Clinical impact of prostate specific antigen (PSA) inter-assay variability on management of prostate cancer

Purpose: To evaluate the inter-assay variability of six commercially available prostate specific antigen (PSA) assays, its clinical impact in prostate cancer (PCa) and comparison of automated versus manual assays.

Patients and methods: Sera from 495 patients (425 with PCa and 70 men with Benign Prostatic Hyperplasia (BPH), were measured with six different assays [three automated assays (a-PSA) and three manual ELISA based assay (m-PSA)]. Variability, agreement and bias were measured and compared among assays using Bland Altman plots and Passing and Bablok regression analysis. The possible impact of inter-assay variability on important clinical scenarios was also studied.

Results: All the assays were well correlated (r: 0.88-0.98); however there was significant disagreement and bias between the systems, which were more pronounced among the a-PSA assays. The Bland Altman plot showed that the variability was high between the m-PSA assays and the standard Abbott system with mean difference of 3.8-5.8ng/ml. In contrast, the a-PSA had better agreement with mean difference of 0.8-2.3ng/ml. Beckman Coulter showed the best agreement to the institutional reference (slope-1.097; 95% CI: 1.06-1.14; p<0.05, and intercept-0.20; 95% CI-0.38-0.58; p<0.05, Passing Bablok). It led to significant variability in PCa risk stratification and failure to detect biochemical failure in more than 50% cases.

Conclusions: The discrepancies between the assays lead to significant clinical misinterpretation with risk group migration and detection of biochemical failure post radiotherapy. There are significant discordances between automated and ELISA based assays.