Page 46 - An Evidence Review of Active Surveillance in Men With Localized Prostate Cancer

Patterns of Care Study [POCS], or Veterans Administration studies). Nonrandomized
comparative studies also had to use multivariable methods (regression or propensity-score based)
or instrumental variable methods to estimate treatment effects. Operationally, we required
adjustment at least for patient age for all observational studies; when such studies reported on
mixed populations (localized mixed with more advanced disease), we also required adjustment
for at least one marker of disease severity (e.g., disease stage, tumor grade, Gleason score).
For all Key Questions, potentially eligible studies identified through screening titles and
abstracts (see previous section) were retrieved in full text and were reviewed by a single
investigator using the above listed criteria.

Key Question 5 (Research Needs)
We did not perform a separate literature search for this Key Question, but instead reviewed
the evidence for Key Questions 1 to 4 to identify research gaps.

Data Extraction and Summaries

We considered primary research studies of diverse designs, including published systematic
reviews of primary research studies. We list here the information that was extracted from these
types of evidence. The summary Tables are provided in Appendix C.

Primary Research Studies

We extracted bibliographic information, eligibility criteria, enrollment years, study duration,
and the number of patients included in the final analytic sample. We also extracted additional
information from primary research studies considered relevant to each Key Question.

Key Question 1 (Trends in Incidence, Mortality and Features at
Diagnosis)
From each study that provided information on temporal trends we extracted information that
allowed the reconstruction of trends over time in incidence, mortality/survival, patient-, tumor-,
and system-level characteristics at diagnosis. For parsimony, we grouped the extracted
information in 5-year bins covering the time period of interest (1980-2010/11). When a study
reported multiple estimates of the parameters of interest within a single 5-year bin, we only
extracted information for the year closest to the mid-point of the bin (e.g., if a study reported
incidence rate data for all years between 1980 and 1985, we extracted the incidence rates for the
years 1982 only).
From studies reporting statistical tests for change of the parameters of interest over time, we
extracted the following information (when available): the specific method used for statistical
analysis of trend data, estimates of trend statistics, and p-values for changes in parameters of
interest over time.

Key Question 2 (Definitions of Observational Management Strategies)
To describe the definitions of observational management strategies used in published studies
of such strategies, we extracted information on patient- and tumor-level characteristics used as
eligibility criteria, followup or monitoring parameters, or specific triggers for intervention (active
therapy). We also extracted details on the definition of disease progression used in each study.

11