Hematopoiesis Shows Closer Correlation with Calculated Free Testosterone in Men than Total Testosterone

Abstract

Serum total testosterone (tT)9 includes both free testosterone (fT) (0.5%–3% of total and believed to be the biologically active form) and sex hormone–binding globulin (SHBG)- and albumin-bound forms (1). Guidelines for diagnosing and treating male hypogonadism that emphasize measurement of tT (2) may not address the major variations in SHBG and consequent testosterone bioavailability (testosterone bioavailability [BT] either calculated or assayed directly) that are associated with age, obesity, anticonvulsant therapy, and thyroid and liver disease. For these reasons, recent Endocrine Society of Australia guidelines (3) acknowledge the value of measuring SHBG as well as BT, but nevertheless state that, “There is no evidence that free or bioavailable testosterone levels, which are usually not directly measured but calculated…are a better measure of androgen status than total testosterone level. Moreover, reference intervals for these [BT levels] are even less well defined than those for measured testosterone concentrations. Therefore, they [BT levels] are not recommended for clinical decision-making.” Although commercially available direct fT assays are not always robust, calculated fT (c-fT) values from a mass-action formula based on directly assayed levels of tT and SHBG ± albumin show promise (4). The Vermeulen equation [c-fT = {√[(SHBG − tT + albumin)2 + (2 × albumin × tT)] − (SHBG − tT + ½ albumin)}/0.001 albumin (normal range in our laboratory: 215–760 pmol/L when tT and SHBG are expressed in nmol/L and albumin in g/L)] gives much greater weighting to SHBG compared with albumin levels, aiming to reflect the much stronger affinity of SHBG binding for testosterone, so that, in practice, variation in albumin concentration within the physiologic range exerts a relatively small effect on c-fT. Hence, for simplicity, many laboratories input a fixed value (46 g/L in our lab) to the calculation rather than the actual measured serum albumin concentration. Measuring both serum tT and SHBG and using these 2 values in clinical decision-making but without calculating fT (as per Endocrine Society of Australia guidance) seems counterintuitive. However, it is important to recognize that both total and BT calculated values are surrogates for the various biological end points of androgen action (androgenicity) that remain difficult to quantify. To test our hypothesis that c-fT might provide a better surrogate for androgenicity than tT, we identified an androgen-dependent biological parameter that was (a) important to human health; (b) robust and reproducible with low CV; (c) routinely measured in the clinical setting so data collection is convenient and inexpensive; (d) showed an association below, within, and above the adult male normal range for serum tT; and (e) was independent of aromatization to estradiol. Complete blood count (CBC) was an obvious candidate, since stimulatory effects of testosterone on hematopoiesis are well recognized but poorly understood. Men have higher mean hemoglobin (Hb) concentrations, hematocrit (Hct), and red cell count (RBC) than women, children, and hypogonadal males; intentional (e.g., competitive sports) and inadvertent/iatrogenic testosterone overtreatment can also cause erythrocytosis. Our local population is overwhelmingly Caucasian, and our laboratory processes all samples collected by local primary and secondary care providers. Laboratory screening for hypogonadism and for monitoring testosterone therapy is performed in both primary and secondary care settings, but the Endocrine unit advises on all initiations of testosterone treatment and provides long-term input in respect to dose adjustments. Thousands of men had undergone same-day testing for CBC, tT, and SHBG levels in our laboratory that consisted of morning venipuncture for diagnostic purposes in untreated men and “trough” venipuncture for monitoring men receiving intramuscular testosterone replacement, as per local guidelines. We extracted biochemical and hematological parameters over 4 years on 2,529 consecutive men (aged 18+ years) from our database. We hypothesized that large sample size and low CV for parameters measured would allow meaningful correlations to emerge, despite limited data on confounding factors due to de-identification of data, as mandated by the Information Guardian. Data collected were for Hb, Hct, RBC, SHBG, and tT levels. SHBG (nmol/L) and tT were analyzed on the Roche Modular System (Roche Diagnostics) by a 2-site sandwich immunoassay and electro-chemiluminescence technology. fT was calculated using the Vermeulen equation (4). Mean data (SD) were as follows: tT, 13.5 (7) nmol/L; c-fT, 267 (156) pmol/L; Hb, 14.7 (1.6) g/L; Hct, 0.44 (0.04); and age, 54 (17) years. c-fT and tT were nonparametrically distributed and were therefore log-transformed before performing correlation analyses (Stata Statistical Software, 2013 Release). Ln-c-fT correlated (P < 0.001) with Hb (rho 0.304), Hct (rho 0.300), and