Background

Vitamin D is an essential nutrient for bone health [1] and possibly for other extra-skeletal health outcomes [1, 2]. The main sources of vitamin D are dietary intake and dermal synthesis during sunlight exposure [3]. Vitamin D (along with calcium and zinc) has been prioritized by the Food and Agriculture Organization (FAO) together with the World Health Organization (WHO), as part of the update of their 2004 nutrient requirements for children aged 0–3 years [4, 5]. Dietary Reference Values (DRV) for vitamin D, as estimates of the dietary requirements for the vitamin, are crucial from a public health perspective in providing a framework for the prevention of vitamin D deficiency and optimizing vitamin D status of individuals [6]. With the vitamin D DRV update in mind, a recent FAO-WHO-commissioned systematic review and meta-analysis evaluated circulating 25-hydroxyvitamin D (25OHD), parathyroid hormone and other newer potential biomarkers of vitamin D status (such as free and bioavailable 25OHD, 24,25-dihydroxyvitamin D, C3-epimer of 25OHD, and vitamin D3) in terms of their use in defining dietary requirements for vitamin D in young children [7]. The systematic review concluded that circulating 25OHD is a robust and reliable marker of vitamin D status in infants and children [7].

In setting DRVs for vitamin D, there is a need to clarify the relationship of serum 25OHD and the reference level of the critical indicator(s) of health outcomes for nutrient adequacy, taking into consideration sex, life-stage and vulnerable groups [8]. This serum 25OHD threshold, in turn, is used to establish the recommended vitamin D intake which maintains a stated percentage individuals above this threshold, and thus ensuring adequacy. For infants and children, the FAO-WHO prioritized the risk of rickets as the critical indicator amongst other skeletal and extra-skeletal health outcomes [9]. Rickets is a softening and weakening of bones at the growth plate, which can lead to painful and long-term health consequences [10], including potentially life-threatening complications [11]. It can be diagnosed based on clinical signs, biochemical tests and radiographies [10]. Several authorities and expert bodies have established vitamin D recommendations that indicate a minimum recommended serum 25OHD level, based on minimizing the risk of developing rickets in children, or osteomalacia in adults [3, 12]. However, there is a lack of consensus on this minimum 25OHD threshold, with values varying from 25 up to 50 nmol/L (see Table 1). Differences between these recommended serum 25OHD thresholds could be explained by differences in the body of evidence considered, variability in the vitamin D assays [13], and the characteristics of the populations, such as calcium intake [14] and sun exposure [15].

Table 1 Internationally reported minimum serum 25OHD thresholds to ensure adequate bone health used to set vitamin D DRVs in children (ordered by increasing serum 25OHD threshold)
Full size table

The present systematic review and individual participant data (IPD) meta-analysis was commissioned by the FAO-WHO with the key objective of determining a serum 25OHD threshold, based on the risk of rickets, to inform the setting of the vitamin D DRV for young children. In particular, emphasis was placed on the determination of a serum 25OHD threshold in the setting of adequate dietary calcium intake. This is important because of the DRV convention that setting a vitamin D intake requirement is based on the assumption that the intake of calcium and all other nutrients is adequate [3, 12, 16]. Of note, other authorities and expert bodies thus far were unable to include this aspect in their consideration of serum 25OHD thresholds.

Methods

The present systematic review and meta-analysis, including IPD analyses, follows the guidance provided as part of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA)-IPD statement [17]. Approval by a research ethics committee to conduct the IPD meta-analysis was not required because the aim of this secondary analysis was consistent with the ethical approval received for the individual studies. The current analysis was conducted on anonymized data.

Eligibility criteria

Studies involving generally healthy (apart from rickets) children below 4 years of age with total serum 25OHD levels (referred to as 25OHD henceforth) and radiologically confirmed active rickets were included. Studies in which the presence or absence of rickets were diagnosed only clinically or biochemically, but not radiologically, were excluded (to lower risk of misdiagnosis). Children 4 years and above or with conditions, such as low birth weight, prematurity, hereditary rickets, vitamin D resistant or dependent rickets, were excluded. If only serum 25OHD2 or 25OHD3 was measured, the study was excluded. The following study designs were included: cross-sectional, cohort, case–control, case report, case series, surveillance studies, before-after studies, and trials. Conference abstracts, systematic reviews, commentaries, and editorials were excluded. There were no restrictions related to the geographical location or the language.

Search strategy

A systematic search of Embase was conducted on 7 June 2022. The search strategy is shown in Appendix 1. The search was supplemented with a manual screening of the reference lists of included articles, reviews and key international vitamin D DRV reports from other authoritative bodies [3, 12, 18,19,20,21]. Study selection was conducted in duplicate by two reviewers.

Data collection processes, data items, IPD and data protection

Information on the characteristics of the study and their participants, the 25OHD measurement methods, as well as the method of estimation of calcium intake were extracted by one reviewer and verified by a second reviewer. Aggregate- and individual-level data (where available) for serum 25OHD and calcium intake were extracted. For before-after studies and trials with vitamin D supplementation, only the baseline data were extracted.

In the case of those identified priority studies that reported having measured calcium intake as well as serum 25OHD, collaboration, in the form of IPD sharing, was requested. The authors of each study were contacted by e-mail (up to a maximum of 3 times). For willing collaborators, data were initially de-identified at source before encryption and transfer by e-mail. In line with recently published principles and recommendations in relation to the sharing and reuse of IPD [22], data within the individual data files were used to establish an overall anonymized data file, as follows: only data on the prioritized IPD variables within the transferred files were included, there were no personal identifiers included. The anonymized data file was held in Excel® V15.30 (Microsoft Corporation, USA).

Data analysis

The statistical analyses were conducted in the graphical user interface RAnalyticFlow (version 3.1.8) with R (version 3.6.3). Serum 25OHD values were transformed into the common unit of nmol/L and calcium intake into mg/d, using the conversion factors 2.496 mol/g for 25OHD and 24.95 mmol/g for calcium. If means and standard deviations were not reported, they were estimated using medians, interquartile ranges, confidence intervals, standard errors, t values, P values, F values [23]. If data was only available in plots, it was extracted using PlotDigitizer [24]. Non-detectable levels of serum 25OHD were imputed using the midpoint between the detection level of the assay and zero.

The data distributions of the study-level estimates and individual-level data were plotted in histograms and outliers reviewed. Data were subjected to random effects multi-level meta-analyses, and ninety-five percent confidence intervals (95% CI) were computed. Studies that could not be meta-analyzed were summarized in a narrative manner.

The odds of having rickets at different serum 25OHD thresholds were calculated. The sensitivity (i.e. percentage of the population with disease correctly identified by the threshold) and specificity (i.e. percentage of the population without the disease correctly identified by the threshold) of different serum 25OHD thresholds to detect rickets were calculated and plotted as a receiver operating characteristic (ROC) curve. The maximal Youden index was calculated and used to determine at which serum threshold the sensitivity and specificity were maximized and thus represents the maximum potential effectiveness of a biomarker like serum 25OHD.

The sensitivity and specificity analyses were performed on the IPD subset of individuals with adequate calcium intake (as newly defined by FAO-WHO, i.e., Average Nutrient Requirement (ANR) values for 0–6 months-old, 210 mg/d, 7–11 months-old, 330 mg/d, and 1–3 year-olds, 490 mg/d) (Personal communication from Dr Jason Montez, WHO Scientist) as well as on the entire IPD dataset (irrespective of dietary calcium intake). To assess the robustness of the results, further sensitivity analyses were conducted. One sensitivity analysis was done including only IPD data with known adequate calcium intakes assessed by multiple 24h recalls. To be able to include IPD data where calcium intake was not reported, an additional sensitivity analysis was conducted with imputed missing calcium intake data. Where calcium intake was missing, it was assumed to be adequate in infants exclusively breastfed and in children with a diversified diet, including dairy products, and assumed to be insufficient in infants below 6 months of age with mixed feeding and in children above 5 months exclusively breastfed, with a low and null dairy intake, special unbalanced or vegan diet.

Results

Study characteristics

From a total of 1112 records identified within the search, a total of 120 studies with 5412 participants (mean age 17 months) were included (see Fig. 1). The majority of the studies were case reports (N = 39) and case series (N = 40), followed by case–control studies (N = 19) and trials (N = 19), cohort studies (N = 2) and a cross-sectional study (N = 1). The studies were conducted in all regions of the world, except Latin America. The countries in which most of the studies were conducted were the United States of America (N = 22), Nigeria (N = 14), India (N = 12), and Turkey (N = 11). The studies covered latitudes from 60.5°N to 40.9°S (mean 31.4° N). In the majority of the studies (79%), the skin pigmentation of the participants was dark. While most of the studies did not report which method was used to measure circulating 25OHD (N = 63), the remainder reported the use of competitive binding radioimmunoassay (N = 44), chemiluminescence immunoassay (N = 8), liquid chromatography-tandem mass spectrometry (LC–MS/MS) (N = 4), or high-performance liquid chromatography (HPLC) (N = 1). Only two studies reported participating in a vitamin D assay standardization program. The characteristics of the included studies are shown in Table 2.

Fig. 1
figure 1

Flow chart for study selection

Full size image
Table 2 Study characteristics of studies of rickets in young children
Full size table

Individual data on serum 25OHD was reported for 65 studies with 930 participants (mean age 31 months old, range 0–47 months old), of which 75% had radiologically confirmed rickets and 25% did not have rickets. Sixteen studies reported having measured calcium intake. Upon request for IPD on serum 25OHD and calcium intake, 11 studies agreed and provided the data (666 participants, meaning that 71.6% of the participants in the IPD dataset had a corresponding calcium intake measured) [25,26,27,28,29,30,31,32,33,34]. The remaining five studies did not respond or were not able to provide the data [35,36,37,38,39]. Among the 11 studies for which data was provided, calcium intake was estimated by multiple 24h recalls in 8 studies, by a single 24 h recall in one study, by a 3-day food record in one study, and a food frequency in one study. Using the FAO-WHO’s age-specific ANR values, 23% of participants had adequate calcium intake and 77% had insufficient calcium intake.

Study-level meta-analysis

The meta-analysis of all the studies (irrespective of study design and calcium intake) showed that children with rickets had a mean serum 25OHD of 23 nmol/L (N studies = 77, 95% CI 19–27), whereas children without rickets had a mean serum 25OHD of 62 nmol/L (N studies = 19, 95% CI 55–70). When restricting the meta-analysis to case–control studies (N studies = 16), the children with and without rickets had a mean serum 25OHD of 32 (95% CI 23–40) nmol/L and 64 nmol/L (95% CI 56–73) nmol/L, respectively. When restricting the meta-analysis to case reports, case series and trials, mean serum 25OHD in children with rickets was 17, 20, 26 nmol/L, respectively.

Eight studies were not meta-analyzable, because they did not report mean or median serum 25OHD. In one study [40], refugee children showing up at the hospital were screened for rickets. Of all the children screened, 28.5% had nutritional rickets, 40% had serum 25OHD < 30 nmol/L and 9% have serum 25OHD in the range 30–49 nmol/L. In another study [41], 47% of the infants had a serum 25OHD below 25 nmol/L and of those 72% had radiographic evidence of rickets. A case series [42] of children diagnosed with nutritional rickets found that 79% had serum 25OHD below 50 nmol/L. Another case series [43] of children diagnosed with nutritional rickets found that 62% had serum 25OHD below 25 nmol/L. A surveillance study [44] found that among children with serum 25OHD below 50 nmol/L, 77% had radiological changes associated with rickets.

Two trials [45, 46] compared 6-month-old exclusively breastfed infants of women who received vitamin D supplementation or no supplementation during postpartum. The first trial [45] found that 44% and 75% of children from non-supplemented mothers had serum 25OHD below 25 and 50 nmol/L, respectively; whereas 8% and 25% of children from vitamin D-supplemented mothers had serum 25OHD below 25 and 50 nmol/L, respectively. In the unsupplemented and supplemented groups, 3.4% and 3.6% of the children developed radiological rickets, respectively. In the second trial [46], the equivalent estimates for the development of radiological rickets in unsupplemented and supplemented children were 4% and 2%, respectively. One trial [47] found that at baseline, at 3–5 days of age, that 3% and 6% of infants with serum 25OHD < and > 27.5 nmol/L showed wrist ossification centers. At 6 months of age, after vitamin D supplementation with 100, 200 or 400 IU/d, none of the children showed any radiological signs of rickets [47].

IPD meta-analysis

Based on individual data and irrespective of dietary calcium intake (n = 930, mean 289 mg/d, median 230 mg/d), the serum 25OHD in 700 children with rickets ranged from non-detectable to 180 nmol/L, with a median of 23 nmol/L and mean of 29 nmol/L (95% CI 27–31). The distribution of the serum 25OHD in children with rickets is shown in Fig. 2. More than half (55%) of the children with rickets had serum 25OHD below 25 nmol/L, 62% below 30 nmol/L, 79% below 40 nmol/L, and 87% below 50 nmol/L. In 230 children without rickets, the median serum 25OHD was 57 nmol/L, with a mean of 62 nmol/L (95% CI 58–66).

Fig. 2
figure 2

Serum 25OHD [nmol/L] distribution in children with rickets (n = 700)

Full size image

The odds of having rickets increased exponentially as serum 25(OH)D concentrations decreased below 50 nmol/L, and dramatically so when concentrations fell below 30 nmol/L (see Fig. 3). The sensitivities and specificities of different serum 25OHD thresholds to detect rickets are shown in Fig. 4 and Supplemental Table 2. A sensitivity and specificity of 80% were reached at serum 25OHD concentrations of 42 and 38 nmol/L, respectively. The serum 25OHD threshold at which the sensitivity and specificity were maximized, i.e. the maximal Youden index, was at 40 nmol/L (sensitivity 79% and specificity 77%).

Fig. 3
figure 3

Odds of developing rickets at different serum 25OHD thresholds [nmol/L] (n = 930)

Full size image
Fig. 4
figure 4

Sensitivities (red line) and specificities (blue line) for different serum 25OHD thresholds [nmol/L] (n = 930)

Full size image

Including only children with adequate dietary calcium intake

When considering only the children with adequate calcium intakes (n = 640, mean 580 mg/d, median 522 mg/d), the odds of having rickets increased exponentially as serum 25(OH)D concentrations decreased below 60 nmol/L, and dramatically so when concentrations fell below ~ 25 nmol/L (see Fig. 5). The sensitivities and specificities of different serum 25OHD thresholds to detect rickets are shown in Fig. 6 and Supplemental Table 2. A sensitivity and specificity of 80% were reached at serum 25OHD concentrations of 32 and 28 nmol/L, respectively. The thresholds at which the sensitivity and specificity were maximized (i.e. Youden index) was 28 nmol/L. When including only studies with known adequate calcium intakes estimated from multiple 24 h recalls, the Youden index was at 33 nmol/L. When including studies with known adequate calcium intakes as well as assumed adequate calcium intakes (imputed), the Youden index was at 30 nmol/L.

Fig. 5
figure 5

Odds of developing rickets at different serum 25OHD thresholds [nmol/L] in children with adequate calcium intake (i.e. 0–6 months old ≥ 210 mg/d, 7–11 months old ≥ 330 mg/d, 1–3 years old ≥ 490 mg/d) (n = 154)

Full size image
Fig. 6
figure 6

Sensitivities (red line) and specificities (blue line) for different serum 25OHD thresholds [nmol/L] in children with adequate calcium intake (i.e. 0–6 months old ≥ 210 mg/d, 7–11 months old ≥ 330 mg/d, 1–3 years old ≥ 490 mg/d) (n = 154)

Full size image

Discussion

In terms of the establishment of a DRV for vitamin D for young children, identification and selection of an appropriate serum 25OHD threshold is critical. This serum 25OHD concentration should protect a majority of children against increased risk of nutritional rickets and thus form a basis for derivation of a recommended dietary intake which will allow young children to maintain serum 25OHD concentrations at or above this threshold. It is not intended as a clinical threshold diagnostic for rickets.

Expert authorities charged with the establishment of vitamin D recommendations have thus far generally relied on reported individual baseline serum 25OHD concentration data in case reports, and mean/median 25OHD concentration data from studies of other designs, without trying to meta-analyze data to set a serum 25OHD threshold [3, 12, 18]. For example, the Scientific Advisory Committee on Nutrition (SACN) in the UK concluded that individual and mean serum 25OHD concentrations of children with rickets were < 25 nmol/L in the majority of studies (44 included) in their DRV exercise in 2016 [18]. The US Institute of Medicine (IOM) in 2011 identified 13 studies in their DRV exercise, and while 6 studies reported mean or median serum 25OHD concentrations < 30 nmol/L in children with rickets, the remaining studies reported mean serum 25OHD concentrations > 30 nmol/L [range 36–50 nmol/L] [12]. The European Food Safety Authority (EFSA) expert panel concluded that there is no risk of vitamin D deficiency rickets with serum 25OHD concentrations at or above 50 nmol/L and adequate calcium intake [3]. In the present extensive systematic review, the meta-analysis of study-level data also showed that young children with radiologically confirmed rickets had a mean serum 25OHD concentration of 23 nmol/L (95% CI 19–27).

A key limitation in the interpretation of such study-level data is the fact that they could be confounded by dietary calcium, especially as many of the studies were from developing countries where calcium intakes may be low [12, 18]. Thus, whether the rickets in these studies was caused solely by vitamin D deficiency and/or by low calcium intake is not clear. The present work sought to address this key knowledge gap by obtaining IPD from those studies that measured calcium intake as well as serum 25OHD in children with rickets. This data enabled an analysis of sensitivities and specificities in relation to odds of rickets at different serum 25OHD thresholds, and consequently, the estimation of the maximal Youden index, which is a measure of the potential effectiveness of a biomarker and an index used for setting optimal thresholds on medical tests. The analysis suggested the serum 25OHD threshold at which the sensitivity and specificity were maximized, i.e. the maximal Youden index, was around 28 nmol/L in children with adequate calcium, whereas this increased to 40 nmol/L in the entire sample which included children with insufficient calcium intakes. The latter would be more reflective of the types of datasets that IOM, SACN and EFSA would have based their threshold decisions upon. If dietary calcium intake is low, and serum calcium concentrations decrease, the compensatory metabolic response is an accelerated conversion of 25OHD to 1,25-dihydroxyvitamin D (via parathyroid hormone), which normalizes serum calcium concentrations [18]. This increased 25OHD catabolism leads to an increased vitamin D requirement. The IOM have suggested that when calcium intakes are inadequate, vitamin D supplementation to the point of serum 25OHD concentration up to and beyond 75 nmol/L has no effect [12]. The present findings based on empirical data from young children with adequate calcium intakes are consistent with the suggestion by the IOM, as well as other agencies briefed with the development of vitamin D DRVs, that in the face of adequate calcium intake, the risk of nutritional rickets increases below a serum 25OHD concentration of 30 nmol/L and is minimal (although not absent) when serum 25(OH)D concentrations range between 30 and 50 nmol/L [3, 12, 18]. Another report that explored the interaction between 25OHD and calcium intake from a single study, also found that the risk of rickets increased below 40 nmol/L or even higher in children with the lowest calcium intakes [14].

Two major intertwined strengths of this review are the meta-analysis of IPD, to complement the study-level meta-analyses, and that differences in calcium intake could be accounted for and the analysis be restricted to children with adequate calcium intake. The comprehensive search of the literature for the relevant studies ensured that all those studies with measured calcium intakes were identified. IPD data for 25OHD was available for 65 studies (n = 930) out of 120 studies (n = 5412). IPD was requested from the 15 studies that reported having measured calcium intakes and was obtained from 10 of these studies. The comparison of sensitivities and specificities of different thresholds allowed the identification of an optimal minimum serum 25OHD threshold. The present work also emphasized rickets as confirmed radiologically, to reduce the risk of misdiagnosing children with or without rickets.

The limitations of this review were that, due to resource and time constraints, only one online database was searched systematically. However, to ensure no important and relevant studies were missed, the reference lists of other reviews were screened. Another limitation of this review, in common with all DRV exercises to-date, is the potential variability in the serum 25OHD measurement data amongst included studies that used different analytical methods. In fact, only two of the studies included reported participating in a vitamin D assay standardization program. The measurement of serum 25OHD can vary widely between assays and participation in a vitamin D assay standardization program is recommended [13]. In addition, it was not possible to assess whether there might be different serum 25OHD thresholds on the basis of ethnicity or ancestry because the majority of the studies were of dark-skinned participants. The available data was not reported in a sufficiently consistent matter to be able to take into account sun exposure and geographical location.

In conclusion, the present IPD-level meta-analyses suggest that a minimum serum 25OHD threshold of ~ 28 nmol/L and above would represent a low risk of nutritional rickets for the majority of children with an adequate calcium intake. However, a higher 25OHD threshold is likely necessary to prevent rickets in populations with low dietary calcium intakes, which includes the geographic areas of Africa and South Asia, where rickets remain widespread. This threshold while useful within a vitamin D DRV process, as indicative of the risk of disease, it is not intended as a clinical threshold diagnostic for rickets.