Context Despite widespread use of serum prostate-specific antigen (PSA) testing
to detect prostate cancer, the relative effectiveness of different PSA screening
strategies is unknown.
Objective To compare prostate cancer mortality, PSA testing rates, and biopsy
rates using various PSA screening strategies, including the standard strategy
of annually testing men aged 50 through 75 years.
Design and Setting A Monte-Carlo simulation based on a Markov model was used to simulate
the natural history of prostate cancer using different starting ages, testing
intervals, and PSA thresholds for prostate biopsy. Age-specific PSA levels
and prostate biopsy detection probabilities were determined from population
data and surgical series.
Main Outcome Measures Numbers of prevented prostate cancer deaths, PSA tests, and prostate
biopsies per 1000 men aged 40 through 80 years, compared among 7 different
strategies vs no screening.
Results Compared with annual PSA testing beginning at age 50 years, the strategy
of PSA testing at ages 40 and 45 years followed by biennial testing beginning
at age 50 years was estimated to simultaneously reduce prostate cancer mortality
and number of PSA tests and biopsies performed per 1000 men. Specifically,
compared with no screening, the standard strategy prevents 3.2 deaths, with
an additional 10,500 PSA tests and 600 prostate biopsies, while the earlier
but less frequent strategy prevents 3.3 deaths, with an additional 7500 PSA
tests and 450 prostate biopsies. Strategies that lowered the PSA threshold
for prostate biopsy to below 4.0 ng/mL or strategies that used age-specific
PSA levels were not more efficient than use of a PSA threshold of 4.0 ng/mL.
These 2 findings remained true under all sensitivity analyses performed to
test assumptions of the model.
Conclusion Recognizing that the efficacy of PSA screening is unproved, the standard
strategy of annual PSA screening beginning at age 50 years appears to be less
effective and more resource intensive compared with a strategy that begins
earlier but screens biennially instead of annually.