A major effort is under way in the United States and Canada to screen men over the age of 50 for prostate cancer. Each year, hundreds of thousands of men undergo digital rectal examination and measurement of their serum prostate-specific antigen. Partly as a consequence of increased screening, radical prostatectomy rates increased sixfold between 1984 and 1990. 1

The ability of prostate-specific antigen (PSA) and transrectal ultrasound (TRUS) to detect prostate cancer at an early stage has raised hopes that screening might reduce mortality from prostate cancer, now the second leading cause of cancer death among men.2

In 1994, however, the early detection of prostate cancer remains of uncertain benefit. No randomized controlled trial of screening has been carried out, and evidence from uncontrolled studies is unconvincing. Screening is very costly and the benefits of treating localized prostate cancer remain unproven. To illuminate the screening question, we performed a decision-analytic cost-utility analysis of prostate cancer screening. We asked two questions.

• Given the available evidence, what is the net clinical benefit of screening for prostate cancer?
• What is the economic burden of screening for prostate cancer?

Methods

Using combinations of digital rectal examination (DRE), TRUS and PSA, we considered four potential programs. Each program included a single episode of screening. We first considered screening with DRE alone, followed by ultrasound-guided biopsy for a palpable nodule or induration. We also considered a policy of using PSA alone, followed by confirmatory DRE and TRUS if the initial PSA was above a specified threshold.

As a third screening option, we modelled the combination of DRE and PSA as the initial screen with a confirmatory TRUS prior to biopsy if the PSA alone was elevated. Finally, we considered the combination of PSA, DRE and TRUS with ultrasound-guided biopsy for any single abnormal test as an upper bound of sensitivity, cost and invasiveness. We compared these four strategies to a conservative policy of treatment when symptoms developed, without cancer screening per se.

To simplify the decision problem, we made the following assumptions about how cancer would be treated: clinical stage A (impalpable disease) and clinical stage B disease (palpable tumour on DRE) identified by screening would be treated with radical prostatectomy. Clinical stage C disease would be treated with external beam radiotherapy, and clinical stage D disease would be treated with orchiectomy when diagnosed.

Our model captures the clinical effects and costs of short- and long-term treatment complications. Short-term complications include rectal injury, wound infection, pulmonary embolus, deep vein thrombosis, anastomotic leak and stricture, systemic sepsis, acute cystitis and acute proctitis. Long-term complications include impotence (complete and partial), incontinence (complete and partial), obturator nerve injury, chronic cystitis, chronic proctitis and death.

We modelled long-term prognosis using Markov cohort models, in which clinical events are characterized as transitions between health "states."4 Separate models were constructed for each combination of initial clinical stage and treatment modality.

Data

Table 1 shows selected costs, short-term morbidities and utilities for long-term events and health states. We estimated the probability of short- and long-term complications of surgery and radiotherapy from published case series. Long-term complications are stratified by severity (e.g. impotence and incontinence) and adjusted for prevalence of pre-existing conditions in the population (e.g. impotence).

Stage-specific disease progression and mortality rates were estimated from published case series. We adopted the definitions of disease progression and prostate cancer death used in individual studies, even though definitions were variable.

Utilities for chronic health states were elicited from a group of 10 physicians including urologists, radiation oncologists and internists. We constructed scenarios describing impotence, incontinence and metastatic disease. We used "Gambler," 5 an automated graphical utility assessment tool, to elicit the utility of these health states using the time trade-off method. We adjusted for the short-term morbidity of treatment and its complications by subtracting quality-adjusted life expectancy penalties from calculated quality-adjusted life expectancy results.

Direct variable costs for inpatient services were obtained from the Clinical Cost Manager (Transition Systems, Inc., Boston, Massachusetts) at the New England Medical Center.

Results

When quality of life is considered in addition to survival, our model predicts that screening produces a net loss of quality-adjusted life expectancy (3-13 days). Thus, though cancer mortality is reduced and morbidity from metastatic disease is diminished by screening, these effects are more than offset by the short- and long-term morbidity of therapy. Losses are smallest at age 50 and become progressively greater at ages 60 and 70.

Economic Results
All programs result in a net increase in costs; no program is cost-saving when compared to no screening.For all screening programs, initial screening and treatment costs exceed savings from reduction of local and metastatic complications of disease.

Because each screening program resulted in loss of quality-adjusted life-years (QALYs), all programs were dominated (greater cost, less health benefit) by no screening. Therefore, when quality of life is included in the economic analysis, no screening program could be considered to be economically attractive.7,8

Sensitivity analyses were performed for all central variables. At the baseline estimate of surgical efficacy, the qualitative result of our analysis did not change over plausible ranges of individual biologic, quality-of-life and economic variables. Compared to no screening, all programs showed a small gain in life expectancy, but a net loss of quality-adjusted life expectancy and increased costs across wide ranges of values for disease prevalence, probability and utility of post-operative impotence and incontinence, prevalence of impotence and costs of screening for and treating prostate cancer.

The results of the analysis, however, were affected by varying the efficacy of surgical treatment. The results did not change as we increased or decreased by 50% the post-surgical probability of disease progression to either local recurrence or metastatic disease. However, using the most optimistic estimate of surgical efficacy (best cancer mortality rate in surgically treated group, worst rate in initially untreated group), all screening programs increased quality-adjusted life expectancy by a small amount (less than three days) for men aged 40-70. At this optimistic estimate of treatment efficacy, screening with PSA alone was again the most economically attractive program with an incremental cost-utility ratio of $42,000 per QALY for 50-year-old men. No other screening programs were attractive (incremental cost-utility ratio less than $100,000 per QALY) for 50-year-olds, and no program was attractive for men aged 60 or over.

Discussion

While controversy surrounds the issue of net health benefit, there is no doubt about the effects of screening on costs. At an individual level, screening results in increased cost per person screened. The aggregate effect is astronomical: published estimates of the annual cost of screening for prostate cancer and treating those with localized cancer in the US range from $11.9 to $27.9 billion.9-11

Because screening is already common in both Canada and the US, these figures overstate the marginal costs of universal screening in 1994. They do, however, provide an estimate of the magnitude of resources that would be consumed by large-scale screening, and that might be saved by abandoning the practice.

The expected benefit of screening for each individual will vary from that predicted for the population as a whole. Combinations of co-morbidity, risk attitude, valuation of sexual function, probability of iatrogenic impotence and time dependence of utilities may make screening more or less desirable for selected individuals. At our baseline estimate of treatment efficacy, however, change in any one of these patient-related variables does not make screening attractive. This is true even for black men, with their higher prevalence and worse natural history of disease.

For all men, the central issue in determining the desirability of screening remains the net benefit of treatment. Screening for prostate cancer will not yield an overall health benefit unless treatment of early stage prostate cancer reduces cancer mortality at an acceptable cost in treatment-related morbidity. Neither the improvement of screening tests nor the restriction of screening to high prevalence populations will improve the desirability of screening if treatment itself is not of net benefit. The value of recurrent screening that attempts to pick out very early stage tumours also depends on the net benefit of treatment. This is reflected in our analysis as sensitivity of the analytic result to assumptions regarding the efficacy of treatment, and insensitivity to other variables.

Yet the quality of evidence underlying our (and all other) estimates of treatment benefit is low. Few studies have evaluated the effects of radical prostatectomy or external beam radiotherapy on short- or long-term quality of life. With a single exception,12 all studies of the effect of radical prostatectomy on survival are uncontrolled and subject to bias. The interpretation of existing studies is complicated by heterogeneity in the definitions of clinical outcomes. Disease progression and cause-specific death are variably reported; clinical stage and tumour grade are interpreted according to a plethora of classification systems.

Finally, even reliable estimates of treatment benefit derived from high-quality randomized controlled trials would not allow definitive prediction of screening benefit using decision or other models. Potential biases associated with all cancer screening programs (length bias, patient selection bias) introduce uncertainty into the linkage between disease detection and patient benefit.13,14 Because screening biases lead to overestimating the benefits of screening, true benefit is probably less than or equal to predicted benefit. Randomized trials of screening remain the only method for reliably ascertaining the net effect of screening. Until the results of such trials are available, projections based on the existing information are the best we can do. Our analysis, based on the evidence available in 1994, suggests that screening for prostate cancer cannot be justified as a rational health policy.

(a) Dr Krahn's presentation was an abridged version of an article previously published in JAMA (Sept 14, 1994-Vol 272, No 10, pp 773-80; copyright 1994, American Medical Association). It is published here with permission from JAMA.

References

2. Boring CC, Squires TS, Tong T. Cancer statistics 1992. CA 1992;42:19-38.

3. Krahn MD, Mahoney JE, Eckman MH, Trachtenberg J, Pauker SG, Detsky AS. Screening for prostate cancer: a decision analytic view. JAMA 1994;272(10):773-80.

4. Beck J, Pauker S. The Markov process in medical prognosis. Med Decis Making 1983;3:419-58.

5. Gonzalez EF, Eckman MH, Pauker SG. "Gambler": a computer workstation for patient utility assessment [abstract]. The 4th Annual Meeting of the Society for Medical Decision-making; 1992.

6. Weinstein MC. Clinical decision analysis. Philadelphia: WB Saunders, 1980.

7. Laupacis A, Feeny D, Detsky A, Tugwell P. How attractive does a new technology have to be to warrant adoption and utilization? Tentative guidelines for using clinical and economic evaluations. Can Med Assoc J 1993;148:921-4.

8. Detsky AS, Naglie IG. A clinician's guide to cost-effectiveness analysis. Ann Intern Med 1990;113:147-54.

9. Optenberg SA, Thompson IM. Economics of screening for carcinoma of the prostate. Urol Clin North Am 1990;17:719-37.

10. Dorr VJ, Williamson SK, Stephens RL. An evaluation of prostate-specific antigen as a screening test for prostate cancer. Arch Intern Med 1993;153:2529-37.

11. Kramer BS, Brown M, Prorok P, Potosky A, Gohagan J. Prostate cancer screening: what we know and what we need to know. Ann Intern Med 1993;119:914-9.

12. Byar DP, Corle DK. VACURG randomized trial of radical prostatectomy for stages I and II prostatic cancer. Urol 1981;17(4 Suppl):7-11.

13. Eddy DM. Finding cancer in asymptomatic people: estimating the benefits, costs, and risks. Cancer 1983;51:2440-5.

14. Black WC, Welch HG. Advances in diagnostic imaging and overestimations of disease prevalence and the benefits of therapy. N Engl J Med 1993;328:1237-43.

Author Reference

[Previous][Table of Contents] [Next]