Acute Respiratory Tract Infection and 25-Hydroxyvitamin D Concentration: A Systematic Review and Meta-Analysis

Observational studies and randomised controlled studies suggest that vitamin D plays a role in the prevention of acute respiratory tract infection (ARTI); however, findings are inconsistent and the optimal serum 25-hydroxyvitamin D (25(OH)D) concentration remains unclear. To review the link between 25(OH)D concentration and ARTI, we searched PubMed and EMBASE databases to identify observational studies reporting the association between 25(OH)D concentration and risk or severity of ARTI. We used random-effects meta-analysis to pool findings across studies. Twenty-four studies were included in the review, 14 were included in the meta-analysis of ARTI risk and five in the meta-analysis of severity. Serum 25(OH)D concentration was inversely associated with risk and severity of ARTI; pooled odds ratios (95% confidence interval) were 1.83 (1.42–2.37) and 2.46 (1.65–3.66), respectively, comparing the lowest with the highest 25(OH)D category. For each 10 nmol/L decrease in 25(OH)D concentration, the odds of ARTI increased by 1.02 (0.97–1.07). This was a non-linear trend, with the sharpest increase in risk of ARTI occurring at 25(OH)D concentration < 37.5 nmol/L. In conclusion, there is an inverse non-linear association between 25(OH)D concentration and ARTI.


Introduction
Acute respiratory tract infection (ARTI) is very common, with most people experiencing at least one episode of ARTI each year [1]. ARTI includes upper respiratory tract infection (URTI) and lower respiratory tract infection (LRTI). The common cold and influenza are the most common ARTIs globally; the highest rate occurs during the winter months in temperate areas with little seasonal change in tropical regions [2]. During epidemic months, influenza can affect 20% to 50% of people worldwide [2]. Bronchitis and pneumonia are the two most common infections of the lower airways, and mortality due to LRTI is high [3]. In 2015, more than 2.8 million deaths worldwide were due to LRTI; children and the elderly are the most affected groups [3].
Vitamin D is critical for skeletal health and may play a role in other health outcomes, including infection. Vitamin D status is estimated by measuring the serum concentration of 25-hydroxyvitamin D (25(OH)D); this varies seasonally and the lowest concentration in the winter/spring months coincides with the highest ARTI incidence, suggesting a link between vitamin D and ARTI [4]. This is supported by laboratory studies demonstrating an important role of vitamin D in the immune system. Vitamin D promotes the elimination of pathogens and suppresses prolonged inflammatory responses [5][6][7].
It enhances the production of antimicrobial peptides such as defensins and cathelicidins, which offer natural protection against microbial pathogens [8].
Many studies have investigated the link between 25(OH)D serum concentration and ARTI and the effect of vitamin D supplementation on ARTI. Two recent meta-analyses of observational studies considered associations between 25(OH)D concentration and ARTI incidence in children [9,10]. One included case-control studies in children aged ≤ 5 years and found higher odds of 25(OH)D deficiency in those with LRTI [9]. The other found an inverse association between prenatal maternal 25(OH)D concentration and risk of ARTI in the offspring [10]. The most recent systematic review and meta-analysis including older children and adults was published in 2015 as a conference abstract only-it included 19 observational studies and found a significant inverse association between 25(OH)D concentration and risk of ARTI [11]. A systematic review of 25 observational studies and 14 randomised-controlled trials (RCTs) was completed in 2013 [12]. It concluded that there was an inverse association between 25(OH)D concentration and risk of ARTI, but did not include a meta-analysis [12]. Some observational studies have found an inverse link between 25(OH)D concentration and severity of ARTI, as measured by duration of the illness, hospitalisation and severity index [13][14][15][16]; however, there has been no meta-analysis of these findings.
Results from RCTs investigating the effect of vitamin D supplementation on ARTI are inconsistent. Two meta-analyses found a significant benefit of vitamin D supplementation on ARTI [17,18] while another three did not [19][20][21]. There are indications of a greater protective effect in people with marked vitamin D deficiency [20,22], but the trials were unable to indicate an optimal concentration of 25(OH)D. We therefore conducted a systematic review and meta-analysis of observational studies to evaluate the link between serum 25(OH)D concentration and the risk and severity of ARTI in adolescents and adults. Findings from this meta-analysis will provide insights into the influence of vitamin D on ARTI risk and severity, and give an indication of the optimal 25(OH)D concentration for ARTI prevention and management.

Materials and Methods
The study protocol was registered with the PROSPERO International Prospective Register of Systematic Reviews prior to commencement [23].

Search and Screening Strategy
PubMed and EMBASE databases were searched from their inception until 12th June 2019. Keywords were chosen from the Medical Subject Headings (MeSH) terms in PubMed and explosion (exp) of EMTREE terms in EMBASE. Essentially, we searched for the terms "vitamin d" or "25-hydroxyvitamin D" or "25OHD" or "25(OH)D" or "hypovitaminosis D" in combination with "respiratory tract infection*" or "respiratory infectio*" or "respiratory diseas*" or "pneumonia" or "influenza" or "bronchiolitis" or "common cold". The complete search strategies are shown in the supplementary material (Supplementary File S1).
All titles, abstracts and full text were independently screened by two investigators, Hai Pham and Aninda Rahman. Any discrepancies were resolved through consultation with Rachel Neale and Mary Waterhouse.

Inclusion/Exclusion Criteria
Inclusion criteria: The eligibility criteria were: (1) observational studies including cohort, case-control and cross-sectional studies; (2) published in English; (3) full text available; (4) reported the association between circulating 25(OH)D concentration and ARTI risk or severity; (5) the target population was healthy adults or adolescents aged 12 years and older.
Exclusion criteria: Studies investigating the link between 25(OH)D concentration and tuberculosis or chronic lung conditions such as chronic obstructive pulmonary disease and asthma were excluded.

Definition of Outcome
Primary outcome: The primary outcome was the risk of ARTI, defined as an acute infection of the respiratory tract in either the lower or upper airway or with the location not specified. ARTI was either self-reported via surveys or symptom diaries, or clinically confirmed with or without evidence from X-rays or laboratory tests.
Secondary outcome: The secondary outcome was the severity of ARTI, defined according to the duration of the illness, hospitalisation, admission to an intensive care unit, symptom severity score or index, or mortality.

Quality Assessment
We used the customised Newcastle-Ottawa scale (NOS) to assess the quality of each study [24]. The NOS tools are slightly different for each study design but generally include three main categories, namely: (1) selection of participants; (2) control for confounders; and (3) measurement of exposure or outcome (Supplementary File S2).

Statistical Analysis Methods
We used STATA 13 (StataCorp, College Station, Texas, USA) and SAS 9.4 (SAS Institute Inc., Cary, NC, USA) for statistical analyses. We included studies in the meta-analysis that reported either:
Sufficient data to derive two by two tables of ARTI risk, comparing the lowest versus the highest 25(OH)D category.
We included the estimate from the most fully adjusted model for each study in the meta-analysis comparing the risk of ARTI in the lowest versus highest 25(OH)D category. When a study reported results using different 25(OH)D thresholds, we used 50 nmol/L as it was the most commonly used threshold. Two studies reported HRs as the measure of association between 25(OH)D concentration and risk of ARTI [25,26]; we included these two studies in the meta-analysis, considering the HR as an approximation of the OR. A random-effects model was used to pool the results.
We used a method described by Greenland and Longnecker to estimate trends across categories of exposure to calculate the effect of each 10 nmol/L decrease in 25(OH)D concentration on the risk of ARTI [27,28]. This method estimates the covariances between multivariable-adjusted odds ratios using the number of cases in each exposure category. We used variance least-squares regression to compute the trends in two studies in which the number of cases in each exposure category was not reported [25,29]. The representative value for each 25(OH)D category was assigned using either the midpoint of the range or by subtracting or adding the half width of the adjacent exposure category for open-ended categories [30]. We also used fixed effects to fit a restricted cubic spline model with 3 knots to evaluate a potential non-linear dose-response association between 25(OH)D concentration and ARTI risk [28].
For the overall analysis of severity, we defined a severe ARTI to be one that had a moderate-to-high severity score or resulted in death (no studies based on other severity outcomes were included). We also performed a separate meta-analysis to assess the association between 25(OH)D concentration (lowest versus highest category) and ARTI mortality.

Assessment of Heterogeneity and Publication Bias
Heterogeneity across studies was assessed using I 2 statistics. We conducted meta-analyses comparing the risk of ARTI in the highest versus the lowest 25(OH)D category stratified by: (i) infection location if reported (URTI or LRTI); (ii) outcome measurement methods (self-reported or clinically confirmed); (iii) mean 25(OH)D concentration in the study population (<60 nmol/L or ≥60 nmol/L; cut point chosen to maximize the number of studies included in each subgroup); and (iv) factors considered as confounders (fully adjusted model or crude/non-fully adjusted model) to explore potential sources of heterogeneity. We used funnel plots and the Egger test to test for publication bias. Figure 1 illustrates the study selection process. The PubMed and EMBASE searches yielded a total of 1589 records after removing duplicates. We screened the full text of 33 studies and excluded nine that did not meet the selection criteria. Of the 24 studies included, 10 reported the association between 25(OH)D and the risk of ARTI [25,29,[31][32][33][34][35][36][37][38], eight reported on the severity of ARTI [14,16,[39][40][41][42][43][44], and six reported on both the risk and severity [13,15,26,[45][46][47].  Figure 1 illustrates the study selection process. The PubMed and EMBASE searches yielded a total of 1589 records after removing duplicates. We screened the full text of 33 studies and excluded nine that did not meet the selection criteria. Of the 24 studies included, 10 reported the association between 25(OH)D and the risk of ARTI [25,29,[31][32][33][34][35][36][37][38], eight reported on the severity of ARTI [14,16,[39][40][41][42][43][44], and six reported on both the risk and severity [13,15,26,[45][46][47].

Study Characteristics
The characteristics of the included studies are summarised in Tables 1 and 2. There were three case-control studies, 13 cross-sectional studies and eight prospective cohort studies. The relationship between ARTI risk or severity and 25(OH)D status was reported either by categories of exposure or per unit increase in exposure. The cut point for 25(OH)D categories varied across studies; the upper limit of the lowest category ranged from 25 to 95 nmol/L, and the cut point for the highest category ranged from 25 to 120 nmol/L (Tables 1 and 2). Serum 25(OH)D concentration was measured using liquid chromatography-mass spectrometry (LCMS) in four studies and non-chromatography assays in 19 studies. One did not specify which method was used to measure 25(OH)D concentration [44].

Study Characteristics
The characteristics of the included studies are summarised in Tables 1 and 2. There were three case-control studies, 13 cross-sectional studies and eight prospective cohort studies. The relationship between ARTI risk or severity and 25(OH)D status was reported either by categories of exposure or per unit increase in exposure. The cut point for 25(OH)D categories varied across studies; the upper limit of the lowest category ranged from 25 to 95 nmol/L, and the cut point for the highest category ranged from 25 to 120 nmol/L (Tables 1 and 2). Serum 25(OH)D concentration was measured using liquid chromatography-mass spectrometry (LCMS) in four studies and non-chromatography assays in 19 studies. One did not specify which method was used to measure 25(OH)D concentration [44].   ARTI risk was measured by (i) self-report via survey questionnaire or symptom diaries (n = 10), or (ii) clinical diagnosis with or without confirmation from X-ray or a laboratory test (n = 6). ARTI severity was measured using a severity index (n = 6) such as CURB or CURB-65 (confusion, uraemia, respiratory rate, low blood pressure, age ≥ 65), pulmonary severity score (PSI), or symptom severity score (SSC); duration of the illness (n = 10); admission to hospital or intensive care units (n = 4); or mortality (n = 6). All 24 studies had a moderate-to-high quality score (Table S1). ARTI risk was measured by (i) self-report via survey questionnaire or symptom diaries (n = 10), or (ii) clinical diagnosis with or without confirmation from X-ray or a laboratory test (n = 6). ARTI severity was measured using a severity index (n = 6) such as CURB or CURB-65 (confusion, uraemia, respiratory rate, low blood pressure, age ≥ 65), pulmonary severity score (PSI), or symptom severity score (SSC); duration of the illness (n = 10); admission to hospital or intensive care units (n = 4); or mortality (n = 6). All 24 studies had a moderate-to-high quality score (Table S1).

Association between 25(OH)D Concentration and Risk of ARTI
We included 14 of the 16 studies that assessed the association between 25(OH)D concentration and risk of ARTI in the meta-analysis comparing the risk of ARTI in the lowest versus the highest 25(OH)D category (N participants = 78,127) and 10 studies were included in the trend analysis (N participants = 69,048). Five studies reported risk of ARTI in at least three categories of 25(OH)D concentration and the number of cases by exposure category; these were included in the evaluation of whether there is a non-linear relationship between ARTI risk and 25(OH)D concentration (N participants = 37,902).

Association between 25(OH)D Concentration and Severity of ARTI
Five studies (N participants = 1495) were included in the meta-analysis of the odds of severe ARTI or mortality combined, comparing the highest and the lowest 25(OH)D category and four studies (N participants = 1422) were included in the mortality meta-analysis. The pooled ORs for severity/mortality combined and mortality separately were 2.46 (95% CI 1.65-3.66; I 2 = 49.8%; p = 0.093) and 3.00 (95% CI 1.89-4.78; I 2 = 66.7%; p = 0.029), respectively ( Figure 5). The duration of ARTI was also inversely associated with 25(OH)D concentration; low 25(OH)D concentration was associated with a more prolonged ARTI in 7 out of 10 studies (Table 3).

Association between 25(OH)D Concentration and Severity of ARTI
Five studies (N participants = 1495) were included in the meta-analysis of the odds of severe ARTI or mortality combined, comparing the highest and the lowest 25(OH)D category and four studies (N participants = 1422) were included in the mortality meta-analysis. The pooled ORs for severity/mortality combined and mortality separately were 2.46 (95% CI 1.65-3.66; I 2 = 49.8%; p = 0.093) and 3.00 (95% CI 1.89-4.78; I 2 = 66.7%; p = 0.029), respectively ( Figure 5). The duration of ARTI was also inversely associated with 25(OH)D concentration; low 25(OH)D concentration was associated with a more prolonged ARTI in 7 out of 10 studies (Table 3).

Association between 25(OH)D Concentration and Severity of ARTI
Five studies (N participants = 1495) were included in the meta-analysis of the odds of severe ARTI or mortality combined, comparing the highest and the lowest 25(OH)D category and four studies (N participants = 1422) were included in the mortality meta-analysis. The pooled ORs for severity/mortality combined and mortality separately were 2.46 (95% CI 1.65-3.66; I 2 = 49.8%; p = 0.093) and 3.00 (95% CI 1.89-4.78; I 2 = 66.7%; p = 0.029), respectively ( Figure 5). The duration of ARTI was also inversely associated with 25(OH)D concentration; low 25(OH)D concentration was associated with a more prolonged ARTI in 7 out of 10 studies (Table 3).

Studies Excluded from the Meta-Analysis
Two studies were excluded from the meta-analysis of ARTI risk, and nine studies were excluded from the meta-analysis of ARTI severity. Reasons for exclusion and the main findings of the excluded studies are provided in Table S2. Studies that reported mean (SD) or median (interquartile range (IQR)) of 25(OH)D concentration in people with severe versus non-severe ARTI or mean (SD) severity score by 25(OH)D category were excluded. As different scales were used to measure ARTI severity across studies, we could not pool the mean difference in severity score between the highest and the lowest 25(OH)D category. Overall, results from 6/9 excluded studies showed an inverse association between 25(OH)D concentration and severity of ARTI measured by duration of the illness, severity score, or the number of days absent from duty (Table S2).

Subgroup Analysis and Publication Bias
We found evidence of significant heterogeneity (I 2 = 78.8%, p < 0.001); therefore, we conducted stratified meta-analyses to explore potential sources of heterogeneity ( Figure 6A-D). The association between 25(OH)D concentration and LRTI was stronger than for URTI, but the former had significant heterogeneity ( Figure 6A). Similarly, the association was stronger for studies with mean 25(OH)D

Studies Excluded from the Meta-Analysis
Two studies were excluded from the meta-analysis of ARTI risk, and nine studies were excluded from the meta-analysis of ARTI severity. Reasons for exclusion and the main findings of the excluded studies are provided in Table S2. Studies that reported mean (SD) or median (interquartile range (IQR)) of 25(OH)D concentration in people with severe versus non-severe ARTI or mean (SD) severity score by 25(OH)D category were excluded. As different scales were used to measure ARTI severity across studies, we could not pool the mean difference in severity score between the highest and the lowest 25(OH)D category. Overall, results from 6/9 excluded studies showed an inverse association between 25(OH)D concentration and severity of ARTI measured by duration of the illness, severity score, or the number of days absent from duty (Table S2).

Subgroup Analysis and Publication Bias
We found evidence of significant heterogeneity (I 2 = 78.8%, p < 0.001); therefore, we conducted stratified meta-analyses to explore potential sources of heterogeneity ( Figure 6A-D). The association between 25(OH)D concentration and LRTI was stronger than for URTI, but the former had significant heterogeneity ( Figure 6A). Similarly, the association was stronger for studies with mean 25(OH)D concentration < 60 nmol/L than in those with mean 25(OH)D concentration ≥ 60 nmol/L, but again with significant heterogeneity ( Figure 6C). There was only a small difference in the pooled OR between studies that used a fully adjusted model and those that reported estimates from crude models or those with less complete adjustment, but the latter had significant heterogeneity (I 2 = 87.2%, p < 0.001) ( Figure 6D). concentration < 60 nmol/L than in those with mean 25(OH)D concentration ≥ 60 nmol/L, but again with significant heterogeneity ( Figure 6C). There was only a small difference in the pooled OR between studies that used a fully adjusted model and those that reported estimates from crude models or those with less complete adjustment, but the latter had significant heterogeneity (I 2 = 87.2%, p < 0.001) ( Figure 6D). The funnel plot shows evidence of significant publication bias (p = 0.024) (Figure 7), suggesting that studies with small sample size and insignificant findings were not published. The funnel plot shows evidence of significant publication bias (p = 0.024) (Figure 7), suggesting that studies with small sample size and insignificant findings were not published.

Discussion
In this systematic review and meta-analysis, we observed significant associations between 25(OH)D concentration and both risk and severity of ARTI, but with significant heterogeneity and evidence of publication bias. There was a non-linear association between 25(OH)D concentration and risk of ARTI, with evidence of a more marked increased risk for 25(OH)D concentration below 37.5 nmol/L. Findings from our meta-analysis of ARTI risk, which includes the largest number of participants, are consistent with results from previous reports. A meta-analysis published in 2015 as an abstract (n = 44,301) reported an increased risk of ARTI in those with 25(OH)D < 50 nmol/l compared with those ≥ 50 nmol/L. The odds ratio was higher than we found (2.63 vs. 1.83); their analysis included children as well as adults and they did not present an estimate for adults and adolescents only [11]. A systematic review including 25 observational studies (10 in adults) reported significant inverse association between 25(OH)D concentration and risk of ARTI, but no formal meta-analysis was conducted [12]. The link between 25(OH)D concentration and risk of ARTI was also reported separately for children and LRTI; a meta-analysis showed a higher prevalence of vitamin D deficiency (< 50 nmol/L) in children with LRTI (n = 550) [9]. This is the first study to our knowledge which performed a meta-analysis of the association between 25(OH)D concentration and severity of ARTI. We found a stronger association between 25(OH)D concentration and severity of ARTI than with risk of ARTI (OR 2.46 vs. 1.83). The result accords with findings from a systematic review that highlighted the potential link between vitamin D deficiency and severe LRTI in children, but no meta-analysis was included [9]. It is difficult to metaanalyse the association between 25(OH)D concentration and severity of ARTI because of variability across studies with respect to variability in assay methods used to measure serum 25(OH)D; the cut points used to categorise 25(OH)D concentration; the scales used to measure severity; and the measures used to estimate effect. We could only include five studies in the severity meta-analysis so were unable to assess the association within subgroups or investigate a potential non-linear trend.
Results from RCTs are inconsistent. Some trials found that vitamin D supplementation reduced the risk of ARTI [48][49][50] while others did not [51][52][53][54]. The inconsistency may be due to differences in study design, vitamin D supplement doses and regimens, and different baseline 25(OH)D concentration. The most recent meta-analysis, using individual participant data, found that vitamin D supplementation reduced the risk of ARTI more strongly in people with 25(OH)D concentration < 25 nmol/L [22]. Our findings suggest that supplementation may be of most benefit in people with a

Discussion
In this systematic review and meta-analysis, we observed significant associations between 25(OH)D concentration and both risk and severity of ARTI, but with significant heterogeneity and evidence of publication bias. There was a non-linear association between 25(OH)D concentration and risk of ARTI, with evidence of a more marked increased risk for 25(OH)D concentration below 37.5 nmol/L. Findings from our meta-analysis of ARTI risk, which includes the largest number of participants, are consistent with results from previous reports. A meta-analysis published in 2015 as an abstract (n = 44,301) reported an increased risk of ARTI in those with 25(OH)D < 50 nmol/l compared with those ≥ 50 nmol/L. The odds ratio was higher than we found (2.63 vs. 1.83); their analysis included children as well as adults and they did not present an estimate for adults and adolescents only [11]. A systematic review including 25 observational studies (10 in adults) reported significant inverse association between 25(OH)D concentration and risk of ARTI, but no formal meta-analysis was conducted [12]. The link between 25(OH)D concentration and risk of ARTI was also reported separately for children and LRTI; a meta-analysis showed a higher prevalence of vitamin D deficiency (< 50 nmol/L) in children with LRTI (n = 550) [9]. This is the first study to our knowledge which performed a meta-analysis of the association between 25(OH)D concentration and severity of ARTI. We found a stronger association between 25(OH)D concentration and severity of ARTI than with risk of ARTI (OR 2.46 vs. 1.83). The result accords with findings from a systematic review that highlighted the potential link between vitamin D deficiency and severe LRTI in children, but no meta-analysis was included [9]. It is difficult to meta-analyse the association between 25(OH)D concentration and severity of ARTI because of variability across studies with respect to variability in assay methods used to measure serum 25(OH)D; the cut points used to categorise 25(OH)D concentration; the scales used to measure severity; and the measures used to estimate effect. We could only include five studies in the severity meta-analysis so were unable to assess the association within subgroups or investigate a potential non-linear trend.
Results from RCTs are inconsistent. Some trials found that vitamin D supplementation reduced the risk of ARTI [48][49][50] while others did not [51][52][53][54]. The inconsistency may be due to differences in study design, vitamin D supplement doses and regimens, and different baseline 25(OH)D concentration. The most recent meta-analysis, using individual participant data, found that vitamin D supplementation reduced the risk of ARTI more strongly in people with 25(OH)D concentration < 25 nmol/L [22]. Our findings suggest that supplementation may be of most benefit in people with a 25(OH)D concentration ≤ 37.5 nmol/L, with some benefit up to 60 nmol/L, although these values should be considered with caution in light of the different assays and markedly varying cut point used across studies.
We observed high heterogeneity, and this mostly persisted in subgroup analyses. Despite this, the direction of effect was consistent; lower 25(OH)D concentration was associated with increased risk or severity of ARTI. There are indications of publication bias, indicating that small studies that showed no significant association were either not identified or not published. Publication bias has also been observed in meta-analyses of RCTs [17,22]. It is thus possible that the benefits of vitamin D for reducing risk or severity of ARTI have been over-estimated.

Conclusions
Our study is the largest to date and to the best of our knowledge, is the first to include a meta-analysis of the association between 25(OH)D concentration and severity of ARTI. Our findings suggest an important role of vitamin D in prevention of ARTI risk and severity, particularly in people with low 25(OH)D concentration. However, it is challenging to identify an optimal 25(OH)D concentration or a concentration below which supplementation would be of benefit due to the lack of consistency in both 25(OH)D assays and reporting across studies. It is important to improve consistency of reporting, as well as assays, to enable the field to move forward.

Conflicts of Interest:
The authors declare no conflict of interest.