Introduction
Vitamin D deficiency, defined as serum 25-hydroxyvitamin D
3 (25 (OH)-D) concentrations below 50 nmol/L, is highly prevalent in Dutch critically ill patients [
1]. Several studies in critically ill patients report associations between vitamin D deficiency, a disturbed parathyroid hormone (PTH)-vitamin D axis and increased mortality [
2]-[
5]. A biological basis how hypovitaminosis D may cause mortality could be hypocalcaemia. Hypocalcaemia is a well-known abnormality in critically ill patients in the course of sepsis and rhabdomyolysis [
6]. Second, vitamin D regulates both innate and adaptive immune systems. Vitamin D deficiency leads to immune dysregulation and has been proposed as an underlying pathogenic mechanism of infections [
7]. Third, vitamin D deficiency is associated with increased markers of systemic inflammation associated with multi-organ failure [
8]. Moreover hypovitaminosis D reduces, despite maximal upregulation of PTH levels, formation of 1,25-dihydroxyvitamin D
3 (1,25 (OH)-D) at the tissue level. This may be critical in mediating the beneficial pleiotropic functions of vitamin D, involving innate immunity, mucosal barrier and endothelial function. Recently, a systematic review and meta-analysis including observational and interventional studies on vitamin D in non-critically ill patients, suggests an association of deficiency with cardiovascular diseases, diabetes, and all-cause mortality in the former but not in the latter studies [
9]. In non-critically ill patients, of a prior meta-analysis of 18 randomised controlled trials, intake of supplementary doses of vitamin D was associated with a 7% decrease in mortality [
10].
Therefore, we conducted a systematic review to pool the available data and to study the possible effect of vitamin D deficiency in critically ill patients on the incidence of infection, sepsis and association with mortality.
Discussion
This study represents the first comprehensive systematic review and meta-analysis focused on studies in which the effects of vitamin D deficiency in critically ill patients on occurrence of infection, sepsis and mortality rates are described. These results show that vitamin D deficiency (<50 nmol/L) is associated with increase in infection rate, sepsis, 30-day mortality and in-hospital mortality in adult critically ill patients, worldwide.
The association between vitamin D status and immunity has been already supported by a number of studies [
31]-[
33]. However, results in healthy volunteers exposed to experimental human endotoxaemia suggest a lack of this association between vitamin D and inflammatory cytokine levels [
34]. Therefore, it is suggested that the differences in the ability to produce vitamin D, may contribute to a difference in the susceptibility to microbial infection. Our study adds to the discussion on the association by an objectively derived pooled risk. The relation between 25 (OH)-D levels and sepsis has been described previously [
8]. Vitamin D-deficient patients are at higher risk for blood culture positivity, which may contribute to higher sepsis rates [
3]. However, Cecchi
et al. found no clear relationship on outcome between lower vitamin D levels in septic patients when compared with a matched cohort [
16]. The most recent study by Moromizato
et al. included in our analysis specifically showed a threshold of 25 (OH)-D less than or equal to 40 nmol/L to be associated with sepsis [
26]. Thus our results are in agreement with the hypothesis that vitamin D deficiency is a contributor to sepsis.
Some published studies [
17],[
35] suggested an association between vitamin D deficiency and mortality in critically ill patients. In the study by Van den Berghe
et al. both 25 (OH)-D and 1,25 (OH)-D levels were lower among non-survivors in critically ill patients [
17]. Matthews
et al. noted in their surgical ICU cohort that most deaths occurred at vitamin D levels less than 32 nmol/L and that no deaths occurred at levels higher than 65 nmol/L [
30]. The CopD study done by Durup
et al. reported a reversed J-shape relation between 25 (OH)-D and all-cause mortality, suggesting that too much and too little are deleterious. A serum 25 (OH)-D of 50 to 60 nmol/L was associated with the lowest mortality risk [
36]. The results of this meta-analysis suggest that vitamin D levels below 50 nmol/L, increase 30-day mortality and in-hospital mortality with 76% and 79% respectively. To date, only four randomised trials in adult critically ill patients have been published, which were designed to study normalisation of vitamin D levels and its possible adverse effects such as hypercalcaemia and hypercalciuria [
17]-[
20]. These studies were not sufficiently powered to investigate the effects of vitamin D normalisation and potential benefits on hard outcomes such as incidence of severe infections and/or ICU mortality (S11 Table 3 in Additional file
1). The recently published
Lancet review supports the relation between 25 (OH)-D deficiency and all-cause mortality in observational studies [
9]. The discrepancy with the interventional studies could be due to underpowered numbers, low dosages or short duration of supplementation. The role for supplementation is unclear and appropriate dose-response studies with 1,25 and 25 (OH)-D must be done. Therefore the authors’ conclusion about low vitamin D as merely a marker of disease has to be confirmed in prospective interventional studies.
Data from biochemical and molecular studies indicate that vitamin D, in particular its active form 1,25 (OH)-D, has a much wider role than only the maintenance of calcium homeostasis and bone health. Sufficiency of vitamin D activity can thus also be defined by sufficient autocrine and paracrine production of 1,25 (OH)-D at serum 25 (OH)-D levels of at least around 75 nmol/L [
37],[
38]. This active form is responsible for most, if not all, of the biological and pleiotropic effects including antimicrobial actions and immunomodulatory effects of vitamin D [
7],[
39]. The study by Zittermann
et al. demonstrated the superiority of predicting mortality by 1,25 (OH)-D as compared to 25 (OH)-D, supporting the assumption that adequately circulating 1,25 (OH)-D levels may play a role for survival [
40]. Marshall
et al. emphasised measuring 1,25 (OH)-D instead of total 25 (OH)-D as well; they postulated that the disease processes regulate vitamin D metabolism so that the low 25 (OH)-D levels observed in disease may be merely a biomarker of disease severity [
39]. Unfortunately, the observational articles meta-analysed in this manuscript do not consider 1,25 (OH)-D on outcomes.
Two studies [
3],[
5] had a time lag between admission into the ICU and vitamin D blood sampling. To illustrate the importance between 25 (OH)-D time of measurement and admission, the authors conducted a sensitivity analysis considering patients with 25 (OH)-D drawn before or after 90 days prior to hospital admission. This sensitivity analysis showed that the association that was found between vitamin D on outcomes was not modified by time lag.
Our study has several limitations. First, we included both prospective and retrospective studies in this meta-analysis, which is a matter of debate. In retrospective studies the control for confounders is difficult. We have addressed this by adding a subgroup analyses per outcome based on design, see S10 in Additional file
1. However, in the prospective study data the effect on mortality is lower, whereas the sample size in the prospective studies may have been insufficient to show an association. The five retrospective papers [
3]-[
5],[
25],[
26] contain large sample sizes enabling multivariate analysis for mortality ruling out confounders such as age, gender, race, glomerular filtration rate (GFR), C-reactive protein (CRP), season, disease severity and so on. Second, the studies included are observational so that a causative link between hypovitaminosis D and outcomes cannot be established. Additionally, the variability in measured 25 (OH)-D levels is probably multifactorial. It is possible that a random single 25 (OH)-D measurement in ICU patients does not appropriately reflect the vitamin D status [
41]. Furthermore, alterations in vitamin D binding protein, fluid shifts [
42] and assay variability with coefficients of variation ranging from 6% to 13% [
43] may limit applicability of single measurements on outcome prediction used in most of the included studies. The different cutoff levels used by different studies are based on study endpoints (for example, fracture or osteoporosis) done in the general population. The applicability of these cutoff levels in the critically ill is unclear, especially because cutoff values between bone- specific and pleiotropic endpoints are different. There is heterogeneity in the definitions of infection and sepsis used in the included studies, sepsis was defined varying from positive blood cultures [
3] to SIRS criteria together with a source of infection [
23] but the I
2 test did not show heterogeneity. We only found some heterogeneity (
P = 0.08) in the forest plot combining studies with infection as outcome.
Competing interests
The authors declare they have no competing interests. None of the authors received financial support for this study.
Authors’ contributions
KdH, ABJG and AS designed the study; KdH and ME performed the meta-analysis. KdH prepared the manuscript. ABJG supervised, and AS and HRHG edited the manuscript. All authors read and approved the final manuscript.