| ObjectiveThe limitation of serum PSA as a prostate-cancer specific diagnostic test has well-understood. Multiplex urine-based assay which has emerged outperforms single biomarker (e.g. PSA) for predicting prostate cancer, whereas the combination of what kinds of biomarkers has to be fully optimized. Our aim is to determine whether a strategy of combining gene-based (e.g. PCA3 and TMPRSS2:ERG), protein-based (e.g. ANXA3) and metabolite-based (e.g. SAR) makers in urine could optimize a multiplex model for detecting prostate cancer.MethodsA total of 86 untreated prostate cancer patients and 45 patients with no evidence of malignancy (NEM) were enrolled in this study.The expression patterns of PCA3, TMPRSS2:ERG and urine PSA were evaluated in urine using real time PCR, ANXA3 was tested using Western blot, and SAR was examined by liquid chromatography-mass spectrometry.Univariate logistic regression analysis was used to assess correlation between single biomarker and prostate cancer diagnostic status and then generate corresponding model, and multivariate logistic regression analysis to exclude insignificant markers and refine the final multiplex model. The diagnostic performance of each model was assessed using ROC analysis.The age, serum PSA, prostate volume, biopsy Gleason score, clinic T stage (cT stage) and risk group of patients were recorded. The correlations of each model with these clinicopathological characteristics were calculated by Spearman rank correlation analysis and Pearson correlation analysis.Results1. Univariate logistic regression analysis revealed that Odds Ratio (OR) of PC A3, TMPRSS2:ERG, ANXA3 and SAR in the subgroup of patients with PSA4～10ng/ml were 1.123 (95%CI:1.054～1.198),1.294 (95%CI:1.100～1.522),0.916 (95%CI: 0.873～0.962) and 1.040 (95%CI:1.009～1.072), respectively; and OR of those in the subgroup of all patients were 1.075 (95%CI:1.038～1.115),1.220 (95%CI: 1.099～1.354),0.918 (95%CI:0.881-0.957) and 1.034 (95%CI:1.011～1.058), respectively, suggesting that all of PCA3, TMPRSS2:ERG and SAR were risk factors of prostate cancer, ANXA3 was protective factor. In addition, multivariate logistic regression analysis indicated that PCA3, TMPRSS2:ERG, ANXA3 and SAR, which showed significant associations with cancer presence in univariate logistic regression analysis, were further included in the final multiplex model in both cohorts.2. In the cohort of patients with PSA from 4 to 10ng/ml, ROC analysis showed that the area under the curve (AUC) of each diagnostic model derived from PC A3, TMPRSS2:ERG, ANXA3, SAR and the final panel were 0.733,0.720,0.716,0.659 and 0.840, respectively. In the population of all patients, the AUCs of each diagnostic model were 0.739,0.732,0.728,0.665 and 0.856, respectively. Additionally, the performance of diagnostic model from the final panel significantly outperformed that of PCA3 with highest AUC value in the single markers in both groups (p=0.018 and p=0.008, respectively).3. Each of diagnostic models from PC A3, TMPRSS2: ERG, ANXA3, SAR and the final panel did not correlated with age, serum PSA, prostate volume and biopsy Gleason score in both populations (all p>0.05), except for PCA3 versus serum PSA (p=0.045) and TMPRSS2:ERG versus serum PSA (p=0.026) in subgroup of all patients in the group of all patients, and PCA3 versus serum PSA (p=0.012) in subgroup of NEM patients in the group of all patients. The probabilities of diagnostic models from PC A3 and TMPRSS2:ERG increased and ANXA3 decreased continuously from low-cT stage to high-cT stage and from low risk group to high risk group in both cohorts (all p<0.05). Non-significant correlations of diagnostic model from SAR and the final panel with cT stage and risk group were found in both groups, with the exception of diagnostic model from the final panel versus risk group (p=0.041) in the population of patients with PSA4～10ng/ml.ConclusionsIn this study, we confirmed that all of PC A3, TMPRSS2:ERG and SAR were risk factors of prostate cancer, and ANXA3 was protective factor. The performance of diagnostic model from the final panel including PC A3, TMPRSS2:ERG, ANXA3 and SAR significantly outperformed that from any of single biomarkers (e.g. PC A3, serum PSA). Only limited association of the model from the final panel with risk group except for age, serum PSA, prostate volume, biopsy Gleason score and cT stage was detected. The strategy of combining gene-based, protein-based, metabolite-based biomarkers in urine is the important guiding significance to optimize the combination of what kinds of biomarkers in multiplex urine-based assay. |
PDF Full Text Request