NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Harris R, Lohr KN, Beck R, et al. Screening for Prostate Cancer [Internet]. Rockville (MD): Agency for Healthcare Research and Quality (US); 2002 Oct. (Systematic Evidence Reviews, No. 16.)

  • This publication is provided for historical reference only and the information may be out of date.

This publication is provided for historical reference only and the information may be out of date.

Cover of Screening for Prostate Cancer

Screening for Prostate Cancer [Internet].

Show details

3Results

This chapter presents results from our review of the scientific literature pertaining to the 9 key questions listed in Chapter II and identified in the analytic framework. We divide the discussion into subsections as dictated by the topic. Evidence tables providing more details about the design, conduct, results, and quality of the studies reviewed for this report are found in Appendix B; the specific evidence tables are identified in the relevant sections below. Citations to specific publications in the evidence tables represent the articles from a given study that document specific information in the table itself; other papers from the same study that were not used for specific data in evidence tables are not cited there but may be used in the text and cited in this chapter.

Key Question 1: Efficacy of Screening in Reducing Mortality from Prostate Cancer

The first key question, indicated by the overarching arrow in the analytic framework (Figure 1), can be addressed in 2 ways: directly by data from randomized controlled trials (RCTs) or case-control studies of screening for prostate cancer, or indirectly by associating ecologic, population-level data regarding increases in prostate screening with reductions in mortality at the expected time in the expected population. Thus, we undertook 2 separate searches to examine these issues.

The earlier review by the USPSTF found no RCTs of screening and only a single case-control study (which showed no effect of screening by digital rectal examination [DRE] on prostate cancer mortality). No ecologic data were available at that time. 2

For the first search, we accepted only RCTs or case-control studies examining the effect of screening on prostate cancer mortality. We found 1 RCT and 2 case-control studies. Details on these studies can be found in Evidence Tables 1A - 1C (Appendix B).

Randomized Controlled Trial

Labrie et al. completed the first RCT of prostate cancer screening. 20 In 1988, the investigators randomized 46,193 men ages 45 to 80 years registered in the electoral rolls of Quebec City and in the Quebec provincial health registries to 2 groups (ratio of 2:1 favoring invited group). One group was invited to be screened with a prostate-specific antigen (PSA) test (cutpoint = 3.0 ng/ml) and DRE. By the end of 1996, about 23% of the invited group and 6.5% of the not-invited group had actually been screened. This low adherence rate reduces the power of the study to detect a true difference that could be attributable to screening.

The authors analyzed the study by combining all men from both the invited and the not-invited groups who were actually screened, comparing their prostate cancer mortality with that for the men in both groups who had not been screened. They calculated a 69% reduction in prostate cancer mortality from screening.

Using data presented in the paper, an intention-to-treat analysis can be conducted. Among the 30,196 men in the invited group, 140 deaths from prostate cancer occurred (4.6 per 1,000); among the 15,237 men in the not-invited group, 73 deaths occurred (4.8 per 1,000). Because of low adherence to screening in the invited group, or because of lack of efficacy of screening, this study does not provide evidence to support the practice of prostate cancer screening.

Two other RCTs of screening for prostate cancer, both initiated in 1994, are ongoing. The National Cancer Institute's "Prostate, Lung, Colorectal, and Ovary" (PLCO) Trial is randomizing 74,000 men ages 60 to 74 years at 10 study sites to annual screening for 4 years with DRE and PSA compared with usual care. The European Randomized Study of Screening for Prostate Cancer (ERSPC) is randomizing 190,000 men ages 50 to 75 years in 7 countries to screening with PSA, DRE, and transrectal ultrasound (TRUS) or usual care. In 1998, the ERSPC investigators changed their screening approach to PSA alone, with a cutpoint of 3.0 ng/ml. Neither of these studies will have data on mortality from prostate cancer for several more years.

Case-Control Studies

The 1996 USPSTF review reported on a case-control study that found no evidence that DRE prevents late-stage prostate cancer (odds ratio [OR] = 0.90; 95% confidence interval [CI], 0.5-1.7). 21 Since that time, 2 additional nested case-control studies have provided mixed results. All 3 studies had similar designs and were well conducted.

Richert-Boe et al. conducted their case-control study among patients of a large health maintenance organization (Kaiser Permanente Northwest). 22 Cases were 150 patients who were 40 to 84 years of age when their prostate cancer was diagnosed and who died of the disease. Investigators selected 2 controls per case, matched for age and membership in the health plan. They then examined medical records to determine whether a previous DRE had been done and whether the DRE was performed for screening or diagnostic reasons. A similar number of cases and controls had had a screening DRE during the 10-year study period (OR = 0.84; 95% CI, 0.48-1.46).

Jacobsen et al. conducted a similar study among residents of Olmsted County, Minnesota, using the unified data system of the Rochester Epidemiology Project. 23 Investigators identified 173 patients who had died of prostate cancer as their cases and matched them to 346 patients as controls (2 per case). As in the previous studies, this research team used medical record reviews to determine whether each patient had had a DRE and whether it had been done for screening or diagnostic reasons. DREs performed during the year immediately before diagnosis were eliminated, because the investigators thought that these could well have been done for diagnostic rather than screening reasons. Control subjects had had more DREs between years 2 and 10 before diagnosis than case subjects (OR = 0.51; 95% CI, 0.31- 0.84), indicating a protective effect of DRE.

The Jacobsen et al. results were robust to a number of different analyses, such as excluding cases (and their matched controls) whose deaths may not have been due to prostate cancer, excluding DREs performed in the presence of symptoms that may have indicated prostate cancer, and adding a comorbidity index as a potential confounder. When they examined the data by year of most recent DRE, the investigators found the same odds ratio for every year up to 6 years before diagnosis but not for more distant DRE.

The reasons for the differences in the results from these otherwise similar studies are not clear. One group has suggested that eliminating DREs in the year before diagnosis in the Jacobsen et al. study adds bias, 24 but a major problem in all 3 studies is distinguishing between screening and diagnostic DRE. 25 We also note that all 3 studies are small, and all are consistent with an effect of DRE of up to 50% reduction in prostate cancer mortality.

We found no case-control studies of PSA screening. This can be explained at least in part by the fact that insufficient time has elapsed since the introduction of PSA as a screening test (late 1980s). Thus, its impact on prostate cancer mortality would be difficult to assess. Such studies are planned, however. 26

Ecologic Data

For the second search, we accepted only studies of prostate cancer surveillance over time that associated indicators of screening with mortality. We found 7 such studies,9,28,30,31,33,42,205 although only 2 actually relate screening and mortality in a quantitative manner. An eighth study used national data to model the effect of changes in screening on changes in mortality, given various assumptions about the natural history of prostate cancer. 33 Details can be found in Evidence Table 1C (Appendix B).

The 2 quantitative studies are from investigators at the National Cancer Institute (NCI), using incidence data from the Surveillance, Epidemiology, and End Results (SEER) system, together with mortality data from the National Center for Health Statistics.9,27,28 They document several trends within the United States: a dramatic increase in age-adjusted prostate cancer incidence that accompanied increased screening in 1989, the peaking of incidence rates in 1992, and the subsequent decline. In 1991, the incidence of distant-stage disease began to decline (for all races and all SEER areas), with a decline in localized and regional disease beginning in 1992. The reduction in the incidence of distant-stage disease has been dramatic: an annual reduction of 17.9% since 1991 in white men.

With regard to prostate cancer mortality, between 1969 and 1987 the age-adjusted rates gradually increased for both whites and African Americans. Between 1987 and 1991, the mortality rates increased at an accelerating rate, 11% increase for white men and 14% increase for African American men. In 1991, mortality rates for whites began to decline and, in 1992, rates for African Americans followed suit: 16.1% decline for white men between 1991 and 1997 and 10.9% decrease for African American men from 1993 to 1997. (Preliminary data showed a continued decline in 1998.) Mortality rates declined in all age groups at about the same time.

The NCI investigators considered several potential factors to explain these trends. One possibility is screening with PSA. PSA testing began to increase at about the time of increasing prostate cancer incidence. A study of Medicare data found that the percentage of white men older than age 65 who had received a PSA test in the previous year increased from 1.2% in 1988 to about 40% in 1994. 34 The pattern of increased incidence followed by decreased incidence, decreasing late-stage disease, and then reduced mortality is what one would expect from the application of an effective screening program. In addition, the fact that mortality started to decline for all races and age groups at about the same time (i.e., calendar period effect) lends support to a global effect that affected all groups similarly.

One problem with ascribing the ecologic trends to screening is the timing of the decline in mortality. With cancers such as prostate, considered to be generally slow growing,35,36 the time between the application of an effective screening program and an expected reduction in mortality is a matter of many years, whereas in this case a decline in mortality was seen only 2 to 3 years after widespread screening. Using an NCI model, investigators found that PSA screening could explain this trend only with several assumptions. 33 If one assumes that PSA screening reduces prostate cancer mortality by 20%, and if the mean lead-time is no longer than 3 years, then the fall in mortality can be explained largely by screening. Earlier estimates of the mean lead-time of screening for prostate cancer, however, had been 5 years or longer. 35

The argument that the decline in mortality can be attributed to PSA screening would be stronger if one could show that the decline is largest in areas with more screening. To date, data are conflicting about this issue.29,37,38

Another problem with concluding from the ecologic data that screening is effective is the presence of alternative explanations for the trends. Investigators have offered 3 general hypotheses. The first 2 explanations agree that PSA likely accounts for the increased incidence of prostate cancer but offer different explanations for the decline in mortality. The third explanation postulates the existence of unknown factors.

The first alternative explanation (attribution bias) suggests that misattribution of deaths to prostate cancer that are actually caused by other conditions may account for the trends outlined above. This possibility is suggested by several facts: (1) death from prostate cancer often occurs in older men with multiple comorbid conditions; (2) studies have found inconsistencies between medical record review and death certificate causes of death in men with prostate cancer;8,39 and (3) the mortality curve for prostate cancer closely parallels the incidence curve (both its rise and fall). If the percentage of deaths attributed mistakenly to prostate cancer is stable, then one would expect that the prostate cancer mortality rate would increase and decrease in close approximation with the prevalence (and thus with the incidence) of prostate cancer in the population. 28 Studies to investigate misclassification of prostate cancer deaths are under way.

The second alternative explanation is that improved treatment has reduced mortality. During the late 1980s and early 1990s, 3 major treatment changes emerged: (1) rates of radical prostatectomy increased; (2) luteinizing hormone-releasing hormone (LHRH) agonists and antiandrogen agents were developed, and this allowed for improved androgen deprivation without castration; and (3) refinements were made in radiation therapy, such as 3-D conformal radiation.

The third possible explanation for these puzzling trends is that changes in one or more unknown risk factors are increasing both the incidence and the mortality from prostate cancer. This alternative seems less plausible than the others for several reasons: the rates have later declined, the changes occurred so dramatically, and the trends affected all age groups at the same time (whereas risk factor changes typically affect some groups before others). Other experts have noted, however, that mortality from several cancers has been declining recently and that this trend is not completely understood. 40

Further international analyses of reductions in prostate cancer mortality have been published. Quebec and Canada as a whole, 30 as well as England and Wales, 32 have experienced decreases in the mortality rate from this disease in a pattern similar to that seen in the US SEER data. 30 Within the United States, a population-based analysis from Olmsted County, Minnesota, also shows similar findings. 31

A recent ecologic analysis from Austria found that prostate cancer mortality in Tyrol, an area with a free screening program, began to drop below that of the rest of the country a few years after screening began.41,42 This finding could be attributed to the screening program, changes in treatment that accompanied the program, attribution bias, or some combination of the three.

Summary of Results on Efficacy of Screening

We found a single RCT of PSA screening with low screening adherence and poor data analysis; 3 well-conducted nested case control studies (2 since the second Guide to Clinical Preventive Services appeared in 1996) of DRE screening with mixed results; and ecologic evidence that is suggestive but not conclusive of a benefit of screening, largely because of the timing of mortality trends and the presence of alternative explanations. If screening is effective, we are not able to determine to any degree of precision from these data the magnitude of the benefit.

Key Question 2: Yield of Screening for Prostate Cancer

The second key question, indicated by arrow No. 2 in the analytic framework (Figure 1), deals with the yield of screening for prostate cancer. Ideally, we would like to determine what type of prostate cancer is the most appropriate target for screening, the prevalence of this type of cancer, and the sensitivity and specificity of available screening tests for detecting this type of cancer. We first consider 2 methodologic issues involved in these questions: our knowledge about the appropriate target of screening and the optimal reference standard test for use in defining the sensitivity and specificity of screening tests. We then consider estimates of the sensitivity and specificity of PSA screening and, by comparison, the accuracy and reliability of other screening strategies. Finally, we examine studies of the yield of large screening programs. (Evidence Tables 2A-2B)

Methodologic Issues

Cancers to Target

Prostate cancer has a heterogeneous natural history. Autopsy studies have found occult prostate cancer in some 30% of men ages 50 and older who have died of other conditions. 13 Although the lifetime risk of being diagnosed with prostate cancer is about 16%, the lifetime risk of dying of this disease is about 3.4%. 14 The discrepancy between these numbers is similar to the discrepancy noted earlier between the annual number of men diagnosed with prostate cancer and the number dying from it. It shows that, although some prostate cancers cause suffering and death, others are clinically unimportant, i.e., they would never cause symptoms within the life span of a typical man.

Ideally, screening would target only those cancers that are destined to cause clinically important disease. What is not clear is how to distinguish clinically important from clinically unimportant prostate cancer.

Most clinicians and researchers have defined clinically important cancers as those that are localized (i.e., intracapsular or organ-confined) and have either a large enough volume or a high enough grade that they appear to have the potential to grow beyond the prostate. Theoretically, this type of cancer can be cured by prostatectomy or radiation therapy. For example, Schroder et al., 43 in a population-based screening study, found that 62% of men with prostate cancer detected by having a PSA between 4.0 ng/ml and 9.9 ng/ml had cancer confined to the prostate gland, and 32% were both organ-confined and had a Gleason score of 7 or greater, indicating a high potential to grow.

In this model of screening, clinically unimportant cancers are both intracapsular and have no characteristics associated with further growth. In most screening studies, investigators find only a small percentage of cancers that meet these criteria. For example, in a screening study of volunteers utilizing PSA and DRE, Catalona et al. 44 found that only 8% of cancers detected by screening were organ confined, well differentiated, and involved 1 prostate quadrant. 45

The size of the discrepancy between diagnoses and deaths indicates that this model cannot be exactly correct. Not everyone with clinically important cancers by these criteria dies of prostate cancer; the criteria for defining clinically unimportant cancers are too restrictive.

Statistically, the characteristics chosen by these researchers to define clinically important (or unimportant) cancer are correct. Pathologic stage at diagnosis, histologic grade of the tumor, tumor volume, patient age, and PSA level are associated with prognosis. Men with good combinations of all these factors have an excellent prognosis (and their cancers are likely not clinically important); men with bad combinations of all of these factors have a poor prognosis (and their cancers are likely clinically important).

Unfortunately, the great majority of men with screen-detected prostate cancer fall between these extremes; their prognosis remains uncertain.46-50 Research has yet to define factors for this large "intermediate" group that discriminate well between those men with cancers that are destined to cause suffering and death and those men with cancers that will cause no or minimal symptoms. 51

In addition to refining the above model with better criteria, we might consider another model. Although some organ-confined prostate cancers are clinically important, a second group of cancers may also be important in the sense of being responsive to earlier treatment. These are tumors that have invaded locally beyond the capsule (i.e., "extracapsular" or "locally invasive"), with or without metastases to distant structures. If treatment of these cancers could delay their progression, even if the men were not cured, screening might still provide substantial benefit. Recent trials showing the effectiveness of androgen deprivation treatment for locally advanced cancers should at least raise the question of whether such cancers should be considered an appropriate target for screening. These points are taken up again in the discussion of Key Question 6, below.

Thus, the definition of clinically important prostate cancer, the target of screening, is not yet settled. Among the unresolved issues are which organ-confined cancers are most important to find and treat and whether some extracapsular tumors should be regarded as appropriate screening targets. The lack of a clear, evidence-based definition of clinically important cancer makes it impossible to determine the extent to which screening detects clinically unimportant cancer, a critical issue in the screening controversy, and this problem in turn complicates calculating the potential benefits and harms of screening.

Those who claim that screening detects only a small number of clinically unimportant cancers argue that PSA and DRE are sensitive enough to detect clinically important cancer, but not sensitive enough to detect clinically unimportant cancer. A retrospective analysis from the Physicians' Health Study has been cited in support of this idea. 35 In this study, conducted before widespread PSA screening, investigators drew blood and froze the sera at the beginning of the study. After the study, they analyzed the sera for PSA. The PSA results for men who had been diagnosed with prostate cancer (N = 366) were compared with those of a matched control group. Using a PSA cutpoint equivalent to 4.0 ng/ml to define abnormal, the investigators found that the PSA test was more sensitive for cancers that were labeled "aggressive" (i.e., extracapsular or higher grade) than for those that were labeled "nonaggressive" (i.e., intracapsular and lower grade). Although provocative, this study still leaves unanswered the question of how many of these detectable prostate cancers were clinically important, as about 90% of men with prostate cancer did not die of this disease over the follow-up period. 35

Further research is needed to define better the factors that discriminate between different prognostic types of prostate cancer. In the meantime, the percentage of prostate cancers detected by screening that would never cause serious clinical symptoms and the prevalence of cancers that would cause such symptoms are both unknown.

Reference Standard

To calculate the sensitivity and specificity of a screening test, ideally, one should compare the results of a screening test with a standard reference test that has been applied uniformly among all those screened. The usual reference standard used in prostate cancer screening studies is transrectal needle biopsy of the prostate, but this test is rarely done in the absence of a positive screening test.

Even if it were done uniformly in screening studies, prostate needle biopsy is an imperfect reference standard for 2 reasons. First, it misses some cancers; from 10% to 20% of men who had had a negative initial series of biopsies have cancer on repeat biopsy series.52-56 Thus, some men categorized as not having cancer actually do have it, falsely lowering the measured sensitivity.

Second, in clinical practice and research, a "biopsy" is actually from 4 to 6 (or more) biopsies. Multiple biopsies are taken, most from normal-appearing areas of the prostate. An analysis of this practice concluded that up to 25% of apparently PSA-detected tumors and more than 25% of apparently DRE-detected tumors were likely in fact to have been detected by serendipity, that is, an incidental finding from a blind biopsy. 57 Thus, some men who are categorized as having cancer detected by screening actually have serendipity-detected cancer. This again falsely increases sensitivity.

Another possible reference standard is longitudinal follow-up. Men who do not develop clinical prostate cancer over an extended period of time did not have clinically important cancer. Other than the retrospective Physicians' Health Study described above, we found no study that used longitudinal follow-up of screened and nonscreened populations as a reference standard.

Accuracy of Screening

The most common screening tests for prostate cancer are PSA and DRE. TRUS has been largely abandoned as a primary screening modality because of its high cost, inadequate reproducibility, and inadequate sensitivity. We discuss here the operating characteristics of the PSA, the DRE, and proposed variations on the PSA test.

Because of the problems of an imperfect reference standard that is not uniformly applied and the lack of an evidence-based definition of what cancers should be the target of screening, the sensitivity and specificity of screening tests for prostate cancer cannot be determined with precision. Cancers that are actually present may be missed; cancers may be detected serendipitously and attributed to screening; and cancers that have no clinical importance may be detected and counted as true positives rather than as false positives.

Screening with PSA

The Physicians' Health Study avoids some of the bias of the problematic reference standard by employing longitudinal follow-up as a reference standard. 35 Nevertheless, DRE screening may well have occurred (as this study involved a group of physicians with ready access to health care). The sensitivity of a PSA of 4.0 ng/ml or higher for detecting aggressive prostate cancers that appeared within 2 years of screening was about 91%; the sensitivity for detecting nonaggressive cancers within the same period was about 56%. The sensitivity for cancers appearing within 4 years was 87% for aggressive cancers and 53% for nonaggressive cancers. Among men who were not diagnosed with prostate cancer over 10 years, 9% had an initial PSA of 4.0 ng/ml or greater (i.e., specificity of 91%).

With the methodologic concerns above, other studies provide estimates of sensitivity for PSA with a cutpoint of 4.0 ng/ml of 63% 58 to 83%. 59 A recent population-based screening study estimated the sensitivity of this strategy to be 73%. 43

Mettlin estimated the specificity of PSA (cutpoint of 4.0 ng/ml) to be about 90% for the first screening round, 58 and Jacobsen et al. found declining specificity with age: 98% for men in their 50s to 81% for men in their 70s. 59

PSA has a lower specificity among men with larger prostate glands. This includes the large number of older men with BPH. One study of 4 carefully chosen populations found, for example, that the likelihood ratios for various PSA levels were much lower among men with BPH than among men without BPH. 60 Thus, the PSA does not distinguish cancer as well among men with BPH as among those without BPH.

Because of the reduced specificity in older men with BPH, some experts have proposed that the PSA cutpoint be adjusted for age, with higher cutpoints for older men and lower cutpoints for younger men. 61 Age-adjusted cutpoints might be different in African American and white men. 62 Clearly, this change would increase cancer detection among men in their 50s (i.e., increase sensitivity and reduce specificity) and reduce detection among men in their 70s (i.e., decrease sensitivity and increase specificity). One study found that this strategy had little impact on overall specificity and missed more cancers than the non-age-adjusted PSA strategy. 63 Another multi-site study found that age-adjusted PSA cutpoints did improve specificity, but at the cost of a large reduction in sensitivity among older men. They found that other screening strategies (such as percent free PSA [%fPSA], see below) were superior in maintaining overall accuracy. 64 Oesterling argues, however, that improving sensitivity in younger men and specificity in older men actually improves the likelihood that the test will reduce prostate cancer mortality. 65

To improve the detection of clinically important cancers, some have proposed decreasing the cutpoint for defining an abnormal PSA for all men from 4.0 ng/ml to 3.0 ng/ml (or even 2.6 ng/ml).66,43,67 Studies of patients with PSA between 2.6 ng/ml or 3.0 ng/ml and 4.0 ng/ml who were screened with DRE or TRUS, or who were offered biopsy at this lower PSA value, have found that 12% to 23% of these patients have prostate cancer. 68 Decreasing the cutpoint for PSA from 4.0 ng/ml to 2.6 ng/ml or 3.0 ng/ml would increase the percentage of men undergoing biopsy by an absolute 6% to 11%.67,69-71

Some have argued that many of these cancers are clinically important and thus need to be detected and treated.69,72 The value of this increased detection, however, is unknown. In one study, 80% of the cancers detected among men who had PSA values between 2.6 ng/ml and 4.0 ng/ml (and who had surgery) were pathologically organ confined; 17% were low volume and low grade (the authors' definition for clinically unimportant). 68

Screening with Variations on the PSA

PSA Density

Because of the problem of reduced specificity in older men with BPH, Benson proposed that the PSA be adjusted for prostate volume as measured by TRUS.73,74 This test, called the PSA density (PSAD), is expressed in ng/ml PSA/cc prostate gland. Higher values indicate a higher probability of prostate cancer. Cutpoints from 0.078 to 0.15 have been used in the research literature.

Several research groups have tested the PSAD at various cutpoints; none found a large advantage beyond simple PSA testing.64,75-78 Because the PSAD is more expensive and logistically difficult, and because some investigators have found that the TRUS lacks reproducibility, 79 the PSAD test has fallen out of favor as a primary screening test.

Percent free PSA (%fPSA)

In the serum, PSA circulates in 2 forms: free and complexed with such molecules as alpha-1 antichymotrysin. Men with prostate cancer tend to have a lower percentage of their PSA in the free form compared with men without prostate cancer.80,81 Thus, %fPSA (in which higher values are more "normal") has been proposed as a new test to improve the specificity of the total PSA assay. Various cutpoints have been used; an abnormal test, indicating possible cancer, may be a value below 15% 82 to 30%. 83

Several studies have examined the %fPSA test. Its major use in research has been to increase the specificity of screening by distinguishing between men with PSA between 4.0 ng/ml and 9.9 ng/ml who should be biopsied and those who should not. Using various cutpoints, from 20% to 40% of biopsies could potentially be avoided, although 2% to 15% of cancers would then be missed.82,84-88

Catalona et al. 44 proposed %fPSA as a second stage-screening test for men with PSA between 2.6 ng/ml and 4.0 ng/ml.45,68 These investigators were able to define a cutpoint that would hypothetically detect 90% of cancers while avoiding 18% of biopsies.

Like lower total PSA, higher %fPSA has been associated with better stage and histologic markers of prognosis among men with prostate cancer. 64

An important question with %fPSA, however, is how useful it actually is in clinical practice. To be useful, a negative %fPSA would have to reduce the probability of prostate cancer to a low enough level that men would be willing to forego biopsy.

A systematic review of studies examining %fPSA found that, using the authors' cutpoints for an abnormal test, a man with a PSA of between 4.0 ng/ml and 9.9 ng/ml would still have a probability of prostate cancer of 8% after a negative %fPSA test. Although this additional information represents a decrease in the man's probability of having prostate cancer, it is uncertain whether the reduction goes low enough for most men to forego biopsy. In practice, therefore, the test may or may not reduce the biopsy rate. 89

A similar test, the amount of PSA complexed to alpha-1-antichymotrysin (complexed PSA) has also been shown to enhance specificity relative to total PSA, especially at lower levels of PSA.90-93 Again, the issue remains whether this increased specificity is adequate to reduce the number of biopsies in actual practice.

PSA Velocity

In a small study, Carter noted that men with prostate cancer have a greater increase in their PSA over time than men without cancer. 94 Thus, he proposed that the annual rate of increase in PSA (PSA velocity) be considered as a way of increasing the specificity of the PSA test, using a cutpoint of PSA increase at or greater than 0.75 ng/ml per year. In other studies, this degree of change was neither sensitive nor specific for detecting cancers found by other screening tests.63,95 Because of intraindividual variation, PSA velocity is most useful in men who have 3 or more PSA determinations each separated by a year.63,96,97

Screening with DRE

DRE has a lower ability to detect cancer than PSA. A meta-analysis of DRE studies of unselected populations screened by both PSA and DRE found a sensitivity of 59% (64% for the 4 best studies). 98 A recent study not included in the meta-analysis found that, although DRE found some cancers in men with PSA levels below 4.0 ng/ml or even 3.0 ng/ml, these cancers were usually small and well differentiated. 99 In another large screening study of volunteers, the overall cancer detection rates were as follows: DRE alone, 3.2%; PSA (cutpoint 4.0 ng/ml) alone, 4.6%; the combination, 5.8%. 44 A Canadian screening study found that about 90% of detected cancers would have been found by PSA screening alone. The investigators calculated that 344 men with a normal PSA would need to be screened to find a single prostate cancer. 100

A final factor that dilutes the usefulness of DRE is its limited reproducibility. In 1 small study, the kappa of agreement among 8 urologists, fellows, and residents was 0.22. 101

Studies of the Yield of Large Screening Programs

We found 6 screening studies of large populations using either PSA or a combination of PSA and DRE as the screening test followed by multiple-core prostate biopsy as the diagnostic standard.20,43-45,67,102-106 Each study reported on a single screen among men, most of whom had not previously been screened. One study recruited volunteers from the areas around 6 medical centers; 45 44 the other 5 were population-based studies of men accepting an invitation to be screened. All studies included men beginning at age 45 to 55 and ending at age 67 to 80. Other studies have screened large populations but have used different screening strategies (e.g., American Cancer Society-Prostate Cancer Detection Project [ACS-PCDP]). 58 Using the results of these studies, we can estimate the yield of a screening program for men of different ages (Figures 2-4).

Figure 2: Ages 50-59 - estimated yield of screening with PSA (or PSA and DRE) (prevalence screen).

Figure

Figure 2: Ages 50-59 - estimated yield of screening with PSA (or PSA and DRE) (prevalence screen).

Figure 4: Ages 70-79 - estimated uield of screening with PSA (or PSA and DRE) (prevalence screen).

Figure

Figure 4: Ages 70-79 - estimated uield of screening with PSA (or PSA and DRE) (prevalence screen).

The percentage of participants with a PSA of 4.0 ng/ml or higher ranged from 6.5% (in a younger cohort from Spain) to 14.8% (in an older population of white volunteers from the United States); the percentage with PSA of 3.0 ng/ml or higher ranged from 14% (Finland) to about 20% (Quebec and Rotterdam). The additional percentage of men who had an abnormal DRE but a PSA less than 4.0 ng/ml ranged from 2.2% (Spain) to 11.0% (US volunteers). The total percentage of men with either a PSA greater than or equal to 4.0 ng/ml or a positive DRE is between 8.7% (Spain) and 25.8% (US volunteers).

These results varied by age group.44,45,102,104,105 The percentage of men with a PSA of 4.0 ng/ml or more, for example, was about 3% for men in their 50s and rose to 11% to 17% for men in their 70s. Among the US volunteers, about 15% of men in their 50s and 40% of men in their 70s had either an abnormal PSA or positive DRE. 45

Many men with abnormal screening tests had prostate biopsies; some had prostate cancer detected. The percentage of biopsies that detect cancer and the percentage of men screened who have cancer detected both depend on the prevalence of detectable cancer in the population screened, and thus these figures increase with age. The percentage of men screened who have cancer detected also depends on the percentage of men with an abnormal screening test who consent to having a prostate biopsy. Thus, studies of older, previously unscreened populations with high biopsy rates have a higher cancer detection rate.

In these 6 studies, the percentage of biopsies that detected cancer ranged from about 10% 43 to about 30%.20,67,102,103,106 For men in their 50s, this percentage ranged from about 6% 105 to about 20%; 104 for men in their 70s, to nearly 30%.44,45,107

The percentage of all men screened who were found to have prostate cancer ranged from about 1.2%104,105 to 4.5%. 43 For men in their 50s, this percentage ranged from 0.2%104,105 to 2.0%;44,45 for men in their 70s the range was 3.0% 104 to 7.2%.20,44,45,67,103,106

All 6 studies reported some information on staging (either clinical or pathologic) or histologic grading of the tumors detected by screening PSA.20,44,45,67,102-106,108 In 2 studies, screen-detected tumors were 60% to 70% clinically organ confined.20,67,103,105,106 Three other studies found that, of those men who had prostatectomy after cancer detection by screening, about 70% were pathologically organ confined.44,45,109,110 Two studies67,103,105,106 reported that 8.4% to 12.1% of screen-detected prostate cancers were metastatic at diagnosis;20,67,103,106 1 reported that less than 1% were found to be metastatic on later screening rounds.20,106,67

The percentages of screen-detected tumors that were well differentiated (i.e., Gleason score 2-4) varied widely, ranging from 1%43,104 to 67%. 105 The percentage of screen-detected cancers that were poorly differentiated varied less: from 5%43,104 to 10%.20,67,103,106 Other center-based studies have found that a small percentage of screen-detected cancers are well differentiated and that the great majority (up to 95%) are moderately differentiated.44,111,112

Earlier series of newly diagnosed prostate cancer not detected by screening had shown that 50% or more were at the extracapsular stage and that a higher percentage of the tumors were poorly differentiated.1,113

Variation in Yield with Different Screening Intervals

Two studies provided information about how the rates of positive screening tests and cancer detection vary by repeated annual testing.20,67,106,114 The percentage of men with a PSA of 4.0 ng/ml or greater was 10% to 12% on the initial screening round and dropped to about 6% to 10% on later rounds.20,114 The cancer detection rate decreased from 3.4% to 4.0% on the initial screening round to between 0.6%20,67,103,106 and 2.4% 114 in later rounds. A smaller percentage of cancers in later rounds than in earlier rounds was detected by DRE alone. 106 In 1 study, the percentage of men with a PSA of 4.0 ng/ml or greater who had prostate cancer was about 26% for the first screening round and about 6.2 % for subsequent rounds. 20

Other studies provide information on testing strategies other than annual. Carter et al. used data from the Baltimore Longitudinal Study of Aging, including men ages 55 years and older, to examine the rate at which PSA increased to a level at which a cancer may become incurable (i.e., PSA of 5.0 ng/ml or greater). 115 They found that no man with an initial PSA of less than 2.0 ng/ml experienced an increase of PSA to 5.0 ng/ml or greater within 2 to 4 years. About 27% of men with a baseline PSA of 2.1 ng/ml to 3.0 ng/ml, and 36% of men with a baseline PSA of 3.1 ng/ml to 4.0 ng/ml, had increases in their PSA to 5.0 ng/ml or higher within 2 years. Thus, the authors reasoned that the 70% of the population with a PSA of less than 2.0 ng/ml need not have a PSA more often than every 2 years.

Similarly, a modeling study found that biennial screening of men after age 50 provided nearly the same potential benefit with many fewer biopsies. 15 The investigators also found a small potential benefit in doing 2 tests during the decade of the 40s. Finally, the Physicians' Health Study found that the sensitivity of PSA for prostate cancer appearing in the future did not decline appreciably for the first 4 years after screening. 35

Summary: Yield of Screening

Many uncertainties cloud the yield of screening for prostate cancer. We are not clear about what type of cancer should be detected to have an impact on patient outcomes. Thus, we are not clear about the target for screening. The reference standard test for determining whether cancer is present may find some cancers that are not associated with the screening test and may miss others that may or may not be clinically important. Because of these problems, research has not been able to determine the operating characteristics of screening with precision.

PSA screening with a cutpoint of 4.0 ng/ml clearly detects many prostate cancers; lower thresholds detect more cancers at the cost of more false positives and more biopsies. False-positive screening tests are most common in the setting of men with BPH, a common condition among older men. At least 2 tests (e.g., %fPSA and complexed PSA) reduce the number of false-positive screening tests. Whether these tests can or will have a major impact on clinical decisionmaking remains uncertain. DRE detects some cancers missed by PSA.

In direct studies of the yield of screening programs using PSA and DRE, 10% to 25% of men above age 50 have a positive test. Older men have a larger percentage of positive tests. Overall, 1.2% to 4.5% of men have prostate cancer in an initial screening. In later annual screenings, from 1% to 2.5% have prostate cancer. Older men have higher cancer detection rates.

About 70% of cancers detected in the first round of screening are pathologically organ confined; this percentage increases with later annual rounds of screening. The extent to which the earlier detection of these cancers leads to improved outcomes is uncertain.

The yield of screening in terms of cancers detected declines with repeated annual testing. If screening for prostate cancer does reduce mortality, then biennial screening may give nearly as much benefit as annual screening, especially for those with baseline PSA of less than 2.0 ng/ml.

Key Question 3: Harms of Screening

The third key question, indicated by the first curving downward arrow on the analytic framework (Figure 1), deals with the harms of screening for prostate cancer. These harms can be considered in 2 categories: the psychological effects of the screening process and the physical effects of screening and the clinical evaluation for men with positive screening tests. (Evidence Table 3)

Psychological Effects of Screening

The Rotterdam section of the ERSPC trial, a well-conducted ongoing RCT of the effects of screening (with PSA, DRE, and TRUS) on prostate cancer mortality, examined the psychological effects of the screening process. 116 The investigators administered 3 general quality-of-life questionnaires (including the Medical Outcomes Study Short Form-36, or SF-36) to 600 participants and 235 nonrespondents at different times through the screening process and then compared pretest and posttest data for different groups.

Among men who had a negative screen (n = 381 usable responses), the investigators found a small improvement in mental health scores, a small decrease in anxiety, and no other differences on 3 validated general quality-of-life measures. After a negative biopsy, men who had had a false-positive screening test (n = 160 usable responses) reported small improvements in bodily pain and general health perceptions and a small decrease in anxiety.

For the entire group during the screening process, anxiety was highest among men who had an initially high "trait" anxiety score. After screening, anxiety decreased for men with an initially low trait anxiety score but remained high for men with an initially high trait anxiety score. We do not know whether anxiety levels for these men decreased after a longer period after screening. We also do not know whether patients who had a negative biopsy were informed that they still had a 10% to 20% chance of having prostate cancer (see Key Question 2).

The authors concluded that they had documented little evidence of important psychological harms from the screening process. They noted that this could be because such problems are few or because their measures were not specific to the issue of prostate screening. Others have found that specific measures are best for documenting the psychological effects of screening. 117

Physical Effects of Screening

Essink-Bot et al. also examined patient reports of physical problems encountered in the screening process and the clinical evaluation of positive tests. 116 Fifty-two percent of men experienced either discomfort or pain from the DRE, 29% from the TRUS. Among men who had a biopsy, 90% reported pain or discomfort from the procedure; 38% reported that the pain lasted after the biopsy, but only 2% said that the pain lasted longer than 1 week. Four percent had used painkillers. Four percent also had experienced a fever of 38 degrees Celsius or higher, and 3% had visited a physician because of complications from the biopsy. About 5% reported moderate to extreme interference with daily activities.

Rietbergen et al. used data from the Rotterdam screening program to examine the side effects of needle biopsy of the prostate. 118 Of 1,687 men who had had a biopsy, 7 (0.4%) had to be admitted to the hospital from complications, especially infection.

Summary: Harms of Screening

Evidence about the harms of screening is scant. The screening process is likely associated with some increase in anxiety, but the number of men affected and the magnitude of the increased anxiety are largely unknown. Some screening procedures cause transient discomfort; biopsy of men with positive screening tests is associated with discomfort lasting longer than 1 week in a small percentage of men. Less than 10% of men have ongoing interference with daily activities after biopsy, and less than 1% suffer more serious complications, including infections.

Key Question 4 to 7: Efficacy of Treatment

General Approach

The second edition of the Guide to Clinical Preventive Services found little evidence to support the effectiveness of any treatment, compared with no treatment, for clinically localized prostate cancer. 2 It cited a single RCT with multiple flaws comparing radical prostatectomy with expectant management, which had reported no difference in survival over 15 years of follow-up.119,120 The previous Guide cited observational data showing a low prostate-cancer-specific mortality in untreated men with clinically localized cancer. Finally, it cited a structured literature review of nonrandomized studies that concluded that determining the efficacy of various treatments for clinically localized prostate cancer was not possible. 121

To address various treatment efficacy questions, we reviewed the 23-year follow-up of the earlier RCT (for Key Question 4), searched for any other RCTs and for large, well-conducted observational studies that would provide relevant information on the efficacy of treatment, and reviewed more recent observational data that might refine survival estimates (Key Question 7).

Key Question 4: Efficacy of Treatment with Radical Prostatectomy

Since 1991, radical prostatectomy (RP) has been the most commonly used treatment for clinically localized prostate cancer. It is the initial treatment for more than one-third of newly diagnosed patients. 1 The procedure is usually performed with curative intent on men who have a life expectancy of at least 10 years.

Randomized Controlled Trials

One older RCT, by Iverson et al., compared RP and expectant management for clinically localized prostate cancer; 122 another RCT, by Akakura et al., compared RP and external beam radiation therapy for locally advanced cancer, including some patients with clinically localized disease. 123 One more recent RCT compared RP with expectant management ("watchful waiting") in men with clinically localized but clinically detected prostate cancer. 124 We found no other RCTs comparing treatments for clinically localized prostate or locally advanced prostate cancer in which one arm received RP and the other did not. (Evidence Table 4)

In the Iverson et al. RCT, the research team randomized 142 men with newly diagnosed clinically localized prostate cancer being treated in 15 Veterans Administration hospitals in the United States between 1967 and 1975 to RP or expectant management. 122 Because of lack of funds, the study did not follow patients from 1978 to 1994, when the survival status of all patients was ascertained. Although vital status could be determined for 111 participants (78%), the investigators could not accurately determine cause of death. Randomization had failed to balance several important prognostic factors, such as age and stage. After an average follow-up of 23 years, the investigators found no difference in overall survival between the RP and expectant management groups. Because of the high loss to follow-up, the problems with assessment of outcomes, and the relatively small size of the study, few consider these results definitive.

The Akakura et al. RCT included 95 men with prostate cancer that was palpable on DRE and that either involved both lobes or was palpably extracapsular. 123 Thus, some of the men likely had clinically localized cancer and some had extracapsular cancer. All men received 1 of several androgen deprivation therapies (including diethylstilbestrol, LHRH agonists, orchiectomy, or a nonsteroidal antiandrgen) before and after treatment and then were randomized to either RP or external beam radiation therapy. Five-year prostate-cancer-specific survival was 96.6% in the RP group and 84.6% in the radiation group (p = 0.024). The degree to which this trial represents results from treatment of clinically localized disease is not clear. The effect could well be attributed to an effect on locally advanced disease.

The more recent RCT, by Holmberg et al, 124 randomized 695 men with newly diagnosed prostate cancer to RP or watchful waiting. Only 5% of these cancers were detected by screening, and about 75% were palpable on rectal exam. Of the men assigned to RP, fewer then 8% had positive nodes at surgery. After a followup of 6.2 years, 4.6% of men assigned to RP had died of prostate cancer, compared with 8.9% of men assigned to watchful waiting (absolute difference: 4.3%; relative hazard 0.50; 95% CI 0.27-0.91). This difference in prostate cancer-specific mortality appeared only after 5 years of followup; there was a small trend favoring the RP group in all-cause mortality, but this difference was not statistically significant between groups.

Although this RCT was well-performed, it does not provide direct evidence concerning the efficacy of RP for those cancers detected by PSA screening. As these cancers are likely different from those detected clinically, one should be careful about extrapolating evidence from the cancers in this trial to screen-detected cancers. Also, the additional lead time from screening means that, even if RP is effective for screen-detected cancers, the benefit in prostate-specific mortality would only appear some years after it appeared in this trial (8 years in the trial). The effect of RP on all-cause mortality for any group of clinically localized cancers remains in doubt.

At least 1 RCT of RP compared with expectant management for clinically localized prostate cancer, mostly detected by screening, is ongoing. The U.S. Prostatectomy Intervention Versus Observation Trial (PIVOT) hopes to randomize 1,000 men up to 75 years of age with any histologic grade of localized prostate cancer and a life expectancy of at least 10 years. The trial started in 1994 and is scheduled to continue for 12 to 15 years of follow-up.

Observational Studies

We examined 6 case series of RP with at least 10-year survival data published since 1994. Three were from single institutions (Mayo Clinic,125,126 Johns Hopkins,50,127 and Duke University); 128 2 were analyses of SEER data;129,130 and 1 was a multi-institutional pooled analysis from 8 medical centers. 131 Overall, 10-to-15-year disease-specific survival was 80% to 97% for all analyses for men with well and moderately differentiated tumors. For poorly differentiated cancers, 10-to-15-year disease-specific survival was 60% to 80%.

Lu-Yao and Yao analyzed SEER data together with an age-matched control group. 130 Overall 10-year survival for men who had had RP for well-differentiated prostate cancer was 77% (control group, 65%); for men with moderately differentiated cancers, survival 10 years after RP was 71% (control group, 64%); for men with poorly differentiated tumors, survival was 54% (control group, 62%). After adjustment for the younger age and lower stage of men receiving RP compared with radiation or watchful waiting, 10-year disease-specific survival for the RP group was not different from the radiation or the watchful-waiting groups for men with well-differentiated cancers. Disease-specific survival was only slightly higher for the RP group for moderately differentiated tumors (RP, 87%; radiation, 76%; watchful waiting, 77%); it was much higher for men with poorly differentiated cancers (RP, 67%; radiation, 53%; watchful waiting, 45%).

Summary of Efficacy of Treatment with Radical Prostatectomy

Three RCTs compared any other treatment with RP for clinically localized prostate cancer. One older trial, comparing RP with expectant management, had major methodologic flaws and does not provide definitive results. Another, comparing RP with radiation therapy, was small and included a large percentage of men with locally advanced rather than clinically localized cancer. The more recent RCT, comparing RP and watchful waiting, was larger and well-conducted, but the participants had clinically-detected rather than screen-detected prostate cancer. The results of this trial indicate that, after 8 years, RP reduces prostate cancer-specific mortality but not all-cause mortality. It is likely that any benefit from RP in screen-detected cancer would take even longer to appear. Clearly, further studies are needed before we can draw valid conclusions about the efficacy of RP for screen-detected, clinically localized disease.

In the 6 large observational studies of outcomes after RP, 5 had no internal controls and the other had only age-matched controls. All studies found high (80% to 97%) long-term, 10-year disease-specific survival after RP for well or moderately differentiated cancers and somewhat lower (60% to 80%) disease-specific survival for men who had had RP for poorly differentiated tumors. The 1 study with an internal control group attempted to adjust for differences among the treated populations, finding a small advantage for RP compared with radiation or watchful waiting for men with moderately differentiated cancer and a larger advantage for men with poorly differentiated cancer.

All the observational studies have 2 important weaknesses: (1) the survival rates may be a reflection more of the patients and the tumors than the treatment; 132 and (2) none of these studies specifically included men whose prostate cancer had been detected by screening, so whether any results apply to a screened population remains unclear. With these weaknesses and the lack of convincing RCT evidence, we conclude that the efficacy of RP treatment for localized prostate cancer is unknown.

Key Question 5: Efficacy of Treatment with Radiation

Radiation therapy is the second most commonly used treatment for nonmetastatic prostate cancer; it is the most common treatment for men ages 70 years to 80 years. 1 Two types of radiation therapy are most commonly used and will be reviewed here: external beam radiation therapy (EBRT) and brachytherapy, the insertion of radioactive pellets directly into prostate tissue.

External Beam Radiation Therapy

Research continues to examine the optimal manner of delivery and dose of EBRT for prostate cancer in various stages with various characteristics. Some evidence indicates that 3-dimensional (3-D) conformal radiation, in which computerized tomography is used to guide the radiation beam directly to the prostate rather than adjacent structures, may provide better cancer control with fewer side effects than standard EBRT. Using 3-D conformal techniques, clinicians may be able to deliver higher radiation doses that may be more effective, especially in higher risk patients. 133 Much of this research is not sufficiently mature, however, to determine the impact of these new approaches on patient health outcomes.

Randomized Controlled Trials

The Akakura et al. RCT (see Key Question 4 above) examined the efficacy of EBRT by comparing RP with EBRT in 95 men with either localized or locally advanced cancer. Prostate cancer-specific survival after 5 years was statistically significantly higher in the RP group (RP, 96.6%; EBRT, 84.6%, p = 0.024). 123 We found no other RCT with clinical outcomes comparing EBRT with any other therapy for clinically localized prostate cancer.

Observational Studies

Three large observational studies of men with clinically localized prostate cancer treated with EBRT provide some information about long-term clinical outcomes. The largest study, from the Radiation Therapy Oncology Group (RTOG), included 1,557 men with various stages and grades of prostate cancer treated with EBRT. Prostate-cancer-specific survival after 15 years was 72% for men with clinically localized disease and Gleason score 2 to 6 (well to moderately differentiated), 61% for clinically localized disease and Gleason score 7, and 39% for clinically localized disease and Gleason score 8 to 10. A second large multi-institutional series of patients found a 72% prostate-cancer-specific survival after 12 years of follow-up. 134 Finally, an observational study mentioned earlier for Key Question 4 examined 10-year overall survival among men in the SEER registry who had received various treatments, comparing them to an age-matched control group. For EBRT, the age-matched control group had a 10-year survival of 54%. Survival rates for men with different stages of cancer were as follows: well-differentiated cancer, 63%; moderately differentiated cancer, 48%; and poorly differentiated cancer, 33%. 130

Summary of Efficacy of External Beam Radiation Therapy

One small RCT comparing EBRT with any other therapy for clinically localized prostate cancer found a benefit for RP over EBRT in 5-year survival. Three large observational studies provide information about long-term survival among men treated with EBRT for clinically localized disease. As with the observational studies of RP, one can determine neither the independent effect of the treatment (as compared with the type of patient or the type of cancer) nor the extent to which these studies include patients who are comparable to those who would be detected by screening. With these weaknesses and the lack of convincing RCT evidence, we conclude that the efficacy of EBRT treatment for localized prostate cancer is unknown.

Brachytherapy

Although implantation of radioactive pellets directly into a cancer, or brachytherapy, has been used to treat gynecologic malignancies for some years, this technique has found widespread use in treating prostate cancer only in the past 10 to 15 years. It is most frequently used either alone in men with well differentiated cancer or in combination with EBRT in men with more aggressive cancer. The technique continues to evolve, and research to define its clinical efficacy is still in its infancy. Because it is a simple 1-time outpatient procedure for patients, and because some have the perception that it has fewer side effects, it has become a popular choice of treatment in some areas. The technique is, however, technically difficult and its applicability in community practice is as yet unknown. 135

No RCT with clinical outcomes compared any treatment for prostate cancer with brachytherapy. Two observational studies with 100 patients or more reported clinical outcomes of treatment of clinically localized prostate cancer treated with brachytherapy. One study reported 90% to 100% 5-year survival for 157 patients treated with radioactive gold seeds. 136 Another study found that 15 years after treatment with radioactive iodine seeds, 43% of 126 patients had died of prostate cancer. 137 These investigators also observed that patients selected for this therapy more recently had less aggressive disease (i.e., lower stage and grade).

Summary of Efficacy of Brachytherapy

We found no RCT evidence on the efficacy of brachytherapy, and no large observational data provides useful information about this issue. As for EBRT, we conclude that the efficacy of brachytherapy for clinically localized prostate cancer remains unknown.

Key Question 6: Efficacy of Treatment with Androgen Deprivation

Prostate cancer is often an androgen-dependent disease, and thus androgen deprivation has long been one approach to therapy. In the past, this treatment modality has taken the form of orchiectomy or estrogen therapy, primarily for men with metastatic disease. These therapies had a number of undesirable side effects, including psychological effects in the case of orchiectomy and cardiovascular effects in men given estrogen.

Newer approaches to androgen deprivation therapy (ADT) include drugs (e.g., flutamide or bicalutamide) that block peripheral androgen receptors and drugs that are LHRH analogues (LHRH agonists; e.g., goserelin or leuprolide). This latter group of drugs works by stimulating the release of luteinizing hormone from the pituitary gland, leading to a transient increase in testosterone production by the testes. Paradoxically, when used clinically, LHRH agonists result in a "down regulation" of pituitary receptors, thus markedly reducing testosterone production to the level of a castrate man. LHRH agonists have been used clinically since the late 1980s.

Randomized Controlled Trials

Three RCTs compared clinical outcomes among at least some men with clinically localized prostate cancer who were treated with either ADT or any other treatment. (Evidence Table 5) Lundgren et al., also discussed in conjunction with the efficacy of watchful waiting (Key Question 7) below, compared outcomes among 228 men randomized to estrogen, estramustine (a nitrogen mustard derivative of estradiol with both cytotoxic and androgen deprivation properties), and watchful waiting. Among men followed for 10 years, about 12% of the estrogen group, 22% of the estramustine group, and 35% of the deferred therapy group had developed metastases. (read from Figure 2) During the followup period, about 12% of men in the estrogen group, 18% in the estramustine group, and 28% in the deferred treatment group died from prostate cancer (p = 0.03). Overall survival, however, was similar in all groups. 138

Two other RCTs among men treated with EBRT found that ADT with either orchiectomy 139 or estramustine 140 either increased overall survival 139 or reduced clinical recurrence. 140 In both studies, improved outcomes occurred primarily among men who had positive lymph nodes.

We also examined RCTs comparing ADT with any other treatment for men with locally advanced prostate cancer (i.e., extracapsular but not metastatic disease). Four recent RCTs of ADT (using LHRH agonists) as adjuvant to EBRT or RP (compared with EBRT or RP alone) found statistically significantly improved overall survival (10% to 20% absolute difference) in men who received ADT.141-146

For example, in 1 study overall survival at 5-year follow-up was 79% in the group receiving an LHRH agonist plus EBRT and 62% in the group receiving EBRT alone (p = 0.001). 141 In the only study that added an LHRH agonist to RP, after 7 years 15% of men who received the LHRH agonist and 35% of the men treated only with RP had died. 146

One further RCT of immediate versus deferred ADT (with either orchiectomy or LHRH agonists) without other treatment found improved survival (8% absolute difference) for the immediate ADT group in men newly diagnosed with locally-advanced prostate cancer. 147

Summary of ADT Efficacy

We found little evidence that ADT improves clinical outcomes among men with clinically localized prostate cancer. The studies performed to date on this issue, however, have included a large number of men with more advanced disease. Because the overall prognosis for men with clinically localized disease is often good (see Key Question 7 below), studies of any additive effect of ADT would necessarily require a large number of men followed for some years. We did find strong evidence that ADT, especially in the form of LHRH agonists, does improve clinical outcomes, including overall survival, among men with locally advanced prostate cancer who have already received either EBRT or RP.

Key Question 7: Efficacy of Treatment with Watchful Waiting

One critical issue in screening for prostate cancer is whether aggressive treatment of clinically localized prostate cancer with one of the modalities above produces better outcomes than does simple "watchful waiting." Watchful waiting, also termed "expectant" or "conservative" therapy, implies that no therapy is given initially but that the patient is followed for evidence of progressive or symptomatic disease. Treatment may then be offered only for men experiencing progressive disease. Evidence Table 6 provides details about the following studies.

Randomized Controlled Trials

Two RCTs compared watchful waiting to aggressive therapy for clinically localized prostate cancer: the VA study by Iverson et al described for Key Question 4 122 and an open-label RCT of hormonal therapy by Lundgren, described for Key Question 6. 138 Both studies were small and had methodological flaws.

Lundgren's RCT, begun in 1978, randomized 285 men (mean age 70 years) with clinically localized prostate cancer into 1 of 3 groups: estrogen, estramustine, or deferred treatment. 138 Some 24% of randomized patients were lost to follow-up or excluded for various reasons; randomization was unbalanced; and cardiovascular mortality in the estrogen group was high. During the observation period, prostate cancer-specific mortality was significantly worse in the deferred treatment group (28% compared with 12% and 18% for estrogen and estramustine, respectively, p = 0.03), although overall survival was not statistically different among the groups (40% versus 47% and 46%, respectively).

Observational Studies

Generic Issues

In the absence of compelling RCT evidence, we searched for large cohort studies dealing with the survival of men with clinically localized prostate cancer who were treated expectantly. Six such studies, published since 1994, provide information about the natural history of untreated clinically localized prostate cancer (see Evidence Table for Key Question 7). However, few or none of the prostate cancer cases in these studies had been detected by screening PSA; an unknown number had been detected by screening DRE. Thus, we do not know the extent to which these data are applicable to the screening-detected tumors of today.

Some cases had been detected by a surgical procedure, transurethral resection of the prostate (TURP), which was more commonly done in the past than it is now. 148 In performing this procedure, surgeons retrieved small "chips" of prostate tissue, some of which contained small foci of prostate cancer. Many experts suspect that a large percentage of such cancers are not clinically important. What is not clear is the extent to which current screening strategies detect these "minimal" cancers. If current screening does not detect such cancers, then some of the cancers in these older studies could have a better prognosis than screen-detected cancers of today, making the studies less applicable to today's situation. For example, while many of the prostate cancers detected by TURP were well-differentiated, many fewer cancers detected by PSA screening are well-differentiated. 149

Imaging procedures used today (e.g., computerized tomography and magnetic resonance imaging scans) are much more sensitive in finding advanced disease than the examinations that were used when many of the cancers in these studies were detected. Thus, at least some of the cancers in these studies that had been denoted as clinically localized may instead have been locally advanced or even metastatic. This factor would tend to lower the survival of patients in these studies relative to the survival of patients with typical screen-detected cancers today.

In sum, competing selection biases in these studies may affect their results, although we cannot determine the net direction and magnitude of any bias.

Study-Specific Review

Five of the 6 studies were large, retrospective cohort studies, 1 using SEER data from the United States, 130 1 using data from the Connecticut Tumor Registry,47,150 and 3 using population-based data from Sweden or Denmark.48,151,152 The sixth study was a pooled analysis of individual data from 6 other nonrandomized studies of survival of untreated men with localized prostate cancer. 46

With respect to disease-specific survival rates (i.e., survival rates in which men who die of other diseases are censored), the 5 studies with information on tumor grade show very favorable 10-year to 20-year rates for men with well differentiated, clinically localized prostate cancer who had been treated with watchful waiting.46-48,130,150,152 Lu-Yao et al. for example, found that these men had the same survival as men without prostate cancer. 130 Men with moderately differentiated, clinically localized cancer had worse disease- specific survival than men with well-differentiated disease. Disease-specific survival at 15 years for men with moderately differentiated cancer was 74% to 83%, about 5% 46 to 15% 152 percentage points lower than men with well-differentiated disease. Lu-Yao found that 10-year overall survival was an absolute 11% lower (38% compared with 49%) for men with moderately differentiated prostate cancer than for age-matched controls without prostate cancer. 130

Albertsen et al. found great heterogeneity within the group of moderately differentiated tumors, meaning Gleason score of 5 to 7.47,150 Among men with Gleason score 5, from 6% (ages 50 to 59 years) to 11% (ages 70 to 74 years) had died of prostate cancer 15 years after diagnosis. Among men with Gleason score 6 cancer, 18% (ages 50 to 59 years) to 30% (ages 70 to 74 years) had died of prostate cancer. Among men with a Gleason score 7 cancer, 42% (ages 70 to 74 years) to 70% (ages 50 to 59 years) had died of prostate cancer. Thus, men with Gleason score 7 cancers had a greater probability of dying of prostate cancer than men with Gleason score 5 or 6 tumors. In addition, age had only a small effect on the probability of dying of prostate cancer for men with lower-grade tumors (i.e., Gleason score 2 to 6), but older men had a lower probability of dying of cancers with Gleason score 7. This was also true for Gleason score 8 to 10 cancers (probability of death from prostate cancer was 60% for men ages 70 to 74 and 87% for men ages 50 to 59 years).

These data are particularly important, as most men with screen-detected cancers today have moderately differentiated histology. As noted earlier, some of these men have a good prognosis whereas others have a poor prognosis.

For poorly differentiated cancers, the studies agree that the prognosis for men with clinically localized cancer treated expectantly is grim: Lu-Yao et al. found that men with poorly differentiated but clinically localized tumors had a reduction in overall 10-year survival of an absolute 30% compared with age-matched controls without prostate cancer (17% compared with 47%). 130 Disease-specific survival after 15 years in the other studies for men with poorly differentiated cancer ranged from 13%47,150 to 44%. 48

Results from the Brasso et al. study are difficult to compare with the other studies. 151 These researchers selected only men who had survived for 10 years after their diagnosis, gave no disease-specific survival rates, and provided no information on tumor grade.

Summary: Efficacy of Treatment with Watchful Waiting

We found no convincing RCT evidence of the efficacy of watchful waiting compared with other treatments for clinically localized prostate cancer. Four retrospective cohort studies and a pooled analysis of 6 other cohort studies showed that men with well-differentiated, clinically localized prostate cancer have excellent long-term survival, with little or no reduction in survival compared with similar men without prostate cancer.

With regard to moderately differentiated cancer, these cohort studies found a definite reduction in disease-specific and overall survival compared with the survival of men with well-differentiated cancers, although the magnitude of this reduction varied among groups and among studies. The most detailed analysis of this group found that men with Gleason 7 tumors had a substantially worse disease-specific survival than men with Gleason 5 tumors. 47 All studies agree that men with poorly differentiated cancers have low long-term disease-specific survival.

If the men in these studies are representative of contemporary men with screen-detected cancer, their data can be useful in determining the most appropriate target for screening. For example, one would not target well-differentiated, clinically localized prostate cancer for early detection and treatment. The major concern with these studies, however, is the extent to which selection biases of uncertain direction and magnitude limit their generalizability to the current population of men with screen-detected cancers.

A primary interest in reviewing these studies is to determine the outcomes for men with moderately differentiated prostate cancer, because this is the type of cancer most commonly detected by screening. The studies show that survival varies for this group of patients; some men have a good prognosis, others a poor prognosis. Clarifying this variation is an important research priority.

Summary: Efficacy of Treatment

No treatment has been shown to be effective in improving clinical outcomes for men with prostate cancer confined to the prostate. Among this group, outcomes are worst for men with poorly differentiated tumors and best for men with well-differentiated tumors. The largest number of prostate cancers detected by screening is moderately differentiated; men with these tumors have a mixed prognosis.

Androgen deprivation therapy (ADT) is effective in prolonging survival among men with prostate cancer outside the capsule; this conclusion comes from studies in men who were (presumably) not detected by screening.

Key Question 8: Harms of Treatment

Because harms of treatment are experienced by the men themselves, we examined evidence that measured patients' perceptions of their function rather than assessments by physicians or investigators. Because it is difficult to interpret a proportion of men who are experiencing a dysfunction independent of some comparison, we examined evidence that provided some comparison of function, including control groups who had not had prostate cancer treatment, men with prostate cancer treated in a different way, or sequential studies comparing men's function before and after treatment. Because the frequency of harms changes over time after treatment, we examined evidence of harms at least 1 year after treatment. Details about the studies we found are provided in Evidence Table 7. Table 3 provides a summary of harms at least 1 year after different treatments.

Table 3. Harms of treatment*†.

Table

Table 3. Harms of treatment*†.

Radical Prostatectomy

In 2 studies of acute adverse effects of RP relying on large databases, 30-day mortality was 0.7% (in a VA population ages 45 to 84 years) 153 to 1.0% (in a Medicare population, ages over 65 years); 154 the latter study found that men older than 80 had a 30-day mortality of 4.6%. Major cardiopulmonary complications occurred in 1.7% in the VA population 153 and in 7.4% in the Medicare population for men ages 70 to 74 years. 154

The primary long-term adverse effects that have been associated with RP include erectile dysfunction, urinary incontinence, and bowel symptoms. Advances in the technique of performing RP, including delineation of the anatomy of the dorsal vein complex and pelvic plexus, enabled clinicians to spare important structures, which in turn might reduce complications following RP. 127 Thus, although most of the literature we found concerns standard RP, we especially examined articles that reported results of the newer "nerve-sparing" procedure.

Erectile Dysfunction

One meta-analysis of 40 studies through 1995 compared erectile dysfunction in men after RP or after EBRT. 155 The probability of maintaining erectile function was 0.42 after RP and 0.69 after EBRT (p < 0.0001). The RP probability was similar to that reported in a previous literature review. 121

Twelve studies (some with several publications) published since the meta-analysis met our inclusion criteria.156-172 Two studies compared sexual function among men treated by RP and age-matched population controls.168,171 In 1, 79% of men who had had an RP and 46% of controls reported poor or very poor sexual function. 171 In the other, 82% of men who had had an RP and 63% of controls reported that they were distressed because of decreased sexual function. 168

Seven studies gave the same men questionnaires before (or soon after) and 12 to 24 months after RP to assess the impact of the procedure.157-172 One of these studies is the Prostate Cancer Outcomes Study (PCOS) in which patients with prostate cancer are ascertained from 6 SEER areas and sent questionnaires at various times after treatment. In this study, 41.9% of men at 24 months after RP reported that sexual function was a moderate to large problem. When asked about function before surgery, 17.9% said that sexual function had been a problem (difference about 24 percentage points). This difference in the negative impact of RP on sexual function varied by age. About 50% of men younger than age 60 years suffered a decline in sexual function (from 92.6% before surgery to 39% afterward), whereas about 30% of men ages 75 to 79 years suffered a decline (48.6% before RP to 19.1% afterward).

In a study that gave men questionnaires before and after RP, the percentage of men reporting that erections were usually inadequate for sex increased from 32% before the RP to 93% 12 months after surgery (difference about 60%). 163 In neither this study nor a similar one 162 did sexual function differ between men who had nerve-sparing surgery compared with men who had standard surgery.

In another study that followed men sequentially over time, from an academic center that helped develop the nerve-sparing RP,156,172 Walsh found that 18 months after nerve-sparing surgery, 86% of men who had erections adequate for intercourse before surgery maintained their sexual function. Some of these men used drugs or other devices to assist potency, but 84% of these men reported little or no bother concerning sexual function. Others have questioned whether such results are possible in community practice, or whether the patients were a selected subgroup. 156

Three other studies published during 2001 compared erectile dysfunction in the same men before and after RP.157-172 Sexual potency 1 to 2 years after surgery was impaired over baseline in 60% to 80% of men.

Steineck et al 173 conducted a survey of potential harms of RP about 4 years after randomization into Holmberg et al's RCT of RP versus watchful waiting for men with clinically-detected prostate cancer. 124 Erectile dysfunction (80% in RP group, 45% in watchful waiting group) was more frequent in the RP group.

Studies that have surveyed men a single time 12 to 24 months after having an RP, without controlling for prior function or function in the non-prostate cancer population, have generally attributed a higher level of sexual dysfunction to RP. For example studies by Fossa and Schrader-Bogen found 70% to 100% of men had erectile problems after having an RP.166,167

Summary: Erectile Dysfunction after Radical Prostatectomy

We found that at least 20%, and perhaps as many as 70%, of men who have had an RP in the general community suffer worsened sexual function 1 year later as a result. The evidence is mixed about whether the newer nerve-sparing RP reduces this burden outside of excellent academic centers. (Table 3)

Urinary Incontinence

Most of the same studies mentioned above that examined sexual function also considered urinary function. The 2 studies that compared function between men who had had an RP and an age-matched control group found that the difference in incontinence potentially attributable to RP was 15% (frequent dribbling or no control: 21% in RP group compared with 6% in control group) 171 to 50% (leakage: 65% in RP group compared with 14% in control group). 174

Four of the 5 studies that evaluated change in men's urinary function longitudinally found that an additional 25% to 37% of men were wearing pads for urinary incontinence 12 to 24 months after having an RP.156-158,161 One of these studies, the PCOS, found that the effect on urinary function varied by age. The additional percentage of men who had incontinence more than twice each day 24 months after RP compared with before surgery was about 8% for men ages 60 years and younger, and about 36% for men ages 75 to 79 years. 161

The fifth study was from an academic institution using the nerve-sparing surgery technique, finding that only 7% were wearing pads at 18 months after nerve-sparing RP. 156

In the Steineck et al survey 173 within the Holmberg RCT of RP, 124 urinary leakage (49% in RP group, 21% in watchful waiting group) were more frequent in the RP group. Urinary obstructive symptoms, however, were more common in the watchful waiting group (44% in watchful waiting, 28% in RP group). Bowel function, anxiety, depression, and subjective quality of life were similar in the 2 groups. 172

In 2 studies without a control group assessing urinary function only once after RP, 12% of men reported severe urinary leakage 167 and 19% reported that urinary problems affected their quality of life "quite a bit". 166

Summary: Urinary Dysfunction after Radical Prostatectomy

In a variety of studies, we found that from 15% to 50% of men who had had an RP in the general community suffered substantial urinary problems afterward. We found little evidence about whether the newer nerve-sparing RP reduces this burden outside of excellent academic centers. (Table 3)

Harms of External Beam Radiation Therapy

We will first examine the evidence concerning the harms of EBRT followed by a review of the harms of brachytherapy. We will especially look for evidence concerning recent developments in these fields, especially conformal EBRT and TRUS-guided brachytherapy.

Erectile dysfunction

We found 1 meta-analysis of 40 studies, mentioned above, 155 that found that the probability of maintaining sexual function after EBRT is 0.69 (compared with RP, 0.42). None of these studies were published after 1995.

Three studies168,169,171,175 examined sexual function among men treated with EBRT and age-matched controls without prostate cancer.168,169,171,174,175 These studies found that 20% to 40% more men who had had EBRT suffered sexual dysfunction compared with the control group.

Seven studies examined sexual function over time after EBRT, either by repeated measures or by asking about previous function.157,162,163,170,171,176-179 All 7 showed that 20% to 45% more men had erections inadequate for intercourse 12 to 24 months after EBRT than at baseline. One of these studies found that more men older than 70 years suffered a decline in sexual function than men under age 70 (32% compared with 23%).

Two of these studies included men who had received conformal radiation therapy; 1 study found the same degree of decline in sexual function as other studies of conventional treatment 176 and the other found no decrease in sexual function over 12 months after conformal radiation. 178

Two other studies used a single questionnaire after EBRT to assess sexual function. Each found that about 50% of men had erections inadequate for intercourse.166,167

Summary of Erectile Dysfunction from External Beam Radiation Therapy

From 20% to 40% of men who had no erectile dysfunction before EBRT developed dysfunction 12 to 24 months afterward. (Table 3)

Urinary Incontinence

Three studies171,180 compared urinary function in a group of men who had had EBRT with a population control group.168,171,180 One found no difference in urinary symptoms between men who had had EBRT and controls. 171 The other 2 studies found that the prevalence of severe urinary problems was 12% 180 to 16% 168 higher among men who had had EBRT than among controls.

Five studies examined urinary function over time among men who had had EBRT.157,163,170,177-179 Among those men who had had no urinary symptoms at baseline, from 2% to 8% developed urinary incontinence severe enough to wear pads after EBRT.

Three studies surveyed men concerning urinary function at least 1 year after EBRT.166,167,181 From 12% 167 to 36% 181 of men had frequent urinary incontinence. One of these studies compared standard EBRT with 3-D conformal EBRT, finding a statistically significantly lower prevalence of frequent urinary leakage (36% compared with 29%, p = 0.044). 181

One RCT compared the side-effects of conformal and conventional EBRT. 182 This study does not actually meet our review criteria as it used physician rather than patient assessments of outcomes. It found no difference between the 2 approaches to EBRT in urinary function.

Summary of Urinary Dysfunction from External Beam Radiation Therapy

From 2% to 16% of men who had no urinary incontinence before EBRT developed dysfunction 12 to 24 months afterward. It is not clear whether conformal EBRT reduces the frequency of this side-effect. (Table 3)

Bowel Dysfunction

We found 3 studies168,171,180 that compared bowel function in men who had had previous EBRT with a control population.168,171,174,180 Compared with controls, about 10% to 25% more men who had had EBRT reported marked bowel problems, often increased frequency and urgency of bowel movements.

Five studies assessed bowel function over time in men who had had EBRT.157,163,170,177-179 From 6% to 18% of men who had not had previous bowel problems reported substantially increased bowel problems from 12 to 24 months after EBRT. One small study reported that men who had had conformal EBRT had fewer problems than men who had had conventional EBRT. 178

Two studies surveyed men about bowel function at least 1 year after having EBRT.166,168,171,180,181 They found that 11% to 17% had major problems with bowel function. One of these studies 181 also found that only 4% of men who had had conformal EBRT reported similar problems.

One RCT 182 using physician rather than patient assessment of outcomes found little difference between conventional and conformal EBRT in the development of severe bowel problems. 182

Summary of Bowel Dysfunction from External Beam Radiation Therapy

From 6% to 25% of men who had no bowel dysfunction before EBRT reported marked problems 12 or more months afterward. The evidence is mixed about whether conformal EBRT reduces the frequency of this side effect. (Table 3)

Harms of Brachytherapy

We found 7 studies that assessed potential harms of brachytherapy by patient reports with a validated instrument. Four of these examined scores on validated measurement instruments longitudinally183-186 while the other 3 were cross sectional in design.160,187,188 The studies used one of 2 isotopes (iodine - 125 or palladium - 103) with a variety of doses.

Two longitudinal studies examined sexual function before and at least 1 year after brachytherapy treatment.184,186 One study found that, among men who were potent before treatment, about 21% were impotent and 36% had reduced erectile function 3 years after treatment. 184 The second study found that 35% of men treated with brachytherapy alone had not returned to pre-treatment sexual function 18 months after treatment. 186

Three additional studies, all cross-sectional, assessed sexual function at 7 to 18 months after brachytherapy. One found that the percentage of men who reported erections adequate for intercourse declined from 73% before brachytherapy (measured by patient recall) to 55% after 12 months. 188 In the second study 43% of men reported adequate erections 9 months after brachytherapy. 160 In the third study, 187 investigators measured sexual function with a validated 100 point scale (UCLA-Prostate Cancer Index, higher numbers mean better function), finding that sexual function was 14 points lower than literature controls without prostate cancer, a statistically (p = 0.05) and clinically significant magnitude.

Four studies examined urinary function after brachytherapy.160,185-187,189 Two used longitudinal designs.185,186 These studies found that, although a majority of men will have distressing urinary symptoms in the first months after brachytherapy, from 6% to 12% will have such symptoms 1 year later. Men who had some urinary symptoms before brachytherapy had a higher probability of developing long-standing problems after treatment. 185 Perhaps 25% of men will have some loss of urinary control 1 year after brachytherapy. 186

Two cross-sectional studies used validated scales to assess urinary function at 3 to 7 months after brachytherapy.187,189 In one, the urinary score (I-PSS, lower numbers mean better function) more than doubled from the pre-treatment assessment to 3 months afterward. 189 In another study, urinary scores 7 months after treatment were more than 20 points worse than literature controls (p = 0.001). 187 In the third study, 57% of men had some degree of urinary incontinence 9 months after therapy. 160

Two studies assessed bowel function after brachytherapy. In one, men who had had brachytherapy 7 months earlier were 8 points worse on a 100 point validated scale compared to literature controls (p = 0.05). 187 In the other, about 18% of men reported some degree of diarrhea at 12 months after treatment. 160 Another study found that 19% of men treated with brachytherapy had some persistent rectal bleeding 12 to 28 months after treatment. 190

Summary of Harms from Brachytherapy

We found some evidence that brachytherapy has an impact on sexual, urinary, and bowel function, but insufficient evidence to determine precisely the magnitude of these harms. Our best estimates are that 36% of men will have some erectile dysfunction, 2% to 12% will have some urinary symptoms, and 18% will have some bowel dysfunction 1 year after treatment. (Table 3)

Although it did not meet our criteria (as it has no patient reports), we found 1 large study of procedures during 1991 to 1993 among men in the Medicare population who had had brachytherapy for prostate cancer in 1991. 191 Using claims data, these investigators found that 8.3% of 2,124 men underwent a surgical procedure for bladder outlet obstruction during the follow-up period. In addition, 0.3% underwent colostomy for complications of brachytherapy and 0.6% had a penile prosthesis. About 7% of men carried a diagnosis of urinary incontinence after the procedure.

Harms of Androgen Deprivation Therapy

LHRH agonists constitute the type of ADT that has been most recently studied for treatment efficacy and harms. One systematic review for AHCPR examined endocrine therapy in men with prostate cancer, 192 although whether it required studies to include patient reports of symptoms rather than or in addition to physician or investigator reports is not clear. The review found that 20.8% of men receiving LHRH agonists and 13.3% of men after orchiectomy had erectile dysfunction that prevented intercourse. One other study found that about 20% more men in a group treated with endocrine therapy (type not specified) for prostate cancer had fewer sexual thoughts and lower erectile capacity than a control group without prostate cancer. 168 According to the AHCPR systematic review, about 49% of men receiving LHRH agonists suffered from hot flushes, but less than 5% had gynecomastia. 192

The Prostate Cancer Outcomes Study (PCOS), a national study of men with prostate cancer treated in various ways, has provided information about adverse health outcomes in two reports of men treated with ADT alone for at least the first 12 months after diagnosis. This study used patient reports, but did not include a pre-treatment assessment. The investigators did ask about pre-treatment function at the 6 months post-treatment assessment. They found that 70% to 80% of men who reported previous sexual activity ceased sexual activity after treatment; about the same percentage of men who were potent before treatment were impotent afterward. There were no differences between men who were treated with LHRH agonists and those men treated with surgical orchiectomy. About 25% of men treated with LHRH agonists and 10% treated with orchiectomy reported breast swelling; hotflashes were similar between groups (56.5% for LHRH antagonists and 67.9% for orchiectomy).193,194

One other large national study used the Medicare database to identify men with prostate cancer who had had RP. The study compared self-reported quality of life 7 to 8 years after surgery in men in this group who had had androgen deprivation (some by orchiectomy and some by LHRH agonists) with those who had not. Although most men in both groups had poor sexual function, the androgen deprived group reported greater dysfunction, with only 2% having the ability to have sexual intercourse and 69% having any sexual drive in the previous 30 days. The androgen deprived group also reported lower function in 7 different indices of quality of life (e.g., mental health, activity, worries about cancer, energy, etc) compared to non-androgen deprived men. 195

We found no other studies of 50 or more men taking LHRH agonists that provided patient reports of symptoms.

Anemia and osteoporosis have been reported as potential long-term complications of LHRH agonist therapy.196,197 The frequency and severity of these complications is as yet unclear.

Summary of Harms from Androgen Deprivation Therapy

We found fair evidence that ADT with LHRH agonists reduces sexual function by 40% to 70%, and is associated with breast swelling in 5% to 25% of men. Hot flashes occur in 50% to 60% of men taking LHRH agonists. (Table 3)

Summary for Key Question 8: Harms of Therapy

The sections above have described our findings with respect to organ-specific function for each mode of therapy. All treatments are associated with definite harms, of varying severity and varying frequency. These are summarized in Table 3.

The impact of these symptoms on overall quality of life is complex, however. For example, Litwin et al 169 assessed overall quality of life in addition to individual symptoms in men with localized prostate cancer who had been treated in various ways. They compared quality of life scores among control men and men with prostate cancer within treatment groups. Although they found the same differences in symptoms as our review has found, they found no differences among groups (either among treatment groups or between men with and without prostate cancer) in overall quality of life.

On the other hand, Bokhour et al conducted focus groups with men who had been treated for early prostate cancer 12 to 24 months earlier. About 54% of participants had significant erectile dysfunction; the study documented the manifold effects of this problem on the men's lives, including their "experiences of intimacy with their partners, their relationships with women in social situations, and their self images as sexual beings." 198

Key Question 9: Costs and Cost-Effectiveness of Screening

Several authors have estimated the costs of a screening program for prostate cancer. For example, in 1995 Barry et al. 199 estimated conservatively that first-year costs for a Medicare benefit for PSA screening (including only men ages 65 to 79 years) would be $2.1 billion. Lubke et al. produced estimates of first-year costs of a national screening program using PSA and DRE for men ages 50 to 69 years between $17.6 billion and $25.7 billion 200 (see Evidence Table 8).

Given the uncertainties about the existence and magnitude of benefits, the cost-effectiveness of screening for prostate cancer has been difficult to calculate. A 1993 decision analysis, making optimistic assumptions about benefit from screening and early treatment, found little or no benefit for men with well-differentiated tumors. 201 For men with moderately or poorly differentiated cancers, screening and early treatment could offer as much as 3.5 years improvement in quality-adjusted life expectancy, again using the most optimistic assumptions of treatment efficacy. This model also concluded that, even with optimistic assumptions, men ages 75 years and older are not likely to benefit from screening and aggressive treatment. One major reason for this finding is that any benefits of screening are expected to accrue some years into the future, after many men of this age have died of some other condition. Two subsequent decision analyses have reached the same conclusions. 202 , 203

In 1995, Barry et al. published a cost-effectiveness analysis using very favorable screening assumptions. 199 The marginal cost- effectiveness of screening men age 65 years with PSA and DRE, without adjustment for life quality and without discounting benefits, is between $12,500 and $15,000 per life-year saved. Changing only a few assumptions, however, quickly increased the marginal cost-effectiveness ratio to above $100,000 per life-year saved. Taking into account a decrement in the quality of life associated with the harms of treatment would make this ratio even less favorable. In 1997, these investigators updated their model with more recent data and further assumptions favorable to screening. 204 Their findings were similar.

A more recent model used inputs from US lifetables and prostate cancer mortality rates from the SEER registry to explore the relationship between increased PSA screening and the recent reduction in prostate cancer mortality (see Key Question 1). 33 Assuming that screening reduces prostate cancer mortality by 20% (the level used to calculate sample size in the PLCO trial), then PSA screening could explain the decline in mortality only with a lead time of 3 years or less, much shorter than the 5 years or longer previously thought likely. 35 If lead time is longer than 3 years, then PSA screening can provide at best a partial explanation for the reduction in mortality.

Thus, the cost-effectiveness of screening for prostate cancer depends largely on the efficacy of treatment for cancers detected by screening, and on the length of life of men detected with cancer. If one makes favorable assumptions about efficacy, screening may be cost-effective for men ages 50 to 69 and may have contributed to the recent decline in mortality. If reality is less favorable, then screening could easily result in net harm. The efficacy of treatment for screen-detected cancers is unknown, however, and will not be clear until high-quality RCTs of screening are completed. The models found that men over age 70 to 75 years, or who have a less than 10 year life expectancy, are unlikely to benefit substantially from screening quite apart from efficacy.

Image f3724_F001