Presented at the American Cancer Society Workshop: Review of Current Data Impacting Early Detection Guidelines for Prostate Cancer, Phoenix, Arizona, March 10-11, 1997.
Prostate carcinoma incidence and patient mortality†
The effects of screening and early detection
Article first published online: 20 NOV 2000
Copyright © 1997 American Cancer Society
Volume 80, Issue 9, pages 1857–1863, 1 November 1997
How to Cite
Brawley, O. W. (1997), Prostate carcinoma incidence and patient mortality. Cancer, 80: 1857–1863. doi: 10.1002/(SICI)1097-0142(19971101)80:9<1857::AID-CNCR26>3.0.CO;2-3
- Issue published online: 20 NOV 2000
- Article first published online: 20 NOV 2000
- Manuscript Accepted: 11 JUN 1997
- Manuscript Received: 27 MAY 1997
- prostate carcinoma;
- early detection;
Screening for and the aggressive treatment of prostate carcinoma are controversial, but they are nevertheless being practiced in the U.S. Current clinical studies of the effectiveness of screening will take years to complete. Meanwhile, screening for prostate carcinoma is already having an effect on society.
National and regional trends in prostate carcinoma incidence and data on patient mortality and survival from the Surveillance, Epidemiology, and End Results (SEER) Program of the National Cancer Institute are described in this article. SEER is a population-based cancer data base comprised of nine discrete areas. Fundamental principles of screening are used in this article to explain the impact that prostate carcinoma screening has had in the U.S.
According to the data in the SEER registries, overall prostate carcinoma incidence rates increased at a far greater pace than prostate carcinoma mortality rates during the period 1973-1994. During that period, there was a shift in stage at diagnosis characterized by an increase in local and regional disease, and a decline in distant disease at diagnosis. Overall 5-year survival rates for prostate carcinoma patients also increased. The increase in incidence rates, the shift in stage at diagnosis, and the increase in survival rates are all evidence of increasing early detection. However, these changes are consistent with lead-time bias, length bias, a decline in mortality, and all three could have occurred. In the geographic SEER registries, the prostate carcinoma incidence rates vary markedly. These variations in incidence rates are due to regional variations in practice patterns and screening efforts. On the other hand, the SEER registries have comparable mortality rates. This is evidence of both lead-time bias and length bias.
Substantial regional variations in incidence were found, but regional mortality rates were similar. This is evidence that screening and early detection efforts are resulting in the diagnosis of prostate carcinoma in some men who do not need therapy; thus, prostate carcinoma screening can lead to unnecessary treatment for such men. Furthermore, epidemiologic data do not demonstrate that screening is decreasing mortality. The benefits of screening and early detection, although theoretically possible, are yet unproven, whereas the risks and harms of screening and resultant treatment are definite. Cancer 1997; 80:1857-63. © 1997 American Cancer Society.
Prostate carcinoma is a significant cause of mortality and morbidity, and an effective screening test combined with an effective early stage treatment for the disease is extremely desirable. While prostate carcinoma mortality has been rising for some time, recent increases in the rates at which the disease has been diagnosed in the U.S. are stunning and unprecedented for a noninfectious disease. This dramatic rise is due to use of new screening and diagnostic technologies and is largely an American phenomenon. Other countries have not been so quick to embrace screening and have not experienced dramatic rises in incidence rates. If current trends continue, every American male will have a one-in-six lifetime chance of being diagnosed with prostate carcinoma. This has led to concern that a significant number of men will receive unnecessary treatment. Treatment leads to significant morbidity and complication.
The effectiveness of prostate carcinoma screening is controversial and is likely to continue to be for some time. Indeed, it is likely that even after the initial results of current randomized screening studies are reported, the controversy will continue, no matter what the findings. The study of trends in American prostate carcinoma incidence and mortality rates with the application of basic cancer screening principles may yield some clues regarding the impact of screening, clarify why the effectiveness of screening is unclear, and demonstrate why randomized screening trials are vitally important.
This article presents data collected from the Surveillance, Epidemiology, and End Results (SEER) Program of the National Cancer Institute.1 SEER prostate carcinoma data is assessed herein, and screening principles are used to explain the impact that prostate carcinoma screening has had in the U.S.
SEER data was selected because it provides the most complete information on cancer demographics in the U.S. SEER collects population-based information on all cancers diagnosed among nearly all residents of nine defined areas of the U.S., in which approximately 12% of the American population resides. Projections of cancer incidence and mortality are made by using the SEER data and demographic data of the U.S. Census. Comparison of SEER prostate carcinoma mortality rates with overall U.S. prostate carcinoma mortality rates gathered from state vital statistics agencies show that SEER statistics are representative of the U.S. population as a whole.
The incidence and mortality rates of prostate carcinoma for black and white residents of SEER areas during the period 1973-1994, age-adjusted to the 1970 standard, are shown in Figure 1 (9K). Overall prostate carcinoma incidence and mortality rates approximate the white trends, but 10% of all American prostate carcinoma patients are black, and their incidence and mortality differ significantly. Incidence and mortality rates are given as the number of cases per 100,000 white men or black men in SEER areas per year. Age adjustment to the 1970 standard removed the effect of the generalized aging of the American population. The age-adjusted incidence of prostate carcinoma for white Americans increased by more than 130% between 1973 and 1994 and began decreasing in 1992. The increase for black Americans was 140% between 1973 and 1994, peaking in 1993. Between 1973 and 1994, prostate carcinoma mortality rates increased by 18.6% for white men and by 41.4% for black men.
Shift in Stage at Diagnosis
During the period 1973-1994, the increase in incidence rates was accompanied by a shift in stage at diagnosis. There was a rise in the incidence of disease that was apparently localized or regional at diagnosis and a small decline in disease that was apparently distant at diagnosis. These trends were true for blacks and whites within the SEER areas (data for whites is shown in Fig. 2 (7K)). As the rates of localized and regional disease increased, there was an increase in the proportion of prostate carcinoma patients surviving 5 years or longer. Of men diagnosed in 1973, the proportion surviving 5 years was 61%; for men diagnosed in 1981 and 1989, the proportions surviving 5 years were 73.9% and 87.4%, respectively.
Median Age at Diagnosis
Despite increases in incidence rates and increases in detection of early stage disease, prostate carcinoma has remained primarily a disease of the elderly. The median age at diagnosis in the SEER data base was 72 years for white men in 1980 and remained 72 years through 1993. For blacks, the median age at diagnosis remained 70 years. For every 100,000 white men ages 45-50 years, 16 are diagnosed with prostate carcinoma. For every 100,000 black men ages 45-50 years, 26 are diagnosed with prostate carcinoma. The incidence rates increase dramatically with age. For men ages 75-79 years, the age specific rates are 1410 per 100,000 whites and 1921 per 100,000 blacks.
Regional Differences in Incidence
The data cited above describes residents of the nine SEER areas as a whole. There were differences in incidence among the regions. Connecticut and Iowa had the lowest rates. The highest rates were found for Utah, Detroit, and Seattle-Puget Sound. Figure 3 (10K) shows the differences in incidence among white men in these SEER areas during the period 1973-1994. Although the incidence of disease varied by area, the respective mortality rates for the years 1973-1994 were very similar. Figure 4 (9K) demonstrates the incidence and mortality rates by region for white men during the period 1989-1994 in Seattle-Puget Sound and Connecticut, the two areas with the highest and lowest incidence rates. There were also differences in incidence rates and similarities in mortality rates by area among black men with prostate carcinoma (data not shown).
Trends in Incidence and Mortality
Prostate carcinoma screening, especially screening with serum prostate specific antigen (PSA), is profoundly changing the incidence of diagnosis of the disease in the U.S. There is no doubt that the increase in age-adjusted prostate carcinoma incidence is due primarily to increased use of medical technology. Initial rises were attributed to an increase in the number of transurethral resections for benign prostatic hyperplasia.2 Later increases have been attributed to development and use of the transrectal biopsy gun, transrectal ultrasound, and serum PSA.3 The recent decline in incidence may have been due to a clearing of the prevalent cases. It may also be due to the fact that many organizations began recommending against prostate carcinoma screening in the early 1990s. Indeed, by 1993, most American organizations that published screening recommendations recommended against screening or took the position that its benefits were unproven.4
Principles of Screening
The increases in the incidence rates of localized and regional disease are evidence of increased early detection. The increase in survival is also evidence of increased early detection and possibly of improved treatment as well. These findings should be interpreted using basic screening principles.5 A crucial point is that evidence of early detection is not definitive evidence that screening saves lives or decreases mortality. There are several examples in medical history of situations in which cancer screening caused a stage shift and an increase in survival without saving lives; indeed, there have been situations in which screening was net harmful to the population.6, 7 A favorable stage shift and an increase in survival are consistent with lead-time bias, length bias, and a decrease in mortality. Indeed, all three could occur at once.
Lead-time bias and length bias are not theoretic; their existence has been demonstrated in the study of a number of diseases, including prostate carcinoma. These biases can lead one to believe that there is benefit in a screening test when actually there is none.5 There can even be net harm. Only a screening trial that compares the death rate of a group of people randomly assigned to periodic screening with that of a group randomly assigned to observation can overcome these biases. As yet, no randomized prostate carcinoma screening trial has been completed.
Lead-time bias occurs when a screening test prolongs the time the subject is aware of having the disease but does not prolong the subject's life. For true lead-time bias to occur, the treatment course that is given as a result of early detection does not increase the length of life, but patient survival increases. Most screening studies demonstrate that at least one-third of patients diagnosed with prostate carcinoma and treated with radical prostatectomy will eventually relapse.8 Medical science has not definitively demonstrated that early treatment of these men prolongs their lives. Some of these prostate carcinoma studies may be affected by lead-time bias.
Length bias occurs when slow-growing, less aggressive cancers with good prognoses are detected during screening. A type of length bias is termed "overdiagnosis"; this is the detection of tumors that would never cause death and do not need treatment. Gerber et al.9 demonstrated that length bias does occur in prostate carcinoma screening and early detection. Length bias can be very important in prostate carcinoma because this disease is most commonly diagnosed at ages at which other causes of death may be more frequent. Half of all men diagnosed with prostate carcinoma are age 72 years or older, and this median age has not decreased with the rise in prostate carcinoma incidence rates. Some elderly men may have prostate carcinomas that would be clinically significant and eventually cause death if the patients were not going to die of other diseases before the cancer could progress. These men do not benefit from early detection and treatment of prostate carcinoma. Chodak et al.,10 Johansson et al.,11 and Albertsen et al.12 have demonstrated that a number of diagnosed tumors do not need therapy. Autopsy and cystoprostatectomy studies demonstrate that the reservoir of undetected slow-growing tumors is large. Many of these tumors fulfill the histologic criteria of cancer but will never be clinically significant or cause death.13, 14
Examples from Medical History
Screening, whether beneficial or not, can cause a shift in the stage at diagnosis and an improvement in survival statistics without reducing mortality. In the 1960s, a number of organizations advocated lung carcinoma screening because it had been shown to cause a shift in stage and an increase in patient survival. Accrual to randomized clinical trials of lung carcinoma screening was hindered because of these recommendations; however, several of the lung carcinoma screening trials were eventually completed. Not only did these studies demonstrate that lung carcinoma screening did not decrease mortality, but two suggested that the screening actually increased mortality, especially in populations with significant incidence of comorbid diseases.15, 16 The lung carcinoma screening analogy is appropriate because recommendations for prostate carcinoma screening and lack of understanding regarding the controversy surrounding the screening are hindering subject accrual to definitive clinical trials that may settle the issue of whether prostate carcinoma screening saves lives.
The recent Japanese and Canadian experiences with neuroblastoma screening are also examples of situations in which a cancer screening test caused a favorable stage shift and increased survival without decreasing mortality.6, 7 Studies of screening for catecholamine metabolites have led to the concept that there are two kinds of neuroblastoma: a good form that will never progress and an aggressive, usually fatal form that is unlikely to be found at a curable stage with screening. Neuroblastoma screening has caused a significant number of children to undergo unnecessary treatment.
Regional Differences in Incidence Rates
Screening behavior in the U.S. is heterogeneous, and this provides additional evidence that early detection of prostate carcinoma has elements of lead-time and length bias. Prostate carcinoma incidence rates differed among the nine SEER registries throughout the 1980s and into the 1990s. Figure 3 (10K) shows incidence rates for white males at five SEER sites. Seattle-Puget Sound had the highest incidence rates, Connecticut the lowest. Mortality rates were remarkably similar during the period 1973-1994 (Fig. 4 (9K)). The vast majority of diagnosed prostate carcinomas that cause death do so within 10 years of diagnosis,17 so the different incidence rates for several sites beginning in the early 1980s would be expected to correlate with different mortality rates in the early 1990s. This was not observed (Fig. 5 (13K)), suggesting that higher detection efforts led to the discovery of a higher proportion of more indolent tumors in these areas. In addition, the areas with very similar mortality rates had vastly different patterns of treatment.18, 19 It is likely that screening and aggressive case-finding are leading to the identification of some cases in which treatment is not needed. The SEER population data demonstrates that length bias and overdiagnosis do exist for this disease.
There has been a decline in prostate carcinoma mortality in recent years.20 Among white males, this decline was from 24.7 per 100,000 in 1991 to 23.8 per 100,000 in 1994. The decline was small compared with the rise in incidence rates in past years. It is fair to say that the rate at which white men have been diagnosed with prostate carcinoma increased by more than 80 cases per 100,000 men, whereas the death rate decreased by about 1 death per 100,000 men. This decline in mortality is very small and may be insignificant or an artifact. If the decline is real, it is of note that the decrease in mortality was greater in Connecticut, a state with less screening and a lower incidence than in any other SEER registry or in SEER as a whole. In Connecticut, mortality, in fact, declined from 25.3 to 23 per 100,000 white men.
The Prostate Carcinoma Dilemma
In discussing the difficulties of prostate carcinoma diagnosis and treatment, the late Willet Whitmore once said, "When cure is possible, is it necessary? When cure is necessary, is it possible?"21 According to others, the evidence suggests that there are three kinds of men with prostate carcinoma: 1) those for whom cure is necessary but not possible, 2) those for whom cure is possible but not necessary, and 3) those for whom cure is necessary and may be possible.
The difficulty with prostate carcinoma screening is that we have definite proof that Groups 1 and 2 exist, but the existence of Group 3 is theoretic. Group 1 is large; it may comprise one-third of the men currently diagnosed. We do not know what proportion of men who are diagnosed are treated and cured but do not need to be cured.22 This proportion could be significant; and within the SEER program, it is likely that a larger proportion of men have been cured needlessly in areas with higher screening and incidence rates as compared with men in areas with lower screening and incidence rates. It is truly unfortunate that we can diagnose men with prostate carcinoma but cannot distinguish with great accuracy between those who need treatment and those who do not need treatment. It is even more disconcerting that we are uncertain about whether current treatments for localized disease are effective.
Many have advocated screening and aggressive treatment for black men because of their high mortality rates. The patterns outlined above for white men also hold true for blacks. There is overdiagnosis and overtreatment among blacks just as there is among whites. Compared with white men, black men are more likely to have comorbid diseases that complicate aggressive prostate carcinoma treatment. Although black men potentially have the most to gain from screening, they also are the most likely to be harmed by screening.
Prostate carcinoma early detection efforts have been widely employed in the U.S., generally without appropriate caution. The impact of these early detection efforts has increased overall incidence. Screening has yielded vastly different incidence rates by area but identical mortality rates. This is evidence that screening efforts are causing physicians to treat a large number of men who do not need treatment. The complications and harms of prostate carcinoma treatment are well defined,4, 5, 19 but the benefits are not. In addition, randomized trials that are now underway to determine whether current treatments of localized disease are effective have yet to be completed.
Ultimately, well-designed randomized controlled trials must be completed to determine the true benefits of prostate carcinoma screening and treatment. The medical community must recognize the need for these studies and support them.
It is ironic that many have embraced prostate carcinoma screening even though the potential harms of screening and resultant treatment are evident and definite but the advantages are not. The aim of this article is not to argue that screening is inappropriate and should not be performed; rather, it is to say that the benefits of screening, although theoretically possible, are unproven, whereas the risks and harms of screening and resultant treatment are well established. It is this author's personal belief that men who are offered screening should be informed of the uncertainties surrounding it. Although it truly may cure a few men who need to be cured, this benefit may be achieved at the cost of causing a large number of men with prostate carcinoma to undergo unnecessary treatment and resultant morbidity.
Truthfully explaining to patients what we know and do not know about screening is most appropriate. It is appropriate to express opinions either for or against screening, but opinions should be labeled as such, and it should be explained to patients that the appropriate studies have not been completed. A patient can then make a logical, informed decision to be screened or not screened using his own values. Once a patient has made this decision, his choice should be accepted and supported by his physician and the medical community.
The opinions expressed herein are those of the author and do not necessarily reflect the policies or opinions of the National Cancer Institute or the U.S. Department of Health and Human Services.
- 1SEER Cancer statistics review, 1973-1994. Bethesda: National Cancer Institute, 1997., , , , , .
- 21Consensus Development Conference on the Management of Clinically Localized Prostate Cancer. 1988.