To reduce the number of deaths from prostate cancer, there are four approaches: primary prevention, early diagnosis, effective treatment of curable disease, and improved management of advanced disease. In the foreseeable future, it is unlikely that we will be able to prevent the disease by manipulating environmental or genetic factors or to cure patients with advanced disease while keeping morbidity at an acceptable level. Therefore, in order to reduce the number of deaths from prostate cancer, we are faced with the decisions of how far to go in diagnosing the disease and who should be offered treatment that may lead to cure.

In this issue of Annals, Kramer and colleagues [1] review the status of screening for prostate cancer. They point out that screening based on prostate-specific antigen (PSA) measurements may not be accurate and that the widespread adoption of such screening could lead to increased treatment-related mortality, reduced quality of life for patients who are treated, and a severe economic burden on the medical system. Thus, they believe that widespread screening for prostate cancer should await the results of their randomized study of screening. Looking at the same data from a different perspective, I draw different conclusions.

Are most men undergoing screening? Screening refers to a test done in an asymptomatic group of patients, and diagnosis refers to tests done in those who are symptomatic. Approximately 75% of men older than 50 years have symptoms referable to the lower urinary tract. For these patients, the differential diagnosis includes benign prostatic hyperplasia, prostate cancer, and primary detrusor dysfunction [2]. In the past, many men underwent transurethral resection of the prostate, which provided a histologic diagnosis. However, in recent years, more and more men are seeking medical therapy or less invasive procedures such as transurethral incision of the prostate, balloon dilation, and laser prostatectomy, none of which provides tissue for histologic examination. Thus, a major issue facing many physicians and urologists is not screening but rather the differential diagnosis (benign or malignant disease) in a symptomatic man. Thus, the PSA assay and digital rectal examination are used commonly for this diagnostic (not screening) purpose.

As the authors point out, these tests have reasonable sensitivity and specificity. Originally, many skeptics felt that use of PSA measurements would lead to the diagnosis of the incidental microscopic cancers that are present in 30% of men older than 50 [3]. However, recent evidence suggests otherwise. Studies from two centers indicate that 90% or more of men with nonpalpable cancers who have PSA levels greater than 4 ng/mL will have cancers found at the time of radical prostatectomy that are much larger in size and extent than the incidental microscopic cancers that are present at autopsy ([4, 5]; unpublished data, Epstein J and colleagues). Furthermore, by using sequential screening maneuvers, 70% of patients who are diagnosed will also have cancers that are confined to the specimen and thus are detected at a curable stage ([4, 5]; unpublished data, Epstein J and colleagues). Thus, I do not agree with the authors' estimate of the economic impact, because it fails to recognize that most men in this age group are symptomatic and when they consult a physician they are undergoing diagnostic testing, not screening.

How will PSA measurements be used in the future? When any new test is introduced, it is usually applied to a cross-section of the population. This has been true for the PSA assay and has resulted in arbitrary cutoff points defining normal ranges. However, sequential PSA measurements provide greater specificity in the detection of prostate cancer, and at lower peak levels, than does the use of the arbitrary cutpoints. In the Baltimore Longitudinal Study of Aging, we showed that in patients with PSA levels less than 10 ng/mL, consistent increases of more than 0.75 ng/mL per year in the serum PSA level identified men with prostate cancer an average of 5 years before usual diagnosis with a 72% sensitivity and 90% specificity [6]. Recently, investigators from Washington University confirmed these findings. They used a neural network, a mathematical pattern-recognition paradigm that "learns" complex interactions among input variables and has the ability to recognize subtle patterns of association in multiple data points. This allowed them to analyze the findings from a longitudinal study of 999 men who on initial screening with digital rectal examination and PSA assay were found to be free of prostate cancer [7]. In these men followed at 6-month intervals, a consistent increase in the serum PSA level of 0.72 ng/mL per year over three determinations was the single most predictive value to identify prostate cancer. Investigators from the Mayo Clinic showed that the upper 95% confidence limit for yearly changes in serum PSA levels for healthy white men between 40 and 80 years old is 0.8 ng/mL per year [8]. These data suggest that monitoring the PSA level for a consistent defined increase over time may be the most important practical use of the assay. For example, if a patient's PSA level is 2 ng/mL initially and 6 ng/mL 3 years later, he may not only have cancer but one that is growing.

Once a diagnosis of localized prostate cancer is made, is it really necessary to treat the patient? The authors, using Johansson's findings as support, suggest that the mortality rate for untreated prostatic cancer is quite low [9]. However, Johansson's study has been widely criticized because it included older patients (average age, 72 years) with small, well-differentiated tumors [10]. Most of these men would have been followed expectantly in the United States as well. However, even in this well-selected group of patients who should have done well, at 10 years, 13% had died of cancer and another 50% had developed progressive disease. In making projections, the authors failed to consider the true economic impact of using hormonal therapy in men with progressive cancer. Unless inexpensive doses of diethylstilbestrol (which has cardiovascular side effects) are used, or a bilateral orchiectomy is mandated, the yearly cost of treatment with a luteinizing hormone releasing hormone agonist with or without the addition of an antiandrogen agent ranges from $4000 to $8000 a year. In 2 years, the cost of total androgen ablation would be greater than that of a radical prostatectomy or radiation therapy.

However, rather than focusing on patients with the smallest cancers or quoting mathematical models that suggest little survival benefit for patients undergoing definitive therapy, the authors should have included actual data from the literature. Based on the results of a recent analysis of palpable clinical localized prostate cancer, the development of metastases and death from prostatic cancer was 50% lower in men who underwent radical prostatectomy than in those followed expectantly [11]. To demonstrate a survival benefit from definitive therapy, men must live longer than 10 years. Thus, studies such as Johansson's are truly uninformative, and the inclusion of men older than 70 years in the National Cancer Institute (NCI) screening trial is inappropriate.

It is laudable that the NCI proposes to carry out a randomized trial of screening. This will involve 4 years to recruit patients and 12 years of follow-up. The cost is estimated to be $89 million. Will the NCI study provide the definitive answer? I am not sanguine that this study will yield the answers we need. Men will be screened once a year for only 4 years, which is similar to the screening intervals used in the study that showed a benefit for mammography in breast cancer. However, the doubling time of breast cancer is much more rapid than for prostate cancer, and thus a similar protocol is unlikely to produce similar results [12, 13]. Furthermore, I believe that 4 years is too short a time to test whether a yearly rate of change in the PSA level will prove as valuable as I believe it will. I am also concerned about the lack of mandated treatment. I realize that the treatment of prostatic cancer is controversial. However, in the NCI study, once a diagnosis is made the choice of treatment is left up to the patient and his physician. Although they will be encouraged to use the NCI Physician Data Query database for treatment recommendations, they also may entertain a wide range of unproven treatment options such as cryotherapy, interstitial radiotherapy, or hormonal therapy. For these reasons, I believe that the NCI study should be delayed until the results of ongoing trials provide information about the optimal use of PSA measurements for the identification of significant tumors that need to be treated [6-8]. At that point we would no longer be sailing in uncharted waters. During this delay, the study should undergo major modifications to exclude older patients from the study and to randomize study patients to an effective treatment arm in the event a diagnosis of prostate cancer is made.