A mathematical comparison of techniques to predict biologically available testosterone in a cohort of 1072 men

OBJECTIVE: In the absence of widely available measures of determining free and/or bioavailable testosterone (BioT) physicians may use formulae such as the free androgen index (FAI) to estimate free testosterone. We compared the efficacy of calculated markers of androgen status in predicting serum BioT and hypogonadism. DESIGN: Total testosterone (TT), sex hormone binding globulin (SHBG) and BioT were determined in a large cohort of men. Comparison of calculated androgen levels was performed following endocrine assessment. METHODS: TT and SHBG were determined by ELISA, and BioT was determined by ammonium sulphate precipitation. From these data we calculated FAI and free testosterone using two other published formulae - FTnw (free testosterone as calculated by the method of Nanjeee and Wheeler) and FTv (free testosterone as calculated by the method of Vermeulen). A novel formula was derived to calculate BioT from given levels of TT and SHBG (BTcalculated). The ability of the methods (FAI, FTnw, FTv, BTcalc) to predict BioT were compared using regression analysis. The ability of these markers of androgen status to predict biochemical hypogonadism was compared using area under receiver operator curve (auROC). RESULTS: The equation derived from our data was the best predictor of BioT (R(2)=0.73, P<0.0001) although TT was also a good marker (R(2)=0.68, P=0.0001). In the determination of hypogonadism, of all currently available formulae none were better that the TT (auROC: TT=0.93, FAI=0.72, FTnw=0.91, FTv=0.88) although when TT is borderline (7.5