• © 2005 by American Society of Clinical Oncology

Variations in Prostate Cancer Patterns of Care: Is It the Quality of Care or the Quality of the Data?

A full decade has passed since numerous researchers first noted significant geographic variation in patterns of care in localized prostate cancer.^1-3 Since then, there have been more than 1,000 articles describing outcomes following various therapies for localized disease. In the current issue of the Journal of Clinical Oncology, Krupski et al⁴ document that, despite all of the data on clinical outcomes, these variations still exist. The larger question one must ask is: Why? Is it a problem with the quality of care, the quality of the data, or both?

Krupski et al⁴ used data from 11 state and regional cancer registries participating in the National Cancer Institute's Surveillance Epidemiology and End Results (SEER) program to review patterns of care in men diagnosed with localized or regionally advanced prostate cancer from 1995 to 1999. They noted significant geographic variation in the use of surgery, radiation therapy, and conservative (watchful waiting) therapy. For example, surgery or watchful-waiting rates were significantly higher in the Los Angeles area than in other regions of the United States. Conversely, radiation therapy was employed more commonly in the Detroit area. Interestingly, within registries, there was little variation in treatment rates, indicating local homogeneity in patterns of care. The authors also found that age, race, income, and pathologic differentiation were independently associated with primary surgical or radiation therapy. The authors conclude that their results may suggest “unconscious discrimination on the part of physicians.”

Before we accept this conclusion, we must consider the limitations of the current study. As the authors rightfully point out in their discussion, the SEER data set does not allow for individual level assessment of socioeconomic status (SES). To account for this, the research team ascribed group characteristics to individuals based on 2000 United States census data for individual's county of residence and ethnicity. Another limitation stems from the fact that the SEER data set has no information concerning comorbid conditions. Primary choice of therapy in localized prostate cancer has been previously shown to be associated with the number of comorbid conditions at the time of diagnoses.^5,6 Therefore, if one wants to determine the true effect of ethnicity or SES on primary treatment, comorbid disease must be controlled for, as these factors may all be related.⁷ Although the lack of this information likely affects the results of the current study, it is my belief that, even if the researchers had been able to completely control for comorbidity and had individual level SES data, ethnicity and SES would still have been associated with choice of primary treatment. Assuming that the true strength of the associations may not be as strong as observed in the study but would still be significant, we must ask why these geographic, ethnic, and socioeconomic variations in prostate cancer care exist.

While it is not unreasonable to ascribe these variations to unconscious discrimination on the part of providers, I would suggest that the real cause of the differences in patterns of care is much less insidious. What I believe we are witnessing is not a problem with the quality of care, but rather with the quality of the data. McNeil⁸ noted that uncertainty in medical decision making will result in huge variations in medical practice. This may be due to delayed or obsolete clinical data, results from study populations that are not generalizable to those patients normally seen in the clinic, or varying interpretations of existing data. In localized prostate cancer, all of these factors contribute to uncertainty in medical decision making and are likely to contribute to the variations observed in the current study.

However, more pressing than the potential problems with the quality of the existing information in localized prostate cancer is the overwhelming lack of randomized clinical trial data comparing the commonly used therapies in this disease. To date, there has been only one completed randomized trial comparing surgery to watchful waiting in localized disease.⁹ Beyond this study, we have minimal, if any, randomized trial data directly comparing radiation therapy to surgery or watchful waiting in localized disease. Krupski et al point out the recently abandoned American College of Surgeons Oncology Group trial comparing surgery to brachytherapy as an example of the type of critical study that is needed to determine which patients are best served by particular treatment modalities. The prostate cancer community's ongoing inability to complete randomized clinical trials (or even well-controlled observational studies) comparing primary therapies for localized prostate cancer is an embarrassment to us and a continuing disservice to our patients. Given the lack of evidence to guide decision making in this common malignancy, it is not at all surprising that there is significant geographic, ethnic, and socioeconomic variation in patterns of care.

This is not to say that, were we to complete randomized clinical trials or observational studies that provided valid and useful information, disparities in quality of care would not exist. Consider the case of breast-conserving therapy in localized breast cancer. There is reasonable level I evidence to suggest that segmental mastectomy is equivalent to total mastectomy in overall survival 5 years after diagnosis.¹⁰ Despite this, there is still significant variation in the use of breast-conserving therapy by geographic region, education, and income.¹¹ In this setting, we can comfortably ascribe these variations to true disparities in the quality of care, as opposed to uncertainty regarding the optimal treatment, and can devise appropriate policy interventions to reduce these discrepancies (although there may be additional geographic or sociocultural influences that result in variation in patterns of care that are appropriate and do not represent disparities in the quality of care, assuming clinically relevant outcomes are equivalent). It is to the great credit of the breast cancer community that they have been able to generate the appropriate evidence to eliminate much of the uncertainty in their clinical decision making. Sadly, we have not reached this point in localized prostate cancer. The lack of solid clinical evidence has resulted in the personal opinions and anecdotal experiences of both providers and patients ultimately influencing the choice of primary therapy, as evidenced in the current study. This is not acceptable in the age of evidence-based medicine.