Introduction: Monitoring testosterone (T) levels is increasingly being recommended to guide the effectiveness of androgen deprivation therapy (ADT) in the treatment of advanced prostate cancer (PCa). T levels of less than 20 ng/dl (0.7 nmol/L) on therapy have been associated with better outcomes, with some clinicians advocating medication switch for patients who do not achieve this level. Three main assays for T measurement exist, including radioimmunoassay (RIA), chemiluminescence assay (CLIA), and liquid chromatography-tandem mass spectrometry (MS). CLIA and RIA are commonly used worldwide, however, MS is regarded as the reference standard. We set out to determine the discordance rates of T measurements among men on ADT.

Methods: A retrospective McCain GU biobank (MGB) database review of PCa men on luteinizing hormone-releasing hormone (LHRH) monotherapy for three or more months was conducted. Patients with exposure to second-line hormone agents or chemotherapy were excluded. Corresponding serum samples were identified and split in triplicate for measurement via all three assays. Observational data was reported and T measurements were analyzed for variability, looking for categorical concordance. Over and under-estimation rates were calculated.

Results: Ninety-five patients were included with a mean age of 70 (50–92) years. Eighty (88%) patients were on LHRH agonist therapy. Mean ADT duration was 24.1 (3–144) months. Mean T levels were different, with MS at 11.4 (0.1–282) ng/dL, CLIA at 23.4 (20–204) ng/dL, and RIA at 15.1 (0.1–170.5) ng/dL (p<0.001). Most (95%) patients had T ≤20 ng/dL by MS and CLIA as compared to only 80% by RIA. After subdividing into T categories of ≤20, 20–50, and ≥50 ng/dL, concordance analysis showed that 4.3% and 18.9% of T measured by MS would have a different category result when remeasured by CLIA (Kappa 0.84) or RIA (Kappa 0.50); 16.8% of T measured by CLIA would also have a different category result when remeasured by RIA (Kappa 0.58). Intra-class correlation coefficient between all three measures was 0.83 (95% confidence interval [CI] 0.77–0.88). CLIA and RIA overestimated T in 66% of patients, with T<20 ng/dL measured by MS. Conversely, CLIA and RIA underestimated T in only 2.4%, with T>20 ng/dL measured by MS.

Conclusions: There is significant variability in T measured with RIA, CLIA, and MS. CLIA and RIA overestimated T levels in the majority of patients, leaving a concern of misdiagnosing true castrate patients as being inadequately treated. Clinicians should be mindful of variability in T measurements by assays when using them for decision-making among PCa patients on ADT.