To test antioxidant genes (AOX) expression in circulating cancer cell clusters (CCC). A novel method using molecular, polymerase chain reaction (PCR)-based detection of CCC was applied for predicting prostate cancer and to assess the effect of radical prostatectomy (RP) on reducing CCC and for prognostication of relapse-free survival (RFS), as serum total prostate-specific antigen (tPSA) has limited specificity at 4-10 ng/mL. In all, 240 patients were enrolled in the study, 129 for tumour diagnosis and 111 after RP for disease prognostication. Filtration assay in previously fractionated mononuclear cells (MNC) was used to enrich the CCC and large cells, which were retained in a mesh of 20 microm width. To establish the malignant nature of these cells they were analysed for genomic imbalances detected via PCR-assays of loss of heterozygosity in tumour suppressor loci and of DNA amplification in protooncogen loci. As a screening test in daily practice, real-time reverse transcription (RT)-PCR of AOX was introduced to overcome the laborious and expensive DNA tests. The AOX chosen were glutathione peroxidase (GPX1), Mn-dependent superoxide dismutase 2 (SOD2) and thioredoxine reductase (TXNRD1); selected from 67 marker candidate genes according to sensitivity and specificity data. AOX overexpression in CCC serves as a general marker for solid tumours needing, however, organ markers to relate to the organ of origin. Androgen receptor (AR), PSA and prostate-specific membrane antigen mRNAs served as organ markers for the prostate. Signals were detected in patients' MNC and to a minor level in CCC, rendering to CCC a substantial loss in epithelial features equivalent to a lower grade of epithelial differentiation. Organ markers in the MNC fraction were positive in <85% of AOX testing. The AOX test was tumour predicting (P < 0.001) with a sensitivity of 86%, specificity 82%, positive predictive value 69%, negative predictive value 92%, accuracy 83% and odds ratio (OR) of 28. SOD2 and TXNRD1 expression correlated to tumour size and Gleason score. Objective assessment for the evaluation of the molecular cell markers was achieved by receiver operating characteristic (ROC) curves. The areas under the ROC curve values of the AOXs were 0.7-0.9. RP was followed by a complete clearance of AOX-expressing cells. After RP, a subgroup of patients had residual CCC over-expressing only SOD2 and GPX1 indicating incomplete clearance by RP. Sustained overexpression of SOD2 and GPX1 accounted as risk factors for distant tumour recurrence (P = 0.003) mainly for bone metastases (97% M1b) as evaluated by Kaplan-Meier curves. In univariate analysis the tumour size had a limited effect on the probability of RFS (P = 0.05). In multivariate analysis tumour size, nodal status and Gleason score had no effect. This can partially be attributed to the higher risk level of pathological variables in the AOX over-expressing group but also to ineffective endocrine therapy resulting in marked overexpression of ARs and GPX1, the lead prognosticator gene. The AOX expression level allowed the identification of patients with high progression risk, who have more favourable pathological variables. The AOX testing of CCC is a novel method with excellent prognostic and predictive properties enabling the monitoring of therapies, e.g. effects of RP and endocrine therapy. We speculate that the continuing elevated expression of AOX with special emphasis on GPX1 acts as survival and defence mechanism in CCC required in an atypical environment prone to escape from immune surveillance.