Using PubMed keywords “(“vitamin D”) AND (“human milk” or “breast milk” or “breastmilk” [All Fields])” restricted to humans and from the year of first publication of LC–MS/MS (1987), 306 hits were returned. Only manuscripts with HM concentrations measured by LC–MS/MS were reviewed. The search was conducted in February 2023. An additional manuscript was identified that described a method for vitamin D in HM (31), resulting in 17 original manuscripts. One manuscript focused on sulphated vitamin D metabolites in serum and HM and was not considered further (32). In addition, a study that reported serum 24,25(OH)₂D₃ concentration and HM vitamin D₃ after supplementation (33) was not included since the HM data was the same as that published previously (34). Consequently, 15 manuscripts were reviewed in detail. One additional manuscript was later identified and included in the review (35). Three manuscripts described methods and provided data on a small number of convenience or commercial HM samples (31, 36, 37). Four reports provided only brief method details and data from randomized controlled trials of maternal vitamin D supplementation (34, 38, 39). The remaining manuscripts gave either full or partial descriptions of an analytical method that was applied to observational studies across different population groups at different stages postpartum and in different countries with sample sizes ranging from a few to around 100 individuals.

In the sections below, we review the generation of HM vitamin D concentrations, from the initial sample collection, through analysis and reporting.

Extraction methods for vitamin D metabolites in HM for analysis by LC–MS/MS have included combinations of some or all of saponification, protein precipitation (PP), liquid–liquid extraction (LLE) with different solvents, solid phase extraction (SPE) and semi-preparative HPLC (Table 1) and were recently reviewed (28).

Saponification and PP are used to disassociate vitamin D metabolites from proteins, including vitamin D binding protein (DBP). In addition, saponification is used to break up fat globules that may contain vitamin D metabolites. In direct comparisons, chemical PP led to 1.5 to 2-fold better recovery than saponification (31, 36), possibly due to degradation and cross reactions of compounds during saponification. Different solvents have been used for PP including acetonitrile, methanol and ethanol. Similarly, a wide range of solvents, alone in or combination were used for LLE, partly reflecting optimisation of procedures for different vitamin D metabolites and their different polarities. These have included hexane (39, 48, 53), heptane (42), isopropanol (34), methanol (55), dichloromethane (38) or a mix of hexane:ethyl acetate (9:1) (36, 41, 43).

Descriptions of sensitivity are not always included nor consistently applied and it is therefore difficult to compare sensitivity between methods. Some manuscripts specify a limit of detection (LOD), limit of quantitation (LOQ), or both, but do not always describe how the limits were defined (Table 1). Limits may be presented as units measured in the original sample volume, others in the solvent extract and others in the amount injected onto the chromatography column.

Not all studies described how results below the detection limit were handled during data analysis, a relevant factor that may influence study conclusions. Two studies reviewed here reported using assigned values of either half the detection limit (43) or LOD/sqrt(2) (42) and other studies used assigned values. Others included all the data, but either only reported a range (39) or used the values and reported that their inclusion had no effect on the study conclusions (53). One study did not report results less than the LOQ (31).

Analytical sensitivity can be improved with derivatisation, the chemical alteration of a molecule to improve mass spectrometric detection by increasing ionization efficiency and the molecular weight of the compound to reduce chemical noise. Four manuscripts (39, 41, 43, 53) from two groups reported using DMEQ-TAD (4-[2-(6,7-dimethoxy-4-methyl-3-oxo-quinoxalin-2-yl)ethyl]-1,2,4-triazole-3,5-dione) and two others, PTAD (4-phenyl-1,2,4-triazole-3,5-dione) (31, 36) (Table 1). As expected, analytical sensitivity was higher with the use of derivatisation; LOQs ranged from 0.001 to 0.1 nmol/L (31, 36, 39, 53) and LODs from 0.0003 to 0.001 nmol/L (31, 43) with derivatisation. In studies that did not use derivatisation, reported LOQ ranged from 4.3 nmol/L to 18.2 nmol/L (34, 55) and LOD from 0.025 to 2.5 nmol/L (38, 42, 48). Ferrerio-Vera et al., reported lower LOD and LOQ for underivatised samples than other studies (0.125 nmol/L for 25(OH)D₃ and vitamin D₃) (37).

Despite the knowledge that both vitamin D and 25(OH)D may contribute to anti-rachitic activity (ARA) in HM, not all studies have taken advantage of the specificity of LC–MS/MS methods to measure both metabolites; three studies only measured 25(OH)D (48, 55, 56). The concentrations of vitamin D₂ forms are considerably lower than vitamin D₃ forms (31, 36, 38, 39, 43, 53) and often less than the detection limit. However, Gomes et al. reported similar concentrations of vitamin D₂ and vitamin D₃ in HM, although mean 25(OH)D₃ concentration was double that of 25(OH)D₂ (51).

Studies have presented vitamin D concentrations as either ng/ml, pmol/L or nmol/L. Some studies reported ARA either together with concentration data or alone (35) to describe HM vitamin D content which provides the ability to combine concentrations of 25(OH)D and vitamin D. The ARA assigns higher biopotency to 25(OH)D, such that 1 IU/L ARA equals 25 pg/mL vitamin D and 5 pg/mL 25(OH)D (53). A general issue in some vitamin D literature is the unquantified use of the term ‘vitamin D’ as short-hand for 25(OH)D (e.g., (46)), which may lead to confusion where individual or both forms are of interest.

HM sampling protocols may be an important factor contributing to reported HM vitamin D content. Lipid type and content of HM is known to vary between women, and within and between feeds (13, 57) and may influence vitamin D content (42). Diurnal (58) and seasonal factors (42) also need to be considered since seasonal variation in UVB availability will affect vitamin D status and could confound patterns of vitamin D concentrations observed during gestation and lactation (59). A recent meta-analysis also suggested an association with season and infant age (30). As optimal protocols for HM collection are not established for micronutrient HM content (13, 17) it is essential to report on the protocol for HM collections. Such information, together with data on stage of lactation and milk volume, will aide in the interpretation of HM composition data. Meanwhile, new methods such as dried milk spot sampling, are being investigated (13).

The studies identified in this review demonstrate inconsistency in reporting analytical methods. Partly this reflects the range of types of publication, from very detailed methodological reports, to observational studies or randomized controlled trials. While a number of reports provided sufficient detail to replicate methods and provided data on method validation, others reported minimal method details and/or referenced primary method papers which focussed on serum rather than HM. Such shortcomings in method descriptions makes replication of, or advancements in, methods difficult, if not impossible, at best hindering progress and at worse reporting unreliable data. In a recent example, albeit a HPLC method with UV detection, scant method details are provided, there is no analytical method information and it is not clear whether 25(OH)D or vitamin D, or both were measured (22). Unless the full assay has been published previously for the same matrix, in this case HM, full details of the extraction procedure are required, including starting volume of HM and volumes of chemicals used in the extraction procedures. Chromatography and mass spectrometric parameters should also be included.

While it may be unrealistic to expect a full validation of all methods to a clinical or internationally recognized standard, reporting of certain basic data are required to assess the quality and reliability of the data. This includes LOD/LOQ and a description of how values below these limits are handled; such considerations are frequently under-appreciated and often not discussed in nutritional biomarker literature (60). In addition, assay precision, extraction recovery and how they were derived should be detailed. This is important because in some methods, although isotopically labeled standards were used to correct for recovery, these were not always included from the start of extraction. The ability to include online supplementary material negates word count restrictions and allows the inclusion of full details of analytical methods and method validation parameters. Reporting of data, as discussed by others (17), needs to include actual numbers with mean or median and a measure of the variance, including the range; graphical representation alone is not sufficient. As above, the inclusion of online supplementary material allows summary data in to be included without duplication of results in the main manuscript. Overall, absence of analytical method detail means that study quality cannot be assessed, and consequently data may be excluded from collation for reviews and policy making decisions (17).

Analytical sensitivity is an important challenge for HM analyses because of the observed low concentrations of vitamin D metabolites. This review highlights a wide-range of LOQ, from <1 pmol/L to 18 nmol/L. The advantage of the lower LOQs is likely increased precision and more reliable quantitation at the lower concentrations. In methods with poor sensitivity, many values can be near or below the reported limits. In some cases, reported concentration ranges include values below the LOQ but the precision of these values is not indicated. This is particularly problematic where the LOQ is high, e.g., Oberhelman et al. report an LOQ of 18.2 nmol/L and LOD of 2.5 nmol/L (34). While higher LOQs raise the question on the usefulness and reliability of the method, the requirement of very low values may also be unnecessary, or at least, not physiologically relevant. If the vitamin D₃ concentration of HM was 1 nmol/L (0.38 ng/mL) then 800 mL of HM would provide 0.31 μg/d of vitamin D₃, 30-times less than current US recommendations for infants (400 IU/d (10 μg/d)) (61). The question also arises of how to handle data below the LOQ, particularly when the LOQ is high. Approaches for handling values below the LOD/LOQ include excluding values, use of a surrogate / imputed value (e.g., half the LOD/LOQ, LOD/sqrt (2) (59)) and multiple imputation. Exclusion of data is not recommended since non-detectable values may provide as much information as concentrations that can be measured.

A limitation of all the HM vitamin D assays is the ability to establish accuracy. Over the last two decades, considerable effort has been made to improve the accuracy of serum assays for 25(OH)D. This has involved the development of reference measurement procedures (RMP) and certified reference materials (e.g., NIST Standard Reference Materials), as well as external quality assurance schemes (e.g., Vitamin D External Quality Assurance Scheme, DEQAS), and was only possible with considerable time and resources, such as can be afforded to a biomarker of such importance for clinical and research use. Similar resources are unlikely to be available for HM vitamin D analysis, thus the true value will be difficult to establish. Certified reference materials are available for formula milk with assigned vitamin D_2/3 and/or 25(OH)D₃ values (Table 4) and may be used to establish some degree of confidence in an assay. However, these preparations are unlikely to be representative of HM. Similarly, while accuracy of chemical solutions can be demonstrated, this is unlikely to be representative of accuracy in the biological matrix. The standard addition method may be used to estimate accuracy, but similarly, if the added standard does not behave in the same way as endogenous compound (due to for example, location within lipoproteins, protein binding etc.) then this approach will not provide a true measure. This demonstrates the need to provide as much information as possible about a method, including precision or repeatability, in order to provide confidence in the chosen approach. In house quality control materials to monitor assay performance over time are likely required. Despite the relative ease of HM collection and less restriction on volume than for example, whole blood or serum, commercial HM is not readily available and may be expensive. Access to HM from breast milk banks is possible but can require an extensive ethical approval process, which may not be feasible for method development or quality control material purposes.

In future, researchers should explore the sharing of matrix-matched, internal quality control materials to explore agreement between methods. In this incremental fashion, some consensus may be reached on true value for a given sample and, in the absence of reference method procedures and other high-level approaches, go some way to standardizing measurement, or at least provide a better understanding of the bias between methods.

The data presented here suggest a potential wide range of vitamin D concentrations in HM. In the majority of studies, 25(OH)D₃ and vitamin D₃ concentrations were in the low nanomolar range (<10 nmol/L), similar to concentrations reported using earlier methods (62). However, some studies suggested considerably higher 25(OH)D₃ (55) and vitamin D₂/₃ concentrations (37). Such values appear inconsistent with the majority of studies and may be due to short-comings in analytical methods. Two other recent studies have described similarly high concentrations. One study utilized a HPLC method and reported the unlikely situation where the 2(OH)D₃ concentration was 30 nmol/L higher in HM than in serum in both vitamin D-supplemented and unsupplemented mothers (63). The second study used an automated chemiluminescence analyzer validated for serum but not for HM and described high analytical limits (18 nmol/L) (64, 65). These assays may have suffered from inadequate specificity and cross-reactivity leading to spurious elevated concentrations.

Conversely, a number of other studies reported unmeasurable, zero concentration (46), or unquantifiable vitamin D metabolite concentrations (34) mostly likely due to poor sensitivity, and/or poor extraction recovery. These large differences in concentrations may be due to the many challenges of measuring vitamin D in HM rather than true biological differences. Furthermore, these studies demonstrate that methods for the extraction and analysis of vitamin D metabolites from serum or plasma are unlikely to be suitable when directly applied to the different matrix of HM. Whereas inter-individual differences in fasted serum or plasma composition are relatively small, HM can be highly heterogeneous in respect of fat and protein content and have a different composition across a feed, a day, and across time, that requires a robust extraction methodology to reliably and consistently extract vitamin D metabolites.

The relative importance of different forms of vitamin D in HM is not certain. Not all studies measured both forms, and 25(OH)D₃ and vitamin D₃ are generally not correlated (41). Vitamin D₂ forms do not make a meaningful contribution to the vitamin D content of HM except through significant maternal vitamin D₂ supplement intake. Although an early study indicated 25(OH)D₃ concentrations were 8-fold higher than vitamin D₃ concentrations (23) a subsequent study demonstrated greater vitamin D₂ or vitamin D₃ concentrations than 25(OH)D₂ or 25(OH)D₃ concentrations in HM after supplementation and/or UV exposure. The exception was for the low dose (10 μg/d vitamin D₃) participants where 25(OH)D₃ was more than two-times greater than vitamin D₃ (52). In a case study of a women taking 100,000 IU of vitamin D₂ daily during pregnancy, which led to maternal vitamin D₂ and 25(OH)D₂ serum concentrations of more than 1,300 nmol/L, vitamin D₂ concentrations in HM were over 300 nmol/L, but 25(OH)D₂ concentration was around 20 nmol/L (66). Other studies have stated vitamin D is a greater contributor to ARA than 25(OH)D although metabolite concentration data are not reported (67, 68). Thus, it appears, at least when there is large single bolus dose or regular supply of vitamin D (from supplements or generated through cutaneous synthesis after UV exposure), vitamin D rather than 25(OH)D becomes the major vitamin D form in HM. Consequently, there is much greater variability in vitamin D concentration than 25(OH)D concentration in HM. Evidence from the studies reviewed here are consistent with this notion as there is greater variance in vitamin D₃ concentration than 25(OH)D₃ concentration due to UVB exposure (53) and maternal vitamin D supplementation (38, 39). Conversely, a recent meta-analysis concluded that 25(OH)D is the major contributor to HM vitamin D content (30). However, the meta-analysis included all studies with measured concentrations of 25(OH)D and/or vitamin D without consideration of the reliability of different methods. The study also reported a significant difference in 25(OH)D concentration between analytical methods (LC–MS/MS, HPLC with competitive binding assay, or “Other”) with the “Other” group reporting ~10 times higher 25(OH)D concentration than chromatography-based methods (30), potentially skewing results toward 25(OH)D. Analytical factors related to recovery during extraction, matrix effects and sensitivity may also affect interpretation when comparing between studies. Overall, under conditions of low vitamin D supply, either or both forms can contribute to HM vitamin D content as demonstrated by the studies reviewed here (41, 43). The higher biopotency of 25(OH)D suggests that both forms should be measured in HM. Vitamin D may be more important than 25(OH)D where supply is increased through UVB exposure and/or supplementation. The importance of the parent form of vitamin D to HM total vitamin D content may be related to relatively lower binding affinity vitamin D compared with 25(OH)D. As a result, there is a greater level of free circulating vitamin D that may diffuse across cell membranes into HM. This is in contrast to 25(OH)D that predominantly circulates bound to DBP and that relies on this binding for cellular uptake through the megalin-cubulin pathway (69, 70).

The manuscripts reviewed here show that vitamin D supplementation during pregnancy or lactation can improve the vitamin D content of HM (34, 35, 38, 39), and infant vitamin D status (34, 35). In an earlier study, Wagner et al. provided mothers with either 400 or 6,400 IU/d from 1 month post-partum. Milk ARA did not increase in the 400 IU/d group, but rose steadily in the 6,400 IU to reach a mean ARA of 874 IU/L after 6 months, and mean infant 25(OH)D₃ concentration of 115 nmol/L, a value very similar to that achieved with infant supplementation of 300 IU/d (108 nmol/L) in the same study (67). Similar findings were subsequently reported from a larger randomized controlled trial with maternal doses of 6,400 IU in the USA, whereas the same study found that maternal doses of 2,400 IU did not support infant vitamin D status (62). A recent meta-analysis of maternal vitamin D supplementation and its effect on milk vitamin D content and infant vitamin D status reported a significant increase in milk ARA with vitamin D supplementation and a prediction equation of “ARA = 0.099 × vitamin D dose (IU/D) -30” (71). In addition, the authors calculated that for each 1,000 IU of maternal vitamin D supplementation, infant 25(OH)D concentration increased by 6.8 nmol/L (71). These data suggest that lower doses of maternal vitamin D supplementation are unlikely to provide a milk vitamin D content sufficient to meet a vitamin D intake of 400 IU/d (72, 73). However, it is also uncertain whether the higher doses used in some studies are necessary and thus the maternal supplementation dose required to maintain adequate infant vitamin D status is yet to be confirmed (71, 74, 75).

This review has focussed on 25(OH)D and ergocalciferol (vitamin D₂) and cholecalciferol (vitamin D₃) however, other vitamin D metabolites may also potentially contribute to total vitamin D content. Early reports suggested vitamin D sulphate, a water-soluble conjugate of vitamin D, may be a significant source of vitamin D (76, 77). Although considered biologically inactive (58), hydrolysis of the conjugated form may liberate biologically active metabolites (32). However, non-specific methods and instability of the sulphated form during saponification have been identified as limitations to these methods and subsequent reports with greater specificity suggest the sulphated form is not a major contributor to the vitamin D content of HM (78–81). More recently, a LC–MS/MS method that includes sulphated forms has been developed (32) with mean values for sulphated forms between 87 and 261 pmol/L, comparable to lipid-soluble vitamin D metabolites, but no other data have since been published.

Metabolites downstream of 25(OH)D (1,25(OH)₂D and 24,25(OH)₂D) are unlikely to meaningfully contribute to total milk vitamin D content since in serum, these metabolites circulate at around 1 and 10% of 25(OH)D concentration, respectively. Data from cows milk, even after supplementation, suggest none at all, or insignificant concentrations of 1,25(OH)₂D and 24,25(OH)₂D. The LC–MS/MS method of Gomes et al. included 1,25(OH)₂D₃ and 24,25(OH)₂D₃ and mean concentrations in human milk samples used in the validation of their method and with concentrations greater than the LOQ, were 1.14 and 7.68 pmol/L, respectively (31). Hollis earlier reported concentrations of 13, 135 and 83 pmol/L for 1,25(OH)₂D₃, 24,25(OH)₂D₃ and 25,26(OH)₂D₃, respectively, although these were in the context of relatively high milk 25(OH)D₃ of 780 pmol/L (23). In other reports, 1,25(OH)₂D₃ and 24,25(OH)₂D₃ were not detected (82, 83). Consistent with the data reviewed in this manuscript for 25(OH)D and vitamin D, the limited data available for dihydroxy metabolite concentrations in milk are not consistent, likely due to small sample sizes, different populations with varying vitamin D status, and different analytical methods with varying degrees of specificity and sensitivity. A further compound of potential relevance to the vitamin D content of HM is the C3-epimer of 25(OH)D (3-epi-25(OH)D), an isomer of 25(OH)D. Although the biological relevance of the 3-epi-25(OH)D remains unclear, if it is not chromatographically separated from 25(OH)D then estimates of vitamin D status in serum can be inflated. In particular, it is known that the 3-epi-25(OH)D is generally higher in neonates than other age groups peaking at up to 3 months of age, before declining, and can significantly contribute up to infant status (84–86). The origin of high 3-epi-25(OH)D in neonates is not certain, but appears unlikely to be from the maternal circulation as maternal concentrations are generally similar or lower than those observed in infants and evidence suggests the 3-epi-25(OH)D does not readily cross the placenta (87). While a maternal contribution through lactation cannot be ruled out, particularly where 3-epi-25(OH)D may be increased due to vitamin D supplementation, current thinking is that 3-epi-25(OH)D compounds are a product of immature vitamin D infant metabolism in early life (86, 87). There is little evidence to support appreciable amounts of 3-epi-25(OH)D in HM. Of the studies reviewed here, four stated that 3-epi-25(OH)D was chromatographically resolved (31, 34, 36, 55), and of these, two included validation of 3-epi-25(OH)D quantitation in HM (31, 55), although one presented no concentration data (31). The study that reported disparate values for 25(OH)D₃ concentration in HM, also reported extremely high 3-epi-25(OH)D₃ concentrations in HM, equal to around one-third of the HM 25(OH)D₃ concentration, although maternal plasma 3-epi-25(OH)D₃ was below the LOQ (55).

As highlighted by the range and often multi-stepped procedures for vitamin D measurement in HM, the suitability of a method designed for serum 25(OH)D measurement is questionable. The unique matrix of HM, with high lipophilic compound content, as well as the relevance of both 25(OH)D and vitamin D compounds with differing polarities, means that methods specifically designed for milk are required. The analysis of vitamin D in HM remains challenging, in particular the essential but laborious extraction techniques. Protein precipitation, liquid–liquid extraction and/or saponification remain the major extraction techniques to separate vitamin D from the main other lipophilic compounds in HM. In future, advances in solid phase extraction (SPE) and supported-liquid extraction may provide simplified approaches to purification of vitamin D metabolites (25). Developments in chromatography, including supercritical fluid chromatography, as used by Oberson et al., to measure vitamin D metabolites in HM, purportedly can provide a shorter run time and better chromatographic separation than liquid chromatography (36). The apparent low endogenous concentrations of vitamin D metabolites in HM and/or poor extraction efficiency, often leads to the requirement for large sample volumes for extraction. As discussed above, chemical derivitisation can improve sensitivity and ionization efficiency but only two types have reportedly be used for vitamin D detection in HM, whereas a larger number of derivitisation reagents have been applied to vitamin D metabolites in serum (25). Finally, further improvements in vitamin D analysis may be obtained from developments in mass spectrometer sensitivity and specificity, such as provided by high resolution techniques, in particular hyphenated to liquid chromatography.