RESUMO
OBJECTIVES: Testosterone testing is relevant for evaluating castration adequacy and diagnosis of castration-resistant prostate cancer (PCa). However, the recommended testosterone cut-off of 1.7 nmol/L (50 ng/dL) to define adequate castration is based on consensus and not validated for the automated immunoassays (AIA) used in today's medical laboratories. Furthermore, appropriate population intervals have not been determined by a state-of-the-art assay. We investigated the analytical suitability of this cut-off and the accuracy of the present-day AIAs for testosterone analysis in castrated PCa patients. METHODS: Leftover serum from 120 PCa patients castrated with luteinizing hormone-releasing hormone agonists was analysed for testosterone by five methods: Architect i2000 (Abbott), Access (Beckman), Cobas 6000 (Roche), Atellica (Siemens), LC-MS/MS. For all assays, the castration 95th, 97.5th and 99th percentile upper limits were determined. Furthermore, Passing-Bablok regression, mean bias and Spearman's correlation coefficients were compared to the LC-MS/MS method and total error based on biological variation. RESULTS: All castration upper limits, ranging from 0.472 nmol/L (LC-MS/MS) to 1.25 nmol/L (Access) (95% percentile), were significantly lower than the current castration cut-off (1.7 nmol/L). Slopes of Passing-Bablok regressions comparing the AIA with the LC-MS/MS method ranged from 1.4 (Cobas and Atellica) to 3.8 (Access). The Architect showed the highest correlation with LC-MS/MS (ρ=0.58). All AIA failed to meet the desirable total error criterion. CONCLUSIONS: These results suggest that a lower general testosterone castration cut-off may be more appropriate in evaluating the adequacy of castration in PCa and that present-day AIA lack analytical accuracy to quantify testosterone levels in castrated PCa.