- Top of page
- MATERIALS AND METHODS
The purpose of this investigation was to examine changes in pretreatment prostate-specific antigen (PSA), stage, and grade over the past decade as a function of race and geographic region. A multiinstitutional database representing 6,790 patients (1,417 African-American, 5,373 white) diagnosed with nonmetastatic prostate cancer between 1988 and 1997 was constructed. PSA, stage, and grade data were tabulated by calendar year and region, and time trend analyses based on race and region were performed. There was an overall decline of PSA of 0.8%/year, which was significant (P = 0.0001), with a faster rate of decline in African-Americans (1.9%/year) than for whites (0.6%/year). The odds ratio (OR) for a stage shift was 1.09, which was significant (P < 0.0001), and this shift was greater in whites. The OR for an overall grade shift was 1.15, which was significant (P < 0.0001). Although grade and PSA trends were similar for the different regions, there were significant regional differences in stage trends. The implications are that the face of prostate cancer has changed over the past decade; i.e., the distributions of stage, grade, and PSA (the most important prognosticators) have changed. In addition, the countenances of that face are different for whites and African-Americans. For African-Americans, this is good news: the stage, grade, and PSA distributions are more favorable now than before. For whites, the trends are more complex and more dependent on region. These findings should be used for future clinical and health-policy decisions in the screening and treatment of prostate cancer. © 2001 Wiley-Liss, Inc.
- Top of page
- MATERIALS AND METHODS
Adenocarcinoma of the prostate represents one of the most common cancer diagnoses for which health-care intervention is performed in the United States. Although there appears to be a decline in the incidence after the dramatic increase during the early 1990s [1, 2], the high incidence has nonetheless increased public awareness of the diagnosis. The increase in incidence in recent years has been hypothesized to result partly from the availability and application of a relatively sensitive and specific screening test, prostate-specific antigen (PSA). Although the optimum role of PSA for prostate cancer screening remains a subject of controversy , PSA has an established role in the detection of prostate cancer and appears to have caused a decrease in distant metastases in the screened population . In addition to its use in screening, PSA has a definitive role in determining outcome after intervention [4, 5]. Indeed, PSA level has been approved by the Food and Drug Administration as an aid for the determination of prognosis and management and for the detection of prostate cancer in men aged 50 or older . Clearly, PSA will continue to occupy an important role for prostate cancer diagnosis and management in any health-care delivery system.
The use of PSA as a screening tool has changed the patterns of care, with needle biopsy becoming the predominant first step in diagnosing prostate cancer compared with the past when transurethral prostatectomy (TURP) was the most common method . This has influenced the characteristics of the population diagnosed with prostate cancer over the past decade, and hence, efforts are needed to examine trends in patient and tumor characteristics in the era of PSA-based screening. There has been an increase in incidence during the PSA era, part of which is likely due to increased diagnosing of prevalent cases, as evidenced by a more recent decline . This led to a shift to earlier stage at diagnosis, as well as diagnosis of so-called indolent cases . The lead-time and length-time biases introduced by PSA-based diagnosis have been examined in only a few studies [8, 10]. The Surveillance, Epidemiology, and End Results (SEER) program  found a decline in the incidence of distant-stage disease during the PSA era, with resulting increases in cause-specific and overall survival, and, as indicated before, an increase then a decrease in the incidence of prostate cancer [11–13]. There was also a significant increase in the incidence of localized-stage and low- and intermediate-grade tumors. These initial studies looking at the influence of PSA-based diagnosing suggest changes in incidence, stage, and grade distributions during the PSA era. The SEER analyses, while important in adding to the understanding of changes in these parameters, did not, however, examine PSA trends.
African-Americans have the highest incidence of prostate cancer in the world, and significant white–black differences in the incidence, patterns of care, and outcomes exist in the United States. However, the influence of PSA-based diagnosis upon prostate cancer stage, grade, and PSA patterns among African-Americans has not been addressed in the SEER studies described above, although a small single-institution study  found that PSA, grade, and stage shifted more for African-Americans than for whites over an 8-year period as a consequence of screening. PSA-era racial differences are likely since differences in absolute PSA level between whites and African-Americans in both patients without  and patients with [16–18] prostate cancer have been observed. African-Americans present at a more advanced stage than whites , including in the setting of a military/equal-access health-care environment [20, 21]. There has also been a stage-for-stage tumor volume disparity between whites and African-Americans in careful radical prostatectomy pathological assessment series during the PSA era . PSA disparity was also found in a multiinstitutional Radiation Therapy Oncology Group (RTOG) study , African-Americans with nonmetastatic prostate cancer having a higher serum PSA level at diagnosis than whites. In both the RTOG study and a University of Chicago study , racial PSA differences were associated with socioeconomic indicators: income, education, and insurance status.
Thus far, studies show that (1) more prostate cancers are diagnosed by PSA screening and needle biopsy than before (and this has led to changes in the disease profile at diagnosis), (2) PSA level at diagnosis is a very strong prognostic factor, and (3) racial differences in the patterns of care and disease profile exist, including a higher PSA stage-for-stage among African-Americans. The unknowns are (1) whether any stage, grade, and PSA distribution trends have occurred over the past decade within the nonmetastatic population, which constitutes approximately 85% to 90% of all cases diagnosed in recent years; (2) any differences between whites and African-Americans among these trends; and (3) geographic differences within the United States. These trends will have a significant influence on our understanding of prostate cancer from public-health, clinical-care, public-policy, resource-distribution, and research-focus perspectives.
Our objective was to perform a multiinstitutional, geographically dispersed, pooled analysis to answer the issues raised above. Our hypothesis was that the stage, grade, and PSA distributions have changed over time (i.e., a change in the face of prostate cancer) and that there are likely geographic and racial differences (i.e., a change in the countenances of prostate cancer). To undertake this study, several institutions in each of the major geographic regions in the continental United States with an established database were approached to provide data that allowed for the construction of a multiinstitutional database. The large number of patients in the multiinstitutional database enabled increased statistical power to make more definitive conclusions about PSA, stage, and grade trends than any single institutional database.
MATERIALS AND METHODS
- Top of page
- MATERIALS AND METHODS
Table 1 shows the characteristics (race, stage, and grade) of the nonmetastatic study patient populations from the participating institutions as a function of site. It also shows the geographic region grouping of the participating institutions. As the table demonstrates, the data were taken from multiple institutions representing different geographic regions in the United States. Furthermore, different practice patterns, i.e., university/academic centers, private practice/community centers, and a military/equal-access center, are represented. Using the information in Table 1 and the diagnosis date (i.e., the date of biopsy), a comprehensive database was constructed and used for analyses of trends in pretreatment PSA, stage, and grade.
Table 1. Characteristics of Study Patient Populations from Participating Institutions
| U of C||205||115||90||64||117||24||17||165||11||12|
Pretreatment PSA Analyses
The mean pretreatment PSA by calendar year was charted for the entire group and by race subgroup and geographic region (Table 2a). The following analyses were performed using the logarithm of the PSA (which more closely approaches a Gaussian distribution than the actual PSA): (1) time-trend analyses for the PSA consisting of (a) analysis of the overall group and (b) subgroup analyses based on region, stage, grade, and race [in each case, the null hypothesis was that the log (PSA) time-trend curve has a slope of 0 (i.e., no change in PSA with time)]; (2) a test for a difference in PSA trends between African-Americans and whites [for this analysis, the null hypothesis was that the log (PSA) time-trend curves for both racial groups have the same slope]. The statistical test used for each of the above analyses was the t-test of regression.
Table 2. PSA, Stage, and Grade Information by Calendar Year and Race
|(a) Mean PSA by calendar year and race|
|(b) Number (percentage) of patients in each stage by calendar year and race|
| T1/A||43 (30)||63 (23)||101 (26)||164 (27)||256 (31)||205 (27)||216 (29)||159 (27)||135 (25)||141 (30)|
| T2/B||83 (57)||184 (66)||254 (65)||368 (61)||513 (61)||439 (58)||390 (52)||249 (42)||233 (43)||123 (26)|
| T3-4/C||19 (13)||30 (11)||38 (9)||74 (12)||68 (8)||109 (15)||142 (19)||183 (31)||180 (32)||211 (44)|
| T1/A||19 (37)||12 (23)||23 (24)||38 (28)||32 (18)||42 (19)||42 (17)||60 (30)||47 (28)||20 (32)|
| T2/B||27 (53)||35 (66)||62 (63)||73 (54)||123 (67)||138 (61)||163 (67)||105 (53)||92 (55)||32 (52)|
| T3-4/C||5 (10)||6 (11)||13 (13)||25 (18)||28 (15)||44 (20)||39 (16)||34 (17)||28 (17)||10 (16)|
| T1/A||62 (32)||75 (23)||124 (25)||202 (27)||288 (28)||247 (25)||258 (26)||219 (28)||182 (25)||161 (30)|
| T2/B||110 (56)||219 (66)||316 (64)||441 (59)||636 (63)||577 (59)||453 (46)||354 (45)||325 (45)||155 (29)|
| T3-4/C||24 (12)||36 (11)||51 (11)||99 (14)||96 (9)||153 (16)||181 (18)||217 (27)||208 (30)||221 (41)|
|(c) Number (percentage) of patients in each grade by calendar year and race|
| Grade I||32 (30)||81 (38)||115 (38)||177 (34)||235 (29)||190 (26)||166 (23)||102 (17)||79 (15)||65 (14)|
| Grade II||65 (60)||125 (58)||173 (57)||311 (60)||509 (64)||478 (65)||495 (68)||422 (74)||399 (74)||369 (79)|
| Grade III||11 (10)||10 (4)||15 (5)||32 (6)||54 (7)||62 (9)||70 (9)||50 (9)||59 (11)||34 (7)|
| Grade I||14 (47)||7 (22)||17 (27)||28 (28)||34 (20)||57 (26)||50 (21)||26 (14)||14 (9)||6 (10)|
| Grade II||13 (43)||22 (71)||39 (62)||66 (66)||118 (68)||138 (63)||160 (68)||140 (75)||136 (84)||50 (82)|
| Grade III||3 (10)||2 (7)||7 (11)||6 (6)||21 (12)||24 (11)||27 (11)||21 (11)||11 (7)||5 (8)|
| Grade I||46 (33)||88 (35)||133 (36)||205 (33)||269 (27)||247 (26)||216 (22)||128 (17)||83 (12)||71 (13)|
| Grade II||78 (57)||147 (60)||212 (58)||377 (61)||627 (65)||616 (65)||655 (68)||562 (74)||535 (78)||419 (79)|
| Grade III||14 (10)||12 (5)||22 (6)||38 (6)||75 (8)||86 (9)||97 (10)||71 (9)||70 (10)||39 (8)|
Stage Shift Analyses
Table 2b shows the number (and percentage) of patients in each stage by calendar year and race. Stage shift analyses, which are based on computation of the odds ratio (OR) for a shift to a higher stage, were performed using the information in this table. The following analyses were performed: (1) stage shift analyses consisting of (a) an analysis of the overall group and (b) subgroup analyses based on region and race [in each case, the null hypothesis was that the OR was 1.0 (i.e., that there was no stage shift with time), with an OR of > 1.0 representing a shift to a higher stage, and the statistical test used was the likelihood ratio test of multiple polytomous regression]; (2) a test for a difference in stage shift between African-Americans and whites [for this analysis, the null hypothesis was that both groups have the same stage shift with time, and the corresponding statistical test was also a likelihood ratio test of multiple polytomous regression, with an additional variable (race)].
Grade Shift Analyses
Table 2c shows the number (and percentage) of patients in each grade by calendar year and race. The methodology for the overall and subgroup grade shift analyses and for testing the difference in grade shift between racial groups was identical to that described above for the stage shift analyses.
- Top of page
- MATERIALS AND METHODS
For the overall group, the pretreatment PSA, stage, and grade analyses demonstrated statistically significant trends as a function of time, a change in the face of prostate cancer. In particular, the overall results demonstrate that among patients diagnosed with prostate cancer over the past decade, the PSA at the time of diagnosis is declining and, somewhat surprisingly, there is a shift to a more advanced stage and a higher grade. While this grade migration may be due in part to the trend for pathologists to read disease at a higher grade, the overall PSA time trend is consistent with the expected effect of the implementation of a highly sensitive mass screening tool for detection of disease. Furthermore, the fact that the PSA decline with time is greater for more advanced stage and higher-grade tumors suggests that the widespread use of PSA enables earlier diagnosis of more advanced lesions.
The overall stage trend is consistent with some, but not with other, reports [11, 14]; as there is significant regional variation (discussed below), no nationwide conclusions can be drawn from the overall stage analysis. The overall grade trend is consistent with other reports [2, 25] that in the PSA era the grade is more likely to be moderately than well or poorly differentiated.
Regional similarities and differences exist for the PSA, stage, and grade trends. For the PSA trend, there are regional similarities with a consistent decline regardless of geographic area or type of medical center, with only minor regional variations. For the stage shift, however, there are significant variations by region, as demonstrated in Table 3b. This stage shift is explored further for a shift from T1/2 to T3/4, and the results of this analysis are shown in Table 4. This analysis was performed because a binary analysis is likely to more accurately assess a general clinical stage migration than the categorical analysis shown in Table 3. This is because it is easier to categorize organ-confined disease (T1 and T2) vs. capsular/seminal vesicle invasion (T3 and T4) by digital rectal examination than it is to differentiate T1 vs. T2 lesions. This analysis suggests that the shift to a higher stage is restricted to the West. For the regions that had a shift to a lower stage, this is consistent with the Will Rogers' phenomenon [14, 26] and suggests that both lead-time and length-time biases have come into effect as a consequence of widespread screening. For the region in which there was a shift to a higher stage, the explanation is less clear and inconsistent with the overall trends showing a declining PSA (and hence lower tumor burden) with time. It is possible that in the later years of the study more T1/2 patients underwent radical prostatectomy; hence, both radiation oncology practices (University of California San Francisco and Arizona Oncology Services) participating in this study showed a higher percentage of T3/4 tumors in the later years. Radical prostatectomy patients have a lower stage than radiotherapy patients, and these practice variations may influence presentation stage with time.
Table 4. Results of Binary Statistical Analyses for Stage Shift
| ||Odd ratio for stage shift (95% CI)*||P value for stage shift|
|Overall||1.27 (1.23, 1.31)||<0.0001|
| Military||0.88 (0.80, 0.97)||0.01|
| East||0.93 (0.87, 0.99)||0.04|
| Midwest||0.96 (0.90, 1.02)||0.21|
| South||0.88 (0.69, 1.13)||0.31|
| West||1.54 (1.47, 1.62)||<0.0001|
| African-American||1.04 (0.97, 1.10)||0.25|
| White||1.33 (1.29, 1.38)||<0.0001|
For the grade trend, there is regional similarity; there appears to be a uniform shift to a more advanced grade across all regions, which, as noted above, is consistent with other reports [2, 25] that the percentage of low-grade tumors is decreasing, perhaps due to a decrease in TURP-diagnosed tumors. The grade trend may also be due to fewer prevalent cases being moved to the incident category in later years; that is, the indolent cases that would have otherwise gone undiagnosed were diagnosed by PSA testing in the earlier years of the study. These indolent cases were likely well differentiated, and as this shifting of prevalent cases to incident cases declined due to an exhaustion of prevalent cases, similar to an increase followed by a decline in the newly diagnosed cases in the past 10 to 15 years [11–13], the proportion of moderately differentiated tumors increased in the later years. Grade III (Gleason sum of 7 or more) lesions carry the worst prognosis and generally are treated more aggressively . Thus, we performed a binary analysis similar to the one described for stage above (i.e., grade I/II vs. grade III). The differences between the binary and categorical (grade I vs. grade II vs. grade III) analyses are shown in Table 5. As this table indicates, the statistical power for an increase in grade weakens overall and for all subsets on binary analysis; indeed, only the OR in the military and East remain significant. Because the statistical power for a shift to a higher grade is much stronger for the categorical analysis, an increase in grade II (moderately differentiated) lesions in the PSA era is suggested. Furthermore, the racial subset analysis in Table 5 reveals that this increase in moderately differentiated lesions occurred in the African-American population to a much greater extent than in the white population over the past decade.
Table 5. Grade Shift Statistical Analyses: Categorical vs. Binary
|Odds ratio for grade shift (95% CI)*||P value for grade shift||Odds ratio for grade shift (95% CI)†||P-value for grade shift|
|Overall||1.15 (1.13, 1.18)||<0.0001||1.05 (1.01, 1.10)||0.01|
| Military||1.15 (1.04, 1.26)||0.01||1.26 (1.03, 1.55)||0.02|
| East||1.15 (1.09, 1.21)||<0.0001||1.13 (1.04, 1.24)||0.01|
| Midwest||1.28 (1.21, 1.36)||<0.0001||1.06 (0.94, 1.18)||0.32|
| South||1.06 (0.98, 1.15)||0.11||1.11 (0.97, 1.27)||0.13|
| West||1.09 (1.05, 1.14)||<0.0001||1.02 (0.96, 1.09)||0.46|
| African-American||1.13 (1.07, 1.20)||<0.0001||0.99 (0.91, 1.08)||0.80|
| White||1.16 (1.13, 1.19)||<0.0001||1.07 (1.02, 1.12)||0.01|
Analysis of the regional similarities and differences is important because there are significant regional variations in the treatment of prostate cancer . The military/equal-access site was analyzed independently and (Table 3) shows a highly significant (P = 0.01) PSA decline, no stage shift, and a highly significant (P = 0.01) shift to a higher grade. This is consistent with the overall analysis showing a PSA decline and a shift to a higher grade. The neutral stage shift suggests that the significant regional variations seen in the stage shift may be related to medical-access issues not present in an equal-access medical setting; furthermore, the military practiced screening by digital rectal examination prior to the PSA era, which may explain the neutral stage migration in this population.
Analyses of racial differences in PSA, stage, and grade are quite revealing. There was a greater PSA decline per year in African-Americans compared with whites. This faster rate of PSA decline suggests that the disease profile at presentation is being leveled between the two groups and that access to and utilization of medical resources may explain the higher PSA in African-Americans in earlier years. This is reinforced by the finding of a difference in the pattern of stage migration between the two populations, with more of a shift to a higher stage for whites than for African-Americans. Also, as the binary grade analysis demonstrates above, there is a suggested increase in moderately differentiated tumors in the African-American population over the past decade that is much more pronounced than in the white population. The better disease profile for the African-American group, which more closely approximates that of the white population in the later years of the study, suggests that the disparity in clinical outcome in the past may not be biological but reflective of differences in tumor cell burden at diagnosis and that with continuation of correct diagnostic practices the field may be more leveled. This reinforces other results [14, 24] suggesting that socioeconomic differences are largely responsible for the differences in disease presentation between whites and African-Americans. The composite results of the racial-differences analyses of PSA, stage, and grade suggest that (1) previously reported racial differences were in fact due to socioeconomic differences [29–31], (2) these differences led to outcome differences, and (3) the increased use of PSA as a screening tool appears to ameliorate these differences. For the African-American population, these findings represent a good thing . Although there was a suggestion of this in a small study in a single-institution setting  and in a review of randomized and nonrandomized studies , the current results are the first direct evidence in a geographically diverse multiinstitutional setting.
Although the study is multiinstitutional and represents a wide range of institution sizes, geographic regions, and patient populations, it cannot be considered a truly national study as the institutions were selected; the biases inherent in this type of data collection and analysis are understood. The results of the current study can, however, serve as the basis for a pilot study, as a cooperative group effort, with centralized pathology review and strict, centralized guidelines for patient entry.
Our results are important because the combination of stage, grade, and PSA  determines treatment outcome after radiotherapy or surgery; thus, trends in these parameters, as demonstrated in this nationwide study, have significant implications for issues related to prostate cancer diagnosis and management . Specifically, our results suggest that caution must be taken in interpreting the results of ongoing studies because the disease profile is changing. For instance, the overall outcomes reported recently may, in addition to any therapeutic gain in intervention technique over the past few years, be due to (1) reduced stage-for-stage tumor burden, as demonstrated by the declining PSA with time over this past decade, and (2) the stage migration (Will Rogers) phenomenon (lead-time and length-time biases) demonstrated in certain regions. Although our results indicate that caution is warranted when interpreting treatment results due to the effects of screening, they do reinforce the results of previous studies suggesting that widespread screening is balancing racial differences at disease presentation.