancer Study,11 a large population-based case-control study, suggests that the strength of the cigarette smoking−bladder cancer association may also have increased. In this article, the authors compared the odds ratio (OR) for current smokers relative to never smokers in 3 similar population-based case-control studies performed in New Hampshire in 1994-1998, 1998-2001, and 2001-2004. During the course of the 3 studies, the OR associated with current smoking increased from 2.9 (95% confidence interval [CI], 2.0-4.2) to 4.2 (95% CI, 2.8-6.3) to 5.5 (95% CI, 3.5-8.9). These provocative results suggest that changing cigarette composition over time may be associated with increased bladder cancer risk, analogous to results previously documented for lung cancer. Stronger associations between smoking and bladder cancer could potentially offset the decreased prevalence of smoking in the US population and contribute to the stability of the bladder cancer incidence rates during the past 30 years. However, these findings need replication, particularly in prospective cohort studies.
Population attributable risks (PARs) for tobacco smoking have been estimated to be 50% to 65% in men and 20% to 30% in women.5,12- 15 However, these estimates were based on studies conducted in populations during periods in which the prevalence of smoking was higher in men than in women. Currently, in the United States and in many other countries, the prevalence of smoking is similar in men and women.16,17
Our goal was to estimate the strength of the association between tobacco smoking and bladder cancer and the PARs for smoking among men and women in the large, prospective National Institutes of Health-AARP (NIH-AARP) Diet and Health Study, initiated October 25, 1995, with follow-up through December 31, 2006.
The NIH-AARP Diet and Health Study has been described previously.18 Briefly, a questionnaire was mailed between 1995 and 1996 to 3.5 million AARP members aged 50 to 71 years who resided in 8 states (California, Florida, Georgia, Louisiana, Michigan, New Jersey, North Carolina, and Pennsylvania). Of 617 119 returned questionnaires, 566 401 were completed in satisfactory detail. The NIH-AARP Diet and Health Study was reviewed and approved by the Special Studies Institutional Review Board of the US National Cancer Institute, and all participants gave written informed consent by virtue of completing and returning the questionnaire.
Addresses for cohort members were updated annually in response to participant change of address requests and by matching cohort participants to the US Post Office National Change of Address database. Vital status was obtained by linkage to the Social Security Administration Death Master File and response to mailings. Follow-up time started the date the questionnaire was returned (beginning October 25, 1995) and accumulated until diagnosis of bladder cancer, a move out of the catchment area, date of death, or December 31, 2006, whichever came first.
We identified incident bladder cancers by linking the NIH-AARP Diet and Health Study cohort with the cancer registry databases of 10 states (8 baseline states, plus Arizona and Texas). In a validation study, this approach identified approximately 90% of cancers.19 Bladder cancer cases had an International Classification of Diseases for Oncology (ICD-O)20 site code of C67.0-C67.9 and a transitional cell (urothelial) morphology (ICD codes 8120, 8122, 8123, or 8130).
The baseline questionnaire assessed tobacco use, alcohol intake, demographics, physical activity, and intake of 124 food items. Race/ethnicity (non-Hispanic white, non-Hispanic black, Hispanic, and Asian, Pacific Islander, or Native American) was assessed by self-report and was collected to study whether the association of cancer risk factors differed by racial or ethnic group. Assessment of tobacco use via questionnaire has shown high reproducibility (r = 0.94) and validity (r = 0.92 for women and r = 0.90 for men, relative to serum cotinine levels) in previous methodological studies.21,22 Participants were considered ever cigarette smokers if they had smoked more than 100 cigarettes during their lifetimes. In responding to the questionnaire, ever smokers recorded their typical cigarette smoking intensity using 6 categories of cigarettes per day (1-10, 11-20, 21-30, 31-40, 41-60, and ≥61); former smokers reported years of smoking cessation using 4 categories (stopped within the last year, stopped 1-4 years ago, stopped 5-9 years ago, and stopped ≥10 years ago). We considered those participants who had quit more than 1 year before baseline as former cigarette smokers. A separate question assessed whether participants had regularly smoked pipes or cigars for 1 year or longer.
We completed all NIH-AARP Diet and Health Study analyses using SAS version 9.1 (SAS Institute Inc, Cary, North Carolina), with a 2-sided significance level of P < .05. We calculated age-standardized incidence rates and 95% CIs using 5-year age bands standardized to the entire NIH-AARP Diet and Health Study population.23 The number needed to harm (NNH) was calculated from age-standardized incidence rates.
For relative risks, hazard ratios (HRs) and 95% CIs were calculated using Cox proportional hazards regression models.24 Risk estimates were adjusted for age, education, ethnicity, and pipe or cigar use. Additional adjustment for other possible confounders (alcohol, aspirin and ibuprofen nonsteroidal anti-inflammatory drugs, body mass index [calculated as weight in kilograms divided by height in meters squared], physical activity, self-reported health, intake of fruit, vegetables, or meat, or total energy) did not alter risk estimates. For the less than 3% of the cohort that was missing data for a particular covariate, a separate indicator variable for missing was included in the models.
We tested the proportional hazards assumption by including an interaction term for follow-up time and cigarette use in the Cox proportional hazards regression models and found no statistically significant deviations.
Linear trend tests across categories of cigarette smoking were conducted by assigning participants their appropriate category of cigarette smoking and entering this variable as a continuous term in the regression model. P values were then obtained from the Wald test.
We used the method of Bruzzi et al25 to calculate PARs from multivariate-adjusted β coefficients for ever smoking. The δ method was used to estimate the variance in order to estimate the 95% CIs for the PAR estimates.26
We identified previous US prospective cohort studies that assessed cigarette smoking at baseline and examined the association of current smoking with subsequent bladder cancer incidence by using the following search terms in PubMed and EMBASE ([tobacco or smok* or cig* ] and [cancer or carcinoma or neoplas* ] and [bladder or urinary tract or urolog* or urothelial ] and [cohort or prospective ]) (eFigure). Our search was performed on June 28, 2011, and included all publications in the databases published until then. We did not restrict our search by language. After excluding duplicates, our search yielded 843 articles. Titles and abstracts were reviewed and then we excluded studies conducted in populations outside of the United States or that lacked data on incident bladder cancer, resulting in 60 studies. We reviewed all 60 published manuscripts, excluding 1 published abstract, studies conducted outside the United States, reviews, cross-sectional studies, studies of bladder cancer mortality, studies with overlapping results, and studies lacking risk estimates for current vs never smoking. After these exclusions, 6 publications remained that provided data from 7 cohorts. No further publications were identified upon reviewing the references of these 6 remaining articles. We did not assess study quality and instead chose to include all identified studies in our meta-analysis.
From each article, we extracted data on the authors and year of publication, cohort name, participant sex, mean age, number of never smoking cases and cohort participants, number of current smoking cases and cohort participants, typical amount of cigarettes smoked per day among current smokers, and the relative risk for current, relative to never, smoking. We extracted maximally adjusted risk estimates. In studies that lacked 1 or more extraction variables, we sought this data in previous cohort publications. We did not contact study authors. Two co-authors (N.D.F. and C.C.A.) reviewed each publication to ensure that the data extraction was accurate.
We used Stata version 11.0 (StataCorp LP, College Station, Texas) to perform meta-analysis. Heterogeneity between studies was assessed by the I2 statistic and the Cochran Q test.27 Summary relative risks and 95% CIs were calculated using fixed-effects models (Mantel-Haenszel method). We examined possible publication bias using both Begg and Mazudmar P value for rank correlation28 and Egger weighted regression method.29
Additional analyses were performed including data from previously published studies together with NIH-AARP Diet and Health Study. Possible heterogeneity across studies was examined using the I2 statistic and the Cochran Q test.
Participants with cancer (except nonmelanoma skin cancer) at baseline (n = 51 234), proxy respondents (n = 15 760), those participants who died or who were diagnosed with cancer on the first day of follow-up (n = 13), or those participants who failed to provide information about cigarette use (n = 19 329) or cigar and pipe use (n = 12 537) were excluded, resulting in an analytic cohort of 281 394 men and 186 134 women. Men and women entered the study at similar ages, but men had more formal education, drank more alcohol, ate less fruit and vegetables, were more likely to have ever smoked cigarettes, pipes, or cigars, and to have smoked more than 40 cigarettes per day than women. However, a higher proportion of women than men were current smokers. The median age of smoking initiation was 17 years in the subset of the cohort (118 557 men and 72 030 women) who returned a follow-up questionnaire between October 7, 2004, and December 30, 2004 (Table 1).
During the course of 4 518 941 person-years of follow-up, 3896 men and 627 women were newly diagnosed with bladder cancer. Overall incidence rates were 144.0 per 100 000 person-years (95% CI, 139.4-148.5) in men and 34.5 per 100 000 person-years (95% CI, 31.8-37.3) in women. Cigarette smoking was a strong risk factor for bladder cancer in both sexes (Table 2). Relative to never smokers (69.8 per 100 000 person-years in men and 16.1 per 100 000 person-years in women), former and current smokers had increased risk of bladder cancer in both men (former smokers, 154.6 per 100 000 person-years; adjusted HR, 2.14; 95% CI, 1.92-2.37; NNH, 1179; and current smokers, 276.4 per 100 000 person-years; adjusted HR, 3.89; 95% CI, 3.46-4.37; NNH, 484) and women (former smokers, 40.7 per 100 000 person-years; adjusted HR, 2.52; 95% CI, 2.05-3.10; NNH, 4065; and current smokers, 73.6 per 100 000 person-years; adjusted HR, 4.65; 95% CI, 3.73-5.79; NNH, 1739). The combined risk estimates including both sexes were 2.22 (95% CI, 2.03-2.44) for former smokers (119.8 per 100 000 person-years; NNH, 1250; 95% CI, 1171-1343) and 4.06 (95% CI, 3.66-4.50) for current smokers (177.3 per 100 000 person-years; NNH, 727; 95% CI, 671-794) relative to never smokers (39.8 per 100 000 person-years).
As in previous studies, smoking cessation was associated with reduced bladder cancer risk in both sexes. Participants who quit 10 years or more before baseline had lower incidence rates of bladder cancer than those who quit 1 to 4 years or 5 to 9 years before baseline. Nevertheless, relative to never smokers, relative risks remained increased for men and women who quit even 10 years or more before baseline. In never users of cigarettes, pipe and cigar use was also associated with increased risk of bladder cancer in men (HR, 1.29; 95% CI, 1.07-1.56; 92.5 per 100 000 person-years) relative to nonusers (69.8 per 100 000 person-years; NNH, 4405). Too few women in the cohort smoked pipes or cigars to be analyzed.
Overall, men had 3.71 (95% CI, 3.39-4.06; 144.0 per 100 000 person-years vs 34.5 per 100 000 person-years) times the risk of women for bladder cancer (Table 3). Among stratum of cigarette smoking, risks for men relative to women ranged from 1.99 to 6.62. Increased rates persisted in never smokers in which men (69.8 per 100 000 person-years) had 4.07 (95% CI, 3.34-4.97) times the bladder cancer risk of women who never smoked (16.1 per 100 000 person-years).
The PAR for ever smoking in the NIH-AARP Diet and Health Study was similar in men (0.50; 95% CI, 0.45-0.54) and women (0.52; 95% CI, 0.45-0.59).
We then performed a systematic review and meta-analysis of previously published US prospective cohort studies of current cigarette smoking and incident bladder cancer (eFigure). We identified data from the 7 cohorts (Table 4). In these cohorts initiated between 1963 and 1987, the summary risk estimate was 2.94 (95% CI, 2.45-3.54) with an I2 of 0.0% and the Cochran Q test P value for between-study heterogeneity was .554. We observed no evidence for publication bias by either Egger weighted regression (P = .32) or Begg and Mazumdar rank correlation method (P = .29).
Addition of risk estimates from the NIH-AARP Diet and Health Study to the meta-analysis increased the summary risk estimate to 3.75 (95% CI, 3.43-4.10) and increased the I2 to 48.7%, such that the Cochran Q test P value for between-study heterogeneity became statistically significant (P = .049).
In the NIH-AARP prospective cohort study, cigarette smoking was strongly associated with bladder cancer risk in both men and women, and ever smoking explained a similar proportion of bladder cancer in both sexes, with PARs of 50% in men and 52% in women.
With follow-up occurring between October 25, 1995, and December 31, 2006, current smoking was associated with a relative risk of 4.06 (95% CI, 3.66-4.50) in men and women combined. This risk estimate for current smoking is broadly similar to those estimates observed in New Hampshire case-control data for cases diagnosed in 1998-2001 and 2002-2004 and higher than those for cases diagnosed in 1994-1998. The 1994-1998 cases had an OR of 2.9 (95% CI, 2.0-4.2), whereas the cases diagnosed in 1998-2001 had an OR of 4.2 (95% CI, 2.8-6.3), and the cases diagnosed in 2002-2004 had an OR of 5.5 (95% CI, 3.5-8.9).11 Previously published US prospective cohort studies of cigarette smoking and incident bladder cancer risk in men and women were initiated between 1963 and 1987. The summary estimate from these 7 cohorts was 2.94 (95% CI, 2.45-3.54), which is significantly lower than that observed in our study. These observations parallel those previously reported for lung cancer in which changes in cigarette design have been linked to stronger associations with cigarette smoking.10 Changes in the constituents of cigarette smoke, including apparent increased concentrations of β-napthylamine, a known bladder carcinogen, and tobacco-specific nitrosamines,9 may have strengthened the cigarette smoking−bladder cancer association as well. Another potential explanation is the increased awareness of bladder cancer risk in smokers, which may prompt earlier diagnostic workup. Alternatively, differences between our study and past studies could have been due to chance, although a recently published meeting abstract from the VITamins And Lifestyle Study (VITAL) also indicated an HR of 4.0 (95% CI, 2.9-5.8) for current smoking vs never smoking for incident bladder cancer.45
Although our data suggest that the association of cigarette smoking with bladder cancer has strengthened, incidence rates have stayed largely constant during this same period. However, cigarette composition is just one of the smoking-related changes occurring during this time. Substantial reductions in the prevalence of cigarette smoking have also occurred.17 Our results, and those of the New England Bladder Cancer Study,11 suggest that the strengthening of the smoking-related relative risks, perhaps due to changing cigarette composition, may have offset the effect of declining smoking prevalence, at least to some extent, contributing to relatively stable incidence rates of bladder cancer during the past 30 years. Future work is needed to investigate this hypothesis.
In the NIH-AARP Diet and Health Study cohort, in which the prevalence of smoking is generally similar in men and women, as observed in the US population,16,17 the PAR for smoking was approximately 50% in both sexes. Previous studies have found PARs of 50% to 65% in men and 20% to 30% in women,5,12- 15 but were conducted in populations in which the prevalence of smoking in women was considerably lower than in men.46 In our study cohort and in the general US population,16,17 however, the prevalence of smoking is similar in men and women. This is the first article to our knowledge to demonstrate that the increased prevalence of smoking in US women has led to an increased PAR for smoking, such that the PARs for smoking and bladder cancer are now similar in US men and women.
In addition to bladder cancer, tobacco smoking is strongly associated with increased risk of lung cancer.47 Incidence rates of lung cancer, similar to bladder cancer, are higher in men than in women worldwide.1 Historically higher rates of tobacco smoking in men relative to women likely explain most of the excess lung cancer cases in men. Because the prevalence of tobacco smoking in women has increased,48 incidence rates of lung cancer in men and women have converged in many countries, including the United States.16,48,49 Furthermore, we demonstrated similar incidence rates of lung cancer in men and women of the NIH-AARP Diet and Health Study cohort, both among men and women who smoked similar amounts and among never smokers.50
In contrast with the lung, incidence rates of bladder cancer have not converged in men and women,7 even in countries such as the United States in which men and women now smoke similar amounts.49 In our study, we observed consistently higher incidence rates of bladder cancer in men than women, both among individuals who smoked similar amounts and among never smokers. Our results are consistent with the National Bladder Cancer Study,15 a population-based case-control study conducted in 1978. In this study, as in our study, risk of bladder cancer remained higher in men vs women who were never smokers. Although differences in the prevalence of smoking are likely an important explanation for the excess of bladder cancer in men in many parts of the world where cigarette smoking is substantially more common in men than in women,51,52 our results and those of the National Bladder Cancer Study15 suggest that differences in smoking use do not completely explain higher incidence rates of bladder cancer in US men.2 Higher incidence rates in men could also reflect occupational exposures, because men in general are more likely than women to work in specific occupations that have been traditionally associated with bladder cancer risk (ie, aromatic amine-manufacturing workers, leather workers, painters, truck drivers, machinists, and aluminum workers).5,53- 55 We lacked assessment of occupation in our study; however, bladder cancer risk among men in the National Bladder Cancer Study was attenuated after adjustment for occupational exposures, but remained increased relative to women.15 Alternatively, physiological differences between men and women, such as differences in the levels of sex hormones, could contribute to higher rates in men. Several recent studies provide evidence for associations between menstrual and reproductive factors with bladder cancer,43,56,57 and this is an active area of investigation.
Strengths of our study include assessment of smoking use before cancer diagnosis, very large number of participants and incident bladder cancers, and presentation of both incidence rates (absolute risks) and relative risks. However, several limitations should be noted. We lacked information about the age at smoking initiation for a majority of cohort participants and so could not calculate smoking duration or pack-years. Among the subset of cohort participants (118 557 men and 72 030 women) returning a follow-up questionnaire between October 7, 2004, and December 30, 2004, the median age at smoking initiation was 17 years in both men and women. In addition, smoking status was assessed only at baseline and was not updated during the course of follow-up. Because a number of participants probably quit during follow-up, risk estimates for current smoking in our study are likely to be attenuated. In addition, our results may not apply to other populations, particularly those in other countries that may differ in smoking prevalence and cigarette composition.
In conclusion, tobacco smoking was a strong risk factor for bladder cancer, with PARs of approximately 50% in both men and women. We found higher risk estimates for current cigarette smoking relative to never smoking in the NIH-AARP Diet and Health Study cohort, initiated in 1995, than were reported in previous publications from cohorts initiated between 1963 and 1987. These results support the hypothesis that the risk of bladder cancer associated with cigarette smoking has increased with time in the United States, perhaps a reflection of changing cigarette composition. Prevention efforts should continue to focus on reducing the prevalence of cigarette smoking.
Corresponding Author: Neal D. Freedman, PhD, MPH, Division of Cancer Epidemiology and Genetics, National Cancer Institute, National Institutes of Health, Department of Health and Human Services, 6120 Executive Blvd, EPS/320, MSC 7232, Rockville, MD 20852 (email@example.com).
Author Contributions: Dr Freedman had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Freedman, Silverman, Hollenbeck, Schatzkin, Abnet.
Acquisition of data: Hollenbeck, Schatzkin.
Analysis and interpretation of data: Freedman, Silverman, Abnet.
Drafting of the manuscript: Freedman, Silverman, Abnet.
Critical revision of the manuscript for important intellectual content: Freedman, Silverman, Hollenbeck, Schatzkin, Abnet.
Statistical analysis: Freedman, Silverman, Abnet.
Obtained funding: Schatzkin.
Study supervision: Freedman, Hollenbeck, Schatzkin, Abnet.
Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest and none were reported.
Funding/Support: This research was supported by the Intramural Research Program of the National Institutes of Health, National Cancer Institute, Division of Cancer Epidemiology and Genetics. Cancer incidence data from Arizona were collected by the Arizona Cancer Registry; from Georgia by the Georgia Center for Cancer Statistics; from California by the California Department of Health Services, Cancer Surveillance Section; from Michigan by the Michigan Cancer Surveillance Program; from Florida by the Florida Cancer Data System under contract to the Florida Department of Health; from Louisiana by the Louisiana Tumor Registry; from New Jersey by the New Jersey State Cancer Registry; from North Carolina by the North Carolina Central Cancer Registry; from Pennsylvania by the Division of Health Statistics and Research, Pennsylvania Department of Health; and from Texas by the Texas Cancer Registry.
Role of the Sponsor: The sponsor reviewed and approved final submission but had no role in the design and conduct of the study; in the collection, analysis, and interpretation of the data; or in the preparation of the manuscript.
Disclaimer: The views expressed herein are solely those of the authors and do not necessarily reflect those of the cancer registries or contractors. The Pennsylvania Department of Health specifically disclaims responsibility for any analyses, interpretations, or conclusions.
Additional Information: This article is dedicated to the memory of Arthur Schatzkin, MD, DrPH.
Additional Contributions: We thank the participants of the NIH-AARP Diet and Health Study for their outstanding cooperation.
This article was corrected for errors on October 18, 2011.