Early life vitamin D status and asthma and wheeze: a systematic review and meta-analysis

We conducted a systematic literature search using PubMed, Embase, CINAHL, and CNKI (in Chinese) databases and Google Scholar for studies published up to 19th July 2017 and the Cochrane Central Register of Controlled Trials that reported data on antenatal or early postnatal vitamin D status or intervention and asthma and wheezing in children or adults. Details of the search strategy are provided in Additional file 1: Supplement 2.

The present systematic review focused on the effect of vitamin D level in blood (maternal or cord or infant) and intake (maternal intake during pregnancy or infant intake) on asthma and wheeze. We included randomized controlled trials (RCTs), quasi-randomized controlled trials, non-randomized controlled trials, prospective and retrospective cohort studies that measured maternal vitamin D status during any trimester of pregnancy or cord blood or offspring vitamin D status during infancy, and having asthma or wheeze during childhood or adulthood as outcomes, which (i) were diagnosed by doctors (including parental or self-reports), or (ii) required the use of asthma medication, or (iii) as assessed by the International Study of Asthma and Allergies in Children (ISAAC) questionnaire (Additional file 1: Supplement 3). Two reviewers (SS and WX) independently screened the titles and abstracts and identified potentially relevant publications according to the selection criteria, of which full text was obtained and read (online Additional file 1: Supplement 4). The reference lists of all papers of interest were scrutinized to obtain other relevant articles. Disagreements over inclusion were resolved through consensus, and where necessary, a third reviewer (JL) was involved.

The same reviewers independently extracted data using a standard data extraction form. We collected data on the number of subjects with and without asthma or wheeze in the antenatal or early postnatal vitamin D exposed/intervention and non-exposed/control groups; risk estimates (crude and/or adjusted odds ratios [ORs], relative risks [RRs] and hazard ratios [HRs]) and the corresponding 95% confidence intervals (CIs) at any available age end point. Furthermore, we recorded information on the population, geographical location, inclusion and exclusion criteria, interval of follow up, exposure measurement, outcome definition, measurement and window of assessment, and confounders (more details in Additional file 1: Supplement 5). Study authors were contacted when missing data was an issue. Disagreements on data extraction were checked against the original articles and/or resolved by a third reviewer. The data extracted were entered into Review Manager software and were double checked by two reviewers. Risk of bias was assessed using the tool recommended by the Cochrane Collaboration [17, 18]. Two reviewers independently rated the risk of bias for each study as high, low, or unclear (‘Definitely yes’ in the tool was considered as ‘low’ risk, ‘Definitely no’ as ‘high’ risk, ‘unclear’ comprised ‘probably yes’ and ‘probably no’ as specified in the tools [18]). Disagreements were resolved by discussion or by involving a third reviewer. Each cohort study was classified into one of the following categories: high (more than two criteria not applied/met); moderate (one or two criteria not applied/met or unclear); and low risk of bias (all criteria applied/met).

We pooled the original risk estimates as reported (most being ORs) and used adjusted risk estimates where available. When only frequency distributions were provided, we calculated unadjusted ORs and their 95% CIs from the outcome distribution of exposed and non-exposed groups. Where outcome assessment was done in multiple time points in a study, we chose the one with the longest follow-up period or the time point when the age of the participants was comparable to other studies.

We pooled the risk estimates from each study that compared the risk in the highest vitamin D concentration or intake category with the lowest (which is the referent category), and also the risk in vitamin D sufficiency (≥75 nmol/L [4]) with deficiency (< 50 nmol/L according to the Institute of Medicine (IOM) definition [19]). As it could be difficult to diagnose asthma in children below 5 years of age, we stratified the studies on asthma according to the age of outcome assessment (> 5 years and ≤ 5 years). For exploring the association between blood vitamin D status and childhood asthma, a sensitivity analysis was performed by excluding the studies assessing the outcome in adulthood. We contacted the authors of the studies (n = 8) that had different cut-off values, and asked them to re-analyze their data according to the above categorization scheme. Of the eight authors contacted, four responded and provided additional data [20‐23].

Both fixed-effects and random-effects models were used to summarize the effect sizes. Heterogeneity was evaluated by using the Q (P > 0.10 in the Chi² test for low heterogeneity) and I² statistics (moderate heterogeneity for I² > 30% and considerable heterogeneity for ≥75%). We presented pooled ORs from random-effect models when considerable heterogeneity was observed. Stratified analyses by time window of vitamin D and outcome assessment (for wheeze only), latitude, geographical region, adjustment for family history of allergic disease, seasonality, and smoking status, methods of exposure and outcome ascertainment, and risk of bias were performed to explore the sources of heterogeneity. Funnel plots and Egger’s test were used to explore the possibility of publication bias. Number needed to treat for preventing asthma by increasing from low to high levels of vitamin D or via supplementation was estimated by the baseline asthma prevalence estimates and pooled ORs [24]. Statistical analyses were performed with Review Manager software version 5.3 [25] and STATA version 13.

We have identified 423 potentially relevant publications from PubMed, 855 from Embase, 36 from CNKI database, 105 from CINAHL and 66 from Google Scholar. After excluding duplicates and publication that did not meet the inclusion criteria, three RCTs [26‐28] and 33 cohort studies [20‐23, 29‐57] were included (Fig. 1).

The main characteristics of the eligible RCTs, cohort studies and those studies excluded during full text screening are shown in Tables 1 and 2 and Additional file 1: Table S1, respectively. Two of the three RCTs were single-center trials and recruited low-risk pregnant women. The other was a multicenter trial enrolling pregnant women who had history of allergic disease. The dose of vitamin D administration and outcome definition varied across the three trials (1619 participants) (Table 1). The median ages of outcome assessment in cohort studies that reported association between blood vitamin D level and asthma, vitamin D intake and asthma, blood vitamin D and wheeze, and vitamin D intake and wheeze were 6 years (range 1.5–25), 5 (1.5–31), 5 (1–10), and 3 (1–10), respectively. Only two studies assessed the outcome in adulthood [36, 57]. Of the cohort studies, 26 studies reported data on asthma and 17 on wheeze. Nineteen of these 26 asthma studies measured vitamin D levels (31,940 participants from 13 cohorts were included in the meta-analysis); the rest assessed dietary/supplement intake (41,952 participants). A variety of operational definitions of asthma were used, including parental reports of doctor-diagnosed asthma (n = 7), parental reports of asthma medication use (n = 1), a combination of both (n = 2), parental reports of doctor-diagnosed asthma with wheezing symptom (n = 4), parental reports of doctor-diagnosed asthma and wheezing and/or asthma medication use (n = 5) and doctor-diagnosed asthma or asthma medication use by medical record review (n = 5). Two studies did not report the definition of asthma. All 17 studies on wheeze assessed maternal vitamin D status during pregnancy, with 12 measuring blood vitamin D levels (8123 participants from 8 cohorts were meta-analyzed) and 6 recorded intake (5678 participants). Most wheeze studies defined outcome using parental reports of wheezing symptoms in the past 12 months (Table 2).

Two RCTs [27, 28] reported non-significant trends of vitamin D supplementation during pregnancy on preventing the development of offspring asthma (OR, 0.82; 95% CI 0.50–1.36 and HR, 0.8; 95% CI, 0.6–1.0, respectively). As the intervention and the outcome definition were different, we did not pool the effect size. Apart from having insufficient information to assess allocation concealment, risk of bias was low in both RCTs (Additional file 1: Figure S1).

Figure 2 and Table 3 show the results from cohort studies as well as overall and subgroup summary ORs for the relationship between blood vitamin D and the risk of asthma assessed at > 5 years of age. Fix-effects ORs were not significant for the highest vs. the lowest categories of blood vitamin D concentrations (8 cohort studies, 28,436 participants, Fig. 2a) and for the sufficient vs. deficient groups (5 cohort studies, 23,339 participants, Fig. 2b), with moderate heterogeneity. The pooled estimate did not change materially (OR 0.91, 95% CI 0.73, 1.12, I² = 32%, for highest vs. lowest; OR 1.02, 95% CI 0.84, 1.24, I² = 47% for sufficient vs. deficient groups) after excluding one study assessing asthma in adulthood. Little heterogeneity was observed across subgroups, in studies measuring cord blood vitamin D level, adjusted for family history of allergic disease, or seasonality, those with exposure measured by liquid chromatography-tandem mass spectrometry (LC-MS/MS), and studies with moderate risk of bias (Table 3).

Figure 3 and Table 3 show the results from cohort studies and the overall and subgroup summary ORs for the relationship between blood vitamin D and the risk of asthma assessed at ≤5 years of age. The fixed-effects ORs were 0.81 (95% CI 0.65, 1.01, I² = 0%, 6 cohort studies, 27,776 participants, Fig. 3a) for the highest vs. lowest categories of blood vitamin D concentrations and 0.93 for the sufficient vs. deficient groups (95% CI 0.85, 1.03, I² = 36%, 6 cohort studies, 27,776 participants, Fig. 3b). Although higher maternal blood vitamin D level during pregnancy was significantly associated with lower risk of asthma assessed at ≤5 years, we did not observe any significant associations in the stratified analyses for studies that had more robust methodology (adjusting for further confounders, e.g. family history of atopy, seasonality, smoking status; blood calcifediol measured by the gold standard LC-MS/MS; studies with lower risk of bias) (Table 3). Six eligible cohort studies [38, 39, 43, 44, 46, 47] analyzed blood vitamin D as a continuous variable and we found no association with childhood asthma.

There was no association between vitamin D intake and asthma at ≤5 years (6 cohorts, 35,257 participants) or > 5 years (3 cohorts, 32,023 participants) (Table 4). A significant inverse association between vitamin D intake and asthma at ≤5 years of age emerged when we restricted our analysis to studies that did not adjust for seasonality or had high risk of bias, with low heterogeneity within these subgroups; and between vitamin D intake and asthma at > 5 years of age when we restricted to studies that adjusted for family history of atopy, seasonality or smoking status, or those that assessed outcome in childhood (OR 0.73, 95% CI 0.56, 0.94, I² = 0%, 25,301 participants from 2 studies, for highest vs. lowest) (Table 4). The number needed to treat to prevent one case of asthma was 39 (range 25–93) via maternal vitamin D supplementation (Additional file 1: Table S2).

Studies that measured blood vitamin D concentrations generally had a low risk of bias except for attrition bias due to incomplete outcome data, confounding bias, and detection bias in the assessment of prognostic factors (Additional file 1: Figures S2 and S3). Several studies that assessed vitamin D intake had a high risk of detection bias due to recall of vitamin D intake, confounding bias, and attrition bias due to incomplete outcome data. (Additional file 1: Figures S4 and S5). No evidence of publication bias was found from the funnel plots or Egger’s tests for blood (P = 0.963 for asthma > 5 years and P = 0.655 for asthma ≤5 years) and intake (P = 0.913 for asthma > 5 years and P = 0.410 for asthma ≤5 years) (Additional file 1: Table S3).

All three RCTs [26‐28] used wheeze as primary outcome (one combined with asthma [28]). Two RCTs found no definitive effect of vitamin D supplementation during pregnancy on wheeze among offspring by 3 years of age (OR 0.86; 95% CI 0.49, 1.50; HR 0.76; 95% CI 0.52, 1.12) and one found a borderline statistical significance association (HR 0.8; 95% CI 0.6, 1.0). We did not perform pooled analysis for wheeze because of the major differences in intervention. Risks of performance and attrition bias were considered to be high due to missing outcome data and the lack of blinding in the trial conducted by Goldring et al. [26] (Additional file 1: Figure S1).

We did not observe significant association between maternal vitamin D concentrations during pregnancy and offspring wheeze in cohort studies (8123 participants from 8 studies, OR 0.91, 95% CI 0.76, 1.09 for highest vs. lowest levels, I² = 25%; 2324 participants from 3 studies, OR 0.91, 95% CI 0.70, 1.19 for ≥75 nmol/L vs. < 50 nmol/L, I² = 0%) (Fig. 4). Similar results were observed in stratified analyses (Table 5). Within-group inconsistency disappeared or was attenuated in studies that measured vitamin D intake during pregnancy or adjusted for family history of atopy or seasonality or blood calcifediol measured by LC-MS/MS or studies with moderate risk of bias. The overall fixed-effects OR of wheeze relating to maternal vitamin D intake was 0.66 (95% CI 0.53, 0.82; 5238 participants from 6 studies, I² = 52%, Table 5). The association remained statistically significant when restricting to studies that: assessed outcome at or before 3 years of age and  after 3 years of age, were conducted in a temperate area, adjusted for family history of atopy or seasonality, and those that had lower risk of bias. The Baiz paper [38] did not find an association in early transient, late onset or persistent wheeze, while Camargo et al. [39] reported a significant protective effect by 5 years of age (OR 0.95, 95% CI 0.91, 0.99), but this has not been consistent in earlier time windows reported in the studies by Camargo et al. [39] and Visness et al. [44]

The 12 cohort studies that investigated blood vitamin D levels and wheeze were evaluated to have low risk of bias, in terms of participant selection, exposure assessment, and window of outcome assessment and co-intervention. Some studies had high or unclear risk of confounding bias, detection bias and attrition bias (Additional file 1: Figure S6, a and b). Overall, the studies on vitamin D intake had higher risk of bias, due to their inadequacy in exposure assessment, assessment and adjustment for prognostic factors, and outcome follow-up (Additional file 1: Figure S7, a and b). There was no evidence of publication bias from the funnel plots or Egger’s tests for blood (P = 0.772) and intake (P = 0.954) (Additional file 1: Table S3).