Prostate specific antigen (PSA) is used in the follow-up of prostate

Transcription

1 496 Race Independently Predicts Prostate Specific Antigen Testing Frequency following a Prostate Carcinoma Diagnosis Steven B. Zeliadt, M.P.H. 1,2 David F. Penson, M.D., M.P.H. 3,4 Peter C. Albertsen, M.D. 5 John Concato, M.D. 6,7 Ruth D. Etzioni, Ph.D. 1,2 1 Fred Hutchinson Cancer Research Center, Seattle, Washington. 2 School of Public Health and Community Medicine, University of Washington, Seattle, Washington. 3 Veterans Affairs Puget Sound Health Care System, Seattle, Washington. 4 Department of Urology, University of Washington School of Medicine, Seattle, Washington. 5 Section of Urology, University of Connecticut School of Medicine, Farmington, Connecticut. 6 Veterans Affairs Connecticut Healthcare System, West Haven, Connecticut. 7 Department of Internal Medicine, Yale University School of Medicine, New Haven, Connecticut. BACKGROUND. The goals of the current study were to describe patterns of prostate specific antigen (PSA) surveillance for prostate carcinoma progression in a community-based cohort of patients and to identify independent clinical and sociodemographic factors that predict the frequency of surveillance. METHODS. Patients diagnosed with localized prostate carcinoma from October 1, 1991 to December 31, 1992 in New Haven and Hartford, Connecticut, were identified. Data were collected through standardized outpatient medical record review. Multivariate statistical methods were used to determine the factors that independently predicted the frequency of surveillance. RESULTS. Six hundred fifty-eight men with localized prostate carcinoma were included in the cohort. Forty-five percent of all patients were tested at least once annually, and 69% were tested at least once every 2 years. Multivariate models indicated that African American men were half as likely as Caucasian men to receive annual testing (odds ratio [OR], 0.49; 95% confidence interval [CI], ). Men diagnosed at age 70 years or older were 38% less likely to have annual testing than men diagnosed between the ages of 65 and 69 (OR, 0.62; 95% CI, ). A higher Gleason score and PSA at presentation also were associated independently with higher rates of annual PSA surveillance. CONCLUSIONS. Postdiagnosis PSA surveillance is common, although not universal. African American men were at significantly greater risk for receiving less frequent testing compared with Caucasian men. This disparity in access to care may explain, in part, previously observed racial differences in survival in prostate carcinoma. Further research is needed to identify the reasons for the racial disparity in PSA surveillance and to design interventions to lessen these differences. Cancer 2003; 98: American Cancer Society. Supported by grants CA and CA from the National Cancer Institute, and by grant HS09578 from the Agency for Health Care Research and Quality and grant DF from the Patrick and Catherine Weldon Donaghue Foundation. Dr. Penson was a Level I Veterans Affairs Health Services Research and Development Career Development Awardee and a Robert Wood Johnson Clinical Scholar. Address for reprints: Ruth D. Etzioni, Ph.D., Fred Hutchinson Cancer Research Center, P.O. Box (MP-665), Seattle, WA ; Fax: (206) ; retzioni@fhcrc.org Received January 29, 2003; revision received March 27, 2003; accepted April 2, KEYWORDS: prostate carcinoma, recurrence, prostate specific antigen testing, race, surveillance. Prostate specific antigen (PSA) is used in the follow-up of prostate carcinoma patients to monitor the signs of disease recurrence or progression. 1 Although PSA was approved originally as a tumor marker for disease recurrence, little information is available on its use as a surveillance tool in clinical practice. No studies have documented definitively that surveillance leads to delayed progression or improved survival. 2 In addition, the optimal frequency of PSA surveillance is unknown. A recent survey of 1050 urologists suggests that considerable variation in practice patterns exists. 3 Although respondents most commonly stated that they measured PSA every 3 months during the first year after treatment, every 6 months during the second year, and annually thereafter, considerable variation was re American Cancer Society DOI /cncr.11492

2 PSA Testing following Diagnosis/Zeliadt et al. 497 ported in patterns of care. The variability reported by urologists may be due to the lack of a community standard regarding the frequency of surveillance testing. The American Urological Society 2000 policy statement recommends periodic surveillance but does not specify a PSA testing interval, whereas the National Comprehensive Cancer Network guidelines recommend PSA testing every 6 months for 5 years with annual testing thereafter. 4,5 Patterns of PSA use following a prostate carcinoma diagnosis are of interest for several reasons. Variations in the frequency of surveillance may be an indication of problems with access to optimal followup care. Access to care is often cited as a potential explanation for why the prostate carcinoma mortality rate among African American men is twice the rate among Caucasian men If PSA monitoring proves to be efficacious, racial variation in PSA testing may be a contributing factor to the disparity in mortality seen between African American and Caucasian men. Even if PSA monitoring is of limited effectiveness, it is still important to characterize patterns of use because of implications related to morbidity and costs. The primary goal of the current study was to describe PSA surveillance patterns in a communitybased cohort of men with localized prostate carcinoma diagnosed in 1991 and 1992 and followed through Another goal was to determine whether or not race or other clinical or sociodemographic characteristics are associated independently with the intensity of PSA surveillance. MATERIALS AND METHODS Study Population The study population consisted of all men diagnosed and treated with biopsy-proven prostate carcinoma from October 1, 1991, to December 31, 1992, at one of four hospitals (Yale-New Haven Hospital, St. Raphael s Hospital, St. Francis Hospital, and Hartford Hospital). These hospitals represented the catchment areas of metropolitan New Haven and Hartford, Connecticut. Cases were identified through the Connecticut Tumor Registry. The current study was approved by each hospital s institutional review board (IRB), as well as by the State of Connecticut Department of Health IRB. A total of 988 men with biopsy-proven prostate carcinoma were identified initially. To be included in the study, subjects had to be diagnosed with clinically localized disease, be followed for at least 1 year after diagnosis, and be either African American or Caucasian. The study was limited to men with localized disease for two reasons. First, we believe that this is the group of patients most likely to benefit from PSA follow-up after primary treatment. Second, patterns of PSA testing are likely to vary depending on the stage at presentation and limiting the analysis to men with localized disease reduces this source of potential confounding. Subjects were excluded if their digital rectal examination indicated extraprostatic extension of disease; if imaging tests (including bone scans, computerized tomography scans, or magnetic resonance imaging studies) indicated regionally advanced or metastatic disease; or if they received long-term adjuvant hormone ablation therapy. These criteria resulted in a final cohort of 658 men who were treated by 45 physicians. Data Collection and Ascertainment of PSA Testing Data were abstracted from the outpatient medical records of the physician primarily responsible for treating the patient s prostate carcinoma, supplemented by inpatient record review when needed. Information was collected on age at diagnosis, race, initial treatment, PSA level at diagnosis, disease grade, and the presence of comorbid conditions at the time of diagnosis. The baseline PSA level was defined as the PSA level closest to the date of diagnosis but before the initiation of treatment. All subsequent dates and values of PSA tests were abstracted, as well as any symptom information and the dates and types of subsequent therapies. Information regarding race was abstracted from the patient s medical record. Baseline PSA was classified into 1 of 4 categories: 4 ng/ml, ng/ml, ng/ml, and 20 ng/ml. Comorbidity at the time of diagnosis was ascertained using the Charlson index. 11,12 Statistical Analysis The frequency of PSA use was characterized using two methods. A longitudinal measure of use was defined, which consisted of the number of tests conducted during each year of observation following the date of diagnosis. The date of diagnosis was chosen rather than the treatment date to ensure consistency among all patients including men who selected watchful waiting. A profile measure of use was also defined, which characterized men as annual or biannual users for the duration of their follow-up. Annual use was defined as receipt of at least 1 PSA test during the first year after diagnosis and at least 1 PSA test during every 12- month period beginning with the date of their first test. Similarly, biannual use was defined as at least 1 test during every 24-month period. All statistical analyses were performed using Stata Version 7 SE for Windows (Austin, TX). Univariate t tests were used to compare means and chi-square tests were used to compare African American and Caucasian men in

3 498 CANCER August 1, 2003 / Volume 98 / Number 2 terms of clinical and demographic characteristics, differences in follow-up time, and time from diagnosis to the first PSA test. The association between these factors and choice of initial treatment was evaluated. Longitudinal measure of use Generalized estimating equation (GEE) regression was used to assess the impact of covariates on the longitudinal measure of use. 13 For this measure, the unit of analysis was the person-year of observation. For each 12-month period (person-year), patients were included only if they were followed for the entire period without the initiation of secondary hormone therapy. The dependent variable was the number of tests in each 12-month period, and this variable was assumed to follow a Poisson distribution. Independent variables included time from diagnosis, initial treatment, disease stage and grade, race, and age at diagnosis. Two-way interaction terms were used to evaluate whether the effect of time differed by initial treatment or other variables. To summarize the longitudinal data and to account for censored observations, nonparametric techniques were used. 14 Death, loss to followup, or secondary treatment were all considered to be censoring events in this analysis. Estimates of the mean number of tests received over time were plotted by time following treatment. Confidence intervals (CIs) were obtained using bias-corrected bootstrapping methods with 100 replications. 15 Profile measure of use For this measure, the unit of analysis was the individual. Each subject s entire surveillance history was assessed. For annual testing, subjects had to have at least 1 test during each 12-month period in which they were alive without the initiation of secondary therapy. Logistic regression was used to evaluate the importance of demographic and clinical covariates in predicting annual testing. Variables were selected a priori and included the following: age, initial treatment, Charlson comorbidity index score, PSA level at diagnosis, Gleason score, and race. Alternative definitions of regular PSA testing were considered, including receipt of at least 1 test every 2 years, as well as testing annually for the first 2 years and then biannually thereafter. RESULTS The median follow-up time for the entire cohort was 6.6 years (range, years). The clinical and demographic characteristics of the cohort are presented in Table 1, as are the same characteristics stratified by race. African American men made up 7% of the study population and were younger and had more advanced disease at diagnosis (as evidenced by stage at diagnosis and PSA level) compared with Caucasian men. In addition, comorbid conditions were more frequent at the time of diagnosis among African American men compared with Caucasian men. As expected, patients who selected radical prostatectomy were younger and had fewer comorbid conditions and lower PSA levels at the time of diagnosis. The men who selected radiotherapy had higher-grade disease and more comorbid conditions. The patients who selected watchful waiting generally were older than those who selected surgery or radiation. During the study period, 122 deaths occurred among the study population. Twenty-three (34%) deaths were in the watchful waiting cohort, 83 (25%) deaths were in the radiotherapy cohort, and 16 deaths (6%) were in the surgery group. The average time to the first PSA test after diagnosis was 6.2 months for the surgery group, 10.8 months for the radiotherapy group, and 13.7 months for the watchful waiting group. Differences between the surgery cohort and both the watchful waiting and radiotherapy groups were statistically significant (P and P 0.001, t test). The difference between the watchful waiting cohort and the radiotherapy cohort was not statistically significant (P 0.121). The average time to the first PSA test was 12.1 months and 8.9 months among African American and Caucasian men, respectively (P 0.060). The number of study participants who were eligible for monitoring in each period following diagnosis and the number of PSA tests they received each year are presented in Table 2. For each of the 12-month periods, 22 29% of the men did not receive any PSA monitoring, whereas greater than 70% of the men received at least 1 test annually. Many men also received more than 1 test annually. Forty-six percent of men received 2 or more tests the first year after diagnosis, which declined steadily to 24% in the 7th year after diagnosis. Among the 424 men who were eligible for monitoring and who were followed through Year 5, 28 (7%) had not received any tests by that time. The total number of men decreases each period due to censoring as we include only men who have not yet initiated secondary treatment and have complete follow-up information available for the entire 12-month period. The patient characteristics associated with the annual profile measure of PSA testing are presented in Table 3. Of the 658 men, 298 (45%) were classified as being tested at least annually during the entire followup period. Alternative definitions produced the following results: 369 (56%) men received testing once a year for the first 2 years and biannually thereafter and 457 (69%) men received testing at least once every 2 years. Univariate statistics (Table 3) suggested that

4 PSA Testing following Diagnosis/Zeliadt et al. 499 TABLE 1 Characteristics of the Study Population Characteristic Caucasian men (%) African American men (%) P value Total 614 (93) 44 (7) Age at diagnosis (yrs) Mean SE Range (77) 28 (64) Initial treatment Radical prostatectomy 232 (38) 20 (45) Radiotherapy 317 (52) 21 (48) Watchful waiting 65 (11) 3 (7) PSA at diagnosis (ng/ml) 4 55 (9) 2 (5) (42) 10 (23) (25) 13 (30) (20) 17 (39) Unknown 23 (4) 2 (5) Gleason score (10) 6 (14) (74) 27 (61) (14) 10 (23) Unknown 8 (1) 1 (2) Charlson comorbidity index (63) 18 (41) (26) 18 (41) 2 69 (11) 8 (18) Deaths during follow-up period 115 (19) 7 (16) Average follow-up time SE (yrs) Mean time to first monitoring test SE (mos) SE: standard error; PSA: prostate specific antigen. TABLE 2 Number of Subjects with Complete Follow-up per Year and Number of Prostate Specific Antigen Tests for Each 12-Month Period after Diagnosis Yr following diagnosis No. of subjects (%) who had: No tests 150 (23) 129 (22) 127 (24) 136 (28) 124 (26) 99 (29) 63 (28) One test 203 (31) 211 (35) 197 (37) 181 (37) 182 (39) 125 (40) 107 (48) Two tests 191 (29) 192 (32) 173 (32) 140 (29) 129 (27) 89 (26) 40 (18) Three or more tests no. 114 (17) 69 (11) 41 (8) 33 (7) 35 (7) 17 (5) 14 (6) Total no. of subjects race, age, PSA level at diagnosis, Gleason score, and comorbidity were associated with testing frequency. A trend also was observed among treatment choice. For example, annual testing was most common among men who selected radical prostatectomy and was least common among the watchful waiting group. Table 3 also shows the frequency of secondary therapy in the study population. Secondary hormonal therapies were initiated more frequently in men who received annual testing. Of the 298 men who received annual testing, 94 (32%) subsequently received secondary therapy, compared with 40 of 386 (10%) men who received less frequent testing. The results of the multivariate logistic regression model are presented in Table 4. Controlling for all other factors, African American men were half as likely to receive annual testing as Caucasian men were (odds ratio [OR] 0.49, 95% CI ). Age also retained significance as a predictor of annual testing. Evaluation of biannual rather than annual testing yielded similar results with the exception of race (OR 1.29; ). It is noteworthy that the definition of biannual testing classified only 11 (25%) African American men as not receiving testing at this frequency, which resulted in a lower statistical power to detect significant differences by race. The results of the regression model for the longi-

5 500 CANCER August 1, 2003 / Volume 98 / Number 2 TABLE 3 Annual Prostate Specific Antigen Testing by Patient Characteristics Characteristic Annual monitoring (%) Infrequent or no monitoring (%) P value Total 298 (45) 386 (55) Race Caucasian 284 (46) 330 (54) African American 14 (32) 30 (68) Age (yrs) (48) 83 (52) a (52) 82 (48) (40) 195 (60) Initial treatment Radical prostatectomy 125 (50) 125 (50) Radiotherapy 149 (44) 189 (56) Watchful waiting 24 (35) 44 (65) PSA at diagnosis (ng/ml) 4 21 (37) 36 (63) a (44) 150 (56) (47) 89 (53) (53) 65 (47) Unknown 5 (20) 20 (80) Gleason score (35) 44 (65) a (45) 264 (55) (55) 44 (45) Unknown 1 (11) 8 (88) Charlson comorbidity index (49) 208 (51) (41) 104 (59) 2 29 (38) 48 (62) Subsequent secondary hormone treatment 94 (70) 40 (30) a PSA: prostate specific antigen. a Significant at the 0.05 level. tudinal measure of use are presented in Table 5. The current analysis defines the dependent variable to be the number of tests received each year after diagnosis, thereby allowing us to model how the frequency of testing varies over time. Results indicate that on average, African American men received 3 tests for every 4 tests received by Caucasian men (relative rate [RR], 0.76; 95% CI, ). Older age was associated with 13% fewer tests each year (RR 0.87; 95% CI, ). The frequency of testing declined over time, with 15% fewer tests in Years 3 5 compared with the first 2 years after diagnosis (RR 0.85; 95% CI, ) and nearly 24% fewer tests after Year 5 (RR 0.76; 95% CI, ). Higher PSA levels at diagnosis and higher Gleason scores also were associated with more frequent testing. Figure 1 shows the estimates of the average number of PSA tests received by any given time point following diagnosis. The results are presented by treatment type and by race and take into account censoring due to death, loss to follow-up, and initiation of second treatment. These results are not adjusted for any other patient or disease characteristics. Men treated initially with radical prostatectomy had more tests than men treated with radiotherapy and both groups had more tests than the men who selected watchful waiting (Fig. 1). These differences persisted over time. By 5 years from the time of diagnosis, men who selected watchful waiting had 4.3 (95% CI, ) PSA monitoring tests, men who were treated with radical prostatectomy had 6.9 (95% CI, ) tests, and men who were treated with radiotherapy had 5.0 (95% CI, ) tests. Age is associated closely with testing frequency and an age differential between African American men and Caucasian men was observed in this dataset, so the race analysis was restricted to men younger than age 75 years. DISCUSSION Although the value of PSA testing for disease recurrence or progression in men with prostate carcinoma remains uncertain, such testing has become a routine

6 PSA Testing following Diagnosis/Zeliadt et al. 501 TABLE 4 Odds of Receiving at Least One Prostate Specific Antigen Test Annually (Profile Measure): Multivariate Model Characteristic Odds ratio 95% CI Race Caucasian 1.0 African American 0.49 a (0.24, 0.97) Age (yrs) (0.55, 1.38) a (0.41, 0.94) Initial treatment Radical prostatectomy 1.0 Radiotherapy 0.84 (0.45, 1.59) Watchful waiting 0.91 (0.61, 1.36) PSA at diagnosis (ng/ml) (0.77, 2.59) (0.84, 3.02) a (1.10, 4.17) Unknown 0.42 a (0.11, 0.99) Gleason score (0.84, 2.49) a (1.11, 4.24) Unknown 0.22 (0.03, 1.98) Charlson comorbidity index (0.50, 1.05) (0.37, 1.06) PSA: prostate specific antigen; CI: confidence interval. a P values are significant at the 0.05 level. part of postdiagnostic care for many men with the disease. Despite this common practice, there is a scarcity of data regarding how the test is being used in the community and regarding the clinical factors that may influence its use. The current study uses primary data from 658 men diagnosed with localized prostate carcinoma at 4 hospitals that represent the catchment areas of metropolitan Hartford and New Haven, Connecticut, making it one of the first reports of community-based patterns of PSA surveillance in prostate carcinoma follow-up. Our study confirms that considerable variation exists in PSA testing practice after a prostate carcinoma diagnosis. Younger age, worse pathologic differentiation (higher Gleason score), and higher PSA at baseline were associated with more frequent PSA testing in the follow-up period. The choice of initial treatment also was correlated with PSA testing frequency. For example, men who elected watchful waiting received fewer PSA tests than did men who received radiotherapy or surgery. Most importantly, however, African American men were found to be half as likely as Caucasian men to receive annual PSA monitoring. TABLE 5 Relative Annual Rates of Prostate Specific Antigen Testing Each Year following Diagnosis: Longitudinal GEE1 Model Characteristic Odds ratio 95% CI Race White 1.0 African American 0.76 a (0.70, 0.82) Age (yrs) (0.83, 1.05) a (0.78, 0.97) Initial treatment Radical prostatectomy 1.0 Radiotherapy 0.84 a (0.75, 0.93) Watchful waiting 0.83 a (0.70, 0.98) PSA at diagnosis (ng/ml) (0.99, 1.36) (0.99, 1.39) a (1.15, 1.65) Unknown 0.59 a (0.39, 0.92) Gleason score (0.99, 1.35) (0.99, 1.45) Unknown 0.84 (0.43, 1.66) Charlson comorbidity index (0.85, 1.04) (0.73, 1.02) Time from diagnosis (yrs) a (0.80, 0.90) a (0.70, 0.82) GEE: generalized estimating equation; CI: confidence interval, PSA: prostate specific antigen. a P values are significant at the 0.05 level. These results were confirmed by a longitudinal model, which demonstrated that for every four follow-up PSA tests received by Caucasian men, African American men received only three. The significance of these findings is underscored by the fact that more intensive testing was associated with less favorable prognostic factors, which are observed more frequently among African American men Currently, there is no information to connect follow-up PSA testing with improved survival, nor are there data to suggest the ideal frequency of testing. Nevertheless, should PSA surveillance be efficacious, the differential in testing frequency that we observe may be one additional factor contributing to the unexplained gap in mortality rates, which are nearly twice as high among African American men compared with Caucasian men. 20 A debate is ongoing as to whether racial differences in prostate carcinoma outcomes are due to disparities in access to care or biologic differences in

7 502 CANCER August 1, 2003 / Volume 98 / Number 2 FIGURE 1. Mean number of prostate specific antigen (PSA) monitoring tests over time by treatment and race (unadjusted). Asterisks indicate men who were diagnosed when they were younger than 75 years old. RP: radical prostatectomy; XRT: radiotherapy; WW: watchful waiting. disease. Extensive data are available that demonstrate that African American men present later in the disease process with more advanced-stage disease In addition, numerous authors have concluded that, even when controlling for stage at presentation and other related covariates, African American men are less likely to receive aggressive therapy than are Caucasian men These reports support the hypothesis that racial differences in outcomes are due to disparities in access to care relating to the diagnosis of and initial treatment for the disease. This notion is supported by additional studies showing that within equal-access healthcare systems, survival by stage at diagnosis is similar for Caucasians and African Americans. 6,9,10,27,28 Nonetheless, studies of pathologic features of tumors in African American and Caucasian men also have suggested that there are racial differences in the biology of prostate carcinoma tumors that explain observed differences in outcomes. 29 In addition, other studies, including a large Health Maintenance Organization study, have found race to be an independent predictor of outcomes following a prostate carcinoma diagnosis, suggesting that biologic differences may play a role. 30,31 The current study does not provide direct evidence that disparities in access to care account for the differences in survival that have been noted between African American and Caucasian men newly diagnosed with prostate carcinoma. It does, however, document another potential mechanism through which differences in access to care might affect outcomes in men with prostate carcinoma, namely, the receipt of regular surveillance for the recurrence or progression of the disease. Although the impact of postdiagnosis PSA testing on survival currently is unknown, health care providers and policy makers should be aware that racial disparities exist in this area and that additional effort is needed to ensure equal access to PSA surveillance. One strength of the current study is the use of community-based data collected from the practices of 45 health care providers. However, it also has several limitations. Extensive efforts were made to capture data on all patients treated for prostate carcinoma in the greater Hartford and New Haven areas, but the cohort was not truly population based. For example, patients who resided in the community but chose to obtain care outside of these four hospitals were not included in the study. Nonetheless, the four hospitals were chosen because they are the only inpatient facilities within greater Hartford and New Haven. Accordingly, we expect that the majority of patients diagnosed with prostate carcinoma in these cities were included in the study. In addition, our study cohort was limited to men initially diagnosed and treated between October 1, 1991, and December 31, 1992, in an effort to maximize follow-up. It is likely that differences exist in follow-up between men diagnosed in the early 1990s and men diagnosed today, which would limit the generality of the findings. An additional limitation is that we did not model the influence of changing PSA levels on test frequency. Patients who have suspicious levels at one point in time may subsequently receive more frequent testing in the future, whereas patients with undetectable or unchanging levels may be more likely to receive less frequent testing. We restricted our analysis to men who did not initiate secondary treatment within the first year following diagnosis, thereby eliminating men with highly suspicious posttreatment PSA levels at the start of follow-up. However, a goal for future studies is to investigate the correlation between suspicious results and the frequency of subsequent tests. Finally, no data were collected on socioeconomic factors such as insurance status, income, and education, which may have important implications in the intensity of follow-up care. These factors have been shown to be correlated with race. 32 Consequently, our racial disparity findings may

8 PSA Testing following Diagnosis/Zeliadt et al. 503 be a result of disparity in socioeconomic factors, which were not measured in the current study. In summary, the current study documents that PSA surveillance for prostate carcinoma recurrence and/or progression is used widely in the community but that its use is not uniform. An important finding is that racial disparities exist in PSA surveillance, with African American men being half as likely as Caucasian men to receive annual PSA monitoring. Further research is needed to identify the barriers to PSA surveillance among African American men and to design interventions to improve access in these patients. REFERENCES 1. Polascik TJ, Oesterling JE, Partin AW. Prostate specific antigen: a decade of discovery what we have learned and where we are going. J Urol. 1999;162: Seidenfeld J, Samson DJ, Aronson N, et al. Relative effectiveness and cost-effectiveness of methods of androgen suppression in the treatment of advanced prostate cancer. Washington: Agency for Health Care Policy and Research, Oh J, Colberg JW, Ornstein DK, et al. Current followup strategies after radical prostatectomy: a survey of American Urological Association urologists. J Urol. 1999;161: American Urological Association. Prostate-specific antigen (PSA) best practice policy. Oncology (Huntingt). 2000;14: , , 280 passim. 5. NCCN. National Comprehensive Cancer Network practice guidelines in oncology: prostate cancer 2003 [monograph online]. Version Available from URL: nccn.org [Accessed 24 March 2003]. 6. Optenberg SA, Thompson IM, Friedrichs P, Wojcik B, Stein CR, Kramer B. Race, treatment, and long-term survival from prostate cancer in an equal-access medical care delivery system. JAMA. 1995;274: Fowler JE Jr., Terrell F. Survival in blacks and whites after treatment for localized prostate cancer. J Urol. 1996;156: Presti JC Jr., Hovey R, Bhargava V, Carroll PR, Shinohara K. Prospective evaluation of prostate specific antigen and prostate specific antigen density in the detection of carcinoma of the prostate: ethnic variations. J Urol. 1997;157: ; discussion, Freedland SJ, Jalkut M, Dorey F, Sutter ME, Aronson WJ. Race is not an independent predictor of biochemical recurrence after radical prostatectomy in an equal access medical center. Urology. 2000;56: Freedland SJ, Amling CL, Dorey F, et al. Race as an outcome predictor after radical prostatectomy: results from the Shared Equal Access Regional Cancer Hospital (SEARCH) database. Urology. 2002;60: Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis. 1987;40: Albertsen PC, Fryback DG, Storer BE, Kolon TF, Fine J. The impact of co-morbidity on life expectancy among men with localized prostate cancer. J Urol. 1996;156: Zeger SL, Liang KY. An overview of methods for the analysis of longitudinal data. Stat Med. 1992;11: Ghosh D, Lin DY. Nonparametric analysis of recurrent events and death. Biometrics. 2000;56: Efron B, Tibshirani R. Bootstrap measures for standard errors, confidence intervals, and other measures of statistical accuracy. Stat Sci. 1986;1: Fowler JE Jr., Bigler SA. A prospective study of the serum prostate specific antigen concentrations and Gleason histologic scores of black and white men with prostate carcinoma. Cancer. 1999;86: Moul JW, Sesterhenn IA, Connelly RR, et al. Prostate-specific antigen values at the time of prostate cancer diagnosis in African-American men. JAMA. 1995;274: Vijayakumar S, Karrison T, Weichselbaum RR, Chan S, Quadri SF, Awan AM. Racial differences in prostate-specific antigen levels in patients with local-regional prostate cancer. Cancer Epidemiol Biomarkers Prev. 1992;1: Moul JW, Connelly RR, Mooneyhan RM, et al. Racial differences in tumor volume and prostate specific antigen among radical prostatectomy patients. J Urol. 1999;162: Edwards BK, Howe HL, Ries LA, et al. Annual report to the nation on the status of cancer, , featuring implications of age and aging on U.S. cancer burden. Cancer. 2002;94: Natarajan N, Murphy GP, Mettlin C. Prostate cancer in blacks: an update from the American College of Surgeons patterns of care studies. J Surg Oncol. 1989;40: Mettlin CJ, Murphy GP, McGinnis LS, Menck HR. The National Cancer Data Base report on prostate cancer. American College of Surgeons Commission on Cancer and the American Cancer Society. Cancer. 1995;76: Polednak AP, Flannery JT. Black versus white racial differences in clinical stage at diagnosis and treatment of prostatic cancer in Connecticut. Cancer. 1992;70: Shavers VL, Brown ML. Racial and ethnic disparities in the receipt of cancer treatment. J Natl Cancer Inst. 2002;94: Jones GW, Mettlin C, Murphy GP, et al. Patterns of care for carcinoma of the prostate gland: results of a national survey of 1984 and J Am Coll Surg. 1995;180: Klabunde CN, Potosky AL, Harlan LC, Kramer BS. Trends and black/white differences in treatment for nonmetastatic prostate cancer. Med Care. 1998;36: Roach M III, Krall J, Keller JW, et al. The prognostic significance of race and survival from prostate cancer based on patients irradiated on Radiation Therapy Oncology Group protocols ( ). Int J Radiat Oncol Biol Phys. 1992;24: Hart KB, Wood DP Jr., Tekyi-Mensah S, Porter AT, Pontes JE, Forman JD. The impact of race on biochemical disease-free survival in early-stage prostate cancer patients treated with surgery or radiation therapy. Int J Radiat Oncol Biol Phys. 1999;45: Powell IJ. Prostate cancer in the African American: is this a different disease? Semin Urol Oncol. 1998;16: Robbins AS, Whittemore AS, Van Den Eeden SK. Race, prostate cancer survival, and membership in a large health maintenance organization. J Natl Cancer Inst. 1998;90: Grossfeld GD, Latini DM, Downs T, Lubeck DP, Mehta SS, Carroll PR. Is ethnicity an independent predictor of prostate cancer recurrence after radical prostatectomy? J Urol. 2002; 168: Dale W, Vijayakumar S, Lawlor EF, Merrell K. Prostate cancer, race, and socioeconomic status: inadequate adjustment for social factors in assessing racial differences. Prostate. 1996;29: