Background: As reference laboratory methods for measuring free testosterone (FT) by equilibrium dialysis (ED) are laborious, costly and nonautomatable, FT levels are often calculated (cFT) rather than measured. However, the predictive accuracy of such estimates in routine use relative to laboratory measurements is not well defined. We provide a large-scale evaluation of the predictive accuracy for different FT formulae compared with laboratory ED measurement and an analysis of clinical factors that may influence accuracy.
Methods: The accuracy of five different cFT formulae (two equilibrium binding, three empirical) based on immunoassays of total testosterone (TT) and SHBG was evaluated by comparing those estimates with FT measurement by ED in 2159 serum samples from men at a single research laboratory over several years.
Results: cFT formulae show systematic discrepancies from the two equilibrium-binding formulae. One empirical formula overestimated FT relative to ED measurements, whereas two newer empirical cFT formulae were more concordant. These discrepancies persisted after correction for serum albumin and were not influenced by obesity, ethnicity or gonadal status.
Conclusions: Commonly used cFT formulae significantly overestimate FT relative to laboratory measurement by ED in male serum samples. The accuracy of the formulae is not influenced by correction for serum albumin, obesity, ethnicity or gonadal status. Such inaccuracy relative to the reference method renders some cFT estimates unreliable for evaluating androgen deficiency as recommended by clinical best practice guidelines.