Free Testosterone: Pumping up the Tires or Ending the Free Ride?

The review by Goldman et al. (1) is a strong restatement of the free hormone hypothesis (FHH) with its contingent focus on measurement of so-called “free” testosterone (FT). The FHH has even been recently described as a “central dogma of endocrinology” (2), reminiscent of the hubris of Crick’s (3) central dogma of molecular biology stated as a unidirectional information transfer from DNA to RNA to protein, which proved neither a dogma (an unassailable proposition) nor universally correct. The review provides an outstandingly clear and encyclopedic summation of what is and is not known of the molecular mechanisms of testosterone binding to circulating binding proteins. On the other hand, its coverage of FT measurement and its interpretation are both more selective and tendentious. Above all the review’s overall focus on means rather than ends, on how to do it rather than why, the review belies why the application of FT measurements remain contentious so that less expert readers may wonder what controversy there is that the authors hint at fleetingly. Despite being extant for decades, the use of FT measurement has barely been stated in a testable form and virtually never directly tested as a refutable hypothesis. Rather by inference and repetition as if it is self-evident, it has become entrenched as an enthusiastically wielded yet largely untested concept that goes from one paper to the next without ever seeming to pass through a questioning mind. The FHH originated in a now obsolete 1970s pharmacology theory of adverse drug reactions by which such interaction was attributed to mutual displacement of drugs bound to circulating binding proteins (4). That now discarded speculative concept (5) has been superseded by physiological drug interaction mechanisms, such as molecular receptor binding, cytochrome P450 induction/inhibition, P-glycoprotein, and ion channel blockade (6). Contemporaneously, the 1970s provided fertile ground for introducing the concept of “free” hormones into Endocrinology. The revolutionary invention of radioimmunoassay introduced the notions of free and bound hormones in immunoassay and also hormone receptor theory, both deriving from the biochemical theory of equilibrium binding equations. Although pharmacological theories evolved, FHH persisted in Endocrinology making a remarkable transit from a simple, illustrative heuristic to quasi-axiomatic status without ever undergoing critical empirical testing. Regardless of its flimsy basis, invoking FT has flourished as a panchreston (explain-all), a final arbiter to resolve uncertainty about androgen status. This reached a farcical zenith in papers glorifying FT calculation as by the “Law of Mass Action,” an affectation comparable with describing measuring weight on a miscalibrated scale according to the “Law of Gravity.” In that sense the review is a welcome restatement of a hypothesis whose quasi-axiomatic status has granted it a free pass from engaging with logic or evidence in its application. To provide a more nuanced picture, an alternative view that FT represents a muddled and misleading concept, an explanatory fiction that provides false reassurance in men’s reproductive health care, is outlined briefly. The fatal flaw of FHH theory, not addressed in the review, is its lack of a coherent physiological basis for interpretation of FT. Although the review notes the evidence refuting each of the three key assumptions of interpreting FT—that only unbound testosterone is able to enter tissues to exert androgenic effects, that testosterone bound to circulating proteins constitutes an inert buffer, and that the same mechanisms operate equally in all tissue capillaries—it proceeds as if these defects do not matter. Most critically, the assertion that the small fraction of unbound (or loosely bound) circulating testosterone represents its most active moiety lacks sound logical basis. Even accepting the assertion that testosterone must transit through a “free” compartment on its way to tissues, that moiety would be just as accessible to sites of degradation, where it is metabolically inactivated. On that basis FT may equally reflect the most readily metabolized, least active and most evanescent, fraction of circulating testosterone rather than the most bioactive. Thus it cannot be known a priori whether FT measurement reflects androgen action, androgen degradation or some net balance between them. In this light it is inexplicable, a meme finding a receptive host, how one could arrive at the a priori generic interpretation that FT measurement represents the most active fraction of circulating testosterone for any or all tissues. Even accepting these assumptions as doctrinal verities, the focus on FT represents an excessively narrow, simplistic view of testosterone action that overlooks aromatization and amplification to testosterone’s potent bioactive metabolites, estradiol and dihydrotestosterone, respectively. In this light it is doubtful that measuring FT could be interpreted meaningfully in clinical or biological terms as an arbiter of overall androgen status. Experiments of Nature involving genetic modifications of sex hormone binding globulin (SHBG) and analogous circulating binding proteins also shed doubt on the plausibility of the assertion that FT represents the most active moiety of circulating testosterone. The rare mutations inactivating human SHBG (7–9) convert all circulating testosterone to “free” or “bioavailable” form but display no distinctive phenotype. As possible artifacts of ascertainment bias, further evidence is required to distinguish an authentic phenotype of human SHBG deficiency if one exists. Similarly, increasing reports of genetic polymorphisms of SHBG that alter testosterone’s binding affinity to SHBG may influence measurement of circulating testosterone but without necessarily any reproductive or other health detriments. The frequency of mutations identified in the vitamin K-dependent protein S featuring SHBGlike domains (10) suggest that more SHBG mutations with uncertain impact on androgen physiology may be discovered. Targeted genetic inactivation in the murine reproductive systems has consistently produced informative phenotypes replicating cognate human reproductive functions and disorders with high fidelity (11). However, as FT is an abstraction rather than a molecular species, it is not feasible to direct genetic testing by inactivating FT to determine the physiological impact of its elimination. However, Nature has performed an informative corollary experiment in that mature rodents have no circulating SHBG because mature murine liver does not express SHBG. Yet, the absence of circulating SHBG in mature rodents has no apparent reproductive or other phenotype indicating that SHBG, and by implication FT, is redundant in mammalian reproductive physiology. This is reinforced by experiments humanizing mice to express transgenic human SHBG in the mature mouse liver so that SHBG then appears in the bloodstream. The main effect was that SHBG reduced clearance rate and prolonged half-time of circulating testosterone, confirming analogous evidence in nonhuman primates (12), as well as dampening the extreme pulsatility of circulating testosterone in mice (13), an effect without known adverse impact. Despite dramatically elevated (200-fold) circulating testosterone, there were minimal changes in serum luteinizing hormone, seminal vesicle, and levator ani muscle weights consistent with a minimal hypogonadal phenotype (2, 14). It remains unclear what effects transgenic SHBG would achieve if circulating SHBG levels replicated those in humans rather than being two orders of magnitude greater; analogous massive transgenic overexpression of other reproductive hormones in mice has produced nonphysiological effects (15). In confirming that SHBG is largely if not wholly dispensable for normal mammalian reproductive function, it underlines a salient caution on generic interpretation of FT in relation to androgen action, especially the unfounded claim that FT represents in theory the most “active” moiety of testosterone. Corroborative findings from natural human mutations and genetic murine models that inactivate analogous circulating hormone carrier proteins. These indicate that nonprotein bound, “free” hormones are not necessarily more, but may be a less bioactive moiety of the hormone. Elimination of corticosteroid binding protein (transcortin) rendering all circulating glucocorticoids into a “free” form creates impaired rather than enhanced glucocorticoid action in humans (16) and mice (17). Similar findings are evident with genetic inactivation of vitamin D or vitamin A carrier binding proteins that lead to states of inadequacy of vitamin D (18) or vitamin A (19) rather than intoxication states despite greatly expanded “free” fractions. Another variation is that mutations inactivating human thyroid hormone binding globulin have essentially no clinical phenotype being identified incidentally by anomalous thyroid function tests without clinical consequences (20). Consequently, genetic inactivation of analogous circulating carrier proteins provides no clear basis to believe that FT would be more rather than less bioactive. Rather, the bioactivity of “free” fractions remains inherently unpredictable and empirically are more often less rather than more bioactive than the protein-bound molecule. Despite its tenuous theoretical basis, FT measurement may have empirical utility if it provides evidence on androgen action independent from testosterone. As the gold standard of direct dialysis-based FT measurement is a laborious, exacting methodology lacking standardization and not widely available in current chemical pathology laboratories, this has led to proliferation of empirical formulae to calculate FT, all based on measuring testosterone and SHBG in the same samples. These formulae come in two forms, one being model-based formulae based on equilibrium binding equations with an obligatory requirement for plug-in estimates for testosterone’s binding stoichiometry and affinity for SHBG. The alternative is an assumption- and model-free, purely empirical formulae derived from large databases. Whatever the approach, any calculated FT is necessarily a deterministic (inverse) function of age as it compounds two age-related measures. As such, any variable correlated with age (gray hair, natural teeth, dentures, comorbidities, medications, widowerhood, etc.) is likely to be related to calculated free testosterone without any necessary biological relationship with testosterone. This makes it essential that this calculational device must correspond closely and accurately to directly measured FT or else it introduces a masked confounding by age. Widely used model-based formulae by Södergård et al. (21), Vermeulen et al. (22), and Mazer (23) incorporate two major sources of potential systematic error: assumptions about the stoichiometry and the binding affinity of testosterone for SHBG. Although the review attributes the inaccuracies of model-based formulae to their “linearity,” the discrepancies arise for a different reason: the mistaken assumptions on the stoichiometry of testosterone binding to SHBG. Prior to the crystal structure of SHBG in 2001 (24), the SHBG homodimer was believed to bind only a single testosterone molecule at dimerization junction (25), an assumption incorporated into all model-based FT formulae. However, the SHBG crystal structure showed that each monomer had a testosterone binding pocket so that the homodimer binds two, not one, testosterone molecules. Correcting this erroneous stoichiometry incorporated into the model-based binding equations reduces the marked deviations of model-based formulae from laboratory measured FT (26, 27). Another untested assumption of model-based formulae is the substitution of gravimetric measurement of SHBG by immunoassay—which differ between immunoassay methods—for SHBG binding capacity, which the binding equations actually invoke. Finally, the impact of genetic polymorphisms or acquired disease-related changes in SHBG binding affinity cannot be incorporated into the model-based FT formulae. The review highlights a biophysical model by Zakharov et al. (28) for testosterone binding to SHBG with a novel but incompletely disclosed formula, patented by review coauthors and not yet subject to independent evaluation. The Zakharov biophysical model is a refinement of the equilibrium binding models by fitting two allosterically linked binding sites with distinct affinities, instead of two binding sites with the same affinity per SHBG monomer. Hence, although this work refines the understanding of the molecular binding of testosterone to SHBG, this minor variant of a model-based formula does not overcome the limitation of an obligatory requirement for plug-in binding affinities. Improvement by this thermodynamic model is at least partly attributable to its incorporation of the correct testosterone binding stoichiometry to SHBG. More troublesome problems arise for model-based FT formulae in their requirement for plug-in binding affinities of testosterone for SHBG, which are highly influential and responsible for much of the deviations of model-based formulae from laboratory-measured FT. Measuring such binding affinities, assumed to be population-wide and invariant, is as exacting as dialysis-based direct FT measurement; yet there is a fivefold variation among such estimates used in various equations (23). The alternative, assumption-free fully empirical equations are given scant coverage denying them a fair hearing. These circumvent pitfalls arising from the assumptions requiring use of plug-in estimates for testosterone’s stoichiometry and binding affinity to SHBG. The fully empirical equations are created by regression of dialysis-based laboratory measured FT on measured serum testosterone and SHBG in the same samples using large databases, a methodology open to updating with new data. The most recent empirical formulae derived from a large data set (>4000 serum samples) has been subsequently verified in a different large data set (>2000 serum samples) from another laboratory (27, 29). In both studies, direct head-to-head testing showed that the model-based formulae consistently overestimate FT (27, 29, 30), a finding confirmed independently by others (26, 31–34). Curiously, Zakharov FT calculations produce even higher results than the Vermeulen equation making further independent evaluations of its validity essential. In that context, it is premature for the review to cite a lower confidence limit for calculated FT when the studies on which it is based have given rise to widely differing reference ranges for serum testosterone having reported a lower 2.5th centile of 348.3 ng/dL in one report (35) but subsequently as a 40% lower one (209 ng/dL) (36). Analogous attempts to assess adrenal function using formulae to calculate “free” cortisol from total cortisol and corticosteroid binding globulin have been reported for decades (37, 38) but have proved biased with regard to dialysis-based “free” cortisol measurement (38, 39), as well as providing conflicting findings in liver (39) or adrenal (40) failure. Consequently, such formulae are not recommended for routine clinical use (38). The main claimed utility for FT measurement, and especially its calculational surrogate, has always been the clinical application to infer testosterone deficiency when circulating testosterone levels are not low. By compounding age-related increases in serum SHBG and decreases in serum testosterone, the composite measure FT is rendered still more age-dependent. Hence, anointing FT as a meaningful clinical parameter creates an additional distinct opportunity to justify more testosterone prescribing especially for older men. Already, despite no new approved indications, testosterone prescribing has undergone massive expansion over recent decades with a global 100-fold increase in per capita sales over the last three decades including a 40-fold increase in Canada and tenfold increase in the United States over only the first decade of this century (41) with this expansion largely due to prescribing for age-related hypogonadism in middle-aged men (42). This represents the epitome of disease-mongering, the expansion of disease definition to justify wider marketing of its remedies (43). In this instance the long-standing conventional medical definition of “hypogonadism” as a pathological disorder of the reproductive system (hypothalamus, pituitary, testis) had its boundary elastically stretched, redefined as a conjunction of ubiquitous nonspecific clinical symptoms, common to virtually all medical ailments, combined with a low serum testosterone concentration, regardless of the underlying disorders, if any. In that context, introducing FT widens the definition of age-related “low” testosterone. A major flaw in this approach is that there is no connection between such ubiquitous nonspecific symptoms and circulating testosterone. Although men with androgen deficiency due to pathological hypogonadism display a consistent threshold of serum testosterone with their own distinctive symptoms (44), those symptoms and their blood testosterone thresholds differ widely between individuals. Consequently, grouping individuals according to nonspecific symptoms blurs and attenuates these relationships (45). In the European Male Ageing Study cohort study that purported to define age-related hypogonadism, only weak relationships were observed between sexual symptoms and blood testosterone (46). Those sexual symptoms exhibited high rates of false positive and negatives (>40%), and the proposed biochemical definition depended on introducing the questionable Vermeulen formulae for calculating FT; without it the relationship between sexual symptoms and serum testosterone was nullified by adjustment for age and comorbidities (47). Further, although observational data cannot support causal inference, the study’s conclusions assumed that the low testosterone caused the sexual symptoms, whereas there is increasing evidence for reverse causality (48–51). Ultimately, however, the occurrence of nonspecific symptoms in middle-aged and older men does not mean they are caused by an accompanying low blood testosterone nor that testosterone treatment is appropriate. On the contrary, well-designed placebo-controlled studies, such as the Testosterone Trials (52), are indispensable to determine safety and efficacy of testosterone treatment. So far, the findings are underwhelming (53, 54). This review provides a welcome, comprehensive, and scholarly analysis of testosterone binding to circulating proteins. Its limitation is the application of such knowledge where it reinforces evidence-light orthodoxies including a newly proposed calculated free testosterone calculation, yet to be independently verified. This is a misdirected solution to problems of decisive arbitration on androgen action and, as such, it cannot end the controversy over applying the FT concept. A valid scientific concept requires a sound foundational theory and evidence and being open to testing and refutation not just an unshakeable belief. Introducing the FT into clinical guidelines is particularly hard to understand as using that unproven criterion can only provide false reassurance by merely shifting the uncertainty to an even shakier footing in a subtle bait-and-switch. Clinically, the appeal of FT is a misplaced desire for concreteness deriving from a loss of confidence in clinical judgment on the indications for prescribing testosterone. It seeks an authoritative foundation for the endocrinologist when overt clinical facts fail to provide a therapeutic option in the face of a strident, Internet-confected patient demand for testosterone. However, refashioned, the appeal to FT misconstrues (or denies) a deficiency state arising from a and in FT have to way to that testosterone treatment be to for authentic pathological hypogonadism and there is sound evidence of efficacy and safety in other speculative In the such only to the to the of the which it is based and the of the free hormone hypothesis free testosterone sex hormone binding The has to