An Overview Cost-Utility Analysis of Prostate Cancer Screening
An Overview Cost-Utility Analysis of Prostate Cancer Screening
Cancer of the prostate is now the most common nondermatologic neoplasm in US men, with 240,000 new cases expected to be diagnosed in 1994 and 40,400 men expected to succumb to the disease . Although survivals with localized disease are excellent, in the range of 85% to 90% at 5 years, the prognosis for metastatic disease is quite poor; median survival for metastatic disease is 2 years and has not changed substantially since the introduction of hormonal therapy 50 years ago .
Because of this discrepancy between treatment outcomes for localized disease (which parallel survivals of men without prostate cancer) and treatment outcomes for metastatic disease (in which the majority of men will die from prostate cancer), emphasis over the past 10 to 15 years has concentrated on the potential for early diagnosis and treatment. Other management alternatives include prevention-currently being evaluated by the Prostate Cancer Prevention Trial, a 7-year study of prophylactic finasteride (Proscar) -and new agents for the treatment of metastatic disease. Of these new agents, the most promising may be suramin. Although initial results show that it has activity against prostate cancer, evidence that survival is enhanced remains to be demonstrated . Thus, although efforts are in progress to seek out new methods to deal with this pressing public health problem, at present the most promising approach seems to be through secondary prevention, ie, early detection and treatment.
Prostate cancer screening was first implemented in the 1970s and early 1980s with the use of digital rectal examination (DRE). Results of screening programs employing DRE have varied widely; cancer detection rates range from 0.1% to 7.6%, with generally accepted rates of 1% to 5% . Positive predictive values for DRE range from 6% to 34%, with the most common rates in the 25% range. Because of the poor prognostic implication of extraprostatic disease, perhaps the greatest concern with the use of DRE for early detection has been the detection rates of organ-confined disease. In one of the largest US experiences with a DRE-based screening program, 66% of patients with cancer detected by DRE were upstaged to pathologic stage C (extraprostatic) disease .
With the development and application of prostate-specific antigen (PSA) for early detection and screening, many patients will be diagnosed at an earlier stage than with DRE screening alone. Several studies have shown that such is indeed the case. Catalona and associates at Washington University screened 1,653 asymptomatic men over the age of 50 using a PSA cutoff value of 4.0 ng/mL. Among these men, 37 cases of prostate cancer were detected, for a case-finding rate of 2.2% . Subsequent studies have confirmed this finding in larger groups of men and also have demonstrated that with serial screening, the likelihood of detecting organ-confined disease is increased .
Although the performance characteristics of PSA are now well-defined through large-scale screening programs, the impact of such early detection and treatment on more significant endpoints, such as disease-specific mortality and overall morbidity, remain undetermined.
Prostate cancer is unique among solid tumors in its histologic prevalence and natural history, and because of these and other considerations, the seemingly natural application of screening tests to prevent a lethal disease may have a number of potential hazards that could mitigate the salutary effects.
Prevalence-As early as 1935, it was recognized that the histologic prevalence of prostate cancer far exceeded the number of cases in which patients presented with symptoms. Rich  and Moore  in that year found that microscopic disease could be detected with crude sectioning techniques in 8% to 14% of men aged 50 to 60, and in 11% to 23% of men aged 60 to 70. Subsequent studies in the 1950s and 1960s with more histologic sections found that the rate in men over age 50 was 38%; in men 70 to 80 years of age, 41%; and in men 80 to 90 years of age, as high as 57% [10,11].
Interest in this phenomenon waned for 20 years until Sakr and associates in Detroit analyzed a group of younger men who died due to trauma in that city . The authors found an astounding rate of small prostate tumors: 27% in men aged 30 to 39, and 34% in men aged 40 to 49. The risk of a US male dying of prostate cancer is approximately 3%; thus, if as many as 50% to 70% of men can be expected to develop prostate cancer during their lifetime, it may be the case that some tumors detected through screening are not destined to lead to morbidity or mortality, and therefore treatment is unnecessary.
Natural History-A number of series have reported the results of a policy of watchful waiting (surveillance) for localized prostate cancer. It must be recognized that most of these series have used some form of patient selection to determine which patients are eligible for observation, and that the average age in these series is approximately 72, an age at which comorbidity may play a much higher role in mortality rates than prostate cancer. Nevertheless, the results with surveillance have been impressive.
Adolfsson and Carstensen analyzed 61 patients, all under age 70, for a mean follow-up period of 96 months . Although 82% of men had developed local progression at 10 years of follow-up, only 8% died from prostate cancer during that period. The largest US experience was reported from Memorial Sloan-Kettering Cancer Center in 75 patients with localized prostate cancer . Although the median time to local progression was about 7 years (84 months), of patients with stages B1 (T2a), B2 (T2b), and B3 (T2c) cancer, median survivals were 215, 138, and 197 months, respectively.
Using individual patient data from 828 patients in six published series, Chodak and associates performed a metaanalysis of results of surveillance . The authors found that disease-specific survival 10 years after diagnosis was 87% for both grade 1 and 2 tumors, and 34% for men with grade 3 disease. The most recent evidence on this subject has been provided by Aus . Of a group of 536 patients, 349 had no evidence of metastatic disease at diagnosis and were followed with surveillance alone. In this observation group, of those who survived beyond 10 years, 50% died of prostate cancer.
These data thus suggest that although many men with prostate cancer who are followed without treatment will die of other diseases, with further passage of time (more than 10 years), the risk of prostate cancer mortality may increase substantially. Because of this uncertainty and the poor quality of existing series of untreated patients, the impact of early detection on prostate cancer-specific survival or morbidity from this disease is unknown.
It is intuitively obvious that in the absence of properly designed, prospective, randomized, controlled clinical trials of prostate cancer screening, the benefit from current screening methods is unknown. Indeed, for this reason, the National Cancer Institute is currently enrolling subjects in a prospective trial in which half of participants undergo screening examinations and the other half have usual medical care within the community (the Prostate, Lung, Colorectal and Ovarian Cancer Screening Trial, or PLCO) . Because of the unknown effect of screening, the utility of treatment also remains in question. Nevertheless, a number of authors have used the available literature to perform decision analyses to determine the net utility of prostate cancer treatment.
The first of these analyses was conducted by Wasson and Fleming [18,19]. The authors first performed a structured literature review to determine potential outcomes of treatment or surveillance for prostate cancer, and used these data to conduct their decision analysis. Unfortunately, a number of methodological problems have been noted in this analysis:
1. The natural history of untreated prostate cancer patients included four series of patients with T1a (stage A1) disease (which has been recognized as a completely different and significantly less virulent entity than those tumors generally detected by screening). This resulted in rates of metastatic progression as much as 10-fold less than those in Chodak's metaanalysis.
2. Utility factors were determined by the authors and other physicians rather than by patients themselves.
3. Treatment-related mortality was determined from a Medicare-age population rather than the younger male population who would be most likely to benefit from screening and in whom mortality is as much as 20-fold less .
Nevertheless, these two series concluded that the estimated benefit of treatment for well, moderately, and poorly differentiated prostate tumors is 0.16, 0.75, and 1.3 years, respectively. Using the authors' derived utility rates, quality-adjusted life years (QALYs) gained by treatment are -0.34, 0.33, and 1.00 years, respectively.
Because of the flaws detected in the methodology of this study, Beck and associates conducted a re-analysis, using what they felt were more appropriate probabilities . Their re-analysis used estimates of survival of untreated prostate cancer from both Chodak's metaanalysis and from data from a series of patients treated with brachytherapy (the authors feel that brachytherapy may have had poor efficacy, but this patient series may portray the natural history of the disease). Table 1 summarizes these comparisons. As can be seen, using either Chodak's data or the brachytherapy data, the gains in both total life expectancy and in quality-adjusted life expectancy may be considerable when prostate cancer is treated.