1 Background

Vitamin D is traditionally recognized for its role in maintaining bone health across all age groups [1]. However, recent studies have revealed its critical influence on the immune system, particularly its capacity in modulating immune responses against respiratory viruses [2,3,4]. Vitamin D has also been found to decrease the production of inflammatory cytokines, potentially reducing the morbidity and mortality from such infections [2,3,4].

The emergence of COVID-19 pandemic has brought vitamin D deficiency (VDD) or lower serum 25(OH)D levels to the forefront of medical research, highlighting its strong link with COVID-19 infection [5,6,7]. Numerous studies [5,6,7] have evaluated the impact of VDD, predominantly in Caucasian populations experiencing severe forms and death from COVID-19 [5], highlighting the potential importance of vitamin D as a secondary preventive strategy to mitigate disease [8]. Vitamin D has also shown promise in preventing COVID-19 but not received attention to the same extent.

The COVID-19 pandemic has revealed a disproportionate excess of deaths in ethnic minority groups, particularly in Northern US States [9], notably among Black Americans and those of South Asian origin [5]. This disparity has been linked to lower vitamin D levels among them, compared to those with lighter skin [10], supporting the hypothesis that VDD contributes to a disproportionately higher risk of COVID-19. The magnitude of this risk could even be greater among South Asians living close to the equator, whose dark skin tones produce lesser vitamin for the same exposure time. In further exploration, a study from India reported a higher risk of VDD associated with the severity of COVID-19 [11], however the evidence remains inconclusive on the prophylactic benefit of vitamin D for primary prevention of COVID-19.

VDD is currently identified as a major public health problem in South Asia [10], including Sri Lanka [12]. Therefore, this study intended to assess the magnitude of the risk of lower concentrations or deficiency related to vitamin D in the development and progression of COVID-19 in a South Asian country. This evidence would be of value particularly in resource-limited settings, to reduce the burden on over-whelmed healthcare systems, against the less affordable treatment facilities including vaccines available for the ever-changing virus variants [8]. The findings from this study may serve as a foundation for further investigations into the role of vitamin D in combating novel respiratory diseases which could be widely transmitted through airborne routes, especially in countries with poor housing and sanitation.

2 Methods

A matched case–control study was conducted among adults aged 35–74 years residing in the district of Colombo, Sri Lanka during January-August 2021. The cases were RT-PCR confirmed COVID-19 patients admitted to the two major hospitals that were designated as COVID-19 treating centres in Sri Lanka (National Institute of Infectious Diseases (NIID) and Colombo East Base Hospital, Mulleriyawa) [13]. Controls were apparently healthy adults residing in the catchment area of the two hospitals, without a diagnosis of COVID-19 based on a negative RT-PCR. Those who had been living abroad within the past three months were excluded from the study.

The required sample size was 95 participants per group, as per calculated for a matched case control study [14], in order to estimate an odds ratio (OR) of 2.42 for the risk of VDD associated with COVID-19 reported among South Asians in the United Kingdom [6], probability of an exposure-discordant pair of 50.9%, proportion in the control arm of 47.4% based on the prevalence of VDD in Sri Lanka [12], 10% non-response, Z value of 1.96, type II error of 0.2, and 5% precision. During recruitment, controls were matched 1:1 by the sex of each case. Cases were recruited consecutively from the hospital inward registers, after checking their eligibility and willingness for participation. Controls were recruited consecutively, each one matched by the sex of the corresponding case, from the register that was maintained by the public health staff of the area for COVID-19 contacts who have been released from their 14-day quarantine period following a negative RT-PCR.

Data were collected by MBBS qualified doctors and trained nurses working in the two designated COVID-19 treating hospitals. An interviewer-administered questionnaire was used to assess the basic characteristics including the information on potential confounders, such as comorbidities (presence of hypertension, diabetes, hypercholesterolemia, cardiovascular and chronic respiratory diseases confirmed by documental evidence, such as diagnosis cards and clinic records), smoking status and alcohol consumption. Body weight and height were measured using standard protocols to determine the body mass index (BMI). The severity of COVID-19 status was determined in cases using the standard classification [15].

A non-fasting venous blood sample of 4 mL was obtained from cases upon hospital admission and from controls upon release from their 14-day quarantine period. It was collected into a plain tube under aseptic conditions and transported at room temperature to the laboratory. Serum 25(OH)D level was measured in nmol/L by a chemiluminescence method using ‘DiaSorin LIAISON 25-OH D, Stillwater, Minnesota, USA assay’. This is a direct competitive immunoassay which detects total vitamin D2 and D3 levels; and is shown to be valid and reliable (intra-assay coefficient of variation (CV) of 4.9% and inter-assay coefficient of 5.4%, indicating acceptable level (< 10%); dynamic range between 4 and 150 ng/mL; sensitivity of ≤ 4 ng/mL and specificity of 100% for both vitamin D2 and D3 [16]. The tests were performed by trained medical laboratory technicians under the supervision of an expert on medical laboratory accreditation. Individuals having serum vitamin D < 50 nmol/L were identified as having ‘VDD’ and others as ‘no VDD’ [1].

The study was conducted in accordance with the Declaration of Helsinki and approved by the Ethics Review Committee of the Faculty of Medicine, University of Colombo (protocol number EC-20-EM18). Informed written consent was obtained from all participants after providing them with an information sheet in the language of their preference. Their participation was voluntary and the right to withdraw from the study at any time without penalty and effect on medical care or loss of benefits was informed. Privacy and confidentiality were ensured. Blood collection was carried out by a trained nurse as per the World Health Organization Guidelines on best practices in phlebotomy. Those who were found to have low vitamin D levels were referred to the closest endocrine clinic of state sector hospitals and were advised on locally available vitamin D-rich food that can be consumed and sensible sun exposure.

2.1 Data analysis

The Statistical Package for Social Sciences (SPSS) version 22 was used for data analysis. Data normalcy was assessed using the Kolmogorov–Smirnov test. Continuous data normally distributed were presented as mean and standard deviation (SD) and non-normally distributed data as median and interquartile range (IQR). Categorical data were presented in percentages.

A univariate analysis was performed between cases and controls (based on COVID-19 status) in relation to their vitamin D level (nmol/L) and other continuous variables using Mann–Whitney U test or independent sample t test; and the differences in relation to their VDD status and other categorical variables using Chi-squared test and crude OR. Unadjusted p-values were determined at 0.05 significance level. Next, to assess the risk of vitamin D level (nmol/L) on COVID-19 infection independent of other factors, a conditional logistic regression model with adjustments for covariates (age, ethnicity, BMI, smoking and alcohol status, presence of co-morbidities) was applied to determine adjusted p values. The dependent variable was the case/control status. Also, to assess VDD as an independent predictor of COVID-19 infection, adjusted ORs were calculated for VDD with adjustments for the same variables in another model.

Another univariate analysis was performed between mild and moderate/severe cases (based on the disease severity) in relation to their vitamin D status (as nmol/L and as VDD) along with other variables. Next, binary logistic regression models were applied to cases in the same manner, with disease severity (moderate/severe versus mild) as the dependent variable to assess the independent role of vitamin D level and VDD on the COVID-19 severity. The models were run using backward likelihood (LR) method and were evaluated using Nagelkerke R, Omnibus test and Hosmer–Lemeshow goodness-of-fit test.

3 Results

A total of 104 cases and 104 sex-matched controls participated in the study. As shown in Table 1, the majority in both groups were educated up to grade 6–11, employed and consuming tobacco and alcohol at the time of recruitment. Cases were older and predominantly of Sinhalese ethnicity, while the controls had fewer persons with co-morbidities. In both groups, stroke and chronic respiratory conditions were present in less than 5%.

Table 1 Demographic and socio-economic characteristics of the cases and controls
Full size table

3.1 Relationship of vitamin D with the COVID-19 status

Serum 25(OH)D level in the sample ranged between 7.5 and 97.5 nmol/L, in a positively skewed distribution, with a mean value of 34.2 nmol/L (SD = 21.2) among cases and 39.8 nmol/L (SD = 17.8) among controls. The cases presented with a significantly lower median vitamin D level (30.0 nmol/L; IQR = 21.2) compared to the controls (35.9 nmol/L; IQR = 28.3) (p = 0.02) (Fig. 1). The other factors significantly associated with COVID-19 status in the univariate analysis were (Table 2) older age, Sinhalese ethnicity, presence of co-morbidities, namely diabetes mellitus, hypertension, asthma, hypercholesterolemia, myocardial infarction, stroke and chronic respiratory disease, and history of tobacco and alcohol consumption during last six months (unadjusted p < 0.05).

Fig. 1
figure 1

Distribution of serum 25(OH)D levels in cases and controls and according to COVID-19 disease and severity status

Full size image
Table 2 Potential risk of vitamin D level (nmol/l) and other factors on COVID-19 infection
Full size table

Box represents the serum 25(OH)D data of participants while the upper and lower boundaries of the box show the 75th and 25th percentile data. The horizontal bar within the box represents the median. Outliers are indicated by their age.

In the regression analysis, the risk of lower vitamin D level retained its significance when adjusted for confounders (adjusted p = 0.026). Only other variable significant was Sinhalese ethnicity (adjusted p < 0.001) (Table 2).

With regards to the relationship of VDD with COVID-19 status (Table 3), a higher proportion of those with VDD were seen among cases (n = 87; 83.7%) compared to the controls (n = 74; 71.2%). This difference was statistically significant (p = 0.031). When adjusted for confounders, it was however not significant as an independent predictor. Those aged 50 and above (p < 0.001) and Sinhalese ethnicity (p < 0.001) were identified as significant predictors.

Table 3 Predictors of COVID-19 infection
Full size table

3.2 Relationship of vitamin D with the severity of COVID-19

Among the cases, the moderate/severe forms of COVID-19 presented with lower serum 25(OH)D levels (median = 28.0 nmol/L; IQR = 20.4) compared to the mild forms (median = 39.8 nmol/L; IQR = 19.0) (Fig. 1). This difference was statistically significant (p = 0.015) (Table 4). However, when adjusted for other potential risk factors, this relationship was not significant. The only risk that was significant in the regression analysis was for diabetes (42.4% among those with moderate/severe form versus 10.5% among mild form) (p = 0.022).

Table 4 Potential risk of vitamin D level (nmol/l) and other factors on the severity of COVID-19 infection
Full size table

As for VDD, it was not significantly different between moderate/severe cases (n = 71; 83.5%) and mild cases (n = 16; 84.2%) in the univariate analysis (p > 0.05) (data not shown in tables).

4 Discussion

In our case–control study, a significantly lower concentration of 25(OH)D was noted among COVID-19 cases compared to the sex-matched controls. This relationship persisted even after adjusting for age, BMI and co-morbidities. In contrast, VDD, though significantly more prevalent in cases, was not an independent predictor of COVID-19 infection. Similarly, low serum concentrations nor deficiency showed an independent relationship with the severity or mortality of COVID-19, despite a significantly lower 25(OH)D level observed in moderate/severe cases compared to those with mild forms. This suggests that vitamin D may serve as a primary preventive strategy, but not as a secondary preventive strategy in COVID-19 infection.

Observational studies, including a few notable studies conducted in Asia [11, 17,18,19,20] and elsewhere have investigated the relationship between vitamin D and COVID-19. Early studies using population-level databases suffered from time-lag bias, with the actual vitamin D status not reflected well in the participants, as it was measured before the COVID-19 testing [5, 17, 21]. Subsequent case–control studies [6], included healthy individuals from population-based cohorts who were not screened for a negative diagnosis of COVID-19 as controls, leading to misclassification bias. Recent studies conducted in non-Caucasian populations in Saudi Arabia [18], China [20] and India [11] aimed to address these methodological issues, using hospitalized patients with minimum selection bias. Our study provides both supporting and contrasting evidence in this global context.

Our study emphasized that lower vitamin D levels were associated with a higher risk of COVID-19. The difference was significant at both unadjusted and adjusted p values (p = 0.02). The findings align with other case–control studies [6, 8, 11, 18], except one study from India showing no difference between the cases and controls (p = 0.757) [11]. Furthermore, of the traditionally known risk factors of COVID-19 [22], only ethnicity (p < 0.001) was retained in our final model, while none of the pre-existing chronic conditions and BMI exhibited the expected associations. As suggested by some studies, the risk for COVID-19 may be multifactorial and related to the clustering of risk factors rather than individual factors [23]. Nevertheless, when interpreting results, it is important to note that all these studies face a common challenge in ascertaining the temporal relationship, where the vitamin D levels were assessed only after the disease onset. Assuming that vitamin D levels remain stable and unaffected by acute viral infections [24], the current evidence favours vitamin D in preventing COVID-19 infection.

Despite having vaccines with modest success, additional preventive strategies remain important, particularly in resource-limited settings, where the availability of and access to vaccines are major challenges during epidemics. In this backdrop, our study emphasizes the potential of vitamin D supplementation, particularly in countries like Sri Lanka, where there are no routine supplementation programmes. Despite being a tropical country with an abundance of sunlight, there are less opportunities for sun exposure, especially in urban areas, due to sedentary jobs and societal preferences for fair skin, thus contributing to VDD [10]. This justifies the need for supplementation, at least for vulnerable groups, such as the elderly and those with immuno-suppressive conditions like diabetes. However, recommendations for supplementation should be made cautiously, as there is currently no national consensus on dosage and route of administration, which could lead to irrational use, such as over-dosing and toxicity, over-prescribing and exploitations by pharmaceutical companies.

Vitamin D supplementation has demonstrated immune modulation properties, reducing the risk of acute respiratory infections by 32–60%, and its preventive and safe impact [25]. This suggests that it could be a cost-effective preventive strategy not only for COVID-19, but also for other respiratory viruses in future pandemics. During the COVID-19 pandemic, healthcare resources in Sri Lanka were primarily directed towards preventing the spread of the disease. Human resources were mobilized for island-wide contact tracing, quarantine and providing isolation facilities to patients [13], while many health institutions were converted to quarantine centres or isolation wards, leaving long-term care for chronic diseases neglected in the country.

With regards to VDD, studies from China [20] and India [11] have shown a strong relationship with COVID-19, using extreme cut-off values such as 62.5 nmol/L [20] and 30 nmol/L [11]. In comparison, our study demonstrated a two-fold risk of VDD for COVID-19 but failed to show it as an independent predictor (OR = 1.9; 95% CI 0.6–5.7; p = 0.14). This could be due to disease development taking place at much lower vitamin D levels in Sri Lankans. This warrants further exploration, especially when determining target groups for vitamin D supplementation.

Our study did not support the growing evidence linking vitamin D levels or deficiency to the severity of COVID-19 [26, 27]. Though significant in univariate analysis, vitamin D lost its significance when adjusted for confounders. Instead, diabetes emerged as the sole predictor of disease severity (p = 0.022), highlighting the bigger role played by it than vitamin D in determining the severity. This was in line with other case–control studies [28]. Changes in glucose homeostasis, immunological status and inflammation are possible pathogenetic links with diabetes mellitus [29]. Given the high prevalence of diabetes in the country [30], controlling diabetes becomes a critical secondary preventive strategy for COVID-19 in Sri Lanka.

Our study did not assess mortality as an outcome due to the absence of deaths among the cases. Sri Lanka reported exceptionally low mortality rates during the pandemic, further highlighting the role of vitamin D in primary prevention.

There were some strengths of this study. Controls were matched by sex of each case, eliminating a major confounder in interpretations. Misclassification bias related to controls, which was a major limitation in previous studies, was minimized by identifying apparently healthy, non-infected controls with a negative COVID-19 diagnosis confirmed at the end of 14-day quarantine period. Vitamin D was measured using valid and reliable assays with standard protocol and cut-off values. There were some limitations. Mortality could not be studied due to 100% survival among cases. Also, the findings apply to hospitalized patients with pre-existing conditions and not to the general community, thus leading to selection bias.

5 Conclusions and recommendations

Serum vitamin D level was significantly lower in COVID-19 patients compared to non-COVID-19 healthy controls after adjusting for confounders, highlighting its influence on the disease development. VDD was also significantly higher in cases than controls but was not shown to be an independent predictor of COVID-19. Though higher in moderate/severe cases compared to mild cases, both vitamin D level and VDD did not assume an independent relationship with disease severity. Instead, diabetes emerged as the sole predictor of severity. Our study underscores the potential of vitamin D as a primary preventive strategy in preventing COVID-19 like infections, but not as a secondary preventive strategy in reducing its severity, especially in regions with high VDD. However, clinical trials are needed to validate these findings and explore the full benefit of vitamin D supplementation.