Drs. Benoit and Naslund venture into the complex arena of medical economics and cost-effectiveness analysis of prostate cancer screeninga task that is made all the more difficult because of the dual paucity of data on costs and effectiveness. Their underlying premises are that cost control is a dominant concern in the prostate cancer screening debate and that cost-effectiveness analyses have been used to justify denial of prostate cancer screening. Both of these assumptions bear scrutiny.
Why Screen for Prostate Cancer?
The analyses presented in the literature by skeptics of widespread prostate screening[1,2] or by those who specifically recommend against it[3-7] focus primarily on the uncertainty over whether routine screening would lead to net benefit or harm. Society must, at this point, weigh theoretical benefits against the known, substantial morbidity incurred by screening and subsequent therapy. The costs of dominant concern are human, not economic. It is not yet a matter of whether, as the authors suggest, society must decide if the years of life saved in these men warrants the use of its limited health care resources. First, society needs to decide if years of life are saved at all, and, if so, at what cost of morbidity and treatment-related mortality. To couch the debate primarily in terms of interest rates and returns on bonds runs the risk of trivializing the central Hippocratic concern of harm without proven benefit.
Where is the evidence that cost considerations have presented a major impediment to prostate cancer screening? In the late 1980s, the United States entered the largest epidemic of any cancer since the start of formal national cancer record-keepingan epidemic largely attributable to prostate-specific antigen (PSA) screening. From 1984 to 1990, Medicare data showed a 575% increase in the use of radical prostatectomy.
The authors state that breast cancer screening was implemented without knowledge of its effect on breast cancer mortality. Actually, widespread breast cancer screening in the United States did not begin until early reports from the randomized Health Insurance Plan (HIP) of New York study showed a statistically significant mortality reduction in the screened arm. In fact, the study was able to show statistical significance relatively quickly because there was virtually no population screening outside of the trial and therefore virtually no contamination effect in the control arm of the study. In contrast, any randomized prostate cancer screening study in the United States has to contend with the possibility of contamination of the control arm due to the national enthusiasm for, and easy availability of, prostate cancer screening. This may delay our ability to assess the net benefits and harms of screening by years.
Drs. Benoit and Naslund do point out some of the complexities in performing cost-effectiveness analyses. It is incorrect, however, to assume that the most cost-effective strategy would be to withhold any treatment of prostate cancer. Even if this were the least costly approach, its effectiveness in terms of life-years saved would be zero, and the ratio of cost to effectiveness would explode to infinity. Hence, such an approach could be the least cost-effective strategy of all.
The authors go on to point out the difficulties associated with discounting future life-years saved. However, discounting life-year benefits is not a matter of interest rates or economic indicators. Rather, discounting life-years saved in the distant future is performed because personal values demand it. Most people would attach considerably more value to an additional year of life if they were otherwise to die tomorrow than they would if they were to die in 25 years. Such valuation is difficult to account precisely, since it depends on an individuals point of view, but that should not undermine the concept of discounting. In the case of prostate cancer, men are in the position of weighing the immediate risk of morbidity (and even small mortality risk) against the potential of extending their life in the more distant future by some unknown amount.
Another complicated area in some models is the assessment of quality-of-life adjustments. These valuations, again, are personal and are difficult to generalize in any statistical model. However, the authors feel that the assignment of a value of 1.0 in some models to men who are disease-free (ie, a value equivalent to full health) is perhaps an underestimate. The authors suggest the sense of well-being and gratitude associated with successful therapy may provide added value. Although many urologists may agree, the valuations are ideally judged from the vantage point of the patient. For a value greater than 1.0, the assumption would have to be that men must undergo radical prostatectomy in order to feel healthier than someone without any medical problems.
Likewise, the issue of possibility of overdiagnosis is both complicated and central to any assessment of screening effectiveness. The authors feel that there would be little chance of overdiagnosis, based on the size and histologic characteristics of screen-detected cases. Aside from the differences in biological behavior between the asymptomatic lesions that surface in a screening program and clinically symptomatic lesions that look similar under a microscope, and the fact that we only know the natural history of the latter, even the screen detection of lesions with lethal potential could nevertheless represent some overdiagnosis. The patient may be destined to die of an unrelated disease before the prostate cancer would ever cause symptoms. Changing the estimates of the amount of overdiagnosis can have considerable impact on the theoretical effectiveness of a screening test.
Despite the uncertainties attached to so many assumptions, I believe the authors have done the reader a valuable service. Some previous models have been based on assumptions derived from complete structured literature searches on the risks and benefits of prostate cancer screening and treatment, as well as its costs.[3,4,6,7,12,13] Benoit and Naslund have shown that selection of a particular case series can substantially change both the numerator and denominator of a cost-effectiveness calculation, and therefore the output of a statistical model in either direction. That is the nature of calculation for any ratio with uncertainties in both the numerator and denominator.
The authors therefore demonstrate both the problems and the utility of current cost-effectiveness calculations for prostate cancer screening. The power of models is that they expose our uncertainties and show how modest changes in assumptions can change the calculated outcome. In this case, one of the most important unknowns is the efficacy of treatment. Problems arise when we try to rely on the output as the final word. Hence, the comparison of their cost-effectiveness conclusions for prostate cancer screening cannot be accurately compared to the ratios for other health strategies, such as hypertension control, coronary artery bypass, screening mammography, and colon cancer screening.
In contrast to prostate cancer screening, those interventions have a proven and quantifiable effectiveness established in randomized controlled trials. Their cost-effectiveness calculations are therefore more reliable and useful in allocating health resources. Some well-intended past decisions to launch screening programs in the absence of proof of benefit have led to national harm. For example, the screening program for neuroblastoma in all newborns in Japan led to a dramatic increase in the diagnosis of, and operations for, neuroblastoma, but this increase in screen-detected disease did not lead to any decrease in incidence of late-stage disease or mortality.[14,15] A similar experience has been reported from Quebec.
I fully agree with Benoit and Naslund that current cost estimates and treatment efficacy for prostate cancer screening are unknown, and that randomized trials are necessary to settle the issue and permit accurate assessment of net risks, benefits, and economic costs. One such trial is in progress in the United States and has enrolled about 45,000 of the targeted 74,000 male study participants. Randomized trials are the fastest route to the answer, and are more reliable than modifying assumptions in currently available statistical models. In the meantime, particularly since several of the parameters in cost-effectiveness models hinge on personal values known only to the man facing the option of screening, we should present the uncertainties to each man and allow him to make the decision.
1. Kramer B, Brown M, Prorok P, et al: Prostate cancer screening: What we know and what we need to know. Ann Intern Med 119(9):914-923, 1993.
2. Woolf S: Screening for prostate cancer with prostate-specific antigen: an examination of the evidence. N Engl J Med 333(21):1401-1405, 1995.
3. United States Preventive Services Task Force: Screening for prostate cancer, in DiGuiseppi C, Atkins D, Woolf SH, (eds): Guide to Clinical Preventive Services. 2nd ed. pp 119-134. Alexandria, VA, International Medical Publishing, 1996.
4. Canadian Task Force on the Periodic Health Examination, in: Canadian Guide to Clinical Preventive Health Care, pp 812-823. Ottawa, Canada, Communication Group, 1994.
5. Coley CM, Barry MJ, Fleming C, et al: Should Medicare provide reimbursement for prostate-specific antigen testing for early detection of prostate cancer? II: Early detection strategies. Urology 46(2):125-141, 1995.
6. Coley C, Barry M, Fahs M, et al: Early detection of prostate cancer. II: Estimating the risks, benefits, and costs. Ann Intern Med 126(6):468-479, 1997.
7. American College of Physicians. Clinical guideline. III: Screening for prostate cancer. Ann Intern Med 126(6):480-484, 1997.
8. Potosky A, Miller B, Albertsen P, et al: The role of increasing detection in the rising incidence of prostate cancer. JAMA 273(7):548-552, 1995.
9. Lu-Yao G, McLerran D, Wasson J: An assessment of radical prostatectomy: Time trends, geographic variation, and outcomes. JAMA 269(20):2633-2636, 1993.
10. Shapiro S, Strax P, Venet L: Periodic breast cancer screening in reducing mortality from breast cancer. JAMA 215:1777-1785, 1971.
11. Black W, Welch H: Advances in diagnostic imaging and overestimations of disease prevalence and the benefits of therapy. N Engl J Med 328(17):1237-1243, 1993.
12. Wasson J, Cushman C, Bruskewitz R: A structured literature review of treatment for localized prostate cancer. Arch Fam Med 2:487-493,1993.
13. Barry MJ, Fleming C, Coley CM: Should Medicare provide reimbursement for prostate-specific antigen testing for early detection of prostate cancer? IV: Estimating the risks and benefits of an early detection program. Urology 46(4):445-461, 1995.
14. Yamamoto K, Hayashi Y, Hanada R, et al: Mass screening and age-specific incidence of neuroblastoma in Saitama Prefecture. Jpn J Clin Oncol 13(8):2033-2038, 1995.
15. Bessho F: Effects of mass screening on age-specific incidence of neuroblastoma. Int J Cancer 67:520-522, 1996.
16. Woods W, Tuchman M, Robison L, et al: A population-based study of the usefulness of screening for neuroblastoma. Lancet 348:1682-1687, 1996.
17. Gohagan J, Prorok P, Kramer B: The Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial of the National Cancer Institute. Cancer 75(suppl 7):1869-2873, 1995.