Assessing Benefit and Risk in the Prevention of Prostate Cancer: The Prostate Cancer Prevention Trial Revisited
http://www.100md.com
《临床肿瘤学》
the Section of Urologic Oncology, Glickman Urological Institute
Cleveland Clinic Lerner College of Medicine, Cleveland Clinic Foundation, Cleveland, OH
Fred Hutchinson Cancer Research Center, Seattle, WA
Department of Clinical Cancer Prevention, The University of Texas M.D. Anderson Cancer Center, Houston
Department of Urology, University of Texas Health Sciences Center, San Antonio, TX
ABSTRACT
PURPOSE: The Prostate Cancer Prevention Trial demonstrated a 25% reduction in period prevalence of prostate cancer in men randomly assigned to 5 mg/d of finasteride. However, widespread use of finasteride for prevention is inhibited by the observed increased risk of high-grade disease. We present a model of risk and benefit that estimates the potential effects of histologic artifact in the assignment of excess risk for high-grade disease and the possible effect of overdetection bias introduced by finasteride-induced volume reduction.
METHODS: The absolute benefit/absolute risk ratio of finasteride use was estimated by calculating the ratio of absolute risk reduction in the finasteride arm to the absolute risk of excess high-grade cancers. This ratio was recalculated for assumptions that 10%, 25%, or 50% of the excess high-grade cancers were due to histologic artifact, and that there was a 25% overdetection bias in the finasteride arm.
RESULTS: For all cancers the absolute benefit/absolute risk ratio increased from 4.6:1 to 5.1:1, 6.2:1, and 9.2:1 for assumptions of 10%, 25%, or 50% histologic artifact, respectively. The ratio increased from 4.6:1 to 8.2:1 for the assumption of 25% overdetection bias, and to 9.1:1, 10.9:1, and 16.3:1 for combined assumptions of 25% overdetection bias and 10%, 25%, or 50% histologic artifact, respectively.
CONCLUSION: The adoption of a prevention strategy hinges on potential benefits weighed against potential risks. This model demonstrates the magnitude of effect for a hypothesized range of histologic artifact and overdetection bias on the assessment of risk versus benefit for finasteride.
INTRODUCTION
The burden of prostate cancer, the most common nondermatologic malignancy in US men since 1984, can be measured by its incidence, prevalence, and disease-related mortality (Table 1). Although mortality from prostate cancer is decreasing, in the last 5 years alone more than 1 million men in the United States have been newly diagnosed with this disease.1 Despite prostate-specific antigen (PSA) –induced stage migration, a high cure rate for localized disease, and improved understanding of prostate cancer biology, most men who develop metastases are still destined to die as a result of the disease, with almost half a million deaths in the United States between 1989 and 2001.1 The burden of this disease can also be measured in other terms. Newly diagnosed patients must endure anxiety and uncertainty relating to both diagnosis and prognosis, and in the PSA era it is possible for a perfectly healthy young man to be diagnosed with cancer, go through a diagnostic work-up, endure the adverse effects of therapy, and suffer a biochemical relapse without ever having experienced a disease-related symptom.
Recognizing these burdens and faced with an apparent epidemic of prostate cancer in the early 1990s, the National Cancer Institute launched the Prostate Cancer Prevention Trial (PCPT), the first large-scale, population-based test of a chemopreventive strategy in men at risk for this disease. The PCPT was based on two observations: androgens are required for the development of prostate cancer, and men with congenital deficiency of type 2 5- reductase are unaffected by benign prostatic hyperplasia or prostate cancer. The PCPT tested the hypothesis that treatment with finasteride, which induces an acquired deficiency in this enzyme, would lower intraprostatic dihydrotestosterone levels and thereby prevent prostate cancer. In the PCPT, 18,882 men 55 years or older with a normal digital rectal examination (DRE) and a PSA 3.0 ng/mL were randomly assigned to treatment with finasteride (5 mg/d) or placebo for 7 years.2 A prostate biopsy was recommended if the annual PSA level, adjusted for the effect of finasteride, exceeded 4.0 ng/mL, or if the DRE was abnormal. The primary end point was the prevalence of prostate cancer during the 7 years of the study, as diagnosed either by for-cause biopsies (abnormal DRE or PSA) or by end-of-study biopsy specified for all men not previously diagnosed.
The main findings of the PCPT were the prevalence of prostate cancer was reduced by 24.8% (95% CI, 18.6% to 30.6% reduction), from 24.4% (1,147 of 4,692) to 18.4% (803 of 4,368) in those randomly assigned to finasteride compared with placebo, with the most marked reduction seen in Gleason sum 6 cancers; the prevalence of Gleason sum 7 to 10 tumors was higher in the finasteride group than in those receiving placebo (6.4% v 5.1% [a 27% increase]; 95% CI, 7% to 50%), with the most marked increase seen for Gleason sum 8 to 10 tumors; the risk reduction in the finasteride arm was seen in both clinically apparent tumors (those diagnosed as for cause because of an elevated PSA or abnormal DRE) and end-of-study biopsies; the risk reduction among risk groups defined by age, family history, race, and PSA were of the same general magnitude; and sexual adverse effects were more common with finasteride, whereas urinary symptoms were more common with placebo.
The mixed results of the PCPT have left open the question of the ultimate risks and benefits of this agent for prostate cancer prevention. The two major barriers to widespread acceptance of finasteride in this setting are the questions of whether the Gleason sum 6 cancers that were prevented are biologically significant (which we define as destined to metastasize or kill the host), and whether the increased prevalence of higher grade tumors is real or artifactual, given that finasteride is known to change the appearance of prostatic epithelium in a way that could bias assignment of Gleason grade.3-5 We address these issues with a simple model of how histologic artifact and increased risk of detection of cancer based on differences in prostate gland volume in the two study arms of the PCPT might modify the analysis of benefit and risk for this agent.
METHODS
The absolute benefit/absolute risk ratio for finasteride use was defined by calculating the ratio of absolute risk reduction in the finasteride arm to the absolute risk of excess high-grade cancers. This ratio was recalculated for assumptions that 10%, 25%, or 50% of the excess high-grade cancers that were seen in the finasteride arm were due to histologic artifact. In addition, the ratio was recalculated assuming a 25% overdetection bias in the finasteride arm. This assumption is based on two observations. The first observation was that finasteride-treated glands were on average 25% smaller than those in the placebo arm, and a similar number of tissue cores (mean, six per patient) were taken on end-of-study biopsy in both arms of the trial. The second observation was that in previous observations, shorter term finasteride use caused a roughly equal volume reduction in both the transition and peripheral zones of the prostate6; the likelihood that with random biopsy and equal sampling volumes a greater proportion of the prostate is sampled in a smaller gland leads to an increased chance of detecting cancer.7
RESULTS
On the basis of the entire cohort, the 24.8% risk reduction in the finasteride arm translates into an absolute benefit/absolute risk ratio of 4.6:1 (ie, for every 4.6 disease occurrences that are prevented, there is only one additional occurrence of high-grade disease). This ratio of 4.6:1 is derived as follows. First, the absolute risk reduction in the finasteride arm is calculated as 24.4% – 18.4% = 6.0% (95% CI, 4.3% to 7.7%). Second, the absolute increased risk for high-grade disease in the finasteride arm is 6.4% – 5.1% = 1.3% (95% CI, 0.3% to 2.3%). Third, the absolute benefit/absolute risk ratio is the absolute risk reduction in the finasteride arm divided by the absolute increased risk of high-grade disease, or 6.0%/1.3% = a 4.6:1 ratio.
Additional observations lead to theoretical calculations about how the occurrence of histologic artifact and increased risk of detection of cancer in the finasteride-treated glands could influence the calculation of the absolute benefit/absolute risk ratio. Considering histologic artifact first, and using the same mathematical logic as described in the preceding paragraph, Table 2 lists the theoretical absolute benefit/absolute risk ratio based on assumptions that histologic artifact accounts for none, 10%, 25%, or 50% of the excess high-grade disease seen in the finasteride arm. For example, if we assume that 50% of the excess risk of high-grade disease is artifact, the risk ratio would be calculated as follows: 6.0%/(1.3% x 0.5) = 9.2, so that only 0.65% more real high-grade occurrences would be diagnosed on the finasteride arm.
Similarly, the effect of increased risk of biopsy detection in the finasteride arm on the absolute benefit/absolute risk ratio can also be estimated (Table 3). In PCPT, glands treated with finasteride were 25% smaller than those in the placebo arm. Because the mean number of biopsy cores taken in each arm was equal (six), a 25% greater rate of detection in the finasteride arm is suggested. For the purposes of estimating this effect, we assume that a 25% volume reduction for the entire gland translated into a 25% increase in the detection rate of cancer. If the finasteride-treated glands had been the same size as a placebo-treated gland, we estimate an absolute risk reduction of 10.6% instead of the observed 6.0%, calculated as the detection rate in the placebo arm minus the detection rate in finasteride arm with assumed 25% increased risk of detection, or 24.4% – (0.75 x 18.4%) = 24.4% – 13.8% = 10.6% (Table 3).
Examination of the data reveals the following information. Assuming no histologic artifact in the assignment of high-grade disease and no increased risk of detection, the observed absolute benefit/absolute risk ratio is 4.6:1 (ie, for every 4.6 disease occurrences prevented, there is one additional occurrence of high-grade disease [row 1, Table 2]). Assuming that artifact accounts for 10% of the observed increase in high-grade disease, the ratio improves to 5.1:1 (ie, for every 5.1 occurrences prevented, there is one additional occurrence of high-grade disease (row 2, Table 2). Assuming that artifact accounts for 25% or 50% of the observed increase in high-grade disease improves the ratio to 6.2:1 or 9.2:1 (rows 3 and 4, Table 2). Assuming that cancer in the finasteride-treated glands was overdetected by 25% relative to placebo, and factoring varying estimates of the rate of histologic artifact, the absolute benefit/absolute risk ratio ranges from 8.2:1 to 16.3:1 (Table 3). Thus, in the best-case scenario, if there is both a 50% artifact in the assignment of excess high-grade disease and a 25% increased rate of detection, for every 16.3 occurrences prevented there is only one extra occurrence of high-grade disease.
These calculations present the most optimistic estimates of benefit and assume that all of the cancers detected by end-of-study biopsy would ultimately be evident clinically. A similar analysis can be performed only for clinically evident cancers detected from for-cause biopsies. The incidence of detected for-cause cancers during 7 years in the placebo arm was 7.2% and incidence in the finasteride arm was 5.9%; the rates of high-grade disease were 1.9% and 2.6%, respectively. The absolute benefit/absolute risk ratio for cancers detected for cause is therefore (7.2% – 5.9%)/(2.6% – 1.9%) = 1.3%/0.7% = 1.9. Assuming various rates (from none to 50%) of histologic artifact accounting for the higher incidence of high-grade disease with finasteride, the absolute benefit/absolute risk ratio for for-cause cancers ranges from 1.9 to 3.7 (data not shown). Assuming the effects of histologic artifact and a 25% increased rate of detection in the finasteride arm, the ratio ranges from 4.0 to 8.0 (data not shown). Estimates of the absolute benefit/absolute risk ratio using only occurrences detected by for-cause biopsies are likely conservative, given that at least some of the occurrences detected by end-of-study biopsy eventually would also have been detected clinically and would have triggered for-cause biopsies.
DISCUSSION
The decision of whether the advantages of taking finasteride outweigh the potential disadvantages of taking it is complicated. Advantages include 24.8% risk reduction for cancer, fewer urinary symptoms, lower risk of intervention for urinary complaints, and avoiding treatment-related adverse effects. Disadvantages include potential excess mortality from extra occurrences of high-grade disease, the cost in both quality of life and dollars associated with excess sexual adverse effects, and the actual cost of finasteride itself. Overlooked in the debate on these risks and benefits are considerations of the burden of cure of this common disease. Recent data on treatment outcomes shed some light on the severity of this burden.
The good news is that screening has resulted in a substantial clinical and pathologic migration to earlier and more curable stages of disease.8 Several studies document 10-year actuarial biochemical disease-free rates of 93% to 97% for men with T1 and T2 tumors treated by radical prostatectomy or radiation therapy, with overall relapse rates of 15% to 20%.9-13 Even patients who experience relapse after radical prostatectomy have an excellent chance of living 10 or more years.14,15 Despite these encouraging results, the burden of therapy remains substantial. A study of complications after surgical therapy for localized disease in an unselected population-based cohort reported that at more than 18 months after radical prostatectomy, 8.4% of men were incontinent and 41.9% reported that their sexual performance was a moderate to large problem.16 In a similarly designed study comparing outcomes after radiation with those after surgery, the radiation cohort reported an impotence rate of 61.5% and a significantly higher incidence of bowel problems.17 Although many single-institution studies have reported better results in highly selected patients, the majority of men treated for localized disease in the community pay a substantial price to be cured, including fear of cancer recurrence that remains substantial for as long as 2 years after treatment.18 It seems self-evident that an effective prevention strategy would spare many men this burden of diagnosis and cure.
The PCPT was designed to determine if finasteride use could diminish the burden of prostate cancer by reducing its prevalence over a defined time interval (ie, reduce its period prevalence). Period prevalence, which includes cancers diagnosed both by for-cause and end-of-study biopsies, was chosen as the primary end point because it avoided the potential confounds of finasteride use on disease detection, including bias induced by changes in PSA, DRE, and gland volume, and those due to nonadherence and unreported use of finasteride in the placebo arm (drop-ins). Although an end point that included only for-cause biopsies might have had more direct clinical relevance, only one that included all cancers diagnosed during the course of the trial best tested the biologic hypothesis that finasteride could affect the development of prostate cancer.19
In this analysis we present a range of possible scenarios of benefit versus risk of the use of finasteride in the prevention of prostate cancer. We use simple mathematical calculations for assumptions related to the potential biases of histologic artifact and reduction in gland volume associated with long-term use of this agent. Our model suggests that the absolute benefit/absolute risk ratio may be substantially higher than baseline if these assumptions are true. We have avoided assigning specific weighted utilities to the previously enumerated advantages and disadvantages of finasteride use in the calculation of this ratio based on the belief that every individual will assign his own weights based on his age, perceived risk of getting or dying from prostate cancer, level of sexual activity, presence of urologic symptoms, and other indefinable values.
These results of our analysis are limited because we did not know with certainty the actual degree to which the increased risk of detection of cancers resulted in our underestimating the risk reduction for finasteride or the importance (if any) of histologic artifact in assigning higher Gleason scores. Both of these assumptions seem plausible and potentially important, however, because finasteride is known to affect glandular architecture and the Gleason scoring system has never been validated in glands treated with this agent4,5; published observations indicate that sextant biopsies are more likely to find cancer in smaller glands7; and magnetic resonance imaging–based reductions were noted in peripheral zone volume equal to those of the transition zone in glands treated with finasteride for only 2 years.6
The main risk reduction in the PCPT was in the period prevalence of Gleason sum 6 prostate cancer. Whether at a rate of 24.8% or higher, preventing grade 6 disease is certainly relevant clinically. At present there are no biologic, clinical, pathologic, or radiographic markers that allow the prediction of biologic significance of an individual tumor. For the majority of men with newly diagnosed grade 6 disease, the lack of a prognostic marker substantially more often than not leads them to deal with their uncertainty regarding biologic significance by choosing therapy rather than observation. In the most recent update of the CapSure database, which includes representative samples of both academic and community urologic practices, 95% of men with newly diagnosed grade 6 tumors chose to be treated immediately.20 The high treatment penetrance for grade 6 tumors suggests that even if we do not know whether they are biologically significant, for individual men and their physicians these tumors are relevant clinically. This likely reflects the prevailing oncologic belief that even if we are uncertain of the need for cure, early-stage and lower grade tumors are the most curable. The magnitude of clinical relevance is illustrated by the large number of patients with biopsy Gleason sum 6 tumors who undergo surgery or brachytherapy by even the most selective academic practices (including our own).21 Thus, viewed in this context of current urologic practice, preventing grade 6 tumors by finasteride has the added advantage of preventing the anxiety, cost, and morbidity associated with treatment. From a public health perspective, preventing the burden of diagnosis and cure in newly diagnosed patients should be added as a positive aspect of the 24.8% reduction in risk of diagnosis.
The most surprising finding of the PCPT was the 24.4% prevalence of prostate cancer in the placebo arm, four times higher than the 6% assumed for the trial design. This discrepancy can be explained by the fact that the 6% assumption was based on Surveillance, Epidemiology, and End Results incidence estimates, which are derived from clinically evident cases and not on disease prevalence in men undergoing biopsy for no reason other than being on the trial.
Interestingly, the incidence of clinically evident cancers detected for cause by elevations in PSA or abnormal PSA was 7.2% at 7 years, roughly that estimated by use of the Surveillance, Epidemiology, and End Results data. We agree with the view that the large number of clinically unsuspected occurrences discovered by end-of-study biopsies serves to highlight prior clinical observations that many more men have prostate cancer than are destined to die from it, that early-stage prostate cancer is frequently overtreated, and that development of markers that distinguish indolent tumors from biologically significant tumors are needed urgently. In fact, it has been suggested that the risk reduction with finasteride use in the PCPT was closer to 10% than 24.8%, based on the observation that fewer men in the finasteride arm underwent for-cause biopsies, and by assuming that the end-of-study biopsies were not relevant clinically.22 We believe that conclusions based only on for-cause biopsies should be interpreted cautiously because of known biases in ascertainment owing to differences in DRE, PSA, and prostate volume induced by finasteride, and issues relating to adherence and drop-ins. Indeed, the use of period prevalence as the primary trial end point, which includes all cancers detected during intervention, best minimizes the influence of these biases.19 Furthermore, the assumption that cancers detected by end-of-study biopsies are not relevant clinically is arguable given recent data from the placebo arm of the PCPT demonstrating that 15% of men with a normal DRE and PSA less than 4.0 ng/mL have prostate cancer,23 and the recommendation by some advocates of screening that the cutoff for triggering biopsy should be lowered to less than 4.0.24 Even when limited to cancers detected only by for-cause biopsies, the current model demonstrates a range of potential benefit/risk ratios for finasteride use of between 1.9 and 3.7.
Notwithstanding the argument that some high-grade tumors detected while the patient received finasteride may be artifactual, there are some plausible biologic hypotheses that suggest that the effect could be real.21 It is likely that until this issue is settled, routine use of finasteride to prevent cancer will not be embraced widely. A group of pathologists and scientists has been convened to investigate this and other issues related to explanation of the trial results. Although some of the biologic hypotheses are testable in laboratory models, we are uncertain whether this question will ever be answered definitively.
Ultimately, the adoption of a preventive strategy always hinges on its potential benefits weighed against the potential risks. Many examples, both pro and con, can be cited to illustrate this equation. The Breast Cancer Prevention Trial demonstrated that tamoxifen can reduce the incidence of both invasive and noninvasive breast cancer and bone fractures in women at increased risk.25 In that trial premenopausal women with a 5-year risk of developing invasive breast cancer greater than 1.67% derived net benefit from using tamoxifen. Although these benefits occurred at the cost of an increased risk of endometrial cancer, thromboembolic disease, and cataracts, tamoxifen is now approved by the US Food and Drug Administration for prevention of breast cancer in high-risk women. More recently, the Women's Health Initiative concluded that routine estrogen and progesterone replacement in postmenopausal women had advantages in reducing the risks of bone fractures and colon cancer, but that the increased risks of breast cancer, heart attacks, and strokes did not justify routine use.26
We offer the current analysis of PCPT results within this context; we do not know the degree to which increased risk of detection of cancers resulted in underestimating the risk reduction for finasteride or the importance of histologic artifact in assigning higher Gleason scores. This analysis is presented to provide a framework for additional discussion of the results of the PCPT; to provide data for individual men to weigh their own risk/benefit options; to contribute to the debate of whether it makes public health sense to recommend finasteride broadly to prevent prostate cancer, and in anticipation of further pathologic and other related data, including a long-term follow-up study limited to those diagnosed with cancer during the trial with survival and time to metastasis as end points, that can be analyzed within this framework. Ultimately, the decision about whether the benefits outweigh the risks of finasteride for prostate cancer prevention will rest on a more comprehensive analysis including data from new studies.
Authors' Disclosures of Potential Conflicts of Interest
Although all authors completed the disclosure declaration, the following authors or their immediate family members indicated a financial interest. No conflict exists for drugs or devices used in a study if they are not being evaluated as part of the investigation. For a detailed description of the disclosure categories, or for more information about ASCO's conflict of interest policy, please refer to the Author Disclosure Declaration and the Disclosures of Potential Conflicts of Interest section in Information for Contributors.
NOTES
Authors' disclosures of potential conflicts of interest are found at the end of this article.
REFERENCES
Jemal A, Tiwari RC, Murray T, et al: Cancer statistics, 2004. CA Cancer J Clin 54:8-29, 2004
Thompson IM, Goodman PJ, Tangen CM, et al: The influence of finasteride on the development of prostate cancer. N Engl J Med 17:215-224, 2003
Civantos F, Soloway MS, Pinto JE: Histopathological effects of androgen deprivation in prostatic cancer. Semin Urol Oncol 14:22-31, 1996
Bostwick DG, Qian J, Civantos F, et al: Does finasteride alter the pathology of the prostate and cancer grading? Clin Prostate Cancer 2:228-235, 2004
Rubin MA, Kantoff PW: Effect of finasteride on risk of prostate cancer: How little we really know. J Cell Biochem 91:478-482, 2004
Marks LS, Partin AW, Dorey FJ, et al: Long-term effects of finasteride on prostate tissue composition. Urology 53:574-580, 1999
Uzzo RG, Wei JT, Waldbaum RS, et al: The influence of prostate size on cancer detection. Urology 46:831-836, 1995
Jhaveri FM, Klein EA, Kupelian PA, et al: Declining rates of extracapsular extension in radical prostatectomy: Evidence for continued stage migration. J Clin Oncol 17:3167-3172, 1999
Clark PE, Levin HS, Kupelian PA, et al: Intermediate-term outcome with radical prostatectomy for localized prostate cancer: The Cleveland Clinic Experience. Prostate J 3:118-125, 2001
Eastham JA, Scardino PT: Radical prostatectomy for clinical stage T1 and T2 prostate cancer, in NJ Vogelzang, PT Scardino, WU Shipley et al (eds): Comprehensive Textbook of Genitourinary Oncology (ed 2). Philadelphia, PA, Lippincott Williams & Wilkins, 1999, pp 722-738
Walsh PC, Partin AW, Epstein JI: Cancer control and quality of life following anatomical radical retropubic prostatectomy: Results at 10 years. J Urol 152:1831-1836, 1994
Catalona WJ, Ramos CG, Carvalhal GF: Contemporary results of anatomic radical prostatectomy. CA Cancer J Clin 49:282-296, 1999
Kupelian PA, Elshaikh M, Reddy CA, et al: Comparison of the efficacy of local therapies for localized prostate cancer in the PSA era: A large single institution experience with radical prostatectomy and external beam radiotherapy. J Clin Oncol 20:3376-3385, 2002
Jhaveri FM, Zippe CD, Klein EA, et al: Biochemical failure does not predict overall survival after radical prostatectomy for localized prostate cancer: 10 year results. Urology 54:884-890, 1999
Pound CR, Partin AW, Eisenberger MA, et al: Natural history of progression after PSA elevation following radical prostatectomy. JAMA 281:1591-1597, 1999
Stanford JL, Feng Z, Hamilton AS, et al: Urinary and sexual function after radical prostatectomy for clinically localized prostate cancer: The Prostate Cancer Outcomes Study. JAMA 283:354-360, 2000
Potosky AL, Legler J, Albertsen PC, et al: Health outcomes after prostatectomy or radiotherapy for prostate cancer: Results from the Prostate Cancer Outcomes Study. J Natl Cancer Inst 92:1582-1592, 2000
Mehta SS, Lubeck DP, Pasta DJ, et al: Fear of cancer recurrence in patients undergoing definitive treatment for prostate cancer: Results from CaPSURE. J Urol 170:1931-1933, 2003
Thompson IM, Tangen C, Goodman P: The Prostate Cancer Prevention Trial: Design, status, and promise. World J Urol 21:28-30, 2003
Harlan SR, Cooperberg MR, Elkin EP, et al: Time trends and characteristics of men choosing watchful waiting for initial treatment of localized prostate cancer: Results from CaPSURE. J Urol 170:1804-1807, 2003
Thompson I, Klein EA, Lippman SM, et al: Prevention of prostate cancer with finasteride: A U.S./European perspective. Eur Urol 44:650-655, 2003
Walsh PC: Editorial comment. J Urol 171:506-507, 2004
Thompson IM, Pauler DK, Goodman PJ, et al: Prevalence of prostate cancer among men with a prostate-specific antigen level < or =4.0 ng per milliliter. N Engl J Med 350:2239-2246, 2004
Punglia RS, D'Amico AV, Catalona WJ, et al: Effect of verification bias on screening for prostate cancer by measurement of prostate-specific antigen. N Engl J Med 349:335-342, 2003
Wolmark N, Dunn BK: The role of tamoxifen in breast cancer prevention: Issues sparked by the NSABP Breast Cancer Prevention Trial (P-1). Ann N Y Acad Sci 949:99-108, 2001
Randal J: The end of an era? Study reveals harms of hormone replacement therapy. J Natl Cancer Inst 94:1116-1118, 2002
Submitted September 2, 2004; accepted April 25, 2005.(Eric A. Klein, Catherine )
Cleveland Clinic Lerner College of Medicine, Cleveland Clinic Foundation, Cleveland, OH
Fred Hutchinson Cancer Research Center, Seattle, WA
Department of Clinical Cancer Prevention, The University of Texas M.D. Anderson Cancer Center, Houston
Department of Urology, University of Texas Health Sciences Center, San Antonio, TX
ABSTRACT
PURPOSE: The Prostate Cancer Prevention Trial demonstrated a 25% reduction in period prevalence of prostate cancer in men randomly assigned to 5 mg/d of finasteride. However, widespread use of finasteride for prevention is inhibited by the observed increased risk of high-grade disease. We present a model of risk and benefit that estimates the potential effects of histologic artifact in the assignment of excess risk for high-grade disease and the possible effect of overdetection bias introduced by finasteride-induced volume reduction.
METHODS: The absolute benefit/absolute risk ratio of finasteride use was estimated by calculating the ratio of absolute risk reduction in the finasteride arm to the absolute risk of excess high-grade cancers. This ratio was recalculated for assumptions that 10%, 25%, or 50% of the excess high-grade cancers were due to histologic artifact, and that there was a 25% overdetection bias in the finasteride arm.
RESULTS: For all cancers the absolute benefit/absolute risk ratio increased from 4.6:1 to 5.1:1, 6.2:1, and 9.2:1 for assumptions of 10%, 25%, or 50% histologic artifact, respectively. The ratio increased from 4.6:1 to 8.2:1 for the assumption of 25% overdetection bias, and to 9.1:1, 10.9:1, and 16.3:1 for combined assumptions of 25% overdetection bias and 10%, 25%, or 50% histologic artifact, respectively.
CONCLUSION: The adoption of a prevention strategy hinges on potential benefits weighed against potential risks. This model demonstrates the magnitude of effect for a hypothesized range of histologic artifact and overdetection bias on the assessment of risk versus benefit for finasteride.
INTRODUCTION
The burden of prostate cancer, the most common nondermatologic malignancy in US men since 1984, can be measured by its incidence, prevalence, and disease-related mortality (Table 1). Although mortality from prostate cancer is decreasing, in the last 5 years alone more than 1 million men in the United States have been newly diagnosed with this disease.1 Despite prostate-specific antigen (PSA) –induced stage migration, a high cure rate for localized disease, and improved understanding of prostate cancer biology, most men who develop metastases are still destined to die as a result of the disease, with almost half a million deaths in the United States between 1989 and 2001.1 The burden of this disease can also be measured in other terms. Newly diagnosed patients must endure anxiety and uncertainty relating to both diagnosis and prognosis, and in the PSA era it is possible for a perfectly healthy young man to be diagnosed with cancer, go through a diagnostic work-up, endure the adverse effects of therapy, and suffer a biochemical relapse without ever having experienced a disease-related symptom.
Recognizing these burdens and faced with an apparent epidemic of prostate cancer in the early 1990s, the National Cancer Institute launched the Prostate Cancer Prevention Trial (PCPT), the first large-scale, population-based test of a chemopreventive strategy in men at risk for this disease. The PCPT was based on two observations: androgens are required for the development of prostate cancer, and men with congenital deficiency of type 2 5- reductase are unaffected by benign prostatic hyperplasia or prostate cancer. The PCPT tested the hypothesis that treatment with finasteride, which induces an acquired deficiency in this enzyme, would lower intraprostatic dihydrotestosterone levels and thereby prevent prostate cancer. In the PCPT, 18,882 men 55 years or older with a normal digital rectal examination (DRE) and a PSA 3.0 ng/mL were randomly assigned to treatment with finasteride (5 mg/d) or placebo for 7 years.2 A prostate biopsy was recommended if the annual PSA level, adjusted for the effect of finasteride, exceeded 4.0 ng/mL, or if the DRE was abnormal. The primary end point was the prevalence of prostate cancer during the 7 years of the study, as diagnosed either by for-cause biopsies (abnormal DRE or PSA) or by end-of-study biopsy specified for all men not previously diagnosed.
The main findings of the PCPT were the prevalence of prostate cancer was reduced by 24.8% (95% CI, 18.6% to 30.6% reduction), from 24.4% (1,147 of 4,692) to 18.4% (803 of 4,368) in those randomly assigned to finasteride compared with placebo, with the most marked reduction seen in Gleason sum 6 cancers; the prevalence of Gleason sum 7 to 10 tumors was higher in the finasteride group than in those receiving placebo (6.4% v 5.1% [a 27% increase]; 95% CI, 7% to 50%), with the most marked increase seen for Gleason sum 8 to 10 tumors; the risk reduction in the finasteride arm was seen in both clinically apparent tumors (those diagnosed as for cause because of an elevated PSA or abnormal DRE) and end-of-study biopsies; the risk reduction among risk groups defined by age, family history, race, and PSA were of the same general magnitude; and sexual adverse effects were more common with finasteride, whereas urinary symptoms were more common with placebo.
The mixed results of the PCPT have left open the question of the ultimate risks and benefits of this agent for prostate cancer prevention. The two major barriers to widespread acceptance of finasteride in this setting are the questions of whether the Gleason sum 6 cancers that were prevented are biologically significant (which we define as destined to metastasize or kill the host), and whether the increased prevalence of higher grade tumors is real or artifactual, given that finasteride is known to change the appearance of prostatic epithelium in a way that could bias assignment of Gleason grade.3-5 We address these issues with a simple model of how histologic artifact and increased risk of detection of cancer based on differences in prostate gland volume in the two study arms of the PCPT might modify the analysis of benefit and risk for this agent.
METHODS
The absolute benefit/absolute risk ratio for finasteride use was defined by calculating the ratio of absolute risk reduction in the finasteride arm to the absolute risk of excess high-grade cancers. This ratio was recalculated for assumptions that 10%, 25%, or 50% of the excess high-grade cancers that were seen in the finasteride arm were due to histologic artifact. In addition, the ratio was recalculated assuming a 25% overdetection bias in the finasteride arm. This assumption is based on two observations. The first observation was that finasteride-treated glands were on average 25% smaller than those in the placebo arm, and a similar number of tissue cores (mean, six per patient) were taken on end-of-study biopsy in both arms of the trial. The second observation was that in previous observations, shorter term finasteride use caused a roughly equal volume reduction in both the transition and peripheral zones of the prostate6; the likelihood that with random biopsy and equal sampling volumes a greater proportion of the prostate is sampled in a smaller gland leads to an increased chance of detecting cancer.7
RESULTS
On the basis of the entire cohort, the 24.8% risk reduction in the finasteride arm translates into an absolute benefit/absolute risk ratio of 4.6:1 (ie, for every 4.6 disease occurrences that are prevented, there is only one additional occurrence of high-grade disease). This ratio of 4.6:1 is derived as follows. First, the absolute risk reduction in the finasteride arm is calculated as 24.4% – 18.4% = 6.0% (95% CI, 4.3% to 7.7%). Second, the absolute increased risk for high-grade disease in the finasteride arm is 6.4% – 5.1% = 1.3% (95% CI, 0.3% to 2.3%). Third, the absolute benefit/absolute risk ratio is the absolute risk reduction in the finasteride arm divided by the absolute increased risk of high-grade disease, or 6.0%/1.3% = a 4.6:1 ratio.
Additional observations lead to theoretical calculations about how the occurrence of histologic artifact and increased risk of detection of cancer in the finasteride-treated glands could influence the calculation of the absolute benefit/absolute risk ratio. Considering histologic artifact first, and using the same mathematical logic as described in the preceding paragraph, Table 2 lists the theoretical absolute benefit/absolute risk ratio based on assumptions that histologic artifact accounts for none, 10%, 25%, or 50% of the excess high-grade disease seen in the finasteride arm. For example, if we assume that 50% of the excess risk of high-grade disease is artifact, the risk ratio would be calculated as follows: 6.0%/(1.3% x 0.5) = 9.2, so that only 0.65% more real high-grade occurrences would be diagnosed on the finasteride arm.
Similarly, the effect of increased risk of biopsy detection in the finasteride arm on the absolute benefit/absolute risk ratio can also be estimated (Table 3). In PCPT, glands treated with finasteride were 25% smaller than those in the placebo arm. Because the mean number of biopsy cores taken in each arm was equal (six), a 25% greater rate of detection in the finasteride arm is suggested. For the purposes of estimating this effect, we assume that a 25% volume reduction for the entire gland translated into a 25% increase in the detection rate of cancer. If the finasteride-treated glands had been the same size as a placebo-treated gland, we estimate an absolute risk reduction of 10.6% instead of the observed 6.0%, calculated as the detection rate in the placebo arm minus the detection rate in finasteride arm with assumed 25% increased risk of detection, or 24.4% – (0.75 x 18.4%) = 24.4% – 13.8% = 10.6% (Table 3).
Examination of the data reveals the following information. Assuming no histologic artifact in the assignment of high-grade disease and no increased risk of detection, the observed absolute benefit/absolute risk ratio is 4.6:1 (ie, for every 4.6 disease occurrences prevented, there is one additional occurrence of high-grade disease [row 1, Table 2]). Assuming that artifact accounts for 10% of the observed increase in high-grade disease, the ratio improves to 5.1:1 (ie, for every 5.1 occurrences prevented, there is one additional occurrence of high-grade disease (row 2, Table 2). Assuming that artifact accounts for 25% or 50% of the observed increase in high-grade disease improves the ratio to 6.2:1 or 9.2:1 (rows 3 and 4, Table 2). Assuming that cancer in the finasteride-treated glands was overdetected by 25% relative to placebo, and factoring varying estimates of the rate of histologic artifact, the absolute benefit/absolute risk ratio ranges from 8.2:1 to 16.3:1 (Table 3). Thus, in the best-case scenario, if there is both a 50% artifact in the assignment of excess high-grade disease and a 25% increased rate of detection, for every 16.3 occurrences prevented there is only one extra occurrence of high-grade disease.
These calculations present the most optimistic estimates of benefit and assume that all of the cancers detected by end-of-study biopsy would ultimately be evident clinically. A similar analysis can be performed only for clinically evident cancers detected from for-cause biopsies. The incidence of detected for-cause cancers during 7 years in the placebo arm was 7.2% and incidence in the finasteride arm was 5.9%; the rates of high-grade disease were 1.9% and 2.6%, respectively. The absolute benefit/absolute risk ratio for cancers detected for cause is therefore (7.2% – 5.9%)/(2.6% – 1.9%) = 1.3%/0.7% = 1.9. Assuming various rates (from none to 50%) of histologic artifact accounting for the higher incidence of high-grade disease with finasteride, the absolute benefit/absolute risk ratio for for-cause cancers ranges from 1.9 to 3.7 (data not shown). Assuming the effects of histologic artifact and a 25% increased rate of detection in the finasteride arm, the ratio ranges from 4.0 to 8.0 (data not shown). Estimates of the absolute benefit/absolute risk ratio using only occurrences detected by for-cause biopsies are likely conservative, given that at least some of the occurrences detected by end-of-study biopsy eventually would also have been detected clinically and would have triggered for-cause biopsies.
DISCUSSION
The decision of whether the advantages of taking finasteride outweigh the potential disadvantages of taking it is complicated. Advantages include 24.8% risk reduction for cancer, fewer urinary symptoms, lower risk of intervention for urinary complaints, and avoiding treatment-related adverse effects. Disadvantages include potential excess mortality from extra occurrences of high-grade disease, the cost in both quality of life and dollars associated with excess sexual adverse effects, and the actual cost of finasteride itself. Overlooked in the debate on these risks and benefits are considerations of the burden of cure of this common disease. Recent data on treatment outcomes shed some light on the severity of this burden.
The good news is that screening has resulted in a substantial clinical and pathologic migration to earlier and more curable stages of disease.8 Several studies document 10-year actuarial biochemical disease-free rates of 93% to 97% for men with T1 and T2 tumors treated by radical prostatectomy or radiation therapy, with overall relapse rates of 15% to 20%.9-13 Even patients who experience relapse after radical prostatectomy have an excellent chance of living 10 or more years.14,15 Despite these encouraging results, the burden of therapy remains substantial. A study of complications after surgical therapy for localized disease in an unselected population-based cohort reported that at more than 18 months after radical prostatectomy, 8.4% of men were incontinent and 41.9% reported that their sexual performance was a moderate to large problem.16 In a similarly designed study comparing outcomes after radiation with those after surgery, the radiation cohort reported an impotence rate of 61.5% and a significantly higher incidence of bowel problems.17 Although many single-institution studies have reported better results in highly selected patients, the majority of men treated for localized disease in the community pay a substantial price to be cured, including fear of cancer recurrence that remains substantial for as long as 2 years after treatment.18 It seems self-evident that an effective prevention strategy would spare many men this burden of diagnosis and cure.
The PCPT was designed to determine if finasteride use could diminish the burden of prostate cancer by reducing its prevalence over a defined time interval (ie, reduce its period prevalence). Period prevalence, which includes cancers diagnosed both by for-cause and end-of-study biopsies, was chosen as the primary end point because it avoided the potential confounds of finasteride use on disease detection, including bias induced by changes in PSA, DRE, and gland volume, and those due to nonadherence and unreported use of finasteride in the placebo arm (drop-ins). Although an end point that included only for-cause biopsies might have had more direct clinical relevance, only one that included all cancers diagnosed during the course of the trial best tested the biologic hypothesis that finasteride could affect the development of prostate cancer.19
In this analysis we present a range of possible scenarios of benefit versus risk of the use of finasteride in the prevention of prostate cancer. We use simple mathematical calculations for assumptions related to the potential biases of histologic artifact and reduction in gland volume associated with long-term use of this agent. Our model suggests that the absolute benefit/absolute risk ratio may be substantially higher than baseline if these assumptions are true. We have avoided assigning specific weighted utilities to the previously enumerated advantages and disadvantages of finasteride use in the calculation of this ratio based on the belief that every individual will assign his own weights based on his age, perceived risk of getting or dying from prostate cancer, level of sexual activity, presence of urologic symptoms, and other indefinable values.
These results of our analysis are limited because we did not know with certainty the actual degree to which the increased risk of detection of cancers resulted in our underestimating the risk reduction for finasteride or the importance (if any) of histologic artifact in assigning higher Gleason scores. Both of these assumptions seem plausible and potentially important, however, because finasteride is known to affect glandular architecture and the Gleason scoring system has never been validated in glands treated with this agent4,5; published observations indicate that sextant biopsies are more likely to find cancer in smaller glands7; and magnetic resonance imaging–based reductions were noted in peripheral zone volume equal to those of the transition zone in glands treated with finasteride for only 2 years.6
The main risk reduction in the PCPT was in the period prevalence of Gleason sum 6 prostate cancer. Whether at a rate of 24.8% or higher, preventing grade 6 disease is certainly relevant clinically. At present there are no biologic, clinical, pathologic, or radiographic markers that allow the prediction of biologic significance of an individual tumor. For the majority of men with newly diagnosed grade 6 disease, the lack of a prognostic marker substantially more often than not leads them to deal with their uncertainty regarding biologic significance by choosing therapy rather than observation. In the most recent update of the CapSure database, which includes representative samples of both academic and community urologic practices, 95% of men with newly diagnosed grade 6 tumors chose to be treated immediately.20 The high treatment penetrance for grade 6 tumors suggests that even if we do not know whether they are biologically significant, for individual men and their physicians these tumors are relevant clinically. This likely reflects the prevailing oncologic belief that even if we are uncertain of the need for cure, early-stage and lower grade tumors are the most curable. The magnitude of clinical relevance is illustrated by the large number of patients with biopsy Gleason sum 6 tumors who undergo surgery or brachytherapy by even the most selective academic practices (including our own).21 Thus, viewed in this context of current urologic practice, preventing grade 6 tumors by finasteride has the added advantage of preventing the anxiety, cost, and morbidity associated with treatment. From a public health perspective, preventing the burden of diagnosis and cure in newly diagnosed patients should be added as a positive aspect of the 24.8% reduction in risk of diagnosis.
The most surprising finding of the PCPT was the 24.4% prevalence of prostate cancer in the placebo arm, four times higher than the 6% assumed for the trial design. This discrepancy can be explained by the fact that the 6% assumption was based on Surveillance, Epidemiology, and End Results incidence estimates, which are derived from clinically evident cases and not on disease prevalence in men undergoing biopsy for no reason other than being on the trial.
Interestingly, the incidence of clinically evident cancers detected for cause by elevations in PSA or abnormal PSA was 7.2% at 7 years, roughly that estimated by use of the Surveillance, Epidemiology, and End Results data. We agree with the view that the large number of clinically unsuspected occurrences discovered by end-of-study biopsies serves to highlight prior clinical observations that many more men have prostate cancer than are destined to die from it, that early-stage prostate cancer is frequently overtreated, and that development of markers that distinguish indolent tumors from biologically significant tumors are needed urgently. In fact, it has been suggested that the risk reduction with finasteride use in the PCPT was closer to 10% than 24.8%, based on the observation that fewer men in the finasteride arm underwent for-cause biopsies, and by assuming that the end-of-study biopsies were not relevant clinically.22 We believe that conclusions based only on for-cause biopsies should be interpreted cautiously because of known biases in ascertainment owing to differences in DRE, PSA, and prostate volume induced by finasteride, and issues relating to adherence and drop-ins. Indeed, the use of period prevalence as the primary trial end point, which includes all cancers detected during intervention, best minimizes the influence of these biases.19 Furthermore, the assumption that cancers detected by end-of-study biopsies are not relevant clinically is arguable given recent data from the placebo arm of the PCPT demonstrating that 15% of men with a normal DRE and PSA less than 4.0 ng/mL have prostate cancer,23 and the recommendation by some advocates of screening that the cutoff for triggering biopsy should be lowered to less than 4.0.24 Even when limited to cancers detected only by for-cause biopsies, the current model demonstrates a range of potential benefit/risk ratios for finasteride use of between 1.9 and 3.7.
Notwithstanding the argument that some high-grade tumors detected while the patient received finasteride may be artifactual, there are some plausible biologic hypotheses that suggest that the effect could be real.21 It is likely that until this issue is settled, routine use of finasteride to prevent cancer will not be embraced widely. A group of pathologists and scientists has been convened to investigate this and other issues related to explanation of the trial results. Although some of the biologic hypotheses are testable in laboratory models, we are uncertain whether this question will ever be answered definitively.
Ultimately, the adoption of a preventive strategy always hinges on its potential benefits weighed against the potential risks. Many examples, both pro and con, can be cited to illustrate this equation. The Breast Cancer Prevention Trial demonstrated that tamoxifen can reduce the incidence of both invasive and noninvasive breast cancer and bone fractures in women at increased risk.25 In that trial premenopausal women with a 5-year risk of developing invasive breast cancer greater than 1.67% derived net benefit from using tamoxifen. Although these benefits occurred at the cost of an increased risk of endometrial cancer, thromboembolic disease, and cataracts, tamoxifen is now approved by the US Food and Drug Administration for prevention of breast cancer in high-risk women. More recently, the Women's Health Initiative concluded that routine estrogen and progesterone replacement in postmenopausal women had advantages in reducing the risks of bone fractures and colon cancer, but that the increased risks of breast cancer, heart attacks, and strokes did not justify routine use.26
We offer the current analysis of PCPT results within this context; we do not know the degree to which increased risk of detection of cancers resulted in underestimating the risk reduction for finasteride or the importance of histologic artifact in assigning higher Gleason scores. This analysis is presented to provide a framework for additional discussion of the results of the PCPT; to provide data for individual men to weigh their own risk/benefit options; to contribute to the debate of whether it makes public health sense to recommend finasteride broadly to prevent prostate cancer, and in anticipation of further pathologic and other related data, including a long-term follow-up study limited to those diagnosed with cancer during the trial with survival and time to metastasis as end points, that can be analyzed within this framework. Ultimately, the decision about whether the benefits outweigh the risks of finasteride for prostate cancer prevention will rest on a more comprehensive analysis including data from new studies.
Authors' Disclosures of Potential Conflicts of Interest
Although all authors completed the disclosure declaration, the following authors or their immediate family members indicated a financial interest. No conflict exists for drugs or devices used in a study if they are not being evaluated as part of the investigation. For a detailed description of the disclosure categories, or for more information about ASCO's conflict of interest policy, please refer to the Author Disclosure Declaration and the Disclosures of Potential Conflicts of Interest section in Information for Contributors.
NOTES
Authors' disclosures of potential conflicts of interest are found at the end of this article.
REFERENCES
Jemal A, Tiwari RC, Murray T, et al: Cancer statistics, 2004. CA Cancer J Clin 54:8-29, 2004
Thompson IM, Goodman PJ, Tangen CM, et al: The influence of finasteride on the development of prostate cancer. N Engl J Med 17:215-224, 2003
Civantos F, Soloway MS, Pinto JE: Histopathological effects of androgen deprivation in prostatic cancer. Semin Urol Oncol 14:22-31, 1996
Bostwick DG, Qian J, Civantos F, et al: Does finasteride alter the pathology of the prostate and cancer grading? Clin Prostate Cancer 2:228-235, 2004
Rubin MA, Kantoff PW: Effect of finasteride on risk of prostate cancer: How little we really know. J Cell Biochem 91:478-482, 2004
Marks LS, Partin AW, Dorey FJ, et al: Long-term effects of finasteride on prostate tissue composition. Urology 53:574-580, 1999
Uzzo RG, Wei JT, Waldbaum RS, et al: The influence of prostate size on cancer detection. Urology 46:831-836, 1995
Jhaveri FM, Klein EA, Kupelian PA, et al: Declining rates of extracapsular extension in radical prostatectomy: Evidence for continued stage migration. J Clin Oncol 17:3167-3172, 1999
Clark PE, Levin HS, Kupelian PA, et al: Intermediate-term outcome with radical prostatectomy for localized prostate cancer: The Cleveland Clinic Experience. Prostate J 3:118-125, 2001
Eastham JA, Scardino PT: Radical prostatectomy for clinical stage T1 and T2 prostate cancer, in NJ Vogelzang, PT Scardino, WU Shipley et al (eds): Comprehensive Textbook of Genitourinary Oncology (ed 2). Philadelphia, PA, Lippincott Williams & Wilkins, 1999, pp 722-738
Walsh PC, Partin AW, Epstein JI: Cancer control and quality of life following anatomical radical retropubic prostatectomy: Results at 10 years. J Urol 152:1831-1836, 1994
Catalona WJ, Ramos CG, Carvalhal GF: Contemporary results of anatomic radical prostatectomy. CA Cancer J Clin 49:282-296, 1999
Kupelian PA, Elshaikh M, Reddy CA, et al: Comparison of the efficacy of local therapies for localized prostate cancer in the PSA era: A large single institution experience with radical prostatectomy and external beam radiotherapy. J Clin Oncol 20:3376-3385, 2002
Jhaveri FM, Zippe CD, Klein EA, et al: Biochemical failure does not predict overall survival after radical prostatectomy for localized prostate cancer: 10 year results. Urology 54:884-890, 1999
Pound CR, Partin AW, Eisenberger MA, et al: Natural history of progression after PSA elevation following radical prostatectomy. JAMA 281:1591-1597, 1999
Stanford JL, Feng Z, Hamilton AS, et al: Urinary and sexual function after radical prostatectomy for clinically localized prostate cancer: The Prostate Cancer Outcomes Study. JAMA 283:354-360, 2000
Potosky AL, Legler J, Albertsen PC, et al: Health outcomes after prostatectomy or radiotherapy for prostate cancer: Results from the Prostate Cancer Outcomes Study. J Natl Cancer Inst 92:1582-1592, 2000
Mehta SS, Lubeck DP, Pasta DJ, et al: Fear of cancer recurrence in patients undergoing definitive treatment for prostate cancer: Results from CaPSURE. J Urol 170:1931-1933, 2003
Thompson IM, Tangen C, Goodman P: The Prostate Cancer Prevention Trial: Design, status, and promise. World J Urol 21:28-30, 2003
Harlan SR, Cooperberg MR, Elkin EP, et al: Time trends and characteristics of men choosing watchful waiting for initial treatment of localized prostate cancer: Results from CaPSURE. J Urol 170:1804-1807, 2003
Thompson I, Klein EA, Lippman SM, et al: Prevention of prostate cancer with finasteride: A U.S./European perspective. Eur Urol 44:650-655, 2003
Walsh PC: Editorial comment. J Urol 171:506-507, 2004
Thompson IM, Pauler DK, Goodman PJ, et al: Prevalence of prostate cancer among men with a prostate-specific antigen level < or =4.0 ng per milliliter. N Engl J Med 350:2239-2246, 2004
Punglia RS, D'Amico AV, Catalona WJ, et al: Effect of verification bias on screening for prostate cancer by measurement of prostate-specific antigen. N Engl J Med 349:335-342, 2003
Wolmark N, Dunn BK: The role of tamoxifen in breast cancer prevention: Issues sparked by the NSABP Breast Cancer Prevention Trial (P-1). Ann N Y Acad Sci 949:99-108, 2001
Randal J: The end of an era? Study reveals harms of hormone replacement therapy. J Natl Cancer Inst 94:1116-1118, 2002
Submitted September 2, 2004; accepted April 25, 2005.(Eric A. Klein, Catherine )