Predictive accuracy and sources of variability in calculated free testosterone estimates

Background Serum free testosterone (FT) concentrations are commonly requested, but because reference FT methods are too laborious various calculational algorithms for FT based on total testosterone (TT) and sex hormone-binding globulin (SHBG) are frequently used. This study provides the first large-scale evaluation of the predictive accuracy and sources of variability for different FT formulae compared with direct laboratory measurements. Methods Using a large data-set of direct FT measurements by centrifugal ultrafiltration, the predictive accuracy of five different formulas for cFT (four existing plus a new formula) is evaluated in 3975 consecutive blood samples. In a second data-set of 124 samples from a reference panel of healthy eugonadal young men, we estimate the relative influence of the five algorithms and eight different TT and two SHBG assays including all available commercial total TT and SHBG assays together with a gas chromatography/mass spectrometry T reference method. Results cFT formulae show wide discrepancies with equilibrium-binding algorithms showing systematic overestimation relative to direct FT measurements, whereas two empirical cFT methods were more concordant. Variations between commercially available TT immunoassays have a strong impact on calculation of FT with TT assays contributing 82.2% of overall variance compared with 13.7% for the cFT algorithms and 4.1% for the SHBG assays. Conclusions If FT measurements are requested and direct measurement impractical, cFT formulae using TT and SHBG immunoassays provide an approximation to direct FT measurement that is strongly dependent on the TT, cFT formula used and, to a lesser extent, SHBG immunoassays.