Page 143 - An Evidence Review of Active Surveillance in Men With Localized Prostate Cancer

defined a priori and undergo minimal change over time or between centers. A second-tier, but
more practical option, is to allow for a range of a priori AS protocols across centers, also
allowing for modifications over time; however, the AS protocols should be broadly similar to
each other and any changes in protocols should be systematically tracked and adjusted for in
analyses. The determination of which patients are potentially eligible for AS should also be made
a priori. Only these patients (whether they ultimately received AS or another treatment) should
be analyzed. To be interpretable, these studies will need to use multivariable analyses, propensity
scores, or other validated methods (e.g., instrumental variable regression) to adjust for the broad
range of factors that affect the decision to use AS. These include, but are not limited to disease
factors (e.g., stage and grade); disease markers (e.g., PSA and imaging); patient demographics,
psychometrics, personality traits, and personal relationships; clinic features and setting; primary
care physician factors; and treating physician factors. We do not believe that retrospective
studies (without a priori definitions of AS, eligibility criteria, or choice of variables of interest)
are capable of having adequate data for unbiased analyses, because patient and tumor
characteristics are strongly associated with initial treatment choice as well as outcomes (i.e., they
are strong confounders of the treatment-outcome association).
Subgroup analyses of either the trials or the prospective comparative studies should be
conducted to look for particular sets of men who may benefit most (or least) from one approach
or the other. Preferably, these subgroups should be considered a priori to allow studies to be
adequately powered for these subgroup analyses, to minimize bias, and to constrain type I error
(false-positive findings). The factors listed in Key Questions 1 and 2 form a good starting point
to consider which subgroups may be of interest. In addition, future studies that could uncover
better bio- and imaging markers of indolent versus aggressive disease, and thus could better
stage patients as having either low or high risk disease, are necessary to better inform which
patients are most likely to benefit from observational versus active treatment.

89