This article is a US Government work and, as such, is in the public domain in the United States of America.
Prostate cancer mortality rates in the United States declined sharply after 1991 in white men and declined after 1992 in black men. The current study was conducted to investigate possible mechanisms for the declining prostate cancer mortality rates in the United States.
The authors examined and compared patterns of prostate cancer incidence, survival rates, and mortality rates among black men and white men in the United States using the 1969–1999 U.S. prostate cancer mortality rates and the 1975–1999 prostate cancer incidence, survival, and incidence-based mortality rates from the Surveillance, Epidemiology, and End Results (SEER) Program for the U.S. population. The SEER data represent approximately 10% of the U.S. population.
Prostate cancer incidence and mortality rates showed transient increases after 1986, when the U.S. Food and Drug Administration approved the use of prostate specific antigen (PSA) testing. The age-adjusted prostate cancer mortality rates for men age 50–84 years, however, have dropped below the rate in 1986 since 1995 for white men and since 1997 for black men. In fact, for white men ages 50–79 years, the 1998 and 1999 rates were the lowest observed since 1950. Incidence-based mortality rates by disease stage revealed that the recent declines were due to declines in distant disease mortality. Moreover, the decrease in distant disease mortality was due to a decline in distant disease incidence, and not to improved survival of patients with distant disease.
Similar incidence, survival, and mortality rate patterns are seen in black men and white men in the United States, although with differences in the timing and magnitude of recent rate decreases. Increased detection of prostate cancer before it becomes metastatic, possibly reflecting increased use of PSA testing after 1986, may explain much of the recent mortality decrease in both white men and black men. Cancer 2003;97:1507–16. Published 2003 by the American Cancer Society.
During the past 20 years, there have been dynamic changes in prostate cancer incidence rates, due in large part to changing medical practices. The first of these changes occurred with the increased use of transurethral resection of the prostate (TURPS) in the early 1970s to the middle 1980s.1 Subsequently, the prostate specific antigen (PSA) test gained U.S. Food and Drug Administration (FDA) approval in 1986 for use in monitoring prostate cancer recurrence and in 1994 for aiding in the detection of prostate cancer. The diagnostic use of PSA after 1986 led to similar prostate cancer incidence rate increases in white men and black men: Incidence rates rose 108% from 1986 to a peak in 1992 for white males and rose 104% from 1986 to a peak in 1993 for black males.2–5 Declines in distant disease incidence rates began in 1991 and occurred while localized and regional disease incidence rates were still increasing.2, 3, 6–11 Declines in prostate cancer mortality after 1991 have been noted.10, 12, 13
Some have indicated that these patterns provide evidence of beneficial effects of PSA testing.10, 14 They note that the patterns of stage specific incidence rates and mortality rates are consistent with stage migration from distant disease to earlier stages of disease due to PSA screening. That is, the data are consistent with the hypothesis that PSA use led to the increased detection of tumors in the localized or regional stage and that some of these PSA-detected tumors would have been diagnosed clinically a few years later in the distant stage if they had not been detected earlier by PSA screening. This stage migration or stage shift is evidenced by the initial increase in localized/regional disease incidence and subsequent decline in distant disease incidence rates.10, 14
Prostate cancer mortality rates among white men through 1995 showed declining trends after 1991.13 However, it was suggested that the mortality decrease may have been due to errors in death certification associated with the initial large increase and subsequent sharp decrease in prostate cancer incidence rates (i.e., if a certain percentage of deaths in men diagnosed with prostate cancer but dying of another cause incorrectly are assigned prostate cancer as the cause of death, then prostate cancer mortality rates will tend to rise and fall with prostate cancer incidence rates).13, 15 The eventual declines in prostate cancer mortality rates below their 1986 prescreening level for white men, however, cannot be explained by death certification errors because prostate cancer incidence rates at the time of the mortality decrease in the 1990s were still well above the level of incidence rates in 1986, prior to the increased use of PSA testing.16 To further our understanding of the nature of the recent decrease in prostate cancer mortality, we examined the latest prostate cancer incidence, survival, and mortality rates for black men and compared their temporal patterns with the patterns observed in white men.
MATERIALS AND METHODS
Data Sources and Descriptions
Incidence and survival rates were obtained from population-based data collected by the Surveillance, Epidemiology, and End Results (SEER) Program of the National Cancer Institute. Data were used on white men and black men with prostate cancer who were diagnosed from 1975 to 1999 among residents of nine geographic areas: Connecticut, Hawaii, Iowa, New Mexico, Utah, Atlanta, Detroit, Seattle-Puget Sound, and San Francisco-Oakland. The annual incidence rates were age-adjusted to the 2000 U.S. population by direct standardization.17 One-year, 3-year, and 5-year prostate cancer specific survival rates by stage at diagnosis were examined for the diagnostic years 1975–1999. Incidence and survival were calculated using the 2002 SEER*Stat CD-ROM program.17
The categories for tumor extent (disease stage) at the time of diagnosis used in this report were localized/regional disease, distant disease, and unstaged disease. The overall incidence rates were for the total invasive tumors, including unstaged tumors but excluding in situ lesions. Localized disease refers to an invasive neoplasm confined entirely to the prostate. Regional disease refers to a neoplasm that has extended beyond the limits of the prostate directly into surrounding organs, tissues, or regional lymph nodes. Distant disease refers to a neoplasm that has spread to remote sites of the body. Unstaged disease indicates tumors for which insufficient information was available to permit accurate assignment of a stage. Since 1995, SEER reports have combined localized and regional disease in calculating incidence rates for prostate cancer.17 Thus, localized and regional disease were combined for every year since 1975 in analyses of prostate cancer incidence.13
Normally, mortality data are restricted to the information on death certificates, such as race, gender, and age at death. However, in population-based SEER cancer registries, the incidence data on individuals are linked to their mortality outcomes. Therefore, it is possible to examine mortality rates by variables determined at diagnosis, such as stage at diagnosis. This special mortality measure is termed incidence-based mortality (IBM).18, 19 In this report, we examine SEER-area IBM rates by stage at diagnosis to determine which stages are responsible for recent declines in mortality rates. To prevent double counting of prostate deaths in the IBM measures, only sole primary or first primary diagnosed prostate cancer were used in the calculation of the IBM rates.
The prostate cancer mortality rates are calculated from data collected by the National Center for Health Statistics, which receives death certificates from the states and compiles mortality data by race, gender, age, year, and cause of death.20 For the current study, only white men and black men in the United States who reportedly had an underlying cause of death of prostate cancer were included. The mortality rates were age-adjusted to the 2000 U.S. population by direct standardization. The age groups used for the incidence and mortality data include ages ≥ 50 years, 50–84 years, 50–59 years, 60–69 years, 70–79 years, 80–84 years, and ≥ 85 years. For white men and black men, mortality rates are reported from 1969 through 1999.
To allow simultaneous adjustment for age at death, calendar year of death, and year of birth, age-period-cohort models were fit to the prostate cancer mortality rates using 1-year age and calendar-period intervals.21 These analyses were based on 34 1-year age intervals, ranging from age 50 years through age 83 years, and 30 1-year calendar-period intervals, ranging from 1969 through 1998. This resulted in 63 2-year birth-cohort intervals, ranging from 1885–1886 through 1947–1948. The significance of changes in the slope of the calendar-period risk curve or the birth-cohort risk curve were evaluated using linear contrasts.
A standard age-period-cohort analysis is performed using Poisson regression with a logarithmic link and linear predictor, αi + πj + λk, for the rate corresponding to age group i, calendar period j, and birth cohort k.21 To evaluate the change in slope of the calendar-period risk curve in 1991, the following difference between two linear contrasts was used:21 3π1997 + 2π1996 + π1995 − π1993 − 2π1992 − 3π1991 − (3π1991 + 2π1990 + π1989 − π1987 − 2π1986 + 3π1985).
The first contrast characterizes the slope of the calendar-period risk curve between 1991 and 1997, and the second contrast characterizes the slope of the calendar-period risk curve between 1985 and 1991. A significant negative value for this parameter indicates that there was a decrease in the slope of the calendar-period risk curve in 1991. Standard errors of the parameter were adjusted for possible over-dispersion when the deviance for the age-period-cohort fit exceeded the number of residual degrees of freedom.22 A more complete description of parameters to identify changes in the slope of calendar-period or birth-cohort risk curves has been reported.23
The introduction of a beneficial medical intervention usually results in a decrease in the calendar-period risk curve in age-period-cohort analyses of mortality rates, because the impact of improved early detection or improvements in treatment tends to reduce mortality in patients of all ages starting in approximately the same calendar year. Changes in exposure to risk factors usually cause changes in the birth-cohort pattern of risk in an age-period-cohort analysis. Thus, the prostate cancer age-period-cohort analysis can examine whether changes in risk factors are contributing to the declining mortality rates in the 1990s (by looking for decreases in the birth-cohort risk curve that would lead to decreasing rates after 1990). If the decrease in mortality in the 1990s is exclusively a calendar-period phenomenon, however, then the most likely explanation for the decrease is improvement in the early detection and/or treatment of patients with prostate cancer.
For white males, age-adjusted U.S. prostate cancer mortality rates for men ages 50–84 years, 50–59 years, 60–69 years, 70–79 years, 80–84 years, and ≥ 85 years are shown from 1969 through 1999 in Figure 1. The age-adjusted prostate cancer mortality rates for men age 50–84 years peaked in 1991 and declined 27% from 1991 through 1999. For men ages 50–59 years, 60–69 years, and 70–79 years, the 1998 and 1999 rates were at their lowest level since 1950. For all men age < 85 years, the mortality rates after 1995 were lower compared with the rates in 1986, when the FDA first approved the use of PSA. Data on black males are reported in Figure 2, and are discussed below.
To further examine the recent decline in prostate cancer mortality rates, an age-period-cohort analysis was performed (see Figure 3). For white males, there was a significant decrease in the slope of the calendar-period effects curve in 1991 (P < 0.0001). The only major change in the birth-cohort effects curve that would have an impact on recent prostate cancer trends for men age < 80 years was an increase in the slope occurring in the 1930s (P = 0.02). Thus, the recent decrease in prostate cancer mortality rates appears to be exclusively a calendar-period phenomenon, suggesting that the decrease reflects a change in medical practice rather than a change in prostate cancer risk factors.
For black men, age-adjusted prostate cancer rates by age are shown in Figure 2 from 1969 through 1999. The age-adjusted prostate cancer mortality rates for black men age 50–84 years leveled off around 1990 and then declined 17% from 1994 to 1999. For black men ages 50–59 years, 60–69 years, and 70–79 years, mortality rates have dropped below their levels in 1986. The rates were lower than any time since 1969 after 1997 for black men age 60–69 years and in 1999 for black men age 50–59 years (rates for black men are not available prior to 1969).
The slope of the calendar-period effects curve from the age-period-cohort analysis of prostate cancer mortality rates among black men (Fig. 3) decreased significantly in 1991 (P < 0.0001). The birth-cohort effects curve for black men is more dynamic than the birth-cohort effects curve for white men (Fig. 3), but the major decrease in slope around the turn of the century (P < 0.0001) occurred too early to explain the decrease in mortality rates in men age 60–79 years in the 1990s (Fig. 2). Thus, similar to white men, the recent decrease in prostate cancer mortality rates is predominantly a calendar-period phenomenon among black men, suggesting that changes medical practice, and not risk factors, account for the decline.
For both black men and white men, examination of IBM rates by stage at diagnosis (Fig. 4A,B) indicates that the mortality declines that began in the early 1990s are due largely to declining death rates for men with distant disease. Mortality rates in men with localized or regional disease did not begin to decrease until 1997 for white men and have decreased only slightly in recent years for black men. Thus, the substantial decreases in prostate cancer mortality beginning in the early 1990s for both white men and black men cannot be explained by trends in mortality from localized or regional disease.
The declines in distant disease IBM rates may be explained by increasing distant disease survival rates and/or declining distant disease incidence rates. For both white men and black men, total and localized/regional disease survival rates have been increasing since the middle 1980s (Table 1). However, distant disease survival rates have changed little for white men or black men, and have been comparable (Table 1). Distant disease is much more lethal compared with localized or regional disease. From 1992 through 1997, the 1-year and 3-year survival rates for men with distant disease were 81% and 49%, respectively, for white men and 81% and 48%, respectively, for black men. In contrast to these low 1-year and 3-year survival rates for distant disease, the 5-year survival rates for men with localized-regional disease were 96% for white men and 93% for black men. Thus, any intervention leading to increased detection of prostate cancer before it becomes metastatic may have a dramatic and relatively rapid impact on prostate cancer mortality rates.
Table 1. Prostate Cancer Specific Survival Rates by Race and Stage at Diagnosis
Panels C and D in Figure 4 show that distant disease incidence rates declined rapidly for both white men and black men after 1991. This decrease began 5 years after the rapid increase in localized and regional disease incidence rates began and before localized-regional disease incidence rates began to decrease (localized-regional disease rates peaked in 1992 for white men and in 1993 for black men). This pattern is consistent with a stage shift due to increased early detection (prior to 1991) in the localized-regional stage of tumors that would have been diagnosed (in the absence of early detection) after 1991 in the distant stage. The pattern is remarkably similar in black men and white men.
Although the declines in distant disease incidence rates from 1986 through 1999 are rather consistent across age groups, the changes in mortality rates are heterogeneous (Table 2). Smaller decreases or even increases in prostate cancer mortality rates were observed in men age ≥ 80 years despite the fact that marked decreases in distant disease incidence rates and in IBM rates for distant disease were observed in these elderly men. In addition, the declines in mortality rates for black men are smaller compared with the declines for white men, even when the decreases in distant disease incidence rates are comparable. IBM rates for localized or regional disease increased in the 1990s for the oldest men, which diluted the impact of the decreasing rates of distant disease on mortality. Length of survival after a diagnosis with malignant disease depends on the quality of treatment received as well as the disease stage at the time of diagnosis. SEER records include the most aggressive surgical treatment received within 4 months of diagnosis. There is a marked decrease with age in the percentage of patients with prostate cancer who undergo radical prostatectomy, and prostatectomy rates are consistently lower in black men compared with white men. The percentage of patients with localized or regional disease who underwent radical prostatectomy in the years 1991–1997 dropped consistently with age, from 64% in white men and 48% in black men age 50–59 years to 0.7% in white men and 0.4% in black men age ≥ 80 years. Conversely, the percentage of patients for whom there was either no surgical procedure or for whom the most aggressive procedure was a biopsy or TURP increased from 30% in white men and 45% in black men age 50–59 years to 93% in both white men and black men age ≥ 80 years.
Table 2. Age-Adjusted Incidence Rates of Prostate Cancer with Distant Metastases at the Time of Diagnosis and Prostate Cancer Mortality Rates by Race
Age at diagnosis for incidence rates and age at death for mortality rates.
The prostate cancer mortality rates for both white men and black men showed marked declines in the 1990s. The decreasing prostate cancer mortality rates are due primarily to declining distant disease mortality rates, which coincide with declining distant disease incidence rates for both black men and white men. The mortality rate decrease began while prostate cancer incidence rates still were increasing and before the mortality rates for localized and regional disease began to fall. Examination of prostate cancer rates from 1990 through 1999 in Asian Americans and Pacific Islanders and in Hispanics also showed declines in distant disease incidence rates and significant declines in prostate cancer mortality (data not shown). These observations are consistent with the hypothesis that the decreasing prostate cancer mortality in the United States is caused by a stage shift resulting from earlier detection of cancer by PSA testing. That is, tumors that, without intervention, would be diagnosed in the lethal, distant stage are being detected early by PSA testing, so that men are diagnosed in the localized or regional stage; the resulting marked improvement in prognosis leads to decreasing mortality rates.10, 14
There have been recent advances in prostate cancer treatment for patients with locally advanced disease that may be making additional contributions to declines in U.S. prostate cancer mortality rates. Beginning in 1997, a number of studies have shown that patients with locally advanced disease live longer if they receive hormone therapy earlier in the course of their disease.24–30 In the SEER Program, locally advanced tumors are coded as regional disease. Thus, the improved treatments for patients with locally advanced disease cannot explain the observed decrease in mortality rates for men with distant disease. The IBM mortality rates in Figure 4A indicate that localized/regional disease mortality rates did not begin to decline until 1997. Improved treatment of patients with locally advanced disease may well be contributing to these additional recent declines in prostate cancer mortality rates.
At the international level, recent trends in prostate cancer rates have shown inconsistent patterns across countries. Not all countries have trends similar to those in the United States.31, 32 Canada, which has shared similar patterns for breast cancer with the United States,33 has had increases in incidence and subsequent declines in mortality similar to those seen in the United States.34 In the United Kingdom, there have been recent declines in prostate cancer mortality rates, but there has been no prostate cancer screening program and no large increase in prostate cancer incidence.35–37 Trends in survival rates in the United Kingdom do not indicate that prostate cancer treatments are affecting the declines.38 Western Australia has a prostate cancer screening program but has not seen declines in prostate cancer mortality rates.39 Conversely, Tyrol, Austria, where screening has been given, shows mortality rate declines similar to those in the United States as well as evidence that a stage shift has led to decreasing mortality.40 In any country, comprehensive analyses of incidence and survival rates by stage at diagnosis and of mortality rates by age are required to make inferences about the possible causes for the declines or lack of declines in that country. Application of such analyses to U.S. data shows a consistent pattern for both black men and white men.
Evaluation of U.S. trends in distant disease incidence rates and prostate cancer mortality rates by age and race (Table 2) suggests that comparisons of prostatectomy rates also should be included in investigation of international trends. Comparisons of U.S. prostatectomy rates by age and race indicate that the impact of decreases in distant disease rates on overall prostate cancer mortality rates can be obscured in the absence of aggressive surgical treatment of patients with prostate cancer that has not metastasized. Thus, countries with less aggressive surgical treatment of patients with localized prostate cancer compared with the United States will observe a smaller decrease in overall prostate cancer mortality rates, even if the level of PSA screening is equal to that in the United States.
Although descriptive studies cannot provide absolute evidence of cause and effect, the analyses presented in this article show that the observed patterns of prostate cancer rates are consistent with a beneficial effect of PSA testing on prostate cancer mortality. The recent decrease in prostate cancer mortality rates has come at the cost of a very large increase in the number of men treated for prostate cancer since 1986.37 With prostate cancer mortality rates in both white men and black men currently at their lowest levels in several decades for many age groups, a complete delineation of the benefits and limitations of PSA testing and subsequent treatments is needed to allow informed decisions about PSA use.41 This issue is particularly important for black men in the United States who still have some of the highest prostate cancer rates in world.