Sawyer R, Berman JJ, Borkowski A, Moore GW. Elevated prostate-specific antigen levels in black men and white men. Modern Pathology 9:1029-1032, 1996
ELEVATED PSA (PROSTATE SPECIFIC ANTIGEN) LEVELS IN AFRICAN-AMERICAN AND WHITE MEN.
Robert Sawyer, M.D., Jules J. Berman, Ph.D., M.D., Andrew Borkowski, M.D., G. William Moore, M.D., Ph.D.
Departments of Surgery and of Pathology, Veterans Affairs Maryland Health Care System; University of Maryland School of Medicine; and The Johns Hopkins Medical Institutions, Baltimore, Maryland
The usual ranges for PSA (Prostate Specific Antigen) are derived from a community-based population of white men, but are employed for screening on all men, on the assumption that different PSA levels among different ethnic groups would be small or would have no clinical significance. However, recently published reports suggest that PSA levels in a racial population may vary directly with the relative risk of prostate cancer within the population. PSA ranges were determined in African-American and white men registered in the Veterans Affairs Maryland Health Care System. The total patient census of 122,602, has near-equal numbers of African-American and white men, and maintains records of race designation for inpatients. Among patients with no known prostate cancer, there were 10,808 patients age 40 or older, and 19,482 PSA tests. Among these patients, there were 3,274 patients identified as African-American, 2,993 patients identified as white not of Hispanic origin, and 4,541 patients of other race or race unknown. The 95th percentile PSA value in African-American men and white men, respectively, for ages 40-49 is 2.80 and 2.01; for ages 50-59 is 5.40 and 4.19; for ages 60-69 is 9.59 and 7.00; for ages 70-79 is 15.45 and 9.40; and for ages 80 and over is 21.05 and 18.25. At every age group, African-American men had a higher range (for the 95th percentile) than white men. The largest difference is found in men aged 70-79, where the ratio of the upper limit of PSA for African-American men compared to white men is 1.6.
key words: PSA, prostate-specific antigen, ethnic, race, black, African-American, screening, range
PSA is an increasingly popular screening test for prostatic cancer. Recently, hospitals have reported a many-fold increase in the use of PSA testing, with a subsequent increase in newly diagnosed prostate cancers (1). African-American men have one of the highest prostate cancer rates worldwide and have a 50% higher age-adjusted incidence of prostate cancer than white men (2). Asian men have among the lowest rates of prostate cancer, and there may be as much as a 120-fold difference between the rate of prostate cancer in the lowest risk group (Shanghai, China) and the highest risk group (African-American men in San Francisco) (2).
Today, the clinically accepted ranges of PSA levels were adapted from data collected from a community-based Minnesota population consisting entirely of white men (3). In discussing the importance of age-specific ranges, the Mayo Clinic group acknowledged the lack of current information regarding age-specific PSA ranges for African-Americans, Asians and Hispanics (4). A study of PSA levels in Asian men found a lower range of PSA levels in this population (5), suggesting that prostate cancer risk may correlate with PSA levels in the population. A recent letter by our group reported that African-American men indeed have higher ranges of PSA than do white Americans (6). This finding was corroborated in an article published by Morgan et al. (7). The present paper details the methods used to arrive at ranges for African-American and white men using a large database.
MATERIALS AND METHODS
The patient population at the Veterans Affairs Maryland Health Care System contains near-equal populations of African-American and white patients. The Decentralized Hospital Computer Program retains data and supports queries relating to all recorded demographic data, and laboratory tests for a patient population exceeding 120,000, covering 3 hospitals and a network of outpatient clinics. The PSA levels of African-American and white men were studied to determine the age-based PSA ranges for this population. Data were collected using the VA File Manager, version 22, used by nearly all 172 Veterans Affairs medical centers nationwide. Centralized registration of patients with unique identification numbers insures that patients are not entered redundantly, and that all patients have database registration. Laboratory information (including any and all PSA tests results and the date of test) is linked with the patient identifier and stored in perpetuity. Patient age at the time of a PSA test is computed as the difference between the test date and the listed date of birth. Race/ethnicity is recorded according to a classification established by the United States Public Health Service, in which all patients are assigned one of six categories, as follows: 1) Black, not of Hispanic origin; 2) White, not of Hispanic origin; 3) Hispanic; 4) Asian or Pacific Islander; 5) American Indian or Alaskan Native; 6) Other. There were 12,548 patients who received one or more PSA determinations (Hybritech) between October 1, 1989, and January 23, 1996, as part of their routine medical care. Among these, 11,125 patients had no diagnosis of prostate cancer, and 10,808 patients were at least 40 years old. Patients had received between one PSA test (6,538 patients) and 19 PSA tests (1 patient), as shown in Table 1, for a total of 19,482 tests. In order not to bias the final results in favor of patients with many PSA determinations, each PSA test was weighted 1/n if the same patient had n PSA tests after age 40. Thus, in a patient receiving one PSA test at age 48 and another PSA test at age 52, each test is weighted as a half-patient. Since some patients had PSA tests taken in consecutive decades of life in our database, this weighting procedure resulted in 'fractional patients' in some decades. For each race (African-American and white) and each decade of life over 40 years old, MEAN PSA VALUE was calculated as the sum of all PSA values, divided by the number of PSA values (appropriately weighted for fractional patients). The MEDIAN PSA VALUE (same as 50th percentile PSA value) was calculated as that value with half the observed PSA values below the median and half the observed PSA values above the median. Similarly, the 90th PERCENTILE PSA VALUE was calculated as that value with 90% of observed PSA values below the 90th percentile and 10% of observed PSA values above the 90th percentile; and similarly for the 95th PERCENTILE PSA VALUE. Because of the large number of observed data points (circa 3,000 for either race), medians and percentiles were calculated directly rather than through estimates on a regression, as used by the Mayo group (4).
On January 23, 1996, the cumulative patient census at the Veterans Affairs Maryland Health Care System was 122,602. Among these subjects, 10,808 had received one or more PSA determinations between October 1, 1989, and January 23, 1996, had no diagnosis of prostate cancer, and were age 40 or older. Among these patients, there were 3,274 patients identified as African-American, 2,993 patients identified as white not of Hispanic origin, and 4,541 patients of other race or race unknown (Table 2). The PSA ranges for these patients are listed in Table 3. For every age group studied, African-American men had equal or higher mean, median, and 90th and 95th percentile PSA tests than did white men. Our study showed 95th percentile PSA value in African-American men and white men, respectively, for ages 40-49 is 2.80 and 2.01; for ages 50-59 is 5.40 and 4.19; for ages 60-69 is 9.59 and 7.00; for ages 70-79 is 15.45 and 9.40; and for ages 80 and over is 21.05 and 18.25 (Figure 1). The upper limit for PSA is defined as the 95th percentile for each age range. In the study by the Mayo Clinic group, this number was calculated for the midpoint of the age range from a regression analysis of their data points within each age range. This statistical representation of the 95th percentile was necessary because of the small population size under study. For example, the 95th percentile group for persons in the 70-79 age group contained only 5 subjects in the Mayo Clinic study, as compared to 788.4 African-Americans and 993.23 white non-Hispanics in the present study. Thus in the current study, there are over 6,000 subjects with known race, permitting a direct determination of the 95th percentile score from the data set.
All the data employed in this study were available from a single source, namely, the Decentralized Hospital Computer Program in the Veterans Affairs Maryland Health Care System. The final patient population was obtained after subtracting away patients with no PSA tests, patients with no race identifier and patients with known prostatic cancer, all data contained within the DHCP. It included patients whose PSA was obtained as part of screening (based on age), as well as patients whose PSA was obtained as part of the clinical work-up of symptoms relating to non-neoplastic urologic conditions. The data obtained are relevant to patients received in a medical system that includes three hospitals and an extensive outpatient clinic system serving a population of military service veterans. As such, our data may not be representative of other patient populations with a different range and prevalence of individuals with urologic complaints. Large, well-maintained patient databases insure that all patients are included uniquely, that all laboratory tests, including surgical pathology, cytology, and autopsy reports (in the case of deaths) are stored and linked to the patient identifier, and in that basic demographics, including race, are available. Such databases are a tremendous resource to the medical community and can be used for epidemiologic studies (8). In studies that collect PSA levels in a population extending over a period of time, single patients accumulate multiple PSA values (Table 1). This necessitates a reasonable method for scoring PSA values in a percentile calculation. This task is further complicated by the need to stratify PSA values by age, as a single patient might have had PSA values sampled in different decades. A recent study of PSA levels in a population published by Morgan et al. ignored all PSA tests other than the first test performed on each patient (7), resulting in the removal of about 30,000 PSA values, representing about 2/3 of their collected data. In this study, we retained each data entry by giving each PSA test a weight of 1/n, where n is the number of PSA tests for the patient. Thus, in a patient receiving one PSA test in one decade of life and another PSA test in the subsequent decade, each test is weighted as half a patient. For patients with PSA tests taken in consecutive decades of life in our database, this weighting procedure resulted in 'fractional patients' in some decades (Table 2). This weighting technique has the advantage of using all the data, and avoids the bias caused by excluding a test result from analysis. Several methods have been proposed for determining the upper limit of a PSA value, i.e., the value at which an immediate clinical action should be considered. A recent abstract that addressed the issue of PSA values based on race examined a large number of cases culled from over 250 centers. Their study reported whites had a higher mean PSA value than did African-Americans, a result that would seem to contradict our findings (9). To determine a useful range of test values for a population, percentile-statistics are preferable to the mean-statistic, since the latter is easily skewed by a few, large outlier values. For example in Table 3, the high mean value for African-Americans at least 80 years old (623.81) is due to a single PSA value over 100,000 ng/ml. In all likelihood, this patient has an undiagnosed prostate cancer. In contrast, the percentile values for the same group of patients are only slightly skewed (circa 1 ng/ml) by this single outlier value. Traditionally the 95th percentile value is used as the upper limit in many clinical laboratory tests, i.e., tolerance of 5% false positive values. In the current study, there are circa 3,000 subjects in either racial group, permitting a direct determination of the 95th percentile score from the data set. The current study demonstrates that African-Americans have a higher range of PSA levels than do white Americans, and has been corroborated by data from another large study involving patients registered at Walter Reed Army Medical Center (7). The biological and clinical interpretation of these observations is unknown at present. A higher range of PSA levels in the African-American population might indicate a high number of undiagnosed prostate cancers among this group, prompting close surveillance of African-American men with PSA levels greater than 4.0 ng/ml (the PSA level currently used to prompt clinical action). Alternately, the higher PSA levels in African-American men might indicate that African-American men without prostate cancer may have high PSA levels, possibly due to a higher incidence of prostatic hyperplasia, increased PSA "leakiness", increased incidence of prostatic infarction, or any non-neoplastic condition that might elevate PSA spuriously. If this were the case, a higher screening threshold should be used for African-Americans, since it can be assumed that many disease-free African-American men have high PSA levels. The correct clinical interpretation of the data presented in this study lies in knowing the number of men in the disease-free group who in fact have undiagnosed cancer. No study to date has addressed this question, and the proper clinical response to a high PSA level in an African-American man is currently a matter of judgment, not experimental evidence.
We are grateful to Dr. Robert Burger, VA Surgery Service, who initiated the project by requesting that Pathology Service provide the PSA ranges for African-American men.
1. Berman JJ, Alonsazana, Brown L, Moore GW. PSA screening for prostate cancer: lack of reduction in Gleason scores. Modern Pathology 7:487-489, 1994
2. Meilke, A.W., and Smith, J.A.: Epidemiology of prostate cancer. Urol. Clin. North Am. 17:709-718, 1990.
3. Oesterling JE, Jacobsen ST, Chute CG, Guess HA, Girman CJ, Panser LA, Lieber MM. Serum prostate-specific antigen in a community-based population of healthy men. JAMA 270:860-864, 1993
4. Oesterling JD, Jacobsen SJ, Cooner WH. The use of age-specific reference ranges for serum prostate specific antigen in men 60 years old or older. J Urol 153:1160-1163, 1995
5. Oesterling JE, Kumamoto Y, Tsukamoto T, Girman CJ, Guess HA, Masumori N, Jacobsen SJ, Lieber MM. Serum prostate specific antigen in a community-based population of healthy Japanese men: lower values than for similarly aged white men. Br J Urol 75:347-352, 1995
6. Sawyer R, Berman JJ, Borkowski A, Moore GW. Prostate-specific antigen in black men. Lancet (letter) 347:1329 (inclusive), 1996.
7. Morgan TO, Jacobsen SJ, McCarthy WF, Jacobson DJ, McLeod DG, Moul JW. Age-specific reference ranges for prostate-specific antigen (PSA) in black men. New Engl J Med 1996;335:304-10.
8. Berman JJ, Moore GW. SNOMED-Encoded surgical pathology databases: a tool for epidemiologic investigation. Mod Pathol, in press.
9. DeAntoni EP, Crawford DE, Ross CA, Hirano D, Blum DS, Berger ER, Stone NN, Eisenberger MA, Gambert SR, McLeod DG, Staggers F. Age and race-specific reference ranges for prostate-specific antigen from a large, community-based study (abstract). Proc Amer Urol Assoc 155:374A (inclusive), 1996
TABLE 1. DISTRIBUTION OF MULTIPLE PSA TESTS PER PATIENT, WITH NO DIAGNOSIS OF PROSTATE CANCER. Number of Number of Cumulative Cumulative PSA Tests Patients Patients Tests 1 ............ 6,538 ...... 6,538 ...... 6,538 2 ............ 2,156 ...... 8,694 ...... 10,850 3 ............ 1,081 ...... 9,775 ...... 14,093 4 ............ 525 ...... 10,300 ...... 16,193 5 ............ 214 ...... 10,514 ...... 17,263 6 ............ 121 ...... 10,635 ...... 17,989 7 ............ 67 ...... 10,702 ...... 18,458 8 ............ 39 ...... 10,741 ...... 18,770 9 ............ 23 ...... 10,764 ...... 18,977 10 ............ 20 ...... 10,784 ...... 19,177 11 ............ 10 ...... 10,794 ...... 19,287 12 ............ 6 ...... 10,800 ...... 19,359 13 ............ 2 ...... 10,802 ...... 19,385 14 ............ 2 ...... 10,804 ...... 19,413 15 ............ 1 ...... 10,805 ...... 19,428 16 ............ 0 ...... 10,805 ...... 19,428 17 ............ 1 ...... 10,806 ...... 19,445 18 ............ 1 ...... 10,807 ...... 19,463 19 ............ 1 ...... 10,808 ...... 19,482 TABLE 2. PATIENT DEMOGRAPHICS FOR THE STUDY SET OF 10,808 PATIENTS Total number of patients in Baltimore VAMHCS database.... 122,602 Patients, at least one PSA test, no prostate cancer...... 11,125 Patients, at least one PSA test, no prost ca, >40 ....... 10,808 African-American ..................................... 3,274 White ................................................ 2,993 Other race or race not stated ........................ 4,541 African-American White Men Men Age 40-49.................. 614.83 ......... 283.50 Age 50-59.................. 504.30 ......... 380.68 Age 60-69.................. 1250.31 ......... 1182.28 Age 70-79.................. 788.40 ......... 993.23 Age >80.................... 116.17 ......... 153.31 Total .................... 3274 ......... 2993 TABLE 3. PSA LEVELS IN AFRICAN-AMERICAN AND WHITE MEN Mean PSA Median PSA 90%ile 95%ile (50%ile) Ages 40-49 black 1.21 1.00 1.72 2.80 white 1.05 1.00 1.70 2.01 ratio 1.2 1.0 1.0 1.4 Ages 50-59 black 1.84 1.00 3.05 5.40 white 1.79 1.00 3.05 4.19 ratio 1.0 1.0 1.0 1.3 Ages 60-69 black 3.97 1.50 6.10 9.59 white 2.26 1.30 4.60 7.00 ratio 1.8 1.2 1.3 1.4 Ages 70-79 black 5.08 2.00 8.70 15.45 white 4.04 1.65 6.20 9.40 ratio 1.3 1.2 1.4 1.6 Ages >80 black 623.81 2.40 12.40 21.05 white 12.30 2.12 9.52 18.25 ratio 50.7 1.1 1.3 1.2