Mendelian randomization analysis demonstrates that low vitamin D is unlikely causative for pediatric asthma

Asthma is the most common chronic childhood disease, affecting approximately 8% of children in the United States. Multiple studies show strong associations between low serum 25-hydroxyvitamin D (25(OH)D) concentration and childhood asthma incidence and severity.1Brehm J.M. Schuemann B. Fuhlbrigge A.L. Hollis B.W. Strunk R.C. Zeiger R.S. et al.Childhood Asthma Management Program Research GroupSerum vitamin D levels and severe asthma exacerbations in the Childhood Asthma Management Program study.J Allergy Clin Immunol. 2010; 126: 52-58.e5Abstract Full Text Full Text PDF PubMed Scopus (384) Google Scholar, 2Chinellato I. Piazza M. Sandri M. Peroni D. Piacentini G. Boner A.L. Vitamin D serum levels and markers of asthma control in Italian children.J Pediatr. 2011; 158: 437-441Abstract Full Text Full Text PDF PubMed Scopus (140) Google Scholar, 3Kolokotroni O. Papadopoulou A. Middleton N. Kouta C. Raftopoulos V. Nicolaidou P. et al.Vitamin D levels and status amongst asthmatic and non-asthmatic adolescents in Cyprus: a comparative cross-sectional study.BMC Public Health. 2015; 15: 48Crossref PubMed Scopus (13) Google Scholar Vitamin D deficiency is highly prevalent in atopic pediatric patients, and vitamin D has been hypothesized to contribute to childhood asthma through regulation of airway reactivity, sensitivity to corticosteroids, or modulation of immune function.4Di Filippo P. Scaparrotta A. Rapino D. Cingolani A. Attanasi M. Petrosino M.I. et al.Vitamin D supplementation modulates the immune system and improves atopic dermatitis in children.Int Arch Allergy Immunol. 2015; 166: 91-96Crossref PubMed Scopus (50) Google Scholar Vitamin D supplementation studies, however, have failed to reveal a benefit for vitamin D in preventing or treating asthma.5Castro M. King T.S. Kunselman S.J. Cabana M.D. Denlinger L. Holguin F. et al.National Heart, Lung, and Blood InstituteEffect of vitamin D3 on asthma treatment failures in adults with symptomatic asthma and lower vitamin D levels: the VIDA randomized clinical trial.JAMA. 2014; 311: 2083-2091Crossref PubMed Scopus (196) Google Scholar, 6Martineau A.R. MacLaughlin B.D. Hooper R.L. Barnes N.C. Jolliffe D.A. Greiller C.L. et al.Double-blind randomised placebo-controlled trial of bolus-dose vitamin D3 supplementation in adults with asthma (ViDiAs).Thorax. 2015; 70: 451-457Crossref PubMed Scopus (81) Google Scholar Recently, Mendelian randomization has been used to help dissect direction and extent of causality identified in association studies between modifiable risk factors and disease. Our overarching goal was to use Mendelian randomization to examine the hypothesis that low 25(OH)D concentration contributes significantly to asthma prevalence or exacerbations. First, we performed cross-sectional analyses to confirm the inverse association between serum 25(OH)D concentration and asthma in our cohort. We then performed a genome-wide association study (GWAS) to confirm that previously identified genetic determinants of vitamin D status were similarly determinative of 25(OH)D concentration in our cohort. Finally, we used the genetic determinants confirmed in our GWAS to perform our Mendelian randomization. The Children's Hospital of Philadelphia Center for Applied Genomics biorepository consisted of 56,835 subjects; 12,842 were used in the Asthma Study Cohort. Asthma severity was inferred from the electronic medical record (see Figs E1 and E2 in this article's Online Repository at www.jacionline.org) per the National Asthma Education and Prevention Program guidelines.7National Asthma Education and Prevention ProgramExpert Panel Report 3 (EPR-3): guidelines for the diagnosis and management of asthma—summary report 2007.J Allergy Clin Immunol. 2007; 120: S94-S138Abstract Full Text Full Text PDF PubMed Google Scholar Our cross-sectional analysis was done using Stata 13.1. Our Mendelian randomization was performed using R (http://www.r-project.org/). Our cross-sectional and Mendelian randomization analyses had greater than 90% power to detect an odds ratio of 1.73 and 1.4, respectively (see Tables E1 and E2 in this article's Online Repository at www.jacionline.org).8Brion M.J. Shakhbazov K. Visscher P.M. Calculating statistical power in Mendelian randomization studies.Int J Epidemiol. 2013; 42: 1497-1501Crossref PubMed Scopus (586) Google Scholar Our cross-sectional analysis was performed on 642 individuals from the Asthma Study Cohort who had serum 25(OH)D concentration evaluated as part of routine clinical care (296 patients and 348 controls). Similar to previous reports,9Mansbach J.M. Ginde A.A. Camargo Jr., C.A. Serum 25-hydroxyvitamin D levels among US children aged 1 to 11 years: do children need more vitamin D?.Pediatrics. 2009; 124: 1404-1410Crossref PubMed Scopus (251) Google Scholar more than half of all subjects (60.6%) in our cohort were vitamin D insufficient (25(OH)D concentration < 30 ng/mL), 27.8% were vitamin D deficient (<20 ng/mL), and 6.2% were severely vitamin D deficient (<10 ng/mL). Unadjusted serum 25(OH)D concentrations were lower in patients with asthma than in controls (25.8 ng/mL vs 28.1 ng/mL; P = .008). In multivariate logistic regression, a 10-ng/mL increase in 25(OH)D concentration was associated with an 18% decreased odds of asthma (P = .02; Fig 1, A). Of asthma cases, 74 subjects (26%) were treated for at least 1 severe exacerbation. Serum 25(OH)D concentration was inversely associated with the frequency of systemic steroids such that for every 10-ng/mL increase in serum 25(OH)D concentration, there was a 34% decrease in the frequency of systemic glucocorticoids for severe asthma exacerbations (P < .002; Fig 1, B). Our cohort reproduced similar associations between 25(OH)D concentration and asthma reported previously.1Brehm J.M. Schuemann B. Fuhlbrigge A.L. Hollis B.W. Strunk R.C. Zeiger R.S. et al.Childhood Asthma Management Program Research GroupSerum vitamin D levels and severe asthma exacerbations in the Childhood Asthma Management Program study.J Allergy Clin Immunol. 2010; 126: 52-58.e5Abstract Full Text Full Text PDF PubMed Scopus (384) Google Scholar, 2Chinellato I. Piazza M. Sandri M. Peroni D. Piacentini G. Boner A.L. Vitamin D serum levels and markers of asthma control in Italian children.J Pediatr. 2011; 158: 437-441Abstract Full Text Full Text PDF PubMed Scopus (140) Google Scholar, 3Kolokotroni O. Papadopoulou A. Middleton N. Kouta C. Raftopoulos V. Nicolaidou P. et al.Vitamin D levels and status amongst asthmatic and non-asthmatic adolescents in Cyprus: a comparative cross-sectional study.BMC Public Health. 2015; 15: 48Crossref PubMed Scopus (13) Google Scholar We performed a GWAS on 1388 white subjects in the Center for Applied Genomics vitamin D cohort (Table I) to confirm that loci previously associated with serum 25(OH)D concentration were associated with serum 25(OH)D concentration in our cohort; our GWAS confirmed significant associations between serum 25(OH)D concentration and single-nucleotide polymorphisms (SNPs) near 2 genes (GC and CYP2R1). Consistent with previous reports in the literature, we found that the gene variants identified that affect serum 25(OH)D concentration are responsible for approximately 10% of the population variation in 25(OH)D concentration (eg, 10-15 ng/mL or 25-37 nmol/L); having variants that positively affect 25(OH)D concentration as compared with those that negatively affect 25(OH)D concentration is the equivalent of a lifetime of supplementation with 1000 to 2000 IU of cholecalciferol. We used the SNP rs10741657 located near the CYP2R1 gene on chromosome 11p15 and the SNP rs2282679 located in the vitamin D–binding protein (GC) gene on chromosome 4p12 to generate a vitamin D propensity score. Scores generated with these SNPs were highly correlated with serum 25(OH)D concentration (odds ratio, 1.92; 95% CI, 1.54-2.39; P = 2.52 × 10−15). We used these scores to perform a Mendelian randomization. In contrast to our cross-sectional analysis, genetic risk scores in our Mendelian randomization analysis were not associated with cases of asthma (β = −0.0000351; P = .85; Table II). We also found no associations between vitamin D genetic propensity scores and severe asthma exacerbations (β = −0.00833; P = .86; Table II).Table IAssociations of SNPs and vitamin D association testingDependent variableGeneAdjusted∗Multivariate linear regression model adjusted for age, sex, body mass index Z score, and season.Nonadjusted†Univariate linear regression model.Beta95% CIP valueBeta95% CIP valuers2282679GC6.51 × 10−14.32 × 10−1 to 8.70 × 10−15.79 × 10−94.60 × 10−12.83 × 10−1 to 6.36 × 10−15.16 × 10−4rs17467825GC6.55 × 10−14.34 × 10−1 to 8.75 × 10−15.73 × 10−94.66 × 10−12.89 × 10−1 to 6.44 × 10−12.32 × 10−10rs1155563GC5.50 × 10−13.38 × 10−1 to 7.62 × 10−13.54 × 10−73.84 × 10−12.11 × 10−1 to 5.57 × 10−14.09 × 10−10rs7041GC4.91 × 10−12.87 × 10−1 to 6.95 × 10−12.40 × 10−65.05 × 10−13.38 × 10−1 to 6.72 × 10−11.11 × 10−8rs10741657CYP2R13.41 × 10−11.03 × 10−1 to 5.79 × 10−14.99 × 10−33.29 × 10−11.33 × 10−1 to 5.26 × 10−13.04 × 10−4rs10766192CYP2R13.73 × 10−11.62 × 10−1 to 5.83 × 10−15.26 × 10−42.03 × 10−13.22 × 10−2 to 3.73 × 10−18.94 × 10−7rs10500804CYP2R13.54 × 10−11.46 × 10−1 to 5.62 × 10−18.68 × 10−42.00 × 10−13.10 × 10−2 to 3.69 × 10−17.28 × 10−2rs7944926DHCR7/ NADSYN12.35 × 10−1−2.08 × 10−1 to 2.55 × 10−18.42 × 10−11.07 × 10−1−7.27 × 10−2 to 2.87 × 10−11.99 × 10−11Score (rs2282679, rs10741657)GC, CYP2R13.09 × 10−12.13 × 10−1 to 4.06 × 10−13.42 × 10−102.38 × 10−11.60 × 10−1 to 3.15 × 10−11.82 × 10−9∗ Multivariate linear regression model adjusted for age, sex, body mass index Z score, and season.† Univariate linear regression model. Open table in a new tab Table IIAssociations of asthma indices and vitamin D genetic risk scoreClinical phenotypeBeta95% CIP valueAsthma cases∗Univariate logistic regression.−3.51 × 10−4-4.36 × 10−3 to 4.32 × 10−3.85Asthma severity†Univariate ordinal logistic regression.2.15 × 10−2−8.77 × 10−3 to 5.31 × 10−2.16Severe exacerbations (no exacerbations vs any)∗Univariate logistic regression.−8.33 × 10−3−1.02 × 10−1 to 8.51 × 10−2.86Severe exacerbations frequency‡Univariate zero-inflated Poisson regression.−2.14 × 10−2−5.06 × 10−2 to 8.60 × 10−3.17∗ Univariate logistic regression.† Univariate ordinal logistic regression.‡ Univariate zero-inflated Poisson regression. Open table in a new tab Increasing serum 25(OH)D concentration was associated with decreased odds of asthma and was further associated with decreased frequency of severe asthma exacerbations in our cross-sectional analysis. This result supports the idea that our cohort is substantially similar to previously evaluated cohorts. In our well-powered Mendelian randomization analysis, however, the genetic vitamin D propensity score was not associated with asthma or severe asthma exacerbations. Recent studies that show no benefit of vitamin D supplementation for preventing or treating asthma5Castro M. King T.S. Kunselman S.J. Cabana M.D. Denlinger L. Holguin F. et al.National Heart, Lung, and Blood InstituteEffect of vitamin D3 on asthma treatment failures in adults with symptomatic asthma and lower vitamin D levels: the VIDA randomized clinical trial.JAMA. 2014; 311: 2083-2091Crossref PubMed Scopus (196) Google Scholar, 6Martineau A.R. MacLaughlin B.D. Hooper R.L. Barnes N.C. Jolliffe D.A. Greiller C.L. et al.Double-blind randomised placebo-controlled trial of bolus-dose vitamin D3 supplementation in adults with asthma (ViDiAs).Thorax. 2015; 70: 451-457Crossref PubMed Scopus (81) Google Scholar are consistent with the Mendelian randomization results we present here. These findings lend support to the idea that low 25(OH)D concentration may not cause asthma, but that instead asthma predisposes to low 25(OH)D concentration. Fig E2Algorithm for the identification of asthma and asthma severity. LABA, Long-acting beta-agonist; LAMA, long-acting muscarinic antagonists; LTA, leukotriene antagonists.View Large Image Figure ViewerDownload Hi-res image Download (PPT)Table E1Power calculations for cross-sectional analysisSample size642Alpha0.05Case/control ratio0.83Vitamin D deficiency prevalence0.6Power0.80.9Detectable odds ratio1.601.73 Open table in a new tab Table E2Power calculations for Mendelian randomizationSample size5080Alpha0.05K (proportion of cases)0.23Variance explained by propensity scores0.1Odds ratio1.31.41.5Noncentrality parameter7.1912.3718.76Power0.740.940.99 Open table in a new tab