Data Extraction
Eligible studies included assessment of death or disease severity in individuals with plasma vitamin D or vitamin D non-use. They should provide odds ratio (OR), risk ratio (RR), or hazard ratio (HR) with 95% confidence intervals (CI). Inclusion was not restricted by study size.
Data extraction was performed by two independent investigators. Data extracted included authors, study design, country of origin, demographic characteristics (age, sex, and sample size), COVID-19 diagnosis, outcomes relevant studies on the study question. Disagreements were resolved through discussion among all authors.

Results evaluation

The primary analysis focused on the outcomes: Mortality in patients with COVID-19 in the low plasma vitamin D levels group compared with plasma sufficient vitamin D levels group or supplement vitamin D group compared with non-supplement vitamin D group. The secondary analysis focused on need for hospitalization, ICU admission, and need for mechanical ventilation.
We performed a stratified analysis by plasma vitamin D level and hospitalized patients. Besides, a sensitivity analysis was performed when necessary omitting each study to detect the influence on the estimate of the overall effect.

Quality assessment and statistical analysis

This systematic review was conducted in accordance with the Preferred Items guidelines for Reporting for Systematic Reviews and Meta-Analysis (PRISMA), and this study has not been registered.
The New-Castle-Ottawa quality scale15 was used to evaluate the quality of the observational studies. The Cochrane risk of bias tool was used for randomized controlled trials16. In the case of non-randomized interventional studies, the ROBINS - I tool was used17.
Studies included in the meta-analysis reported OR, HR, or RR. For studies that did not report these measures of effects, the RR calculation was based on the Cochrane Handbook for Systematic Reviews18.
Effect estimates with the greatest degree of adjustment for potential confounding factors were extracted. HR was considered comparable to RR. For studies that reported OR, a corrected RR was computed as already described19. Pooled RR and 95% confidence interval (CI) were calculated using a fixed or random-effects model according to the homogeneity of the studies. The Cochran Q test and the I2 statistic were used to evaluate the statistical significance and degree of heterogeneity between the studies, respectively. The statistic I2≥50% reveals substantial heterogeneity. Finally, the publication bias was examined by the Egger test. All analyses were performed with Stata/SE v.14.1 software (StataCorpLP, USA).

RESULTS

Characteristics of the selection studies
Four-hundred and six (406) studies were identified through database research. Of these, 353 studies are duplicate articles or were excluded based on predetermined eligibility criteria during title/abstract review. Excluded criteria were ecological studies, case reports, cross-sectional, not human, not on vitamin D level or not vitamin D supplementation at COVID-19. We identified 17 studies13,20,29–35,21–28 (total 3,108 participants) that were eligible for this review (Figure 1), of which 16 were involved in the meta-analysis. Eleven of them was cohort20,21,35,24,26,28–31,33,34, one case-control27, one randomized clinical trial25, and two quasi-experimental studies22,23. It was not possible to extract data from Pizzini et al32 for analysis. The characteristics of selected studies and participants are summarized in Table 1.
Plasma vitamin D level and mortality in patients with COVID-19
The mortality outcome was extracted from 10 studies, but the adjusted analysis used 9. As shown in figure 2, the mortality in patients with deficient plasma vitamin D levels was significantly high when compared to patients with sufficient plasma vitamin D levels (Adjusted analysis, removing Carpagnano et al, RR=1.41, 95%CI 1.13 – 1.69, I2=0.0%). The analysis included Carpagnano et al13 showed no has significant statically (RR=7.41, 95%CI -2.27 – 17.09) with high heterogeneity (I2=99.9%).
In subgroup analysis, hospitalized patients with deficient plasma vitamin D levels also have an increased risk of mortality (RR=1.42, 95%CI 1.14 – 1.71, I2=0.0%). The plasma vitamin D levels adopted in studies was <10 ng / ml by one study, <12 ng / ml in four studies, <20 ng / ml in five studies and <25 ng / ml in one study.
Vitamin D supplementation and risk of deaths in patients with COVID-19
The pooled RR of deaths in COVID-19 patients in vitamin D supplementation versus non-vitamin D supplementation was 0.10 (95%CI 0.07 – 0.28), without significant heterogeneity (I2=0.0%) (Figure 3). Subgroup analysis showed RR = 0.10 (95% CI 0.08 - 0.28) for regular vitamin D supplementation.
Plasma vitamin D level and hospitalization in patients with COVID-19The data were extracted from two studies30,31 that showed that there was no increase in the risk of hospitalization in patients with COVID-19 when the serum vitamin D is low (OR=1.80, 95%CI 0.97-2.64, I2=0.0%) (Figure 4). Mendy et al30 analyzed patients with plasma vitamin D levels lass 12 ng/ml and Merzon et al31 analyzed levels lass 30 ng/ml.Plasma vitamin D level, vitamin D supplementation and ICU admission in patients with COVID-19The data was extracted from three studies21,27,30(Figure 5), two cohorts, and one case-control. The analysis showed that patients with low vitamin D level have was increased risk of ICU admission (RR=1.76, 95%CI 1.03-2.49, I2=0.0%). In the analysis of vitamin D supplementation, the data was extracted from two studies, one randomized clinical trial25 with calcifediol treatment and one cohort35 that analyzed at the use of cholecalciferol (Figure S1). The analysis showed that the treatment group has was a decreased risk of ICU admission (RR=0.04, CI95% 0.07-0.16, I2=0.0%).Plasma vitamin D level and ventilation requirement in patients with COVID-19The data was extracted and pooled from three cohorts (Figure 6). Alguwaihes et al21 analyzed patients with plasma vitamin D level lass 5 ng/ml, Baktash et al24 analyze level lass 12 ng/ml, and Radujkovic et al33 analyzed patients with plasma vitamin D level lass 12 ng/ml and lass 20 mg/ml separately. Comparing the deficient plasma vitamin D level with the control group, the results suggested that there is an increased risk of ventilation requirement (RR=3.58, 95%CI 1.45-5.70, I2=0.0%).Plasma vitamin D level and COVID-19 severity The data COVID-19 severity (Outcomes combined analysis: Death, ICU, and Mechanical Ventilation) was extracted and pooled from three cohort21,29,30 and one case-control27. The analysis showed that patients with low vitamin D level have was increased risk of deaths, ICU admission, or mechanical ventilation (RR=1.77, 95%CI 1.13-2.42, I2=0.0%) (Figure 7).Sensitivity analyses, assessment of heterogeneity
For sensitivity analyses were performed by excluding one study at a time for the mortality outcome. The inclusion of Carpagnano et al13 in any scenario raised the I2from 0.0% to 99.9%, which led us to carry out the adjusted analysis (Figure 2). Carpagnano et al13, despite showing an effect measure for mortality with low vitamin D level (<10 ng / ml), had only 3 deaths in their study (two in the group with vitamin D deficiency). As the others analysis did not show heterogeneity (I2=0.0% to I2=6.2%), a sensitivity analysis was not performed.