Urethral cancer is rare in the United States, with an estimated incidence among men of 4.3 cases per million person-years.1 The most common histologic subtype is transitional cell carcinoma, followed by squamous and adenocarcinoma. Because of its rarity, the clinical characteristics, including best treatment strategies and expected outcomes, are not well understood. Case series from centers treating a large volume of urethral cancer patients can provide some insight; however, even the largest case series contain fewer than 50 patients.2 Findings from case series are also not always generalizable to the average patient.
In this issue of Cancer, Rabbani3 used Surveillance, Epidemiology, and End Results (SEER) data to describe the clinical characteristics and outcomes of male patients with urethral cancer. Among incident cases from1988-2008, 2065 men with incident urethral cancer, without a previous diagnosis of bladder cancer, were identified. Overall and disease-specific, 5-year survival rates were 46.2% and 68%, respectively. The most common histologic subtypes were transitional cell (77.6%), squamous cell (11.9%), and adenocarcinoma (5.0%). Nearly a quarter of patients presented with locally advanced tumors (T2-T4), and 22.1% had nodal involvement. Sixty-one percent of patients underwent simple surgical excision, 10.1% underwent radical resection, and 10.2% received some form of radiotherapy. Among the subset of 453 men with stage T2-T4 tumors, 98 underwent radical surgery alone, 97 underwent radiation alone, and 15 received both; 243 had neither radical surgery nor radiation. In multivariate survival analysis, predictors of poor outcome included advancing age, higher grade, higher local stage, visceral metastases, and nodal metastases. The authors also note that in the subgroup of patients with T2-T4 tumors, those who underwent radical surgical resection had improved cancer-specific survival compared with those who underwent radiation alone or to those who did not receive radical surgery or radiation.
Tumor registries, such as the SEER, provide valuable information that assist our understanding of cancer incidence and survival in the United States.4 SEER is maintained by the National Cancer Institute and currently collects and publishes cancer incidence and survival data from population-based cancer registries covering approximately 26% of the US population. The large size of the population at risk makes this registry ideal for gathering data to study rare tumors, and its population-based ascertainment method allows observations of tumor characteristics and outcomes that can be generalized to “average” patients. SEER is the only national source of cancer incidence and survival data in the United States, and SEER data are considered by most to be accurate and reliable. A detailed description of the SEER program can be found at the SEER website.4 Since its beginning in 1973, the SEER database has been used to answer hundreds of research questions. SEER data are well suited for some research questions and methodologies but not all.
Descriptive studies report observations without a specific hypothesis. Perhaps the most valid use of SEER is to describe cancer incidence rates and subsequent patient survival. These types of descriptive studies are important because monitoring cancer incidence over time is a critical component to the development of research priorities, screening, and prevention strategies. Analyses of survival rates after initial diagnosis allow assessment of changes in treatment effectiveness over time. The prevalence of a specific cancer (the number of living cancer survivors) provides another measure of the burden of cancer on society. Descriptive research questions can also provide an understanding of how cancer incidence, disease severity, and survival vary between cancer sites and among different subsets of the population. For example, age, race, or sex may be associated with differences in initial stage and/or survival. These observations are important first steps in identifying areas for future study or interventions aimed at achieving an earlier diagnosis in population subgroups. Providing a database for this type of descriptive epidemiology is a main reason that the SEER database is maintained. In its regularly published monographs, SEER provides basic descriptive data on all cancer sites. Previously published work has used SEER to describe urethral cancer incidence in men and women in the United States.1
Data from SEER can also be used in analytic studies designed to compare incidence or survival rates on the basis of demographic or tumor characteristics. For example, a researcher may hypothesize, based on evidence from case series, that African American patients have a higher incidence of a certain type of bladder cancer than white Americans.5 A researcher may also hypothesize that patients with different histologic types of bladder cancer have a different survival experience.6 SEER data are well suited to these types of analyses, but there are limitations on the conclusions that can be drawn when an association is found. SEER does not contain data on common risk factors for malignancy (eg, smoking, lifestyle factors, occupational exposures), so the potential causes for any observed variation in incidence in population subgroups are speculative. Furthermore, SEER does not contain data on comorbid diseases and other factors that may influence survival, so risk adjustment cannot be performed, and, therefore, there is the threat of confounding by these unmeasured factors. Despite these limitations, these types of analytical studies, when performed properly, are valuable to the advancement of our understanding of cancer and to the direction of future research.
Finally, SEER is also commonly used to compare outcomes of differing treatment strategies.7 Outcome studies share some of the limitations of the survival studies discussed above, and because comparisons are being made between treatment groups, the threat of confounding is increased. The type of therapy that is offered to a patient is influenced by innumerable factors that are not measured by SEER. Examples of potential confounding variables include comorbid disease, body mass index, socioeconomic status, smoking history, and health insurance status, among others. So-called “confounding by indication” or “selection bias” is common to the findings of studies that use SEER to compare different treatments. To the extent that healthier patients are more commonly selected for a specific type of therapy (eg, surgical treatment) over another, results will be biased toward improved survival in the treatment cohort that is enriched with healthier patients. This improved survival is independent of any treatment effect. This problem has no solution when using the SEER database. Studies that use SEER linkages to administrative data sources (eg, SEER-Medicare) can adjust for some of these factors by using variables derived from the administrative data, but adjustment for all confounding factors that influence both survival and treatment choices is still likely to be incomplete.
So how does the study of male urethral cancer by Rabbani measure up when we consider the strengths and limitations of SEER? First, male urethral cancer is uncommon, so the SEER database is well suited to address many of the issues discussed in the article. Tumor characteristics were presented using descriptive statistics, and for the first time, population-based survival data were presented for urethral cancer. The study found that patients with higher-grade and higher-stage disease have a worse survival outcome compared with those with lower-grade and stage; these findings confirm results from previous case series. The study also documented that male urethral cancer is primarily managed surgically in the United States, but in the subgroup of patients with local stage T2 or greater, more than half do not receive radical resection or radiotherapy. The Rabbani study3 elaborates on this finding by reporting an improved cancer-specific survival with surgery compared with other therapies, and the limitations of this finding are acknowledged. However, because SEER data are limited by the strong possibility of selection bias when they are used to compare treatment outcomes, the study provides little or no solid evidence that one treatment is superior to another.
In conclusion, population-based tumor registries, such as SEER, are invaluable to research that advances our understanding of cancer incidence and survival. These registries are especially useful for studying rare tumors. Registry data can be used to test hypotheses about variations in cancer incidence and survival among population and tumor subgroups, but SEER is not a good data source for comparing different treatments. To make informed judgments about the validity of findings that are based on analyses of SEER data, readers must be aware of the limitations imposed by the database itself.