Given the greater intensity of screening, diagnosis, and treatment documented among cohort members in the Seattle-Puget Sound area than in Connecticut, we would have expected to see lower prostate cancer mortality in the Seattle region over time. However, we found no significant difference in prostate cancer mortality over 11 years of follow up. The confidence interval around the rate ratio for mortality was inconsistent with more than a 5% reduction in prostate cancer mortality in Seattle. Moreover, the overall drop in prostate cancer mortality seemed similar in the two regions (fig ). The results of this study therefore do not support the hypothesis that the intensity of screening and treatment with surgery or radiation was related to the reduction in prostate cancer mortality seen in the two regions.
A recent report has documented decreases in population based prostate cancer mortality in Tyrol compared with other states in Austria after the introduction of mass prostate specific antigen screening in Tyrol.12
Our results are in conflict with these for unclear reasons; although the intensity of screening in Tyrol seems similar to that in the Seattle area (about two thirds of men screened at least once over five years), the intensity of treatment in Tyrol and the intensities of both screening and treatment in the rest of Austria were not presented in that report.
Limitations of the study
Our study has several limitations. Firstly, the cohorts included only men aged 65 and over in 1987. Screening and treatment for prostate cancer may have a larger impact on younger men. However, as prostate cancer death is rare before age 70,3
if recent decreases in prostate cancer mortality in the United States are attributable to screening and treatment with surgery and radiation this impact would almost certainly be seen among Medicare age men.
Secondly, during 1987-90 Medicare did not routinely pay for prostate specific antigen screening tests. Claims may not have always been submitted, therefore, and numbers of prostate specific antigen tests may be underestimated. Nevertheless, the relative rates of testing between the two regions should not be biased, and the differences in biopsies and incidence confirm the higher intensity of screening in the Seattle area.
Thirdly, follow up over 11 years may not have been long enough to see a difference in prostate cancer mortality. However, if the decreases in population based prostate cancer mortality noted in both these areas (fig ) are indeed attributable to screening and treatment with surgery or radiation, the time frame should have been sufficient to see an effect. Nevertheless, further follow up is planned.
Fourthly, we used SEER data to capture new diagnoses of prostate cancer, so records of diagnosis or cause of death for cohort members who moved out of their SEER areas after 1990 were unavailable. A differential rate of death from prostate cancer among men who moved after 1990 could introduce bias. However, Medicare enrolment files indicated that less than 1% of cohort members without prostate cancer moved out of their regions before dying. As only about 3% of these men would be expected to die of prostate cancer, the impact of any differential rate of death from prostate cancer should be minimal.
Fifthly, as the primary outcome measure was prostate cancer specific mortality a differential assignment of cause of death among prostate cancer patients in the two regions could bias comparisons. Earlier studies have shown a relatively high agreement (over 90%) between medical records and underlying causes of death as recorded on death certificates.13
Our group has shown a high level of agreement between medical records and the cause of death as listed on the death certificate among patients with prostate cancer in Connecticut—91% for men dying in 1995 and 82% for men dying in 1985—with no change in the coding of cause of death among prostate cancer patients dying in the pre-prostate specific antigen (1985) and prostate specific antigen (1995) eras.14
Moreover, as part of the current study, we replicated this analysis for patients with prostate cancer who died in the Seattle area in 1995; once again, the agreement between death certificates and medical records on cause of death was greater than 90%.15
Sixthly, prostate specific antigen and biopsy data were incomplete for men enrolled in a Medicare health maintenance organisation in the two SEER areas during the early prostate specific antigen era. The men enrolled in health maintenance organisations were included in the primary analyses of treatment and mortality rates. Perhaps men enrolled in health maintenance organisations were not as intensively screened in the Seattle area, which would tend to dilute any mortality benefit attributable to screening. However, excluding men enrolled in health maintenance organisations from the mortality analysis did not affect the results: rate ratio=1.0 (0.9 to 1.1).
Finally, men in the two cohorts may have differed on important risk factors for prostate cancer. Although only randomised trials can yield truly unbiased estimates of the efficacy of screening and treatment on prostate cancer mortality, their external validity might be limited.16
This longitudinal cohort study, which applied uniform selection criteria at “zero time” for determining eligibility and included a broad representation of the population at risk, may provide complementary and important insights into what might be expected at the population level.
Why has mortality decreased?
This study does not shed light on the reasons for the decline in prostate cancer mortality in these two regions. More effective treatment for advanced disease may have improved mortality in both areas; rates of androgen deprivation, a primary treatment for advanced prostate cancer, could not be tracked with the data available. Additionally, changes in lifestyle or environmental factors that may affect the progression of prostate cancer may have had a role in these mortality trends.
More intensive screening for prostate cancer and treatment with surgery or radiation among a cohort of Medicare beneficiaries in the Seattle-Puget Sound area compared with a cohort in Connecticut has not led to significantly lower mortality from prostate cancer over 11 years. Longer follow up will be necessary to fully assess the outcome of this natural experiment. Meanwhile, ongoing randomised trials assessing the effectiveness of screening and treatment for prostate cancer should be supported.