Is the vitamin D status of patients with COVID-19 associated with reduced mortality? A systematic review and meta-analysis

ABSTRACT To conduct a systematic review and meta-analysis of studies assessing the association between serum vitamin D status and mortality in patients with COVID- 19. We searched PubMed and Embase for studies addressing the association of serum vitamin D levels and COVID-19 mortality published until April 24, 2022. Risk ratios (RRs) and 95% confidence interval (CIs) were pooled using fixed or random effects models. The risk of bias was assessed using the Newcastle-Ottawa Scale. The meta-analysis included 21 studies that measured serum vitamin D levels close to the date of admission, of which 2 were case- control and 19 were cohort studies. Vitamin D deficiency was associated with COVID-19 mortality in the overall analysis but not when the analysis was adjusted to vitamin D cutoff levels < 10 or < 12 ng/mL (RR 1.60, 95% CI 0.93-2.27, I2 60.2%). Similarly, analyses including only studies that adjusted measures of effect for confounders showed no association between vitamin D status and death. However, when the analysis included studies without adjustments for confounding factors, the RR was 1.51 (95% CI 1.28-1.74, I2 0.0%), suggesting that confounders may have led to many observational studies incorrectly estimating the association between vitamin D status and mortality in patients with COVID-19. Deficient vitamin D levels were not associated with increased mortality rate in patients with COVID-19 when the analysis included studies with adjustments for confounders. Randomized clinical trials are needed to assess this association.

Global data from the pandemic has demonstrated a mortality rate of 0.9% in patients with COVID-19 and without comorbidities, but this rate grows progressively with the patients' increasing age and number of comorbidities (5).Studies associating serum vitamin D levels with acute respiratory infections (6) have led to an ecological study, which showed that countries with populations with lower serum vitamin D levels have higher infection rates and mortality associated with SARS-CoV-2 infection (7).Additionally, Isaia and cols.(8) have found a correlation between higher levels of solar ultraviolet (UV) radiation in some regions and lower morbidity and mortality rates related to COVID-19.The authors hypothesized that this finding could be related to the vitamin D levels in the population.
Skin exposure to UV radiation determines the local photoconversion of 7-dehydrocholesterol to previtamin D3 (9).Two forms of vitamin D are available, i.e., D2 and D3, and the primary source of vitamin D3 -which comprises about 80% of the vitamin D stored in the body -derives from UV conversion (10).Vitamin D Vitamin D status and COVID-19 mortality Arch Endocrinol Metab.2023;67/2 formed in the skin or obtained from diet is metabolized in the liver into 25-hydroxyvitamin D (25[OH]D), which is the vitamin form used to determine a patient's vitamin D status.This form is then hydroxylated in the kidneys into the active form of the vitamin, i.e., 1,25 dihydroxyvitamin D (1,25[OH] 2 D).
Vitamin D status is defined according to 25(OH)D levels as insufficient when < 30 ng/mL, deficient when < 20 ng/mL, and severely deficient when < 10 ng/mL (11,12) or <12 ng/mL, according to some authors (13,14).Small observational studies analyzing the association of vitamin D deficiency or insufficiency with COVID-19 outcomes have shown conflicting results (15)(16)(17)(18)(19). Based on these considerations, we conducted a systematic review and meta-analysis to assess the association between vitamin D status and mortality in patients with COVID-19.

METHODS
This systematic review was conducted following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA).This study has not been registered.

Data search
In November 2020, two of the investigators performed a search in PubMed, Embase, and Cochrane Library databases.Studies published until April 24, 2022, were included in the analysis.The following search strategy was used: "vitamin D" AND "coronavirus" OR "coronavirus infections" OR "COVID-19" OR "severe acute respiratory syndrome coronavirus 2".Among the articles identified in the search, only those published in English, Spanish, or Portuguese were included in the analysis.

Study selection
Two independent authors screened the retrieved articles.Disagreements were resolved through discussion among all the authors.The authors read the abstracts of the retrieved articles and, after excluding irrelevant studies, proceeded to read the full article for screening.
For inclusion in the analysis, the studies should meet the following population, intervention, control, and outcomes (PICO) criteria: (A) include hospitalized patients with COVID-19 (must have included throat swabs for direct identification of upper respiratory SARS-CoV-2 infection using nucleic acid real-time reverse transcription-polymerase chain reaction [RT-PCR] to confirm the diagnosis of COVID-19); (B) include patients with COVID-19 older than 18 years and with vitamin D status identified as "deficient" or "insufficient" by measurement of serum vitamin D (25[OH]D) levels close to the date of the COVID-19 diagnosis (specifically, up to 30 days before or up to 7 days after the diagnosis); (C) enroll patients with COVID-19 and vitamin D status identified as "sufficient," who were used for comparison; (D) examined the association between vitamin D status and mortality.Studies in which serum vitamin D levels measured at admission were not mentioned, letters to the editor, case reports, and articles reporting ecological, cross-sectional, animal model, or pediatric studies were excluded from the analysis.

Data extraction
The eligible studies included the assessment of death in individuals with measured serum vitamin D levels.The studies should have informed odds ratios (ORs), risk ratios (RRs), or hazard ratios (HRs) with 95% confidence intervals (CIs).The inclusion in the analysis was not restricted by study size.
The data extracted from the studies included information about the authors, study design, country of origin, patients' demographic characteristics (age and sample size), COVID-19 diagnosis, outcomes, and confounders.Two independent investigators extracted the data using a structured form designed by the authors.Disagreements were resolved through discussion among all authors.

Analysis of results
The analysis focused on the impact of vitamin D deficiency on mortality outcomes in patients with COVID-19.For this, we first performed an analysis stratified by vitamin D cutoff level (< 20 or < 25 ng/mL and < 10 or < 12 ng/mL, according to each study).We then performed another stratified analysis including studies that adjusted analyses for confounding factors versus those that did not perform such adjustments.For the present meta-analysis, we standardized the vitamin D measurement unit as ng/mL, and for those studies with levels presented in nmol/L, we transformed the values into ng/mL (10).We also performed sensitivity analysis by omitting individual studies to detect the influence of each study on the overall effect estimate.

Quality assessment, risk of bias, and statistical analysis
The Newcastle-Ottawa Scale was used to evaluate the quality of observational studies.This scale assigns up to 9 points to each study (4 points for selection, 2 points for comparability, and 3 points for exposure or outcome) (20).Studies scoring at least 6 points are considered to have a low risk of bias.The studies included in the present meta-analysis reported ORs, HRs, or RRs; for studies not reporting these effects, the RR was calculated according to the Cochrane Handbook for Systematic Reviews of Interventions (21).
We extracted the effect estimates with the most significant degree of adjustment for potential confounding factors.We also considered HR comparable to RR.For studies reporting ORs, a corrected RR was computed as previously described (22).Pooled RRs and 95% CIs were calculated using fixed or random effects models according to the study's homogeneity.Cochran's Q test and I 2 statistic were used to evaluate, respectively, the statistical significance and the degree of heterogeneity between the studies.An I 2 statistic value ≥ 50% reveals substantial heterogeneity.Sensitivity analysis was also performed, removing each study from the meta-analysis to investigate the source of heterogeneity.Finally, publication bias was examined using Egger's test and funnel plot analysis.All analyses were performed using the software Stata/SE, v.14.1 (StataCorp LLC, College Station, TX, USA).

Sensitivity analyses, assessment of heterogeneity, and risk of bias
Sensitivity analysis was performed excluding, individually, each study that adjusted for confounding factors regarding the mortality outcome.The Cochran's Q test remained unchanged, and the I 2 varied from 44.1%-78.5% for the cutoff levels of < 10 or < 12 ng/mL and 56.7%-74.5% for the cutoff levels of < 20 or < 25 ng/mL, indicating that the result was not influenced by heterogeneity (Table 2).
Table 3 shows the risk of bias in the analyzed studies.The median quality score of the studies was 8 (range 6-9).The estimated bias coefficient was 0.105, with a p value of 0.053, indicating the absence of smallstudy effects.The funnel plot analysis performed also detected no possible small-study effects (Figure 4).Therefore, the tests provided weak evidence for the presence of publication bias.

DISCUSSION
This systematic review and meta-analysis assessed only studies in which serum vitamin D levels were measured close to the date of the COVID-19 diagnosis.COVID-19.However, many of these studies used the level measured between a few months and many years before the diagnosis of COVID-19 (44,45).Vitamin D has known biological variability, and its serum levels can vary from 13%-26% over 4 months.The levels can also vary with age and with the emergence of comorbidities (46)(47)(48)(49)(50).

Observational studies have associated low serum vitamin D levels and poor outcomes in patients with
Despite the overall analysis and the analysis by cutoff levels of < 20 or < 25 ng/mL showing an association between low serum vitamin D levels and mortality, the analysis by cutoff levels of < 10 or < 12 ng/mL did not reveal increased mortality risk.A possible explanation for the association between mortality and vitamin D deficiency for the < 20 or < 25 ng/mL group but not for the < 10 or < 12 ng/mL group in the meta-analysis lies in the weight of the studies that did not adjust their analyses for confounders (54.8% in the < 20 or < 25 ng/mL group versus 4.4% in the < 10 or < 12 ng/mL group), as discussed below.The subgroup analyses including studies that have adjusted the analysis for age and at least one more confounder did not show such association, both in the overall analysis and in the analysis by cutoff levels.However, one analysis including only studies not adjusting for confounders revealed an association between low serum vitamin D levels and an increased (1.51 times) risk of death, which   suggests that confounding factors may have driven many results in previous studies.Therefore, our metaanalysis suggests that vitamin D status has no causal effect on mortality in patients with COVID-19.
Investigating the causality of the association between vitamin D status and severity of COVID-19 infection, Patchen and cols.studied single nucleotide polymorphisms (SNPs) related to the risk of vitamin D deficiency.The authors found no association between genetically predicted differences in long-term vitamin D nutritional status and poor outcomes in patients with COVID-19 (51).
Hypovitaminosis D shares many risk factors with the severe form of COVID-19.Indeed, older age, obesity, chronic kidney disease, diabetes, hypertension, and cardiovascular disease are some critical risk factors that have been reported as associated with the severity of COVID-19 (52)(53)(54)(55).Evidence indicates that vitamin D deficiency can be caused by older age, obesity, and chronic kidney disease (56)(57)(58)(59)(60)(61).Type 2 diabetes, hypertension, and cardiovascular disease also are associated with vitamin D deficiency (62,63).Thus, many confounders may compromise the analysis between serum vitamin D levels and COVID-19 outcomes, and the results of observational studies must be interpreted with caution.Despite the well-known modulatory role of vitamin D in the immune response, the participation of the vitamin in preventing mortality in patients with COVID-19 may be outweighed by other mechanisms involved in the complex pathophysiology of the disease (64,65).
Other hypotheses must also be considered to explain our results.The liver hydroxylates vitamin D into 25(OH)D (calcifediol), the circulating form of the vitamin, which has a plasma half-life of 3 weeks (the whole-body half-life of this form is 2-3 months) (66).The serum calcifediol levels are generally used to check an individual's vitamin D status.Calcifediol is hydroxylated into its active form -i.e., 1,25(OH) 2 D -which has a plasma half-life of 4 hours; this half-life is related to the presence of this form of the vitamin mainly in the kidneys but also in some sites other than the kidneys, including pulmonary epithelial cells (10,66).The serum level of 1,25(OH) 2 D is roughly 1,000 times lower than that of calcifediol (67).The kidney is the main organ regulating serum 1,25 (OH) 2 D (68).The ACE2 protein, which is present in renal and lung cells, is a target for SARS-CoV-2 to enter these cells (69,70).Renal dysfunction caused directly by the virus or indirectly by the presence of acute renal injury associated with COVID-19 infection may lead to a reduction in the activity of 1-alpha hydroxylase, the enzyme that converts calcifediol to 1,25(OH) 2 D (71,72).
Another possible hypothesis to explain our findings is that tissue damage caused by SARS-CoV-2 in the kidneys and, to a lesser extent, in the lungs may lead to an active decrease in 1,25(OH)D levels, which are responsible for the majority of the biological actions of vitamin D, but does not result in decreased levels of the vitamin form usually measured in serum (calcifediol).Previous studies have shown that serum 25(OH)D levels may not correlate with serum 1,25(OH) 2 D levels in some clinical conditions (73)(74)(75).
Reduced calcium and phosphorus have been found in critically ill patients with COVID-19, which may indicate a reduction in 1,25(OH) 2 D in these patients since the active form of vitamin D is an essential regulator of calcium and phosphorus levels acting in intestinal absorption and renal reabsorption (68).
Despite the well-known autocrine and paracrine production of 1,25(OH) 2 D by immune system cells, it is difficult to identify whether serum 1,25(OH) 2 D levels influence the immune response against SARS-CoV-2.Playford and cols.have shown that serum levels of 1,25(OH) 2 D but not those of 25(OH)D are associated with cardiovascular risk factors, a common outcome in patients with severe COVID-19 (76).Likewise, Nguyen and cols.have found that serum 1,25(OH) 2 D levels are a better predictor of mortality by sepsis than those of calcifediol (77).Notably, an antiviral action has been proposed for 1,25(OH) 2 D against SARS-CoV-2 and other viruses (78,79).
The present study has several strengths.To our knowledge, this is the first meta-analysis of studies that have included the vitamin D status from serum 25(OH)D levels measured on or close to the date of hospital admission in patients with COVID-19 and the first to include studies with analyses adjusted for confounders.The main limitations of the present study were the fact that the analysis included different cutoff levels for serum vitamin D in the same subgroup, that vitamin D levels were measured 30 days before hospitalization for COVID-19 in one of the included studies (43), the presence of substantial heterogeneity, and the observational design of the selected studies.
In conclusion, the results of the present study showed that, overall, vitamin D status was associated with Disclosure: no potential conflict of interest relevant to this article was reported.

Figure 2 .
Figure 2. Association between serum vitamin D levels and mortality in patients with COVID-19 according to 25(OH)D cutoff level.

Figure 3 .
Figure 3. Association between serum vitamin D levels and mortality in patients with COVID-19, including studies that adjusted the analysis for age and at least one more confounder (obesity, hypertension, diabetes, renal disease, and cardiovascular disease) and studies without adjustment for confounders.(A) Analysis performed by 25(OH)D cutoff level, including only studies with adjustments; (B) analysis performed by 25(OH)D cutoff level, including studies without adjustments.

Figure 4 .
Figure 4. Funnel plot using data from 21 studies associating serum vitamin D levels and mortality.Four studies were included twice, as they analyzed two 25(OH)D cutoff levels.
Vitamin D status and COVID-19 mortality Arch Endocrinol Metab.2023;67/2 mortality in patients with COVID-19, but not when the analysis included studies that adjusted measures of effect for confounding factors.Confounders may have led to the conclusion of the detrimental effects of low serum 25(OH)D levels in patients with COVID-19 observed in some previous studies.Large randomized clinical trials are needed to assess the effects of vitamin D levels (including those of 1,25[OH] 2 D) on mortality in patients hospitalized with COVID-19.

Table 2 .
Sensitive analysis for the mortality outcome, including studies that performed adjusted analysis for confounders Abbreviation: 25(OH)D: 25-hydroyvitamin D. * Value for heterogeneity among studies assessed with Cochran's Q test.

Table 3 .
Quality assessment of the selected articles according to the Newcastle-Ottawa Scale