Cancer surveillance series: Interpreting trends in prostate cancer - Part II: Cause of death misclassification and the recent rise and fall in prostate cancer mortality

Background. The rise and fall of prostate cancer mortality correspond close ly to the rise and fall of newly diagnosed cases. To understand this phenom enon, we explored the role that screening, treatment, iatrogenic (i.e., tre atment-induced) deaths, and attribution bias (incorrect labeling of death f rom other causes as death from prostate cancer) have played in recent morta lity trends. Methods. Join point regression is utilized to assess the recen t rise and fall in mortality and the relationship of total U.S. trends to t hose areas served by the National Cancer Institute's Surveillance, Epidemio logy, and End Results (SEER) Cancer Registry Program. Incidence-based morta lity (IBM) is estimated with the use of prostate cancer data from the SEER Program to partition (from overall prostate cancer mortality trends) the co ntribution of cases diagnosed since the widespread use of prostate-specific antigen (PSA) testing starting in 1987, IBM is also used to examine the co ntribution of stage at diagnosis to the recent prostate cancer mortality tr ends. Results, LBR I for cases diagnosed since 1987 rose above the pre-1987 secular (i.e., background) trend, peaked in the early 1990s, and almost re turned to the secular trend by 1994. This rise and fall of IBM track with t he pool of prevalent cases diagnosed within the prior 2 years. IBM for case s diagnosed with metastatic disease fell starting in 1991, while IBM for th ose diagnosed with localized/regional disease was relatively flat, Conclusi ons. The rise and fall in prostate cancer mortality observed since the intr oduction of PSA testing in the general population are consistent with a hyp othesis that a fixed percent of the rising and falling pool of recently dia gnosed patients who die of other causes may be mislabeled as dying of prost ate cancer. The decline in IBM for distant stage disease and flat IBM trend s for localized/regional disease provide some evidence of improved prognosi s for screen-detected cases, although alternative interpretations are possi ble.