Meta-Analysis of Prostate-Specific Antigen and Digital Rectal Examination as Screening Tests for Prostate Carcinoma

Methods

Results

The data are shown in Tables 1, 2, and 3. In some studies the data on sensitivity, specificity, and positive predictive value were published; however, in other studies these were calculated from the original published data. Pooled parameter estimates for measures at the interval or ratio level of measurement could not be calculated without the raw data.

As shown in Table 1, the overall detection rate for prostate carcinoma by biopsy was 1.8%, with a wide range of 0.6% to 3.9%. There was no specific correlation with the country in which the study was conducted. Of the total prostate carcinoma detected using PSA or DRE as screening tests, 83.4% were localized, with a range of 64.4% to 90.0%. The overall ratio of total biopsies performed to actual positive prostate carcinoma detected was 4.0, with a range of 3.7 to 8.9.

The positive predictive value, sensitivity, and specificity values for PSA are shown in Table 2. A total of 10.1% of the population was positive for PSA > 4 ng/mL, with a range of 4.3 to 17.2 ng/mL. The positive predictive value was 25.1%, with a range of 17.0% to 57.0%. The sensitivity of PSA in detecting prostate carcinoma was 72.1%, with a range of 66.7% to 100.0%. The specificity of PSA in the detection of prostate carcinoma was 93.2%, with a range of 63.1% to 100.0%.

The data on positive predictive value, sensitivity, and specificity of DRE are shown in Table 3. Overall, 5.0% of the population had abnormal findings on DRE, ranging between 4.2% to 19.3%. The positive predictive value for DRE in detecting prostate carcinoma was only 17.8%, with a range of 5.0% to 33.1%. The sensitivity of DRE in detecting prostate carcinoma was 53.2%, with a range of 49.0% to 69.2%. The overall specificity of DRE was 83.6%, with a wide range of 18% to 99.5%.

Discussion

There were two major outcomes of this meta-analysis. One was the potential for detecting early-stage prostate cancer with these screening tests, because 83.4% of total cancers detected was localized. The second important outcome was with the comparison of PSA and DRE as screening tools to detect prostate cancer. The overall sensitivity, specificity, and positive predictive value for PSA were 72.1%, 93.2% and 25.1%, respectively; and those for DRE were 53.2%, 83.6% and 17.8%, respectively. When a patient has abnormal PSA levels or DRE findings, the chance of having cancer is 1 in 4 or 5; conversely, when PSA levels or findings on DRE are normal, the chance of missing the cancer is about 10%.

There are two major limitations of the studies available for current meta-analysis. The most serious flaw is that these studies lack a control group that had no screening or treatment.25 This limitation is not unique to prostate carcinoma studies. For many clinical studies it would be unethical to observe patients without providing the current standards of screening and treatment. The other limitation was a lack of biopsy results (the extant reference standard) for all the participating patients, specially those with normal PSA levels and DRE findings, because of the invasiveness of this test. As a result, all specificity, sensitivity, and positive predictive values reported in the published studies are potentially biased by the effect of aggressiveness in performing the biopsies and the variations in the determination of PSA levels and DRE findings. This point is well illustrated by the reported studies using a lower cutoff value of abnormal PSA levels. Two studies26,27 show that of a total population with prostate carcinoma, 14% have a PSA level of less than 3 ng/mL, 23% to 24% have a PSA level of 3 to 4 ng/mL, and 62% have a PSA level of more than 4 ng/mL. By using a currently accepted cutoff level of more than 4 ng/mL, we are able to detect 62% of prostate cancers, because 38% cancer patients have PSA levels of less than 4 ng/mL.

Criteria for a clinically useful screening test are as follows28: (1) the disease must constitute a serious public health problem; (2) the disease must be able to be diagnosed during an asymptomatic, localized phase; (3) the screening test must have appropriate sensitivity, specificity, and predictive value; (4) the potential for cure must be greater among patients with prostate cancer detected by screening; and (5) improved outcomes related to screening must be shown. After these criteria are satisfied, the cost-effectiveness of the screening program must also be justified.

The importance of prostate cancer as a public health problem and that it can be diagnosed during an asymptomatic, localized stage easily satisfy using PSA and DRE as screening tools for the first two of the above-listed criteria; however, there are no clear answers for the remaining criteria.

An autopsy study of Detroit men found unsuspected prostate cancers in 30% of men in their 20s through 40s, and in more than one half of men older than 50 years. The prevalence of these unsuspected prostate cancers is frequently estimated to be about 33%.2 These high rates of unsuspected prostate cancers are in sharp contrast to the 3.64% estimated lifetime risk of dying from prostate cancer, as well as lower pooled data of 1.8% detection rate of prostate carcinoma in the current meta-analysis. Once regional lymph node involvement occurs, the probability of death from prostate cancer is 70%, and 50% of those will die in 2 years. Prostate cancer is a real risk for the aging man, because almost 10% of men older than 50 years are likely to develop clinically serious disease; therefore, it must be detected at an early stage.29

It is necessary to increase the sensitivity to detect more cancers at an early stage. In the current study, the sensitivities of PSA and DRE screening tests were 72.1% and 53.2%, respectively. Reducing the PSA cutoff point from 4 ng/mL to 3 ng/mL can increase the sensitivity, but doing so will reduce further the positive predictive value.30 It is also well known that PSA values for prostate cancer and benign prostate hyperplasia overlap considerably. Between 21% and 47% of men with histologically proven benign prostate hyperplasia have PSA levels of more than 4 ng/mL, and up to 43% of men with prostate cancer will have a PSA level of less than 4 ng/mL. This overlap makes it harder to differentiate benign prostate hyperplasia from prostate carcinoma in the absence of a biopsy. PSA values also increase with age.17

With prostate cancer, the risk of overdiagnosis is likely to be much more relevant than with other cancer screening, because in men aged 55 to 60 years, the risk of death from other causes is considerably higher than from prostate cancer. It is estimated that for every patient who dies of prostate cancer, at least 380 others have prostate cancer that cannot be detected clinically.31,32 The treatment of prostate cancer consists of radical surgery or radiotherapy, and both can cause complications, including a high frequency of sexual impotence, with a relevant frequency of major rectal and urinary dysfunction, as well as 1% to 2% mortality.

In the current study the positive predictive values for PSA and DRE are about 25% and 18%, respectively, which means 1 of 4 or 5 biopsies is unnecessary. Unnecessary biopsies can lead to multiple invasive procedures, anxiety to the patient, related complications, and high cost of health delivery. To reduce further unnecessary procedures, Oesterling33 has proposed transrectal ultrasonography (TRUS) in the patient with elevated PSA levels but benign DRE findings, then biopsy of visible abnormal lesions only. If findings on the DRE are abnormal, the patient should undergo TRUS and then a biopsy, regardless of the value of PSA. From the rate of growth, a small, organ-confined prostate tumor has been estimated to double in about 4 years. Thus, it will take about 15 years for a 1-mL tumor to become life threatening. It would be more straightforward to say that, until there is evidence about effectiveness of screening in decreasing mortality, based on these growth rates, a man would have at least 15 years of life expectancy to benefit from PSA screening.

The cost burden for prostate cancer screening was calculated by the authors from the current data based on fees charged at a urologist’s office in New Jersey. In the United States, there are 30.8 million men older than 50 years who qualify for PSA screening.34 The cost of screening all these men for PSA would be $3.1 billion. Of this population, 10.1% (approximately 3.1 million) will have PSA levels of more than 4 ng/mL. Assuming all patients with abnormal PSA levels are referred to a urologist for further evaluation, the first visit will cost $275 (for office visit, urine analysis, and culture). The second visit will involve sonography of the prostate, bladder, pelvis, and renal organs, as well as guided needle biopsy, which can cost $ 2,170 per patient. The total cost of these two visits will exceed $7.6 billion for 3.1 million men. Of 3.1 million biopsies performed, 75%, ie, 2.3 million, will be negative for prostate carcinoma.

We need more data showing improved mortality or morbidity before investing in these expensive procedures. In fact, such an answer might come after the PLCO (prostate, colorectal, and ovarian) screening study. This NIH/European randomized study of 148,000 patients to determine whether screening reduces mortality is expected to provide definitive results in 2005 to 2008.35

Conclusions

Our study showed that screening using PSA and DRE can detect 83.4% of prostate cancers in an early, localized stage. In contrast, two thirds of patients with prostate cancer have metastatic disease by the time they become clinically symptomatic. We must improve the sensitivity and specificity of the screening methods by using such tools as age-specific cutoff values, determining free and bound forms of PSA, correcting PSA for benign prostate hyperplasia, and standardizing DRE. More precise screening will permit detection of prostate cancer in a younger population and thus achieve compelling improvements in mortality and morbidity.