Establishment of a WHO Reference Reagent for anti-Mullerian hormone

In 2014, the WHO Expert Committee on Standardization (ECBS) approved the project to develop an International Standard for Mullerian Inhibiting Substance, also known as Anti-Mullerian Hormone (AMH) for immunoassay [1]. This was in response to repeated requests [2‐4] from manufacturers and the clinical community who recognised both that there was an increase in the use of serum AMH measurement and also, that there was a discordance between the AMH measurements reported by the small number of immunoassays available at the time. As serum AMH concentration is considered a biomarker which represents the dynamic reserve of developing ovarian follicles, a major clinical use of AMH measurement is to assess female infertility and to gauge the likely response to ovarian stimulation procedures for assisted reproduction. Recently, there has been an increase in the number of manufacturers who offer, or who are in the process of developing, AMH immunoassays. This has led the WHO to approve the development of a WHO Reference Reagent for AMH (NIBSC code 16/190) with the aim of providing a reference preparation to further investigate the calibration and performance of assay methods.

Produced in ovarian granulosa cells, human AMH (NCBI RefSeq NP_000470.2) is a disulphide-linked homodimer comprising two subunits of 560 amino acids, thereby having a formula mass of 113 kDa for the dimer. Proteolytic cleavage of the homodimer forms N-terminal and C-terminal disulphide-linked dimers which remain non-covalently associated. To our knowledge, commercial immunoassays detect total AMH, either uncleaved, or cleaved but with the N- and C-terminal dimers remaining associated. Each subunit is glycosylated and it is likely that this property of AMH, specifically the effect of glycosylation in retarding migration by non-reducing SDS-PAGE, which has resulted in the accepted convention that the formula mass of AMH is 140 kDa [5]. There are two potential N-linked glycosylation motifs in the amino acid sequence of AMH [6] and Pankhurst and McLennan [7] have suggested that the native protein may also contain O-linked glycosylation. As noted by Nelson and La Marca [2], as a glycosylated protein, the molecular mass of AMH is difficult to ascertain. Despite this, a molecular mass of 140 kDa is used for the widely cited conversion of 1 ng/ml = 7.14 pmol/L AMH and immunoassay methods report in mass or concentration units. The above conversion is cited by manufacturers, used by monitoring organisations (such as UK NEQAS) and there is awareness of the units and the conversion in patient groups.

The uncertainty of the true molecular mass of AMH demonstrates a key challenge presented by the accepted convention of reporting AMH immunoassays in mass units despite the measurand being a complex biological molecule with undefined post-translational modifications. The origin of the “nanogram of AMH” is unclear and the methods by which mass values were assigned to the first CHO-derived recombinant AMH immunoassay standards have not been reported [8, 9]. The unclear traceability of the mass assignment of AMH from the original assays to current assays has been described extensively elsewhere [10, 11]. By necessity, the calibration of new assay methods has relied on the use of serum value transfer using panels of human samples to transfer the AMH measurement from an AMH immunoassay that is considered the current market leader or ‘gold standard’ to the calibrators of the new assay. It is not known whether this process, over more than 15 years of assay development, has led to a situation where the current AMH content of a stated “nanogram of AMH” is different from the original measurements that described a nanogram of recombinant AMH. In situations such as this, where there is no formally established metrological traceability, one possible solution is to define an arbitrary unit of AMH immunoreactivity such as a WHO International Unit (IU) which is defined by the content of the current International Standard (IS) as determined through a WHO collaborative study [12]. However, the maturity of AMH measurement and the historic confusion caused by the discordant values reported by the first commercially available assays has led to the use of mass units being the accepted convention. If an IU of AMH immunoreactivity was established, there is a strong possibility that assay-specific conversion factors would be introduced to enable continued reporting of mass units.

However, foregoing the establishment of an IU of AMH immunoreactivity for an alternative approach, presents challenges. Technically, the assignment of mass units to a reference material requires the use of Système International (SI)-traceable physicochemical methods which precludes the use of measurement procedures with a biological component [13]. For proteins, assignment in SI units with unambiguous traceability has been successfully achieved only for small, unmodified proteins such as growth hormone (22 kDa) [14] and insulin (5.8 kDa) [15, 16]. Furthermore, once physicochemical reference methods have been developed to accurately measure AMH, for instance tryptic peptide isotope dilution mass spectrometry [17], there is still the question of which formula mass to use to convert the molar quantities reported by these methods to mass units.

The challenges highlighted above, outline the context in which we proceeded to formulate and manufacture the reference material coded NIBSC 16/190. Building on work previously undertaken to develop and assess a stable, lyophilised formulation of AMH, [11], the batch of ampoules coded 16/190 comprised recombinant CHO-derived AMH, lyophilised from a formulation containing trehalose and casein. Ampoules of 16/190 were then evaluated in a WHO international collaborative study, the results of which are summarised here. Full details of the results of the collaborative study are reported to the WHO ECBS as part of the process to establish a WHO reference material. This report, WHO/BS/2019.2363, is available on-line [18].

By contrast, measurements of the comparator sample B, containing half the content of AMH, when expressed relative to 16/190, varied with the much lower GCV of 12%. This suggests that the 16 valid laboratory methods exhibited a similar proportionality in their detection of recombinant AMH. As such, possible sources of the observed inter-assay variability in content estimates may be differences in antibody affinity, reaction parameters or assay calibration. Given the unclear traceability of the mass unit of AMH and given that widely discrepant estimates were previously shown to be provided by assays using the same monoclonal antibodies, F2B 12/H and F2B 7/A [11], it is possible that the observed variation is due to differences in assay calibration. It is understandable that, in the absence of a global standard, manufacturers aim to align their methods to currently-available immunoassays to ensure consistency in the measurement of patient samples. We are aware of one manufacturer (Ansh Laboratories) who has pursued SI-traceable value assignment of their calibrators [22]. Other assays have used serum value transfer in which the AMH content of a bank of patient samples, as measured by the current market leading assay, is applied to the signal of the new assay, and therefore to the new calibrators. This approach successfully engineers inter-assay agreement between serum measurements but has the potential to artificially assign a mass value to a set of calibrators. If the calibrator AMH and serum AMH are not recognized by the capture and detection antibodies in an equivalent manner, as may be the case, for example, when bovine AMH calibrators are used, the range of estimates observed in this study is not wholly unexpected.

As expected, this has an impact on the commutability of the candidate reference material. To assess commutability, data from the measurement of the AMH content of 20 patient samples were analysed by an approach similar to that recommended by the IFCC Working Group on Commutability [23]. Of the 16 immunoassay methods which met the validity criteria, 16/190 was within the limits of acceptable bias difference for 6 methods. For a further 3 methods, the mean difference in bias of the nominal dilutions tested, was within acceptable limits for only 3 or 4 of the 6 nominal concentrations assessed. This may be explained by experimental variation, non-constant bias or freeze-thaw of test dilutions. For 7 methods, 16/190 was not commutable. Of these, 2 were the same method on different platforms and 1 was the same method as a commutable method but on a different platform.

As discussed above, the practice of serum value transfer to assign AMH concentration values to assay calibrators of different sources may produce a wide range of responses when presented with a single source of reference standard AMH, in this case human, recombinant AMH. More subtle commutability deviations, particularly where a related platform-assay combination has shown 16/190 to be commutable, may be improved with increased familiarity with the material such as in reconstitution, dilution and examination of the storage conditions of sample dilutions. Lack of commutability in a reference material has often been attributed to the source, matrix and processing that is required to achieve long-term stability in a product that can be readily transported worldwide at reasonable cost to all. It is noted that the concentration of the matrix of 16/190 will be extensively reduced once diluted to the concentrations measured in current assay methods. With regard to the preparation of commutable reference materials, Braga and Panteghini [24] state that ‘the qualification of the measurement procedures about their selectivity should be done independently and in advance’. Thus, it needs to be considered if enough is known about current AMH immunoassays, such as the extent of equivalence of antibody binding to native and calibrator AMH, to consider assessment of commutability appropriate at the current time. Defining the relationship would however, require that validated procedures are available to measure the concentration of the immunoassay calibrators independently of the signal derived from serum value transfer. This will require further technological advances in peptide and glycan identification and quantification.

There is an increasing awareness of the importance of traceability in laboratory medicine such that a measurement result can be related to a reference material through a documented unbroken chain of calibrations [24, 25]. The chain is comprised of a hierarchy of calibrators and measurement procedures, commencing with the Primary Reference material which is defined and measured by a primary reference measurement procedure. It is essential that the reference materials provided at all parts of the traceability chain are commutable. For reference materials calibrated in mass units, the reference measurement procedure must use procedures which are traceable to the SI unit of mass. As mentioned previously, this is a challenge for complex analytes such as glycoprotein hormones which exist in multiple isoforms. For these, international conventional calibrators such as WHO International Standards, have been used successfully to harmonize methods.

The unclear origins of the nanogram of AMH and its subsequent transfer between assays by serum value transfer are clearly far removed from the above metrological ideals for traceability in laboratory medicine. Yet, as discussed previously, there are robust reasons to maintain the accepted units of AMH immunoreactivity. In addition, approval of the Elecsys® AMH assay (Roche) as a companion diagnostic in which the reported AMH concentration is used in an algorithm to calculate the patient dose of Rekovelle®, a follitropin delta product (Ferring Pharmaceuticals) further reinforces the current conventions [26, 27]. Also, this development highlights the challenges of introducing a WHO International Standard, should this result in a requirement for an assay, used as a companion diagnostic, to be recalibrated.

While recognising the issues described above, there is a responsibility to support the current use of AMH immunoassays and to provide a reference preparation which can be used to further investigate assay calibration and the performance of methods.

This study demonstrated that the majority of AMH immunoassays available for clinical use, recognised recombinant human AMH when reconstituted from the stable, lyophilised form provided in ampoules of the candidate reference material coded 16/190. Estimates of AMH content in terms of individual assay calibration showed considerable variability which may reflect historical calibration exercises and current approaches used to assign values to method calibrators which may be native, recombinant or non-human AMH. However, valid assays exhibited a similar proportionality of response as demonstrated by the low variability of the ratio of content estimates of a comparator sample to the content estimate of 16/190. It was recognised by WHO ECBS that the commutability data of this study would not support the establishment of 16/190 as a WHO International Standard.. To this end, the WHO ECBS approved the establishment of 16/190 as a WHO Reference Reagent for AMH, human, recombinant. The status of the material as a WHO Reference reagent, rather than a “full” WHO International Standard, reflects the commutability considerations described above and the resulting intended use as a material to investigate assay performance. Until a physicochemical reference method is available, the AMH content of 16/190 will be defined as 489 ng/ampoule which is the robust geometric mean of the estimates from valid assays. The material is available from NIBSC (nibsc.​org).