MiCheck prostate blood test for aggressive prostate cancer designed for the clinical lab setting.

Abstract

229 Background: Clinicians want a more accurate diagnostic test to identify patients for prostate biopsy. We have previously described the development of a Luminex based test for aggressive prostate cancer (MiCheck). To facilitate widespread deployment of MiCheck on standard clinical chemistry analysers such as Roche Cobas and Abbott ARCHITECT/Alinity systems, we have performed additional analytical validation and model development using the MiCheck-01 clinical samples. A novel algorithm, termed MiCheck Prostate, was developed to differentiate aggressive prostate cancer (Gleason 3+4 or above) from non-aggressive (GS3+3) or no cancer patients. Methods: Serum protein biomarkers were measured in 317 samples from the MiCheck-01 clinical trial using either Luminex Multiplex kits, Abbott ARCHITECT or Beckmann Coulter systems. Logistic regression models were used to select best analytes, then Monte Carlo cross-validation was applied to avoid overfitting. A final model was selected from the cross-validation models with best test specificity at 95% sensitivity. The MiCheck-01 samples were later all measured on a standard Abbott ARCHITECT system using commercial ARCHITECT IVD tests. The best cross-validated model was developed and the results compared to those obtained from the mixed analyte platforms using ROC curve analysis. Results: The MiCheck Prostate algorithm was developed to maximise discrimination between aggressive prostate cancer and no cancer or non-aggressive cancer. The MiCheck Prostate algorithm reports a percentage risk of aggressive prostate cancer on biopsy. MiCheck Prostate is a logistic regression algorithm that combines three serum protein markers and one clinical factor and was derived using data from the MiCheck-01 trial. The algorithm gave an AUC of 0.82, with a 48% specificity at a 95% sensitivity cutpoint, with a negative predictive value of 94% for GS3+4 and higher cancers. The samples were later re-measured using all ARCHITECT immunoassays. Good correlations of each analyte across different measurement platforms was observed (Pearson’s R > 0.9). The algorithm was revalidated using the new set of analyte values. AUC of 0.82 for the detection of aggressive CaP, and sensitivity of 95%, specificity of 47% were obtained for the best model, showing no statistical difference to the previous best model derived from the mixed platforms. NPV of 94% for GS3+4 or higher cancers was maintained. Conclusions: The MiCheck Prostate test shows high sensitivity, specificity and NPV for the detection of aggressive prostate cancer. It uses analytes widely available on standard clinical chemistry analysers such as Abbott Architect and Roche Cobas, hence is simple to deploy in the laboratory setting. Studies are ongoing to confirm the MiCheck Prostate algorithm performance using additional clinical sample cohorts.