Evaluation of prostate cancer antigen 3 for detecting prostate cancer: a systematic review and meta-analysis

Previous studies indicate that prostate cancer antigen 3 (PCA3) is highly expressed in prostatic tumors. However, its clinical value has not been characterized. The aim of this study was to investigate the clinical value of the urine PCA3 test in the diagnosis of prostate cancer by pooling the published data. Clinical trials utilizing the urine PCA3 test for diagnosing prostate cancer were retrieved from PubMed and Embase. A total of 46 clinical trials including 12,295 subjects were included in this meta-analysis. The pooled sensitivity, specificity, positive likelihood ratio (+LR), negative likelihood ratio (−LR), diagnostic odds ratio (DOR) and area under the curve (AUC) were 0.65 (95% confidence interval [CI]: 0.63–0.66), 0.73 (95% CI: 0.72–0.74), 2.23 (95% CI: 1.91–2.62), 0.48 (95% CI: 0.44–0.52), 5.31 (95% CI: 4.19–6.73) and 0.75 (95% CI: 0.74–0.77), respectively. In conclusion, the urine PCA3 test has acceptable sensitivity and specificity for the diagnosis of prostate cancer and can be used as a non-invasive method for that purpose.


Methods
This meta-analysis was performed in accordance with PRISMA 4 guidelines, which prefer reporting items from systematic reviews and meta-analyses.

Search strategy and study selection. A comprehensive, computerized literature search was performed
in PubMed and Embase for work published through December 2014 using a combination of the following key words: ["prostatic neoplasm" or "prostate cancer"] AND ["PCA3" or "prostate cancer antigen3" or "dd3" or "upm3" or "aptima pca3"] AND ["diagnosis" or "sensitivity and specificity"]. Then, the reference sections of the identified publications were searched to identify additional potentially relevant articles. Studies included in our meta-analysis had to meet the following criteria: (1) case-control or cohort design; (2) diagnostic test using PCA3 itself or in combination with other biomarkers; and (3) prostate biopsy as the gold standard.
Data extraction and quality assessment. Data were extracted independently by 2 authors (Y.C. and C.L.) and then crosschecked. For each study, the following information was collected: last name of the first author, publication year, study design and ethnicity, age, PSA, sample size and the values of true positive (TP), false positive (FP), false negative (FN), true negative (TN), and area under the curve (AUC) (with 95% CI) if available. When more than one article was published using the same population, we selected the most recent or most informative report. Disagreements between the two authors were resolved by consensus. The quality of the selected studies was assessed using quality assessment of diagnostic accuracy studies (QUADAS) 5 . The QUADAS tool consists of a set of 14 questions, each of which is scored as yes, no, or unclear.
Statistical analysis. For each study, 2 × 2 tables for each test with TP, FP, FN, and TN results were extracted from the original scientific articles. Pooled estimates of sensitivity (Se) and specificity (Sp) and their 95% confidence intervals were calculated as the main outcome measures. Forest plots were used, and methodological heterogeneity was assessed during selection.
The threshold effect is a characteristic source of heterogeneity in the meta-analysis of diagnostic tests and arises when the included studies use different cut-off points to define a positive result of a diagnostic test. The analysis of the diagnostic threshold was assessed through the receiver operating characteristic (ROC) plane and Spearman's correlation coefficient The ROC plane is the graphic representation of the pairs of Se and Sp, and it characteristically shows a curvilinear pattern if the threshold effect exists. Statistical heterogeneity was measured using the χ 2 test and I 2 scores. The I 2 score was used as a measure of the inconsistency between studies in the meta-analysis and was interpreted as low (25-50%), moderate (51-75%), or high (> 75%).
Data were analyzed using the statistical software package Metadisc, version 1.4. The results were synthesized and represented graphically in a forest plot. If heterogeneity was found, the meta-analysis was performed using a random effects model. If there was evidence of the threshold effect, the studies were combined to create a summarized ROC curve (SROC), to calculate an additional measurement of the accuracy of the technique (Q * ) and to obtain the AUC.

Results
A total of 1,648 relevant references were obtained in our systematic search. The results and study selection process are shown in Fig. 1. There were 245 articles requiring full-text review, and 46 studies were included in the meta-analysis  . In addition, one study (by Hensen et al. 52 ) that focused on combined initial prostate cancer biopsy in a North American and European multi-center cohort that overlapped in two studies 13,34 was included in the stratified analysis of initial biopsy. Additionally, the study by Scattoni et al. 48 included initial and repeat biopsy groups that were treated as two data sets. The quality of the selected studies on diagnostic testing was moderate to high according to the QUADAS scale (Table 1).
Based on the studies described above, we retrieved data from 12,295 patients with PCA3 test results and prostate biopsy, of whom 4,225 were diagnosed with prostate cancer. All studies presented the sensitivity, specificity, and cut-off points (25 studies had a cut-off of PCA3 = 35), and most studies presented the ROC curve  Table 1). Among the 46 trials, most were performed in the U.S. and Europe; 5 were performed in Asia (Table 1). The indices of diagnostic validity obtained from the 2 × 2 tables showed that sensitivity ranged from 46.9% to 95%, and specificity ranged from 21.6% to 100%. In the 40 articles that presented the AUC, it ranged from 0.57 to 0.85. A meta-analysis was conducted using the 46 articles mentioned above. The ROC space showed a curvilinear trend, and Spearman's correlation coefficient was 0.612 (P < 0.001), which suggests the existence of a threshold. There was a high degree of heterogeneity in sensitivity (χ 2 = 271.39, P < 0.001), specificity (χ 2 = 735.87, P < 0.001), and diagnostic OR (Cochran-Q = 137.22, P < 0.001); consequently, the diagnostic indices were calculated using a random effects model. Using a forest plot, the overall sensitivity, specificity, positive likelihood ratio , negative likelihood ratio and diagnostic OR were 0.65 (95% CI 0.63-0.66) (Fig. 2), 0.73 (95% CI 0.72-0.74) (Fig. 3), 2.23 (95% CI: 1.91-2.62), 0.48 (95% CI: 0.44-0.52) and 5.31 (95% CI: 4.19-6.73) (Fig. 4), respectively. We used a summary SROC to aggregate data and obtained a symmetrical curve with an AUC of 0.748 (Fig. 5) that represented the technique's diagnostic performance.

Discussion
The clinical value for prostatic cancer diagnoses was not conclusive. According to the European Association of Urology and the National Comprehensive Cancer Network guidelines, the need for prostate biopsy should be determined on the basis of PSA and/or a suspicious DRE 53,54 . However, serum PSA levels can be elevated in benign conditions, and only 25% of men who are clinically suspected of having PCa will have a positive biopsy 55,56 . Thus, other biomarkers with high sensitivity and specificity are needed for screening.
The PCA3 gene was mapped to chromosome 9q21-22, in antisense orientation within intron 6 of the Prune homolog 2 gene (PRUNE2 or BMCC1), spanning a region of approximately 25 kb 57,58 . Ferreira et al. found that PCA3 may modulate PCa cell survival, and PCA3 expression is androgen-regulated via activation of AR-mediated signaling 59 . PCA3 is a non-coding, prostate-specific mRNA that is highly over-expressed in 95% of PCa cells, with a median 66-fold up-regulation compared with adjacent non-neoplastic cells 60 . Because PCA3 does not encode a protein, the only molecule that can be tested is the mRNA; PCA3 mRNA can be measured in urine sediment after DRE. A PCA3 score is the ratio of PCA3 mRNA to PSA mRNA multiplied by 1,000. Although we currently have a good understanding of the role of PCA3 in tumor genes and tissues, the picture is incomplete.
Several studies have indicated that the PCA3 test is useful in reducing the number of negative biopsies 34,39,51 , and more recently the FDA approved the PROGENSA PCA3 assay as a new test for prostate cancer.
Recent advances have included biomarkers such as HPG-1, AMACR, STAMP1, TMPRSS2, ERG, PHI, and P2PSA 31,42,48,[61][62][63] . Some studies have assessed their efficacy by detecting these markers alone or in combination. Although several biomarkers may have specificity that is the same as or higher than that of PCA3, the non-invasive nature of the urine PCA3 test, which is performed after prostate massage, and its good diagnostic performance may make the PCA3 test a better choice for prostate cancer screening.
The current meta-analysis shows the clinical usefulness of this tumor marker in detecting prostate cancer with initial or repeat biopsy. We synthesized the current knowledge about early diagnosis of prostate cancer with PCA3 determination in urine samples. According to the data from 46 studies analyzed, specificity is 0.65 (range 47-95%), which is not adequate; sensitivity is 0.73, which is somewhat lower, and had a minimum value of 21%. The AUC of 0.75 obtained in the SROC curve suggests acceptable performance of the diagnostic test.
In the stratified analysis, a forest plot of initial biopsy showed overall sensitivity and specificity of 0.65 (95% CI 0.63-0.67) and 0.82 (95% CI 0.81-0.83), respectively, and a symmetrical curve with an AUC of 0.80 (95% CI 0.78-0.82) ( Table 2) Table 2), respectively. These results suggest that PCA3 is potentially more suitable for initial prostate biopsy than repeat prostate biopsy. When we stratified the studies by PCA3 cut-off value, the overall sensitivity, specificity, and symmetrical curve AUC values were 0.63 (95% CI 0.62-0.65), 0.74 (95% CI 0.73-0.75), 0.74 (95% CI 0.73-0.76), respectively, for studies with a cut-off value ≠ 35 (Table 2) and 0.70 (95% CI 0.68-0.73), 0.67 (95% CI 0.65-0.69), and 0.77 (95% CI 0.75-0.79), respectively, for studies with a cut-off value = 35 (Table 2). Although the latter group showed better diagnostic performance, it    Our meta-analysis has some limitations. First, the study numbers and the heterogeneity of their approaches influence the accuracy. Although the gold standard (biopsy) was used in all studies, the patient selection, lack of blinding, and different PCA3 cut-off values caused heterogeneity. Second, a potential publication bias may exist, although we tried to avoid this bias by expanding our searches in different databases and by conducting rigorous screening for studies. We evaluated the quality of the articles according to the QUADAS questionnaire. The quality of the studies in terms of diagnostic testing was moderate to high. The relatively small number of trials included in this meta-analysis and significant heterogeneity across the studies may make our conclusion conservative.
According to the current meta-analysis, the PCA3 test shows good diagnostic performance. However, it requires further exploration in well-designed and appropriately-powered trials to determine intermediate and long-term outcomes. Long-term observational studies of health outcomes are also subject to biases. Data sets Sensitivity (95% CI) Specificity (95% CI) Diagnostic OR (95% CI) I 2 , %* AUC (95% CI)