1.1 Vitamin D assay standardization: an update
Laboratory standardization of vitamin D is a necessary element in developing consensus regarding the serum 25-hydroxyvitamin D (25(OH)D) levels to define hypovitaminosis D [
1‐
3]. Standardization is the process whereby all laboratories and assays are brought into alignment with the “true concentration” based on gold standard reference measurement procedures and certified reference materials [
4‐
6]. Failure to utilize standardized 25(OH)D data is a major contributor to confusion surrounding vitamin D status [
7].
Unfortunately, the vast majority of studies published to date include unstandardized 25(OH)D data. Despite the existence of performance testing/external quality assessment schemes (PT/EQA) - e.g. The Vitamin D External Quality Assessment Scheme (DEQAS)- there are still issues that need to be addressed. It is only since the development of certified 25(OH)D reference measurement procedures [e.g. US National Institute for Standards and Technology (NIST) [
8]] and the introduction of The Vitamin D Standardization Program (VDSP) has it been possible to evaluate assay variation in an unbiased way [
9]. This led to converting DEQAS [
10] to an accuracy-based survey where standardized target values were assigned to serum samples using NIST reference measurement procedures [
10]. Recent data from DEQAS have shown that such unstandardized assays are subject to significant assay variation over time [
11,
12].
It has recently been shown that assays with or without standardization can lead to radically different results in individual studies [
13]. In contrast, some studies in which assays were well calibrated originally, retrospective or after the fact, standardization had only a small effect. Good examples of these are the Canadian Health Measures Survey [
14] and three German national surveys [
15] which used the DiaSorin Liaison assay. In the three German surveys, the prevalence of 25(OH)D <30 nmol/L fell from 24% to 16%, 30% to 15% and 27% to 12.5%, respectively, after VDSP standardization. These results were to be expected based upon DEQAS data that previously showed that the DiaSorin Liaison assay reads low. However, in the Canadian Health Measures Survey, the prevalence of levels <30 nmol/L did not change after standardization [
14]. These results lead to two conclusions: 1) the same commercial assay can have very different results depending on the laboratory and 2) because of the very recent development of the NIST reference measurement procedure and the VDSP, and retrospective standardization, it is unclear which data collected in the past were properly calibrated. Therefore, it is important to standardize all national surveys and 25(OH)D results from key studies.
Meta-analyses can suffer the same problem. However, little effort has been made to use only standardized data in meta-analyses. In the one example where standardized data were included, it was falsely claimed that standardization was not important. In an essay by Cashman et al., new results from a previously published paper [
16] were included to evaluate the vitamin D intake necessary to meet the needs of 97.5% of the population to reach a 25(OH)D concentration of 50 nmol/L. In the re-analysis, it was reported that the value was 28.8 μg/day after VDSP assay standardization and 28.4 μg/day using unstandardized 25(OH)D data [
17]. There are two possible explanations for the results: 1) the originally measurements correctly calibrated to VDSP guidelines or 2) this was the result of a series of errors cancelling each other out.
Related to assay standardization there is another potential problem. Certain immunoassays do not function properly in specific physiological/pathophysiological states. For example, due to high vitamin D binding protein concentrations some immunoassays yield inaccurate 25(OH)D results in pregnant women [
18]. A recent paper illustrates the problems with using an untested immunoassay for measuring serum 25(OH)D in pregnant women [
19]. The authors sought to assess the association between 25(OH)D measured at baseline (preconception), time to achieve pregnancy, at 8 weeks’ gestation and live birth outcomes. The interassay CV of the immunoassay was 15.8% at a mean concentration of 37.7 nmol/L, and 13.1% at 103.8 nmol/L for lyophilized manufacturer's controls and 17% for an in-house pooled serum control. The assay was not only poorly calibrated and its poor performance was not noted. Given the fact that bias can be associated with using immunoassays to measure 25(OH)D in pregnant women [
20], the results cannot be easily interpreted. In these circumstances, 25(OH)D should have been measured at baseline (preconception) and at 8 weeks’ gestation with a VDSP standardized liquid chromatography-tandem mass spectrometry (LC-MS/MS). As a result, it is essential that researchers verify that their immunoassay of choice will function properly in the physiological/pathophysiological state under study by first determining that the immunoassay will yield results comparable to a VDSP certified standardized LC-MS/MS assay.
Vitamin D research data are used to develop government policy including public health/clinical guidelines. The utility and success of these guidelines depend on utilizing very accurate and precise 25(OH)D measurements. Only VDSP standardization can provide assurance that the data used to develop public policy are of the very highest quality because using unstandardized data can have long-terms consequences.
An example is the original source data used to define the 25(OH)D concentration at the lower limit of adequacy in the UK as 25 nmol/L [
21]. The source was a 1976 paper with 9 cases of nutritional rickets and a range of 25(OH)D concentrations from 20 nmol/L to 54.9 nmol/L [
22]. The performance characteristics of the assay were unclear and the cut-point of 10 ng/mL were not justified. Unfortunately, these concerns were not evident then, but government guidelines were set. Once government guidelines are set, they are very difficult to change. This has led to an impasse in which subsequent review panels have not been able to recommend a change in the "cut-points" despite the shortcomings of research database [
19,
23,
24]. Thus, a lack of 25(OH)D standardization is at the root of the uncertainty regarding the definition of a vitamin D threshold for rickets, as well as for defining the entire clinical spectrum from deficiency to toxicity. Continuing to publish studies with flawed 25(OH)D data will perpetuate uncertainty about how to define vitamin D deficiency, adequacy, and toxicity.
There are two types of assay standardization: 1) prospective and 2) retrospective. Prospective standardization is the process whereby the initial measurements for a study are standardized using VDSP guidelines. Retrospective standardization of previously measured study samples is possible when appropriately stored serum samples are available [
25‐
27]. In the discussion above, data from the Canadian Health Measures Surveys, the three German national surveys and the meta-analysis data cited by Cashman et al. were all retrospectively standardized. Three examples of prospectively standardized studies are NHANES 2011-2014 [
27], Australian Health Survey [
28], and a recent study on determinants of QuantiFERON-diagnosed tuberculosis infection in Mongolian schoolchildren [
29].
1.2 Threshold for defining vitamin D deficiency/insufficiency with current methods
An evidence-based consensus regarding the 25(OH)D concentration used to define hypovitaminosis D is needed. In the absence of compelling data, at this time, 25(OH)D values below 30 nmol/L should be considered to be associated with an increased risk of rickets/osteomalacia, whereas 25(OH)D concentrations between 50 and 125 nmol/L appear to be safe and sufficient in the general population for skeletal health [
6,
30]. If we consider 30 nmol/L as the threshold for hypovitaminosis D, the question is whether we can apply this threshold using every current method (mostly automated immunoassays)? Comparisons among and between methods show a bias of several immunoassays compared to LC-MS/MS methodologies [
31‐
34].
Within a given methodology, there are several possible causes for differences, such as lot-to-lot variation in manufacturer reagents or differences in subjects included in different studies. This last possibility leads to another important issue: if we can calculate a method-specific threshold, does this method-specific threshold hold for every group of subjects/patients? Several studies show this is probably not the case, as most immunoassays show matrix specific interference found in some physiological (e.g. pregnancy or ethnicity) and pathophysiological (e.g. intensive care unit, osteoporotic and haemodialysis patients) states. This latter point truly confounds one’s ability to compare levels across different clinical settings, even if the methodology is standardized and optimized.
In the recent method comparison studies described above, serum samples from different kinds of subjects were also used. Analysis of these separate groups shows rather large differences among the different groups of subjects.
Thus, if we consider 30 nmol/L as the threshold for hypovitaminosis D, we should recalculate this threshold for the various currently used automated immunoassays. This is complicated by matrix specific interference, as previously mentioned. All this reinforces the points made earlier: acquiring accurate and precise measurements of 25(OH)D in vitamin D research requires evidence that a selected immunoassay functions comparably to LC-MS/MS in the physiological/pathophysiological state under study and that the measurements be VDSP standardized.
1.3 Threshold for defining vitamin D excess
All major agencies that promulgate nutritional guidelines have made recommendations, which also include tolerated upper levels of intake (TUL or UL). The current consensus for the UL for vitamin D in normal healthy individuals, represented by the Institute of Medicine 2011 recommendations [
24], is 4,000 IU/day (100 μg/day). This value is based solely upon the consideration of vitamin D’s actions to regulate calcium and phosphate homeostasis. This UL may not be accurate in consideration of the putative non-calcemic functions of vitamin D. At steady state, an intake of 4000 IU/day corresponds to a mean serum 25(OH)D level of about 125 nmol/L in a normal healthy subject. It should be noted that the Endocrine Society guideline [
35] for those requiring vitamin D therapy for various disease states recommends a UL of 10,000 IU/day (250 μg/day). A disease state, though, is not the same as a healthy state and thus one cannot forecast what the steady state level would be with levels of intake as high as 10,000 IU per day for a given disorder. At its extreme, someone who suffers with a malabsorption syndrome may require amounts even greater than 10,000 IU per day in order to maintain a reasonably normal level of 25(OH)D.
Authoritative agencies have concerns about vitamin D toxicity from long-term, moderate dosing (chronic toxicity) as well as short-term, high-dose therapy (acute toxicity). Consequently, the ULs and the corresponding serum 25(OH)D levels achieved have become more conservative [
24].
Limited human studies involving acute toxicity, mainly from anecdotal reports regarding accidental overdosing with vitamin D3 [
36,
37], suggest that doses of over 10,000 IU/day and serum 25(OH)D levels of around 250 nmol/L can be tolerated in the short term. These studies usually involve minimal data prior to the appearance of the toxicity and consequently, the events which trigger hypercalcemia remain obscure. Of course, well-documented human trials examining the effect of excessive doses of vitamin D are ethically impossible so that there is a paucity of evidence regarding the exact mechanism by which vitamin D intake causes toxicity. From published reports, though, it is clear that vitamin D toxicity can occur via ingestion of over the counter products such as Soladex, that contains over 800,000 IU of vitamin D per dose [
38]. Genetic disorders in which vitamin D is not normally metabolized can lead to excessive amounts of 1,25(OH)2D, despite normal intake of vitamin D [
39].
Physicians providing vitamin D to patients should be reminded that there are genetic and acquired diseases involving dysregulated vitamin D metabolism that can alter vitamin D intake requirements [
34]. In addition to malabsorption syndromes noted above, obesity can be associated with vitamin D being sequestered in fat tissue. In chronic kidney disease (CKD), impaired activation of vitamin D presents the need to provide active metabolites. Similarly, in advanced liver disease, inability to hydroxylate cholecalciferol requires vitamin D forms that are active. Genetic disorders in which vitamin D is not normally metabolized can lead not only to excessive production of active vitamin D and hypercalcemia or kidney stones or nephrocalcinosis [
39,
40], as noted above, but also to excessive metabolism of vitamin D by activation of cytochrome P450 3A4 (CYP3A4) resulting in rickets or osteomalacia [
39,
41‐
43].
With these many uncertainties, it is difficult for agencies to recommend a specific IU, but most have settled on 4000 IU/day as a safe upper intake level for vitamin D [
24]. However, the range of vitamin D intake of between 4,000-10,000 IU/day [
24,
35] will probably remain as a useful safe buffer zone that physicians can use in the short term and will not result in serum 25(OH)D levels of over 250 nmol/L.
Again, this advice relates to normal, healthy individuals. To apply this advice to some of the conditions described in this section will lead to gross undertreatment of patients who need considerably higher daily doses of vitamin D. This standard advice also does not pertain to individuals who are grossly vitamin D deficient, in whom there may well be an indication to increase levels quickly by using higher doses in the short term.
1.4 Should we still be prescribing ergocalciferol?
The two parent forms of vitamin D, namely ergocalciferol [vitamin D2 or 25(OH)D2] and cholecalciferol [vitamin D3 or 25(OH)D3] are widely available and used commonly. In the United States, there is no over-the-counter form of vitamin D2 or D3 in the 50,000 IU dosage; in the prescription form at that dose, only vitamin D2 is available in the USA [
35]. There are methodological challenges to the measurement of 25(OH)D when immunoassays are used related to the coexistence of both circulating 25(OH)D2 and 25(OH)D3. Specifically, immunoassay antibodies may not detect 25(OH)D2 and 25(OH)D3 equally, and the proprietary releasing agent in these automated assays to free 25(OH)D from vitamin D binding protein may not liberate 25(OH)D3 and 25(OH)D2 equally.
To explore these considerations, a pilot study was performed in which residual plasma was collected from routine clinical laboratories [
44]. Sample pools containing predominantly 25(OH)D3 from 20-255 nmol/L and 10 with 25(OH)D2 from 35-197.5 nmol/L were prepared. Eight independent laboratories analysed these pools using their routine 25(OH)D immunoassays. Data for five FDA-approved automated methods were compared to total 25(OH)D determined by an High Performance Liquid Chromatography (HPLC) assay calibrated to NIST assigned values. These pooled specimen results provided regression equations for total 25(OH)D using the various immunoassay methods. When the results are considered based on the primary form present, i.e. 25(OH)D3 or 25(OH)D2, agreement with HPLC results dramatically improved for 25(OH)D3.
To improve immunoassay accuracy, we suggest focusing on 25(OH)D3 to harmonize commercial immunoassays, since 25(OH)D3 is naturally produced and is the dominant supplement form worldwide.
In addition to causing potentially insurmountable immunoassay challenges, we believe that high-dose ergocalciferol is not the best clinical approach to vitamin D repletion. Cholecalciferol supplements (even 50,000 IU) are widely available at low cost. Thus, we believe the common clinical practice of treating vitamin D deficiency by prescribing high-dose ergocalciferol is no longer best clinical practice. Therefore, we suggest that the first-line of treatment is cholecalciferol where possible and that ergocalciferol only is used for vegans and in other patients opposed to using cholecalciferol. However, it must be remembered that when ergocalciferol is used monitoring of 25(OH)D levels will require measurements made with VDSP-standardized HPLC or LC-MS/MS assays.
1.5 Age as a specific threshold determinant in the general population
It is well established that advancing age reduces skin capability to synthesize pre-vitamin D3 [
45]. Moreover, the prevalence of skin cancer in older adults has reached "epidemic" proportions with a resultant array of recommendations (including a Surgeon General's report) advising avoidance of skin exposure to the sun [
46]. As such, it could be expected that older adults would have poorer vitamin D status. Indeed, higher 25(OH)D levels have been reported in children [
47]. Moreover, the Institute of Medicine appears to have acknowledged this point establishing a Recommended Dietary Allowance (RDA) of 800 IU/day for those > age 70 years [
48], higher than the amount recommended for younger populations (600 IU/day). Similarly, the International Osteoporosis Foundation recommends a higher average vitamin D intake of 800-1,000 IU for older adults [
49].
These recommendations are predicated on the expectation that older adults are more likely to be vitamin D deficient. Some studies have reported lower circulating 25(OH)D levels with advancing age [
49,
50]. For example, a meta-analysis of 33,000 subjects (using unstandardized 25(OH)D data) found those aged >75 years to have 25(OH)D values on average 9 nmol/L lower than those aged 65-75 years. However, a more robust approach using a representative sample (NHANES 2007-2010) and, importantly, standardized 25(OH)D data found no evidence to support lower 25(OH)D concentration in older adults overall or when stratified by race/ethnicity [
47]. Indeed, those age >60 years had higher 25(OH)D values than those aged 40-59 years in the entire cohort, including Hispanics and non-Hispanic blacks. Consistent with this, the prevalence of "low" vitamin D status (using 50 and 75 nmol/L as cut-off points) was numerically lower in those aged >60 years than among adults age 40-59 years among all race/ethnic groups [
47]. Additionally, multiple studies find no effect of age on response of 25(OH)D to oral vitamin D supplementation [
51].
Similar to NHANES, data from the national surveys in Finland, Ireland, Germany and Canada do not find dramatic differences in standardized serum 25(OH)D or vitamin D inadequacy prevalence (defined as a 25(OH)D < 50 nmol/L) based upon age [
14,
15,
52,
53]. However, institutionalized older adults may be at higher risk for vitamin D deficiency, presumably due to limited sun exposure and inadequate supplementation [
54,
55].
Thus, despite expectations that older individuals have lower levels of 25(OH)D on average, age as a specific determinant does not seem to be a key factor. It remains to be seen whether age could contribute to vitamin D status when other risk factors are also present.
Consensus Statements:
1.
Existing data are insufficient to define “low” or “high” vitamin D status thresholds with any degree of certainty because of the lack of standardized 25(OH)D measurements in vitamin D research.
2.
The current approach to defining vitamin D status using circulating 25(OH)D concentration with standardized state-of-the-art methodology is recommended.
3.
Due to assay variability, circulating “25(OH)D” as measured by the multitude of existing assays cannot simply be blindly pooled into meta-analyses. Meta-analyses should not be conducted including studies that use non-standardized assay methodology.
4.
For research and for publication of data, 25(OH)D assays should demonstrate standardization or alignment with reference methodology along the lines proposed by the VDSP.
5.
Laboratories should participate in a 25(OH)D accuracy-based proficiency testing program, e.g. DEQAS or College of American Pathologists (CAP).
6.
Some documentation that the 25(OH)D assay methodology functions properly in the setting being studied, (e.g. pregnancy, hemodialysis) is needed. This can be accomplished by comparison of the assay method being used with a standardized method in the physiologic condition being studied. Given existing assay deficiencies assay manufacturers should develop assays that have comparable ability to accurately measure 25(OH)D2 and 25(OH)D3 in various clinical circumstances.
7.
It seems reasonable to recommend that cholecalciferol rather than ergocalciferol is used for vitamin D supplementation for most people.
8.
The risk of developing rickets/osteomalacia is increased at a 25(OH)D concentration of ≤ 30 nmol/L. This threshold may vary depending on other conditions such as calcium and phosphate nutrition, parathyroid hormone (PTH) levels, and season.
9.
The 25(OH)D concentration ranges among normal subjects between approximately 50-125 nmol/L [
56,
57]. With admitted uncertainty, an upper 25(OH)D threshold of 125 nmol/L is advisable.
Research agenda
1.
Determine whether threshold values for 25(OH)D (both low or high) are applicable under various clinical conditions.
2.
Develop a reference measurement procedure and reference materials for free 25(OH)D assessment.
3.
Determine the added value of free 25(OH)D measurements in the assessment of vitamin D status.
4.
Evaluate the utility of the vitamin D metabolome in various physiologic and pathologic conditions.
5.
Determine variables that affect the utilization (absorption/metabolism/biologic action/transport/storage) of vitamin D.
6.
Determine whether vitamin D status (i.e. “sufficient”, “insufficient” and “deficient” levels) influences supplementation adherence/persistence and outcome.