Urinary Bladder Cancer Risk Factors in Egypt—Letter

In the journal, Zheng and colleagues reported a weaker bladder cancer risk for active than passive smoking based on results from a case–control study in Egypt (1). In total, 1,886 pathologically confirmed cases with urothelial cell or squamous cell carcinoma (SCC) of the bladder and 2,716 controls were included. Control subjects were selected using random sampling of households and health units via family health records. Participation rates were higher for controls than cases. Unconditional logistic regression analyses were calculated adjusting for governorate of residence, urban/rural residence, age, education, urinary tract infections other than schistosomiasis, and menopause status (among women). Other risk factors, for example, occupational exposure to aromatic amines, were not assessed.

Besides infection with Schistosoma haematobium, smoking was confirmed as the most important risk factor for bladder cancer. However, rather implausibly, exposure to environmental tobacco smoke was associated with a higher risk as compared with heavy active smoking. Furthermore, the risk for urothelial cell bladder cancer only slightly increased with increasing number of pack-years, SCC bladder cancer risk was only marginally elevated among cigarette smokers, and no dose-dependent risk increase for SCC bladder cancer was seen with increasing number of pack-years. These findings are in contrast to a pooled analysis of 9 European case–control studies (2) and a recent U.S. cohort study (3), which showed a clear dose–response relation for number of pack-years and SCC risk. However, the authors fail to acknowledge and discuss these findings. Also, bladder cancers were not histopathologically graded according to their invasiveness, although a recent case–control study clearly showed that tobacco-related risks are higher for more invasive tumors (4).

The study's results raise numerous questions about the design and potential sources of bias, including the heterogeneous control sampling techniques, or the surprisingly high response among controls. Recruitment and collection of smoking information are not sufficiently explained. How reliable was self-reported active and passive smoking behavior by disease status, education, or other factors? Simultaneous adjustment for place of residence and rural/urban residence may result in overadjustment with dilution of effect estimates toward unity. Furthermore, detailed information on the pathologic classification of the cases would be helpful to explain why 757 cases, which were not included in the present analysis, still await pathologic confirmation. As observational studies are prone to bias, we need to carefully reflect potential limitations when interpreting a study's major findings that are in conflict with the literature.

See the Response, p. 694