Page 62 - An Evidence Review of Active Surveillance in Men With Localized Prostate Cancer

and race-adjusted biopsy rate did not statistically significantly change over the study 9-year
89
period (trend P > 0.2).

70
Number of cores. A single primary study provided information on the number of biopsy cores
obtained during the investigation of suspected prostate cancer cases (Appendix Table C1.11).
The study utilized the CaPSURE database and covered 1997 to 2002. It demonstrated a
significant increase in the mean number of cores examined per patient (from 7.5 in 1997 to 9.8 in
2002). We note that the study excluded patients who were evaluated with less than six cores,
which may have led to underestimation of the change in the number of cores obtained. However,
the study found that the increase in the number of cores over time was significant (+0.41 cores
per patient per year, P <0.001).

Histopathologic grading changes. In a study of prostate cancer patients from the Connecticut
Tumor Registry (1990-92), investigators obtained medical records, pathology reports, and the
original slides used for pathological examination for 1858 (49 percent) of the patients diagnosed
97
during the study period. A single pathologist (blinded to the originally assigned Gleason score)
regraded all slides (2002-04). The contemporary reading of the slides resulted in the assignment
of significantly higher scores compared to the original readings (mean score increase from 5.95
to 6.8; P < 0.001). The study also demonstrated that this reclassification causes an increase in the
Gleason-score adjusted prostate-cancer-specific survival; even in the absence of changes in
treatment efficacy or tumor biology (since the same patient histories were used such changes
cannot explain differences in survival patterns). This observation is often referred to as the
98
b
“Will-Rogers” phenomenon.
We also identified a structured review on the same topic through additional targeted searches
99
(Appendix Table C1.12). None of the studies included in this review (other than the one
discussed above) fulfilled our inclusion criteria.
System Characteristics
Differences in geographical access and other system-level factors. Four studies (2 CaPSURE,
1 POCS, 1 NCDB), covering 1986 to 2003, reported information on changes in the distribution
of patients by system-level factors (Appendix Table C1.13). 65,71,74,77 Three studies (2 CaPSURE,
1 POCS) provided information on trends in the distribution of patients’ insurance status at
diagnosis. 65,71,77 The two studies utilizing CaPSURE data demonstrated a decrease in the
proportion of patients with Medicare coverage at the time of diagnosis over the time periods
covered (1997-2003 and 1989-2001). 65,71 In contrast, the study using POCS data did not
77
demonstrate a change in the distribution of insurance status over time (1998-2002).
One study, using the CaPSURE database reported on trends in the distribution of settings
65
(community versus academic) and geographic regions over time. Comparing 1997-2001 to
1989-97, there was an increase in the number of patients seen in academic settings (compared to
community settings) and an increase in the number of patients originating from Midwestern

b The Will-Rogers phenomenon arises when a member who is in the bottom half of a group with a high average
outcome is reclassified as a member who is in the top half of a group with a lower average outcome, resulting in
increases of the average outcome in both groups. In the prostate cancer case, the phenomenon occurs when the same
biopsy sample receives a different Gleason score if assessed using different scoring criteria at different time points
(for example, in the 1990s versus now). When members of the low risk group with the least favorable histology are
reclassified into the high-risk group (which on average had worse prognosis than the reclassified members), this will
inflate grade-adjusted survival over time for both groups.

27