Table 2.
Limitations and additional characteristics of the included studies
| Study | Sample size in the study | Sample size included in meta-analyses |
Outcomes evaluated in the study | Outcomes evaluated in the meta-analyses | Findings in the study1 | Limitations in extracting data | ||
|---|---|---|---|---|---|---|---|---|
| D-CIMA | D-CSMA | D-CMMA | ||||||
| Abdollahi et al. [27] | 402 | 402 | – | – | I | I | I(+) | VDD was defined as <10 ng/mL, but the sample distribution for VDD was not suitable for statistical analysis; A statistical comparison was made for VDIS <30 ng/mL; Therefore, the data belonging to VDIS were included in the meta-analysis, and the measurement unit was assigned as “other” |
| Baktash et al. [28] | 105 | – | – | 70 | I-M | M | I(+); M(-) | There was uncertainty in the definition of comparison groups for COVID-19 infection; We sent an email to the corresponding author, but we did not receive a reply; Therefore, COVID-19 infection data were not included in the D-CIMA |
| Campi et al. [29] | 361 | 361 | 103 | 103 | I-S-M | I-S-M | I(+); S(+); M(+) | There was no limitation in data extraction |
| Cereda et al. [30] | 129 | – | 129 | 129 | S-M | S-M | S(+); M(+) | The clinical outcomes were severe pneumonia, admission to the ICU, and in-hospital mortality; The ICU admission data were not appropriate for statistical analysis (0 observed cases); Therefore, we extracted data from severe pneumonia for the analysis of severity |
| D’Avolio et al. [21] | 1,484 | – | – | – | I | – | I(+) | The definition of VDD was not available; There was no adequate information dealing with descriptive statistics; VDD was presented as median (IQR) and the findings were presented graphically; Since we extracted the data based on the number of cases, we did not include it in the meta-analysis In the study, the date of the data received from UK Biobank for the COVID-19 (-) control group was long ago; There was no VDD definition and they used quartile values of VDD; Since we extracted the data based on the number of cases, we did not include it in the meta-analysis |
| Darling et al. [22] | 1,303 | – | – | – | I | – | I(-) | In the study, the date of the data received from UK Biobank for the COVID-19 (-) control group was long ago; There was no VDD definition and they used quartile values of VDD; Since we extracted the data based on the number of cases, we did not include it in the meta-analysis |
| De Smet et al. [31] | 2,903 | 2,903 | 186 | – | I-S | I-S | I(+); S(+) | In the study, chest CT was performed for all COVID-19 patients to determine the disease stage; They classified the patients as stage 1 (early stage), stage 2 (progressive stage), and stage 3 (peak stage); We chose severe cases from stage 3 and non-severe cases from stage 1-2 |
| Hastie et al. [23] | 341,484 | – | – | – | I-M | – | I(-); M(-) | There was no adequate information dealing with descriptive statistics; The findings were presented by using HR and IRR values; Since we extracted the data based on the number of cases, we did not include it in the meta-analysis |
| Hernández et al. [32] | 413 | – | 197 | 197 | I-S-M | S-M | I(+); S(-); M(-) | For COVID-19 infection, VDD was presented as mean (±SD); Since the number of cases was not reported, we could not include it in D-CIMA |
| Im et al. [24] | 200 | – | – | – | I-S | – | I(+); S(+) | In the study, the unit of ng/dL was used for laboratory measurements of serum 25(OH)D levels, and VDD was defined as 25(OH)D<20 ng/dL; We converted 20 ng/dL to 0.2 ng/mL; Since we thought there was an inconsistency, we could not include these data in the meta analyses |
| Israel et al. [33] | 576,455 | 187,234 | – | – | I | I | I(+) | Age matching was performed between the case and control groups, but single summary statistics for the population were not presented |
| Karahan and Katkat [34] | 149 | – | 149 | 149 | S-M | S-M | S(+); M(+) | There was no limitation in data extraction |
| Katz et al. [25] | 987,849 | – | – | – | I | – | I(+) | There was no adequate information dealing with descriptive statistics; The findings were presented as ORs; Since we extracted the data based on the number of cases, we did not include it in the meta-analysis |
| Li et al. [35] | 353,299 | 3,502 | 1,082 | – | I-H-S | I-S | I(+); H(+); S(+) | This study defined hospitalization as one record of origin (whether the patient s tested positive or not); In the study, hospitalized, confirmed, and severe COVID-19 cases were compared with community controls; We think that this study design is inappropriate; For this reason, we extracted data on infections from patients who presented to the hospital with suspected COVID-19 and confirmed COVID-19, and for the severity from patients with confirmed and severe COVID-19 |
| Livingston et al. [36] | 104 | 104 | – | – | I | I | I(-) | The definition of VDD was not available; Statistical comparisons were performed for 25(OH)D <34.4 nmol/L; Therefore, the data belonging to this distinction were included in the meta-analysis, and the measurement unit was assigned as “other” |
| Luo et al. [37] | 895 | 895 | 335 | 74 | I-S-M | I-S-M | I(+); S(+); M(-) | There was no limitation in data extraction |
| Macaya et al. [38] | 80 | – | 80 | S | S | S(+) | There was no limitation in data extraction | |
| Maghbooli et al. [39] | 235 | – | 235 | 235 | H-S-M | S-M | H(+); S(+); M(+) | There was no limitation in data extraction; Although the authors defined VDD as 25(OH)D<20 ng/mL, the dataset that could be extracted from the article belonged to the category of 25(OH)D<30 ng/mL; Therefore, it was included in the overall analysis |
| Mardani et al. [40] | 123 | 123 | – | – | I | I | I(+) | We extracted data from their supplementary file according to 25(OH)D<20 ng/mL |
| Meltzer et al. [41] | 489 | 489 | – | – | I | I | I(+) | There was no limitation in data extraction |
| Mendy et al. [42] | 689 | – | 689 | – | H-S | S | H(+); S(+) | The definition of VDD was not available; Therefore, the measurement unit was assigned as “other” in the D-CSMA |
| Merzon et al. [43] | 7,807 | 7,807 | – | – | I-H | I | I(+); H(+) | There was no limitation in data extraction |
| Panagiotou et al. [44] | 134 | – | 134 | – | S | S | S(+) | The authors classified the patients as admitted to the ITU and non-ITU wards; |
| When we extracted data, we chose patients admitted to non-ITU wards as non-severe and those admitted to ITU wards as severe; A statistical comparison was made for VDIS <50 nmol/L; Therefore, the data belong to VDIS were included in the D-CSMA | ||||||||
| Radujkovic et al. [45] | 185 | – | 185 | – | S-M | S | S(+); M(+) | The authors classified patients as inpatients and outpatients; When we extracted data, we chose the outpatients as non-severe and the inpatients as severe according to their definition |
| Vassiliou et al. [46] | 39 | – | – | 39 | H-M | M | H(-); M(-) | There was no limitation in data extraction |
| Ye et al. [47] | 142 | 142 | 60 | – | I-S | I-S | I(+); S(+) | There was no limitation in data extraction |
| Total | 2,277,860 | 203,962 | 3,564 | 996 | ||||
COVID-19, coronavirus disease 2019; D-CIMA, vitamin D and COVID-19 infection meta-analysis; D-CSMA, vitamin D and COVID-19 severity meta-analysis; D-CMMA, vitamin D and COVID-19 mortality meta-analysis; I, COVID-19 infection; S, severity; M, mortality; H, hospitalization; VDD, vitamin D deficiency; ICU, intensive care unit; IQR, interquartile range; CT, computed tomography; HR, hazard ratio; IRR, incidence rate ratio; SD, standard deviation; OR, odds ratio; ITU, intensive therapy unit; VDIS, vitamin D insufficiency.
1
(+), indicates that the association of low levels of vitamin D and I, S, M and H was statistically significant; (-) indicates that the association of low levels of vitamin D and I, S, M and H was not statistically significant.