Fax: (585) 237-1068
Prostate-specific antigen screening for prostate cancer and the risk of overt metastatic disease at presentation †‡
Analysis of trends over time
Version of Record online: 30 JUL 2012
Copyright © 2012 American Cancer Society
Volume 118, Issue 23, pages 5768–5776, 1 December 2012
How to Cite
Scosyrev, E., Wu, G., Mohile, S. and Messing, E. M. (2012), Prostate-specific antigen screening for prostate cancer and the risk of overt metastatic disease at presentation . Cancer, 118: 5768–5776. doi: 10.1002/cncr.27503
See related article:
See related article:
- Issue online: 19 NOV 2012
- Version of Record online: 30 JUL 2012
- Manuscript Accepted: 6 FEB 2012
- Manuscript Revised: 1 FEB 2012
- Manuscript Received: 30 NOV 2011
- prostate cancer;
- prostate-specific antigen;
- metastatic disease;
- incidence rates
The objective of this study was to estimate the total number of patients who would be expected to present with metastatic (M1) prostate cancer (PC) in the modern US population in a given year if the age-specific and race-specific annual incidence rates of M1 PC were the same as the rates in the era before prostate-specific antigen (PSA) testing.
The authors computed the total number of men who presented with M1 PC in the Surveillance, Epidemiology, and End Results (SEER) 9 registries area in the year 2008 (the most recent SEER year) and estimated the number of cases that would be expected to occur in this area in the year 2008 in the absence of PSA testing. The expected number was computed by multiplying each age-race–specific average annual incidence rate from the pre-PSA era (1983-1985) by the number of men in the corresponding age-race category in the year 2008 and adding the products.
In the year 2008, the observed and expected numbers of men presenting with M1 PC in the SEER 9 registries area were 739 and 2277, respectively, with an expected-to-observed ratio of 3.1 (95% confidence interval, 3.0-3.2). If this ratio was applied to the total US population in the year 2008, then the total number of men presenting with M1 PC in that year would be equal to approximately 25,000 instead of the approximately 8000 actually observed.
If the pre-PSA era rates were present in the modern US population, then the total number of men presenting with M1 PC would be approximately 3 times greater than the number actually observed. Cancer 2012. © 2012 American Cancer Society.
In the United States, prostate cancer (PC) is the second most commonly diagnosed malignancy in men (after skin cancer) and is the second most common cause of cancer death in men (after lung cancer), with 217,000 new cases and 32,000 deaths estimated for the year 2010.1 Theoretically, PC mortality could be reduced by early detection through prostate-specific antigen (PSA) screening. The benefits of early detection have been examined in several case-control studies and randomized trials.2-12 Although some case-control studies have suggested that screening may reduce PC mortality or the risk of metastatic (M1) disease, others did not observe this association.2-7 These differences in conclusions potentially may be explained by variation between the studies in the frequency of screening, the frequency of biopsies for elevated PSA levels, and differences in the management of newly diagnosed PC.
The benefits of screening for PC also have been examined in 4 randomized trials in which PC mortality was the primary endpoint.8-12 In the European Randomized Study of Screening for Prostate Cancer (ERSPC), after a median follow-up of 9 years, screening was associated with an absolute reduction in PC mortality of 0.071% (P = .04), with the number needed to screen (NNS) to prevent 1 PC death of 1410.8 In a Swedish trial of PC screening that was conducted in Norrkoping independent of the ERSPC study, after 20 years of follow-up, the investigators reported an absolute risk reduction in PC mortality of 0.30% (NNS = 333) in favor of screening. However, this estimate did not reach statistical significance (P = .065).9 A Quebec study of PC screening and the US Prostate Lung Colorectal and Ovarian (PLCO) cancer screening trial identified no evidence of a reduction in PC mortality in the screening arm relative to control in an intention-to-screen analysis.10, 11 However, both of those trials were affected significantly by low compliance with the assigned intervention. In the Quebec trial, only 24% of men who were randomized to the screening arm accepted the invitation, whereas the PLCO study reported a 52% contamination rate in the control arm because of off-protocol screening.10, 11 The PLCO trial has been reanalyzed with stratification on comorbidities. After 10 years of follow-up, screening was associated with reduced PC mortality for men with no or minimal comorbidities (absolute risk reduction, 0.14%; P = .03; NNS = 723) but not for men with at least 1 significant comorbidity.12 Unfortunately, publication of those findings from randomized trials failed to resolve the controversy surrounding PSA screening. In particular, the US Preventive Services Task Force recently released a draft recommendation against PSA screening for all men that was widely criticized by other experts and professional organizations.13, 14
In addition to randomized trials and smaller observational studies, the association of screening with PC mortality can be examined at the population level using national tumor registry data. In the United States, data from the Surveillance, Epidemiology, and End Results (SEER) Program are used by the National Cancer Institute (NCI) to estimate annual PC mortality rates. According to the NCI, after the introduction of PSA screening, the age-adjusted PC mortality rate decreased each year by an average of 4.1%.1 It is not clear, however, how much of this reduction can be attributed directly to PSA screening, because improvements in local and systemic therapies for PC also occurred during the last 20 years.
In addition to PC mortality, changes in the incidence of M1 PC over time also can be studied. Unlike PC mortality, which can be affected jointly by early detection and treatment, the frequency of M1 PC at initial presentation is not influenced by treatment, although it may be influenced by early detection through PSA screening. It also should be noted that, unlike biochemical failure, overt M1 PC at initial presentation usually is very symptomatic (at diagnosis or soon thereafter) and rapidly fatal, with a median survival of <1 year to 2.6 years, depending on age.15 Hence, substantial reduction in the incidence of this endpoint potentially may represent an improvement in symptom-free and possibly overall survival because of screening rather than just a lead-time effect.
Analyses of recent SEER data indicated that the incidence of presenting with M1 PC increases with age. For example, although they represent only 16% of older men (aged ≥50 years), men aged ≥75 years comprise almost half (48%) of all men who present with M1 PC.15 These age-specific differences were not explained by a lack of screening in the elderly. Indeed, PSA screening of older men has been fairly common during the last decade.16 For example, according to the National Health Interview Survey conducted in the year 2005, the percentages of men who had a screening PSA test in the past year before the survey were 40% for ages 60 to 64 years, 46% for ages 65 to 69 years, 47% for ages 70 to 74 years, 44% for ages 75 to 79 years, 43% for ages 80 to 84 years, and 26% for ages ≥85 years.16 Because the incidence of very advanced PC increases with age even with relatively high screening rates, it is possible that, in the absence of screening, age differences in the risk of presenting with M1 PC would be even more substantial, with much greater risks in the elderly than in younger men. Furthermore, because the absolute number of elderly men in the modern US population is very substantial and continues to increase because of aging of the population, it is possible that, in the absence of screening, the absolute number of men presenting with M1 PC in the modern US population would be much greater than the number actually observed with the current screening practices. The objective of our current study was to explore this hypothesis using data from the SEER Program.
The specific objectives of this study were: 1) to examine age-specific and race-specific annual incidence rates of presenting with M1 PC in the pre-PSA era and to compare them with the rates observed in the most recent years, and 2) to estimate the total number of men presenting with M1 PC that would be expected in the modern US population in a given year if age-specific and race-specific annual incidence rates of presenting with M1 PC were the same as those in the pre-PSA era.
MATERIALS AND METHODS
Data for this study were obtained from the SEER Program.17 Age-specific and race-specific annual incidence rates of presenting with M1 PC were computed for the years 1983 to 2008 using SEER*Stat software. These analyses were based on data from registries that participated in the SEER Program during the entire study period (1983-2008) to assure that changes in the incidence of presenting with M1 PC over time could not be influenced by the addition of new registries in more recent years. Data from the following SEER registries were included in these analyses: San Francisco-Oakland, Connecticut, Detroit, Hawaii, Iowa, New Mexico, Seattle-Puget Sound, Utah, and Atlanta.
For the purpose of this study, M1 PC was defined as: 1) any discontinuous metastases except those described as limited to lymph nodes, or 2) tumors directly extending to pelvic bone or pelvic wall (or beyond), even if initially diagnosed without known, discontinuous spread. Although tumors in this latter category were extremely rare, they had to be included in the definition of M1 PC, because SEER coding rules did not distinguish such advanced tumors from discontinuous, extranodal metastases for cases diagnosed in the years 1983 to 1987.18 We combined both criteria in the definition of M1 PC to use the same definition consistently throughout the entire study period (1983-2008). Similarly, lymph node metastases had to be excluded from the definition of M1 PC, because lymph node involvement could not be classified definitively as regional versus distant in the same way for each year during the study period because of changes in SEER coding rules over time. In particular, superficial inguinal lymph nodes, which were considered distant in the years 1988 to 2008, were coded as regional for those who were diagnosed in the years 1983 to 1987.18-21 To use the same definition of M1 PC consistently throughout the entire study period, we excluded lymph node involvement from this definition. It must also be noted that, according to SEER coding rules, PSA concentration by itself is not a basis for assignment of M1 status. In particular, patients with a PSA concentration above a certain value are not automatically coded with M1 status, without imaging and/or other diagnostic information.
In the current analyses, the numerator for the incidence rate of presenting M1 PC in a given year was defined as the total number of men presenting with M1 PC who were diagnosed in that year in a particular age-race category, and the denominator was defined as the total number of men in the same age-race category. Men with and without prevalent PC were included in the denominator to assure that changes in the incidence of presenting with M1 PC over time were not influenced by an increasing prevalence of screen-detected, nonaggressive PC during the PSA era. This concept is illustrated in Table 1, which provides the example of a hypothetical population of 1000 men at some time (t) under 2 alternative conditions: 1) no screening before t, and 2) screening before t. In this example, the null hypothesis of no screening effect on the risk of presenting with M1 PC is true; because, under screening and under no screening, the same number of men presented with M1 disease. In Table 1, this is accurately reflected by Rate 1, which was computed using the total population size as the denominator. In contrast, the denominator for Rate 2 was computed as the total population size minus the prevalent cases. Because screening increases PC prevalence by detecting a large number of nonaggressive PC cases, Rate 2 under “screening” differs from Rate 2 under “no screening.” This difference, however, is an artifact produced by a difference in the denominators. To eliminate this artifact from analyses, all incidence rates reported in this article were defined according to Rate 1 in Table 1; that is, as the total number of men presenting with M1 PC who were diagnosed in a particular age-race category in a given year divided by the total number of men in that age-race category.
|No. of Patients/Total No. (%)|
|Variable||Without Screening||With Screening|
|Presented with M1 PC||4||4|
|M1 PC rate 1||4/1000 (0.40)||4/1000 (0.40)|
|M1 PC rate 2||4/975 (0.41)||4/800 (0.50)|
Because the PSA test was approved by the US Food and Drug Administration for clinical use in 1986, we used age-specific and race-specific annual incidence rates from years 1983 to 1985 to estimate the average annual incidence rate of presenting with M1 PC in the absence of PSA testing. Data were averaged over a 3-year period to increase the precision of estimation. We did not include data from the earlier years in this calculation, because our aim was to estimate the incidence of presenting with M1 PC in the period just before introduction of the PSA test. An average pre-PSA annual incidence rate was computed for each age-race category as follows. Let C1, C2, and C3 denote the counts of men presenting with M1 PC in a given age-race category in the years 1983, 1984, and 1985, respectively. Likewise, let N1, N2, and N3 denote the number of men in this age-race category in the years 1983, 1984, and 1985, respectively. The average pre-PSA annual incidence rate for this age-race category would be computed as (C1 + C2 + C3)/(N1 + N2 + N3). Similarly, an average annual incidence rate of presenting with M1 PC in the modern PSA era was computed for each age-race category by averaging annual incidence rates for the years 2006 to 2008 (the 3 most recent years of data available from SEER). The effect of screening on the incidence of presenting with M1 PC was expressed as the difference between incidence rates in the pre-PSA era and the modern PSA era. This difference was computed separately for each combination of age and race.
We also computed the total number of men presenting with M1 PC who were diagnosed in the SEER 9 registries area in the year 2008 (the most recent SEER year) and estimated the number of men who would be expected to present with M1 PC in this area in the year 2008 if age-specific and race-specific pre-PSA incidence rates existed in this population. The expected number of cases for each age-race category was computed by multiplying each age-race specific pre-PSA era incidence rate by the number of men in the corresponding age-race category in the year 2008. These products were added to compute the expected number of cases for all age-race categories combined.22 We also approximated the observed and expected number of men presenting with M1 PC in the year 2008 for the entire country. The observed number for the US was approximated by multiplying by 4 the observed number of cases in the SEER 17 registries (ie, all registries participating in the SEER Program since year 2000), which capture approximately 25% of all newly diagnosed PCs in the United States.17 The expected number of cases for the United States was approximated by multiplying the approximated observed number of cases in the United States by the ratio of expected-to-observed number of cases in the SEER 9 registries area.
Time trends in the annual incidence of presenting with M1 PC by deciles of age are illustrated in Figure 1 for white men and in Figure 2 for black men. The temporal trends were similar for both races; however, the absolute risk for black men, on average, was about twice as high as for white men. Conversely, the absolute numbers of men at risk were much smaller among black men than among white men because of the smaller population size. This explains the greater temporal fluctuation (statistical “noise”) observed in the incidence curves in blacks compared with whites. Similarly, in both races, temporal fluctuation increased with age, which is explained by smaller numbers of men at risk in the older age groups. To reduce this statistical “noise,” ages ≥70 years were combined into 1 category in the figures. In both races, among men aged ≥50 years, the incidence of presenting with M1 PC decreased over time during the 1990s, as PSA testing in the general male population became more common. This trend was particularly noticeable in the older age groups.
The average annual incidence rates of presenting with M1 PC during the pre-PSA era (1983-1985) and during the most recent SEER years (2006-2008) are provided in Table 2. Averaging over the 3-year periods was done to increase the precision of estimation for detailed age-specific analysis, as explained above (see Materials and Methods). Note that, in both races, the difference in incidence (the pre-PSA rate minus the PSA-era rate) increased with age until age 85 years; and, even among men aged ≥85 years, the difference in incidence remained greater in magnitude than in most other age groups.
|No. of Patients|
|M1 PC Incidence||95% CI|
|Race||Age Group, y||1983-1985||2006-2008||Incidence Difference||Lower||Upper|
The total numbers of men presenting with M1 PC observed in the SEER 9 registries area during the year 2008 are listed in Table 3 along with the numbers that would be expected in the absence of screening (if the pre-PSA era incidence rates existed in this population). Note that, when all age categories and both races were considered together, the total number of men presenting with M1 PC that would be expected in the absence of screening was 3.1 times greater than the number actually observed (95% confidence interval, 3.0-3.2) (Table 3).
|No. of Men||95% CI|
|Race||Age Group, y||2008 Male Population (SEER 9)||Observed||Expected||Expected-to- Observed Ratio||Lower||Upper|
In the SEER 17 registries, for the year 2008, the total number of men presenting with M1 PC was 2019 (based on the definition of M1 disease provided above; see Materials and Methods). In total, 1988 men (98.5%) had discontinuous, distant metastases. Because the SEER 17 registries capture approximately 25% of all newly diagnosed PC cases in the United States, it can be estimated that approximately 8000 men presented with M1 PC in the United States during the year 2008 (2019 × 4=8076). If our calculated ratio of expected (in the absence of screening)-to-observed cases from the SEER 9 registries area is applied to the entire country, then it can be estimated that approximately 25,000 cases presenting with M1 PC would be diagnosed in the United States in the absence of screening instead of approximately 8000 cases actually observed (8076 × 3.1 = 25,036).
The objectives of this study were to examine age-specific and race-specific annual incidence rates of presenting with M1 PC in the pre-PSA era and to compare them with the rates observed in recent years. We also estimated the total numbers of cases presenting with M1 PC that would be expected to occur in the modern US population in the absence of screening and compared them with the numbers actually observed under the current screening practices. We observed that, for both races, the apparent benefit of screening in terms of the absolute reduction in the risk of presenting with M1 PC increased with age. In particular, differences in the rates during the pre-PSA era and the PSA era in presenting with M1 PC increased with age until age 85 years (Table 2); and, even after this age, the difference in the rate remained greater in magnitude than in most other age groups. Our analyses also suggested that, if the pre-PSA era incidence rates were present in the modern US population, then the total number of men presenting with M1 PC would be approximately 3 times greater than the number actually observed.
When interpreting these findings, it is important to bear in mind 2 significant issues: 1) the possibility of residual confounding, and 2) the lead-time effects. Confounding in this study would occur if reduction in the incidence of presenting with M1 PC observed over time were influenced substantially by factors other than PSA screening. The strongest known determinants of PC risk are age, race, and family history. In the current study, we controlled for age and race directly by stratification. Family history could cause confounding in this study only if the distribution of genetic risk factors changed during the study period within age-race categories. For example, change in the percentage of black men in the SEER 9 registries area during the study period (eg, because of immigration), by itself, would not be sufficient to cause confounding, because adjustment for race was made in the analyses. However, if the patterns of immigration were such that the prevalence of genetic risk factors for aggressive PC changed over time within a given race (eg, among black men), then confounding could occur. Whether this actually took place could not be determined with the available data.
Uncontrolled confounding in this study also may result from changes over time in the distribution of other factors affecting the incidence of aggressive PC. In particular, obesity has been associated with an increased risk of M1 PC in 1 study, and the prevalence of obesity also increased significantly over the last 2 decades.23 Hence, if obesity had a major influence on the incidence of presenting with M1 PC during the study period, then the true benefit of screening may be underestimated in our study. In other words, the true absolute risk reduction because of screening may be greater in magnitude than the risk differences reported in this article. In addition, there were changes in medical imaging techniques over time that may have resulted in an earlier diagnosis of M1 PC for many patients in more recent years. In particular, in the pre-PSA era, computed tomography and bone scans were not as widespread (or of the same quality) as they are today, magnetic resonance imaging and positron emission tomography-computed tomography were not available, and M1 PC often was diagnosed because of symptomatic progression of metastatic lesions. In contrast, in the modern PSA era, M1 PC often is diagnosed on staging imaging for high-risk disease (eg, because of very high PSA). Despite continuously improving capabilities for the detection of M1 PC during the PSA era, the incidence of presenting with M1 PC decreased over time as PSA testing in the general male population became more common (Figs. 1, 2). Nevertheless, it must be recognized that, in addition to changes in the distribution of known or suspected risk factors for aggressive PC and improvements in imaging techniques, changes in the distribution of unknown risk factors also may occur over time. Therefore, our current findings must be viewed primarily as a description of observed time trends rather than as definitive tests of causal hypotheses about screening. In this regard, it is worth noting that, in the European randomized trial of PC screening, a 41% relative reduction in the incidence of presenting with M1 PC was reported (P < .01), which is in line with our observational data.8
The second issue that must be considered in the context of our current findings is the potential lead-time effect resulting from screening. For instance, consider a man who, in the absence of screening, develops screen-detectable, organ-confined PC at age 70 years; develops micrometastases at age 75 years; presents with overt M1 disease at age 80 years; and dies from PC at age 82 years. Screening could prevent this man's presentation with M1 PC at age 80 years, but it may or may not alter his symptom-free and overall survival, depending on the time of screening. If this man were screened between ages 70 and 75 years, then his prostate cancer potentially could be cured. Conversely, screening between ages 75 and 80 years would result in an earlier diagnosis, but not in a cure. In this case, nominal stage at presentation could be M0, if micrometastases remained initially undetected, but overt metastatic disease could still develop by age 80 years. In this later scenario, screening potentially may have no impact on symptom-free or overall survival despite an earlier stage at presentation (the lead-time effect). It is also possible, however, that, even with undiagnosed micrometastatic disease at initial presentation (eg, at age 77 years), metastases could be discovered before age 80 years and treated earlier, thus prolonging symptom-free and possibly overall survival, or at least delaying the complications of M1 disease.24 Furthermore, systemic therapy, which often is given for high-risk disease in the absence of known metastases, may also result in longer symptom-free, disease-specific, or overall survival.25 Similarly, among men with lymph node-positive PC who undergo prostatectomy, many of whom have undiagnosed micrometastases at the time of surgery, earlier (adjuvant) hormone therapy may result in better recurrence-free and overall survival than treatment at the time of disease progression.26
According to Figures 1 and 2, rapid reduction in the incidence of presenting with M1 PC occurred in the early and middle 1990s, soon after PSA screening was recommended by the American Cancer Society and mass PC screening programs were initiated in the United States.27, 28 For example, from 1989 through 1992, hundreds of thousands of men received PSA screening during PC Awareness Week, with a substantial stage shift for newly diagnosed PCs.27 A similar stage shift was reported from other cohorts in the early 1990s.29, 30 This stage shift, which also is reflected in Figures 1 and 2, probably can be interpreted as a mixture of the lead-time effect and the true benefit of screening. Undoubtedly, among men whose presentation with M1 PC was prevented by screening, a certain proportion could experience only a lead-time effect without any improvement in symptom-free or overall survival. Although it is not known how many men belonged to this group, it must be recognized that, unlike biochemical failure, overt M1 PC at initial presentation is usually very symptomatic (at diagnosis or soon thereafter) and rapidly fatal, with a median survival of <1 year to 2.6 years, depending on age.15 Hence, we believe that many men with this disease would likely benefit from earlier diagnosis, either in terms of improved survival or at least from the palliative perspective (eg, prevention or delay of skeletal complications).24
Our analyses suggested that, if the pre-PSA era incidence rates were present in the modern US population, then the total number of men presenting with M1 PC each year would be approximately 3 times greater than the number actually observed. This is equivalent to 17,000 presenting M1 PC cases prevented each year (∼25,000 expected − ∼8000 observed). When interpreting this number, it is important to consider the consequences of screening in terms of over diagnosis and over treatment of nonaggressive PC as well as the financial cost. Unfortunately, these issues could not be definitively investigated with our SEER data. In particular, we could not determine the number of men who would need to undergo a biopsy for an elevated PSA and treated for screen-detected PC to prevent 1 case of presenting with M1 disease. This must be recognized as a limitation of the current study.
Another relevant question that could not be investigated definitively in the current study is the optimal age at which screening should be initiated and stopped. In our analyses, an absolute reduction in the incidence of presenting with M1 PC increased with age until age 85 years; and, even in men aged ≥85 years, the absolute risk reduction was greater in magnitude than in most other age groups. However, age groups in our study represented age in a given chronological year rather than age at the time of screening. It must be recognized that preventing a presentation with M1 PC among men of a given age was achieved by screening these men some years earlier. For example, according to Table 3, among white men residing in the SEER 9 registries area who were ages 80 to 84 years in the year 2008, 347 men would be expected to present with M1 PC if the pre-PSA era incidence rates existed in this population, whereas the observed number of men who actually presented with M1 PC was only 86. Some of these prevented cases potentially were prevented by screening men ages 75 to 79 years, others were prevented by screening men ages 70 to 74 years, and some even may have been prevented by screening men in their 60s. The same considerations apply to the data presented in Table 2. For men in their 40s, no risk reduction was observed for either race (Table 2); whereas, for men in their early 50s, we observed only a small risk reduction among white men and no risk reduction among black men, (Table 2). These findings do not necessarily indicate that PSA screening of men in their 40s is of minimal or no benefit in terms of preventing advanced stage at PC presentation; because such screening, particularly among those with family histories of aggressive PC, may have contributed to the observed risk reduction in some age categories beyond age 55 years (Table 2), although this could not be definitively investigated with our data. Similarly, we could not determine the extent to which screening of men in their late 70s (eg, aged >75 years) contributed to the observed risk reduction in the age categories 80 to 84 years and ≥85 years (Tables 2 and 3). Because of these considerations, we could not make specific recommendations regarding optimal cutoff ages at which screening should be started and stopped, nor could we investigate the optimal frequency of screening.
In summary, the objective of this study was to compare the risk of presenting with M1 PC under the current screening practices versus no screening using historic data from the SEER database. We observed that the apparent benefit of screening in terms of an absolute reduction in the incidence of presenting with M1 PC increased with age. This appears to be a consequence of the much greater risk of highly aggressive PC in elderly men (compared with younger men) in the absence of screening. We believe that these findings are important given the rapidly aging US population and in light of the recent US Preventive Services Task Force recommendation against PSA screening in all men. Our analyses suggest that, if the pre-PSA era incidence rates were present in the modern US population, then the total number of men presenting with M1 PC would be approximately 3 times greater than the number actually observed. We believe that these estimates must be taken into consideration (bearing in mind the limitations of observational data) when public health policy-level recommendations are made regarding PSA screening.
Support for this work was provided by the Ashley Family Foundation.
CONFLICT OF INTEREST DISCLOSURES
The authors made no disclosures.
- 15Prostate cancer in the elderly: frequency of advanced disease at presentation and disease-specific mortality [published online ahead of print October 17, 2011]. Cancer. 2011., , , , .
- 17Surveillance Epidemiology and End Results (SEER) program and population-based research in urologic oncology: an overview [published online ahead of print April 2, 2010]. Urol Oncol. 2010., , , , .
- 18Extent of Disease. New 4-digit Schemes. Codes and Coding Instructions. Bethesda, MD: SEER Program, National Cancer Institute; 1983. Available from: http://seer.cancer.gov/tools/codingmanuals/historical.html. Accessed December 29, 2011., .
- 19SEER Extent of Disease 1988. Codes and Coding Instructions. Bethesda, MD: SEER Program, National Cancer Institute; 1988. Available from: http://seer.cancer.gov/tools/codingmanuals/historical.html. Accessed December 29, 2011., , , et al.
- 20Fritz A, ed. SEER Extent of Disease. Codes and Coding Instructions, 3rd ed. Bethesda, MD: SEER Program, National Cancer Institute; 1998. Available from: http://seer.cancer.gov/tools/codingmanuals/historical.html. Accessed December 29, 2011.
- 21Johnson CF, ed. The SEER Program Coding and Staging Manual 2004, Appendix C. NIH Publication no. 04-5581. Bethesda, MD: SEER Program, National Cancer Institute; 2004.
- 22Introduction to Stratified Analysis. In Rothman KJ, Greenland S, Lash TL. Modern Epidemiology. 3rd ed. Philadelphia, PA: Lippincott Williams & Wilkins; 2008:265-268., .
- 24Immediate versus deferred treatment for advanced prostatic cancer: initial results of the Medical Research Council Trial. The Medical Research Council Prostate Cancer Working Party Investigators Group. Br J Urol. 1997; 79: 235-246.