Page 145 - An Evidence Review of Active Surveillance in Men With Localized Prostate Cancer

baseline anxiety and higher socioeconomic status were associated with decreased probability of
willingness to consent to randomization for AS versus definitive treatment (i.e., these men did
not take a chance and proactively selected AS). The rest of the heterogeneous studies reported on
men who did not receive treatments or initial treatments. Therefore, whether they were on AS or
WW could not be readily discerned. The following patient and clinical variables are potentially
important in increasing the probability that a patient receives WW or AS: older age, presence of
comorbidities, lower Gleason score, lower tumor stage, lower diagnostic PSA, lower risk groups,
or decreased baseline anxiety. The following patient and clinical variables are potentially
important in increasing the probability that a patient interrupts WW or AS to seek definitive
treatments: younger age, higher tumor stage, higher diagnostic PSA, higher PSA velocity, higher
risk groups, or increased anxiety.
As most of these tentative conclusions are drawn from multivariable analyses of large
databases that did not specifically address the factors that affect the offer, acceptance, and
adherence of AS, whether different treatment options were offered to the patients, whether they
accepted those options, and whether they adhered to their initial choices could only be inferred
from whether they received the treatments or not. In addition, retrospective studies (without
a priori definitions of AS, eligibility criteria, or choice of variables of interest) could not provide
adequate data for unbiased analyses, because patient characteristics are strongly associated with
initial treatment choice.
No trial has published results on comparisons of AS with RP, or RT in men with localized
diseases. One trial reported that men on RP had lower mortality than men on WW; one trial
reported that there was no difference in mortality comparing men in RP with men in WW.
Retrospective studies suggest that men on conservative management had a higher prostate
cancer-specific mortality than men treated with RP. Men who had RP had more urinary
complications than men on WW. Retrospective studies also reported that men treated with RT
had lower mortality than men on WW. They also reported higher rates of urinary strictures in
men treated with RT compared with men on WW. It should be noted that confounding is likely
in many retrospective analyses of large databases. The following example is instructive in
illustrating the potential for confounding bias in observational studies of treatment effectiveness.
Giordano 2008 248 replicated a previously published analysis 208 comparing overall survival
between men who received active treatment (RP or RT) with men who were on WW, and
performed additional analyses on mortality from non-prostate-specific causes (e.g., heart disease,
other cancers, and chronic obstructive pulmonary disease). The study confirmed that patients
with prostate cancer who underwent RP had better survival than those on WW but also suggested
that patients treated with RP had improved survival compared to a randomly selected control
population without cancer (matched with the RP population on the distributions of year of
diagnosis and age at diagnosis). Further analysis showed that the patients who underwent RP had
significantly improved their survival for many non-prostate-specific causes of death (including
diabetes-, cardiovascular disease-, and chronic obstructive pulmonary disease-related death) as
well, compared to those who had observational management. Because these causes of death are
unlikely to be affected by prostate surgery, but conditions related to them (for example chronic
angina or low exercise tolerance) are in fact used by urologists to decide fitness for surgery, the
results suggest that the findings of improved overall survival in the RP group compared to the
WW group reported by the previous analysis may have been affected by residual confounding by
covariates that were not measured or not controlled for in the analysis.

91