A prospective validation trial has shown that measurement of urinary prostate cancer antigen 3 (PCA3) noncoding RNA adds value to screening for early detection of prostate cancer, but the data also suggest that the predictive value of PCA3 could be further improved by integrating it into a multifactorial risk calculator.

The measurement of serum levels of PSA has enabled routine screening for prostate cancer, and treatment before the manifestation of advanced disease. However, PSA is a blunt instrument that can fail to differentiate between cancer and nonmalignant prostate pathologies, and between low-grade and high-grade cancer, leading to extensive overtreatment. To discriminate between these possibilities, a finding of an elevated serum PSA level is followed by biopsy of the prostate and histopathological characterization. Biopsy carries risks of sepsis and morbidity, as well as the possibility of false-negative results, which means that initial negative results are not definitive, and repeat biopsies are frequently required. To improve the diagnosis of prostate cancer, urologists must have access to tools that can accurately predict the risk of clinically relevant, high-grade disease, to enable optimal targeting of biopsy and treatment procedures.

PCA3 testing is FDA-approved to facilitate biopsy decision-making, specifically in men with previous negative biopsy results. High sensitivity and specificity have been reported for this test, but validation has been lacking. Wei and colleagues at 11 US institutions have addressed this issue by comparing PCA3 levels with biopsy results in 859 men, 562 undergoing initial biopsy and 297 having a repeat procedure. In all cases, biopsy was indicated by screening characteristics, principally elevated or increasing PSA, or abnormal digital rectal examination (DRE). The relationship between PCA3 scores and biopsy results was examined in several different ways.

In the initial biopsy group, the positive predictive value was 80%, as 140 men had PCA3 >60 (the test uses an arbitrary, unitless scale), and 112 of these had prostate cancer, of any grade. Sensitivity was 0.42, as 112 of 264 men with cancer had PCA3 >60, and specificity was 0.91, as only 28 of 298 men with negative biopsies had PCA3 >60. In the repeat biopsy group, the negative predictive value was 88% (of 135 men with PCA3 <20, 16 had cancer), sensitivity was 0.76 (16 of 67 men with cancer had PCA3 <20) and specificity was 0.52 (119 of 230 men without cancer had PCA3 <20). Predictive values, sensitivity and specificity are all standard measures that enable comparisons of the performance of different markers. However, in an accompanying editorial, Andrew Vickers points out that these values do not relate well to clinical practice, where tests are conducted on patients with unknown disease status, and he adds that statistical measures that do not distinguish between low-grade and high-grade cancers have little clinical benefit.

Credit: GuidoVrola/iStock/Thinkstock

Wei et al. also investigated the potential of PCA3 to reduce unnecessary biopsies. In this cohort, eliminating biopsies for patients with PCA3 <20 would avoid 135 of 297 repeat biopsies, missing four high-grade cancers with Gleason scores >6. Combining PCA3 <20 with PSA <4 ng/ml would only prevent 23 repeat biopsies, but would identify all cases of high-grade disease. This PCA3 threshold would avoid 231 of 562 initial biopsies, leaving 31 high-grade cancers undiagnosed. PCA3 <20 combined with PSA <4 ng/ml would avoid 84 initial biopsies, missing four cases of high-grade disease. Vickers suggests that, rather than using PCA3 as a binary test with an arbitrary threshold, it should be a continuous variable, integrated with relevant clinical factors and taking account of individual treatment preferences.

Wei et al. addressed the use of multifactorial models by adding PCA3 to the variables of the Prostate Cancer Prevention Trial risk calculator (age, race, PSA, DRE, prior biopsy and family history), resulting in significant improvements in receiver operating characteristic area under the curve values for prediction of high-grade prostate cancer in the setting of initial (from 0.74 to 0.78, P <0.001) and repeat (from 0.74 to 0.79, P = 0.003) biopsy.

The results of this study validate the use of PCA3 in the approved setting of repeat biopsy to reduce overdiagnosis of low-grade cancer while optimizing detection of high-grade disease. A similar role in the initial biopsy setting is not recommended, owing to the underdiagnosis of high-grade disease. Combining measures of PSA and PCA3 could improve on the utility of both as individual markers. Ultimately, the greatest value for patients and urologists could be derived from integration of these markers into multifactorial risk assessment models. Just as the use of PSA has evolved to incorporate PSA density and kinetics, the best use for PCA3 might not yet have been determined.