PROSTATE-SPECIFIC ANTIGEN TESTING IN BLACK AND?
WHITE MEN: AN ANALYSIS OF MEDICARE CLAIMS?
RUTH ETZIONI, KRISTIN M. BERRY, JULIE M. LEGLER, AND PAMELA SHAW
Objectives. To describe the trends in prostate-specific antigen (PSA) use and associated cancer detection
among black and white Medicare beneficiaries older than 65 years during the calendar period from January
1991 through December 1998.
Methods. Medicare claims data were linked with cancer registry data from the Surveillance, Epidemiology
and End Results program of the National Cancer Institute. Data from a 5% random sample of men without
a diagnosis of prostate cancer were combined with data from prostate cancer cases diagnosed during the
calendar period from 1991 to 1998. PSA tests conducted after a diagnosis of prostate cancer were
Results. PSA use has stabilized among white men, reaching an annual rate of 38% by 1995 and remaining
at this level through 1998. The annual rate of use among black men reached 31% by 1998, but was still
increasing at that time. By 1996, at least 80% of tests in both blacks and whites were second or later tests.
By the end of 1996, 35% of white men and 25% of black men were undergoing testing at least biannually
or more frequently. In 1996, 83% of diagnoses in whites and 77% in blacks were preceded by a PSA test.
Conclusions. Older black men lag slightly behind older white men in their use of the PSA test; however,
annual testing rates in blacks have yet to stabilize. In both race groups, an overwhelming majority of
diagnoses are associated with a PSA test, whether for screening or diagnostic purposes. Regular screening
rates in blacks are substantially lower than in whites, but the regular screening rates are relatively low in both
race groups. Should PSA screening prove efficacious, efforts to promote regular use among both black and
white men will likely be needed. UROLOGY 59: 251–255, 2002. © 2002, Elsevier Science Inc.
R. Etzioni’s research is supported in part by grants U01 CA88160
and R29 CA70227.
R. Etzioni is an Associate Member of the Public Health Sciences
Division of the Fred Hutchinson Cancer Research Center.
From the Fred Hutchinson Cancer Research Center, Seattle,
Washington; and Applied Research Branch, Cancer Surveillance
Research Program, Division of Cancer Control and Population
Sciences, National Cancer Institute, Bethesda, Maryland
Reprint requests: Ruth Etzioni, Ph.D., Fred Hutchinson Cancer
Research Center, 1100 Fairview Avenue North, MP-665, P.O.
Box 19024, Seattle, WA 98109-1024
Submitted: June 7, 2001, accepted (with revisions): September
© 2002, ELSEVIER SCIENCE INC.
ALL RIGHTS RESERVED
originally approved for monitoring disease pro
gression after the diagnosis of prostate cancer, the
test was quickly adopted for screening purposes.
By the end of 1994, approximately 53% of whites
and 45% of blacks aged 65 and older in 1988 had
had a test.1
ince its introduction, PSA testing has rapidly
disseminated in the U.S. population.1 Although
Racial disparities in the incidence of, and mortal
ity from, prostate cancer, are well-known.2,3 Black
men are more likely to be diagnosed with the dis
ease and to die of it. A number of surveys have
indicated that black men are also less likely to un
dergo PSA screening.4,5 Barriers to screening
among black men have been explored.6,7
The limitations of survey data concerning pros
tate cancer screening histories have recently been
investigated.8 The study showed that patients’ self-
reports of PSA use were discordant with their med
ical record 29% of the time, with patients tending
to over-report recent use of the test. The primary
reasons advanced for these findings were recall
bias due to memory “telescoping” and lack of
This study compared the use of PSA testing in
blacks and whites using a population-based, ad
ministrative database, which consists of a linkage
between the Surveillance, Epidemiology and End
PII S0090-4295(01)01516-3 251
Results (SEER) Program of the National Cancer
Institute,9 and Medicare claims files from the
Health Care Financing Administration.10 By direct
use of claims data, our goal was to avoid the afore
mentioned biases that can potentially accompany
MATERIAL AND METHODS
The SEER cancer registry collects information on cancer
incidence and survival in 13 geographic areas in the United
States, covering approximately 14% of the U.S. population.9
The Medicare claims data include both inpatient and outpa
tient procedures, as well as physician services. We used data
on individuals aged 65 years and older between 1991 and 1998
inclusive who were entitled to Medicare parts A and B at some
time during this calendar period. Health Maintenance Organi
zation enrollees were excluded.
The linked SEER–Medicare data cover the period from 1991
to 1998. Medicare claims data were available for all SEER cases
diagnosed through 1996, as well as for a random 5% of con
trols without cancer (non-SEER cases) residing in the SEER
registry locations. The HCPCS (CPT-4) codes provided with
the Medicare claims data were used to identify PSA tests (CPT
codes 86316 and 84153), needle biopsy (CPT codes 55700
55705), ultrasonography (CPT codes 76942-76943) and
transrectal ultrasonography (CPT codes 76872 and 76991).
Because of the SEER–Medicare linkage, we were able distin
guish prediagnostic from postdiagnostic PSA tests through
Annual Testing Rates. To estimate the annual testing rates,
we weighted controls by a factor of 20 to account for the
relative fractions of prostate cancer cases and controls in the
data. The denominators for the testing rates consisted of all
men alive and without prostate cancer at the start of the year.
The numerators consisted of all men with at least one PSA test
conducted during the year.
To avoid counting follow-up tests conducted in response to
a suspicious result, we classified tests as either initiating or
follow-up tests. We defined an initiating test as either a first
occurrence of a test in the database or a test conducted at least
3 months after a preceding test. A follow-up test was defined as
a test conducted within 3 months of an initiating test. We
defined a single testing episode as consisting of an initiating
test and all associated follow-up tests. To evaluate the sensi
tivity of results to the choice of interval used to define a testing
episode, analyses were also conducted based on a 6-month
Cancer Detection Rates. Because the test results were not
available, we defined PSA-detected cases as those cases diag
nosed within 3 months of a PSA test. Life table methods were
used to account for a loss to follow-up during the 3 months
after a test with censoring owing to death, the end of the
surveillance period, or repeated PSA testing within 3 months.
Relative Frequency of First Versus Second or Later Tests. The
relative frequency of first versus second or later initiating tests
was of interest because this allowed us to estimate the propor
tion of men adopting the technology versus returning for se
rial screening. However, it was not possible to identify which
tests were first tests because our follow-up period began in
1991 (or later, for men turning 65 thereafter). To estimate the
relative frequency of first tests, we used the observation1,11
that the cancer detection rate for the first tests would be sig
nificantly greater than for later tests. Thus, let p be the propor
tion of first tests among all tests performed during the year.
Then, CDR, the cancer detection rate observed in a given year
is equal to pCDR1 � (1 � p)CDR2, where CDR1 and CDR2 are
the cancer detection rates for the first and subsequent tests,
respectively. The CDR is available from our data for calendar
years 1991 through 1996. We used CDR1 values from Legler et
al.1 and CDR2 values based on the second or later tests in our
Patterns of Care After a Test. To summarize the patterns of
care after a PSA test, we estimated the cumulative incidence12
of prostate ultrasonography, biopsy, and repeated PSA testing
within 90 days after a test. The cumulative incidence is pref
erable to the Kaplan-Meier estimator13 in the presence of com
peting risks (ie, other events that may prevent the event of
interest but are not independent of this event). We were inter
ested in the first follow-up procedure conducted after a test.
For any given follow-up procedure, all other candidate proce
dures constituted competing risks. In the case of prostate bi
opsy, for example, repeated PSA testing, ultrasonography, and
death were competing risks.
Frequency of Regular Screening. Thus far, we have addressed
patterns of PSA use from a cross-sectional point of view. How
ever, the data also provided longitudinal information on test
ing histories that can be used to estimate the frequency of
regular screening. We first estimated the frequency of regular
screening among men alive and without a prior prostate can
cer diagnosis in December 1996. We then examined the
screening histories before this time. We considered several
definitions of regular screening, including annual and bian
To qualify as receiving “strictly annual” screening, an indi
vidual had to have received at least one PSA test in 1994, 1995,
and 1996. For “almost annual” screening, an individual had to
have been tested in at least 2 of the past 3 years. For “strictly
biannual” screening, an individual had to have received at
least one PSA test in both 1994 and 1996 or in both 1993 and
1995. We also considered a “modified biannual screening”
definition, which defined regular screening as a test in either
the current year or the prior year, and a test in at least 1 of the
2 years before that. In all these analyses, we used the testing
episode definition. The analyses were repeated for men alive at
the end of 1993 (annual analysis only), 1994, and 1995. This
allowed us to evaluate whether the frequency of regular
screening was changing over time.
ANNUAL TESTING RATES
Figure 1 displays the overall PSA testing frequen
cies by age group, race, and calendar year. Figures
1A and B show the testing rates by 5-year age
groups. By 1998, the annual testing rates were ap
proximately 20% higher for whites than for blacks,
with the exception of men older than 80 years, for
whom the testing rates appeared to be comparable
across race groups. Figures 1C and D put these
results in historical perspective, showing data for
all age groups combined together with the results
from Legler et al.1
CANCER DETECTION RATES
The cancer detection rates declined with time,
but the decline after 1991 was far less extreme than
that observed during the initial few years of PSA
dissemination1 (Table I). Between 1991 and 1996,
the cancer detection rates declined from 4.5% for
UROLOGY 59 (2), 2002
FIGURE 1. Overall PSA testing frequencies by age, race, and calendar year, showing the number of men with at
least one PSA test during a given year, divided by the number of men alive and without a prior prostate cancer
diagnosis at the start of the year. (A) Whites 1991 to 1998. (B) Blacks 1991 to 1998. (C,D) All age groups
combined, together with comparable results from Legler et al.1 through 1994 (previous analysis is analysis of Legler
TABLE I. Estimates of overall CDRs, CDRs associated with first testing episodes,* and CDRs
associated with second or later testing episodes
KEY: CDR � cancer detection rate; CDR1 � CDR associated with first testing; CDR2 � CDR associated with second or later testing.
A 3-month testing episode definition is used.
* Data from Legler et al.1
† Values after 1993 set equal to 1993 values.
‡ Frequency of first relative to second or later testing episodes by race and calendar year.
whites to approximately 1%, and from 6.2% for
blacks to approximately 2%.
second or later episodes. The estimates for the
6-month testing episode definition were similar
but higher; 10% and 13% of testing episodes in
1996 were first episodes for blacks and whites, re-
spectively. We also computed empirical estimates
of the relative frequency of first versus second or
RELATIVE FREQUENCY OF FIRST TESTS
Table I shows that by 1996, the vast majority of
testing episodes in both blacks and whites were
UROLOGY 59 (2), 2002
FIGURE 2. Patterns of care after a test by race and calendar year. Cumulative incidence of specific follow-up
procedures within 90 days of an initiating PSA test (3-month definition). For a given procedure, death due to other
causes, prostate cancer diagnosis, and performance of alternative follow-up procedures acted as competing risks.
(A) Whites. (B) Blacks.
later tests for the years 1994 through 1996. Because
this calculation was based on testing histories that
go back only as far as 1991, the results represent
upper bounds on the relative frequency of first
tests. Among whites, the upper bound was 22.4%,
14.4%, and 10.7% for the years 1994, 1995, and
1996, respectively; for blacks, the corresponding
figures were 28.8%, 21.2%, and 17.5%.
PATTERNS OF CARE AFTER A TEST
Figure 2 shows the frequency of various fol
low-up procedures conducted within 90 days of a
PSA test by race and calendar year. Among the
follow-up procedures conducted, the frequency of
follow-up PSA testing appeared to increase with
time relative to prostate biopsy. This result ap
peared to hold for both blacks and whites and sug
gests that either the frequency of an abnormally
high PSA value is declining or the response to an
abnormal result may be tending toward less inva
FREQUENCY OF REGULAR TESTING
Among men alive at the end of 1996, approxi
mately 35% of whites and 25% of blacks were fol
lowing strict biannual testing, up from 25% and
18% in 1994. For annual testing the results were
17% and 10% in 1996 versus 8% and 5% in 1994
for whites and blacks, respectively. For the more
relaxed definitions, the rates of use were slightly
higher than the corresponding “strict” definitions.
Many of the men tested in 1996 had had a prior test
but did not qualify as being on a biannual or more
frequent screening schedule.
This analysis of PSA testing among Medicare re
cipients provides a comprehensive picture of the
use of the test among elderly black and white men
in the United States through 1998. Annual testing
frequencies have reached a steady state among
whites and are leveling off among blacks. Approx
imately one third of whites and one fourth of
blacks at risk of a prostate cancer diagnosis appear
to be following a regular screening regimen. The
trends in patterns of use among whites are similar
to those among blacks, although the diffusion of
the technology among blacks has been lower and
slower, with the annual frequency of use by 1998
about 15% to 20% lower than in whites.
Our results are consistent with prior surveys of
screening behavior in black and white men4 6 in
that they indicate that black men are less likely to
use the PSA test. However, the actual testing rates
estimated from the SEER–Medicare data indicate
that the discrepancies in frequency of use between
the two racial groups are lower than might be ex
pected given the survey results. We found that, by
1998, approximately 38% of whites and 31% of
blacks were being tested at least once a year, cor
responding to an odds ratio of 73% for testing in
blacks versus testing in whites. In contrast, a recent
analysis of responses to the New York Behavioral
Risk Factor Surveillance System (BRFSS) in 1994
and 1995 estimated an odds ratio of 30% for testing
in blacks versus testing in whites.5 One possible
reason for this difference, in addition to our data
source being an administrative claims database
rather than a survey, is that our results indicate
trends across all SEER areas combined, rather than
within a single area.
The amount of heterogeneity across the SEER
areas in the dissemination of PSA testing is sub
stantial. This was noted by Legler et al.,1 who
showed a correspondence between prostate cancer
incidence patterns and rates of first-time PSA test-
UROLOGY 59 (2), 2002
ing in two SEER areas with quite different PSA use
trends. We repeated our key analyses by SEER area
and found striking differences in testing patterns.
For instance, annual PSA testing rates among
whites were almost uniformly highest in Atlanta,
Detroit, and Los Angeles and lowest in Connecti
cut, Utah, and Iowa, with up to a 50% discrepancy
between the highest and lowest rates in some years.
Virtually all areas showed increasing rates of use
with time, with the exception of Seattle, where use
was virtually flat, following earlier, rapid adoption
of the test.1
A major limitation of the use of claims data to
analyze testing patterns is the lack of information
on the reasons for conducting the test. Although
most SEER cases in our database had had a PSA test
within 3 months before diagnosis, it is not clear
how many tests were true screening tests con
ducted in asymptomatic individuals. This problem
severely complicates attempts to draw inferences
about the effects of PSA screening on the outcomes
of interest like disease mortality. The vital impor
tance of this information argues for the expansion
of surveillance systems to include it at the time of
Our estimates of the frequency of regular screen
ing are relatively high for a screening technology
that has not been proved to save lives and that has
been associated with significant treatment morbid
ity. However, the implication is that more than
60% of white men and 70% of black men alive and
at risk of a prostate cancer diagnosis are not being
tested regularly or at least biannually. Should reg
ular PSA screening prove to be efficacious in the
randomized studies that are under way,14,15 its reg
ular use among older men will need to be pro
ACKNOWLEDGMENT. To Nicki Schussler for assistance with
programming and the development of the SEER-Medicare
data set; and to Holly Hoegh and Marta Induni from the Can
cer Surveillance Section of the California Department of
Health Services for providing us with the California BRFSS
1. Legler J, Feuer E, Potosky A, et al: The role of prostate-
specific antigen (PSA) testing patterns in the recent prostate
cancer incidence decline in the USA. Cancer Causes Control 9:
2. Hankey BF, Feuer EJ, Clegg LX, et al: Interpreting
trends in prostate cancer. Part I. Evidence of the effects of
screening in recent prostate cancer incidence, survival and
mortality rates. J Natl Cancer Inst 91: 1017–1024, 1999.
3. Stanford JL, Stephenson RA, Coyle LM, et al: Prostate
Cancer Trends 1973–1995. NIH Publication No. 99-4543. Be
thesda, National Cancer Institute (SEER Program), 1999.
4. Demark-Wahnefried W, Strigo T, Catoe K, et al: Knowl
edge, beliefs and prior screening behavior among blacks and
whites reporting for prostate cancer screening. Urology 46:
5. Steele CB, Miller DS, Maylahn CM, et al: Knowledge,
attitudes and screening practices among older men regarding
prostate cancer. Am J Public Health 90: 1595–1600, 2000.
6. Myers RE: African American men, prostate cancer early
detection examination uses, and informed decision-making.
Semin Oncol 26: 375–381, 1999.
7. Shelton P, Weinrich S, and Reynolds WA: Barriers to
prostate cancer screening in African American men. J Natl
Black Nurses Assoc 10: 14–28, 1999.
8. Jordan TR, Price JH, King KA, et al: The validity of male
patients’ self-reports regarding prostate cancer screening. Prev
Med 28: 297–303, 1999.
9. National Cancer Institute: SEER Program home page.
Available at: http://seer.cancer.gov/. Accessed January 4, 2002.
10. Potosky AL, Riley GF, Lubitz JD, et al: Potential for
cancer related health services research using a linked Medi
care-tumor registry database. Med Care 31: 732–748, 1993.
11. Catalona WJ, Smith DS, Ratliff TL, et al: Detection of
organ-confined prostate cancer is increased through prostate-
specific antigen-based screening. JAMA 270: 948–954, 1993.
12. Pepe MS, and Mori M: Kaplan-Meier, marginal or con
ditional probability curves in summarizing competing risks
failure time data? Stat Med 12: 737–751, 1993.
13. Kaplan EL, and Meier P: Nonparametric estimation
from incomplete observations. J Am Stat Assoc 53: 457–481,
14. Gohagan JK, Prorok PC, Kramer BS, et al: Prostate can
cer screening in the prostate, lung, colorectal and ovarian can
cer screening trial of the National Cancer Institute. J Urol
152(5 Pt 2): 1905–1909, 1994.
15. Beemsterboer PM, de Koning HJ, Kranse R, et al: Pros
tate specific antigen testing and digital rectal examination be
fore and during a randomized trial of screening for prostate
cancer: European randomized study of screening for prostate
cancer, Rotterdam. J Urol 164: 1216–1220, 2000.
UROLOGY 59 (2), 2002