Mass spectrometry, immunoassay and valid steroid measurements in reproductive medicine and science

In this issue of Human Reproduction, three related Letters to the Editor appear. The first from Keelan et al. objects that testosterone measurements in cord serum reported by Liu et al. (2016) using a direct (non-extraction) testosterone immunoassay are inaccurate relative to the gold standard of mass spectrometry (MS)-based steroid measurements in cord serum, citing their thorough review of biological and analytical factors impacting on cord serum testosterone and estradiol measurement (Hollier et al., 2014). Liu et al. defend the use of a direct testosterone immunoassay with several arguments. A third letter from Warembourg et al. (2016), authors of another paper who also used a direct testosterone immunoassay for measurement in cord serum, concede that they used a convenient, available method but one which they now recognize is unsound and would no longer use. To understand why Keelan et al. are right but Liu et al. are mistaken while Warembourg et al. are on the right track, requires some background on measuring steroids in biological fluids. Accurate, sensitive measurement of steroids in biological fluids is a fundamental pillar of mechanistic understanding in modern reproductive biology and medicine. During the early 20th century, the original quantitative method to measure sex steroids was whole animal in vivo bioassays which were pivotal in guiding purification of all major steroid classes by the mid-1930s. These in vivo bioassays were, nevertheless, cumbersome, laborious and insensitive and therefore impractical for application to medicine although they continue to have important uses in contemporary regulatory pharmacology and toxicology (Marty and O'Connor, 2014). Whole animal in vivo steroid bioassays were superseded by the revolutionary invention of immunoassay in 1959 (Yalow and Berson, 1959, 1960) marking the birth of modern Endocrinology. This powerful innovation had a formative impact that quickly rippled through all biological disciplines, notably pathology as well as reproductive biology and medicine. Virtually all disciplines touched by immunoassay flourished in the light of newfound ability to measure, with exquisite simplicity and unprecedented sensitivity, low levels of bioactive molecules in biological tissues, orders of magnitude lower than was possible with the traditional in vivo whole animal bioassays. Since its inception, the great and durable niche for immunoassay has been its application to measurement of large, immunogenic peptides for which, even now, purely chemical methods are still only emerging as research tools (Woldemariam and Butch, 2014). Immunoassays depend on an antibody generated against the target analyte but with epitope specificity that is inherently uncontrollable so that specificity limitations dependent on the antibody cannot be fully eradicated. Consequently, immunoassays differ from other analytical chemical methods (e.g. haemoglobin, creatinine or cholesterol), which are much easier to standardize universally using commonly available reagents. However, the application of immunoassay to non-immunogenic small molecules like steroids was delayed by a decade (Abraham, 1969; Furuyama et al., 1970) because additional methods had to be developed not only to create antibodies to non-immunogenic small molecules but also to ensure specificity of the resulting immunoassays. The development of antibodies was achieved by chemical conjugation of steroids to larger immunogenic haptens as carriers. A key caveat was that the antibody generated would remain ‘blind’ to the conjugation site on the steroid, creating the molecular basis for cross-reactivity with structurally related steroids. Furthermore, the innovators recognized that steroid immunoassay validity depended critically on pre-assay separation of numerous steroid precursors and metabolites structurally related to the target analyte as well as removing non-specific matrix interference in antigen-antibody and subsequent analytical reaction, crucial steps achieved by organic solvent extraction and chromatography. Hence the pioneering, in-house steroid immunoassays of the 1970–1980s featured the triplet validity criteria for steroid immunoassays—solvent extraction, chromatography and authentic (tritiated or C14 labelled) steroid tracers—so that their findings using such fastidious methodology remain valid today. However, soon the revelations from steroid immunoassays were so striking that demands grew rapidly for their application for use in medical care. This lucrative market opportunity led to a drive for incorporating steroids into multiplex immunoassay platforms (and corresponding single tube kits for research labs). These simplified methods aim to automate measurement steroids in unprocessed human serum/plasma and have come to dominate chemical pathology and research labs over recent decades whereas the original laborious manual steroid immunoassays disappeared from all but large US commercial pathology companies and a few still active academic pioneers (Stanczyk et al., 2010). Conforming steroid immunoassays to automation required methods to shed solvent extraction and chromatography and replaced authentic radioactive steroid tracers with non-authentic, non-isotopic steroid conjugates bearing convenient, but bulky chemical read-outs compatible with multiplex immunoassay platforms. The simplified format, known as direct steroid immunoassay, violates the triplet validity criteria for steroid immunoassay. Predictably, this commodification soon led to reports of method-specific bias, insensitivity and inaccuracy (Schioler and Thode, 1988; Boudou et al., 2001; Taieb et al., 2002; Stanczyk et al., 2003), worst at low circulating steroid levels, findings replicated numerous times and which represent a characteristic of direct steroid immunoassays. These problems with direct testosterone and estradiol immunoassays were formally recognized over the last decade by the US Endocrine Society (Rosner et al., 2007; Handelsman and Wartofsky, 2013; Rosner et al., 2013) which led in 2014 to Instructions to Authors across its suite of leading Endocrinology journals (Wierman et al., 2014) that highlight the limitations of direct steroid immunoassays. These guidelines note that commercial direct steroid immunoassays are validated for unprocessed human serum or plasma only in restricted circumstances, notably for healthy, adult male samples for testosterone and for healthy, pre-menopausal female samples for estradiol. Beyond these settings direct steroid immunoassays require method validation where circulating steroids are low (e.g. children, women, hypogonadal men for testosterone and children, men and post-menopausal women for estradiol) or alternative biological fluids with very differing spectrum of prevailing steroids (including cord serum, cerebrospinal fluid, etc.) or non-human species. For direct testosterone immunoassays, method-specific differences exist between multiplex platform methods even in the favourable setting of healthy adult male serum for which such assays are optimized and validated (Taieb et al., 2003; Wang et al., 2004; Sikaris et al., 2005). By contrast, for samples from children or women, the findings are inaccurate due to insensitivity and poor reproducibility (Taieb et al., 2003; Wang et al., 2004). When applied to non-human serum or tissue extracts, measurement errors become extreme (Handelsman et al., 2015). Similarly, for estradiol immunoassays developed originally to screen for ovarian hyperstimulation in monitoring during ovulation induction or IVF treatment, measuring high circulating estradiol levels did not require high precision whereas method-specific bias, insensitivity and inaccuracy are striking at lower circulating estradiol levels, such as in men (Huhtaniemi et al., 2012; Handelsman et al., 2014), children (Ankarberg-Lindgren and Norjavaara, 2008) and post-menopausal (Schioler and Thode, 1988; Yang et al., 2004; Lee et al., 2006) or aromatase-inhibitor treated (Santen et al., 2007) women. Similar findings are reported for direct progesterone immunoassays which deviate from mass spectrometry-based methods most severely at low circulating progesterone levels (Boudou et al., 2001; Shankara-Narayana et al., 2016). Mass spectrometry (MS)-based steroid measurement has been available even longer than steroid immunoassays (Shackleton, 2010; Stanczyk and Clarke, 2010) and, as a chemical analytical method relying on steroid structural features, has long been considered the gold standard reference method for steroid specificity. Until recently, however, MS-based methods required expensive equipment of limited availability with less sensitivity and throughput. Recent technological advances have produced bench-top MS equipment integrated with liquid chromatography, which retain reference level specificity but are now matching or exceeding sensitivity of immunoassays. While superior performance based on improved technology would be expected to displace older technology, significant commercial obstacles exist to wider availability of steroid MS measurements. These include cost of equipment, need for operator expertise and slower throughput. Most chemical pathology laboratories avoid purchasing expensive equipment for multiplex immunoassay platforms by entering into reagent rental agreements whereby suppliers provide costly equipment without charge in return for locking the labs into purchasing exclusively the suppliers reagents. These marketing agreements create financial deterrents to switching suppliers or to MS-based steroid measurements reinforced by the financial advantages of automation in reduced and deskilled staffing. Interestingly, in the 1970s after the International Olympic Committee banned androgen doping in sports and developed anti-doping labs aiming to detect androgen with methods sufficiently robust to withstand medico-legal challenge by an athlete who might potentially be barred from working in their profession. It was soon concluded that immunoassay technology had inadequate specificity so that all subsequent anti-doping steroid detection tests are based on MS methods. Nevertheless, even four decades later, the growing clinical demand for more accurate and multiple steroid measurement has yet to create a tipping point when chemical pathology labs switch to MS-based steroid measurements as they do for drug and other small chemicals. In a wider context, like all chemical measurement methods, MS-based steroid measurement is subject to analytical error and must conform to external quality control analytical standards to ensure results are comparable between laboratories. The US Centres for Disease Control (CDC) has developed a Hormone Standardization Program (HoST), which is a quality control program aiming to standardize testosterone and estradiol measurement in adult human male and female samples (Vesper et al., 2014). Surprisingly in the light of fixed method-related differences between direct testosterone immunoassays, the CDC program accepted accuracy-based (rather than methodology-based) validation for testosterone measurement which allows for gaming of testosterone immunoassay methods by adjusting their proprietary calibrators to MS-based results. Yet, virtually no direct testosterone immunoassays have been consistently certified in the CDC HoST program. Even such certification would not obviate the need for sound validation of direct testosterone assays when applied to samples other than adult human male or female samples. Turning to the present Letters to the Editor, it is clear that Keelan et al. are right that cord blood testosterone cannot be validly measured by a direct testosterone immunoassay lacking sound method validation in human cord samples (Hollier et al., 2014), a fact conceded by Warembourg et al. However, Liu et al.’s argument that different methods may give different results is simply invalid for quantitative analytical measurements where a pure gravimetric standard and reference methods are available (e.g. testosterone and estradiol) so that results should be accurate and reproducible regardless of the method or laboratory. Alternative methods are valid, if and only if, they give essentially identical quantitative results in agreement with the reference MS-based method. Liu et al. mistakenly claim that lower MS-based testosterone concentrations may be due to procedural losses whereas MS-based testosterone measurement routinely includes stable isotope, internal standards that adjust for procedural losses including extraction. Furthermore, it is an inadequate defence of direct testosterone immunoassay specificity to rely on an unverified manufacturer statement that the antibody had ‘negligible cross-reactivity with other androgens’. Not only did the assay demonstrate cross-reactivity with one of only four related steroids but testosterone would have as many structurally related precursors or metabolites as estradiol, for which Stanczyk estimated >100 structurally related metabolites produced in the body (Stanczyk et al., 2010). Hence cord serum, with its vastly different spectrum of steroids than the adult male serum will have inherently unpredictable patterns of interfering cross-reactivity from a very different spectrum of steroids in cord serum related to the antibody specificity. This contributes to the wide discrepancies reported by Hollier et al. (2014) together with the impact of non-specific matrix effects of unprocessed serum, including effects of haemolysis (Warembourg et al., 2016). Both papers by Liu et al. and Warembourg et al. also used direct estradiol immunoassay, also of uncertain validity in human cord serum and better replaced by MS-based measurements. Defensive chicanery including misguided ad hominem critique are regrettably familiar tactics from those reluctant to accept progress if it requires unwelcome or inconvenient methodological improvement (Taylor et al., 2015). But a major journal aiming to publish timely and durable advances in reproductive medicine cannot become hostage to recalcitrance only solved too slowly by generational change. What should the journal do to avoid such mistakes in the future? The only practicable approach is to strengthen critical awareness of steroid measurement methodology. Both investigators and reviewers should appreciate the validity and limitations of direct steroid immunoassays and ensure that these criteria are met by manuscripts. Most commercial direct steroid immunoassays have published validation to measure testosterone (in adult male sera) or estradiol (in pre-menopausal female sera) which may be cited. For other samples, ad hoc validation for a novel application of direct testosterone or estradiol immunoassays to a new biological fluid or species is necessary. This requires comparison with a reference MS-based measurement of sufficient samples (at least 100) covering the full range to be measured with results appraised by standard method comparison means (Bland–Altman analysis of deviance and Passing–Bablok (non-parametric) or Deming (parametric) regression methods) together with adequate spiked sample recovery and reproducibility metrics for within and between runs. Implementing improved rigour may be achieved by adding a section in the journal's Instruction to Authors highlighting the validity of steroid measurements, asking reviewers to verify that steroid assay methods meet the journal's standards or having specialist manuscript reviewers similar to those used for ethics or statistics to appraise steroid methods. For population, clinical and basic reproductive science research that demands high sensitivity, specificity and reproducibility in steroid measurement and where absolute steroid concentrations are intended to be commutable, the era of convenient, cheap steroid immunoassays is drawing to a close. Multi-analyte MS-based steroid profiling is now ideal for high quality clinical research where absolute accuracy is desirable and will eventually permeate into clinical practice. Nevertheless, the low cost, speed and simplicity of steroid immunoassays mean they will have a continuing role in contexts where the precision required for high quality clinical research is not needed. For example in clinical screening rather than specific diagnosis, for point-of-care methods and low cost laboratory research where relative rather than absolute accuracy may suffice. But the future is bright for wider application of MS-based methodology which not only improves accuracy but importantly widens the scope of steroid measurement to include an almost limitless range of steroid precursors and metabolites. A major handicap of steroid immunoassays was its unavoidable single analyte focus so that each additional steroid required establishing a new immunoassay. This led to circumstances such as for DHT which, nearly 50 years since its discovery as the most potent natural androgen (Bruchovsky and Wilson, 1968), never generated sufficient commercial interest to market an immunoassay so its impact on reproductive health (other than pharmaceutical development of 5α-reductase inhibitors for prostate disease and alopecia (Rittmaster, 1997)) was largely ignored till recent MS-based studies reveal novel association of DHT with men's health (Shores et al., 2014a, 2014b; Yeap et al., 2014; Chan et al., 2017). Multi-analyte MS-based steroid profiling overcomes these limitations whereby a steroid profile can include an almost limitless number of steroid analytes from a single small serum sample and it can also readily incorporate new steroids into an existing profile. This versatility allows for greater flexibility and insight into the flux of steroid pathways when multiple steroids along the pathway can be measured simultaneously such as the newly recognized multiple novel pathways of adrenal and gonadal androgen synthesis (Sharifi and Auchus, 2012; Pretorius et al., 2017). In addition the wider availability of MS-based steroid measurements promises assurance of valid steroid measurements when applied to biological fluids other than serum/plasma from young adult humans, which cannot be provided by direct steroid immunoassays. Like many misadventures, the dark cloud formed by critique of the validity of steroid measurement has the silver lining of clarifying the issue and strengthening the wider understanding of what is required to ensure that steroid measurements are sound and provide the reproducible insight that they are uniquely capable of providing in reproductive medicine and science. Responsible for the content with editorial approval. Nil. None declared.