Statistics
The association between a positive SARS-CoV-2 test compared with a
negative SARS-CoV-2 test result and serum vitamin D levels, IMD, age and
gender was initially studied using unpaired t-test (continuous
variables), rank-sum non-parametric test (ordinal data: IMD) in the
total cohort and subgroups; age, vitamin D by median values and IMD by
the quintile threshold closest to the median
(https://tools.npeu.ox.ac.uk/imd/ - last accessed on 19.05.2020).
Logistic regression analyses were then carried out with SARS-CoV-2 RNA
real time RT-PCR test result as the dichotomous outcome (detected or not
detected) and vitamin D levels and factorised IMD as independent
variables with age and gender included as confounders. The regression
analyses were carried out in the total groups and the previously
mentioned subgroups. All analyses were performed on Stata 14 (StataCorp
LLC, Texas, USA).
Results In the total cohort of 104 individuals tested for SARS-CoV-2, a
positive result was significantly associated with low vitamin D
concentrations and IMD, but not age and gender (Table 1). The total
cohort was then stratified by gender, median age (72 years), vitamin D
levels (34.4 nmol/L) and IMD (quintile 1-4 and 5; median IMD was 35.70,
this figure within quintile 5, the threshold separating quintile 4 from
5 being 34.18). 57.1% of the total cohort had an IMD in quintile 5. In
males and individuals aged >72 years, a positive SARS-CoV-2
test was associated with higher IMD (greater deprivation) and low
vitamin D levels, respectively. Low vitamin D levels were associated
with a positive COVID-19 test in individuals classified as IMD quintile
5, and whilst this was not seen in IMD quintiles 1-4, IMD was higher in
COVID-19 positive individuals with vitamin D levels below the total
cohort median value, this approaching statistical significance (Table
1).
We then performed a logistic regression analysis with COVID-19 positive
diagnosis (reference: negative diagnosis) as the dichotomous outcome.
Low vitamin D levels and IMD quintile 5 (compared with IMD quintiles
1-4) were significantly associated with the outcome whilst age and
gender were not (Table 2: Model 1). Model 2 (Table 2) was a repeat of
Model 1, but with vitamin D stratified by the median value; once again
low vitamin D and IMD quintile 5 were significantly associated with a
positive COVID-19 test. Models 3 and 4 (Table 2) comprised the cohort
stratified by the median vitamin D levels, and IMD quintile 5 was only
significantly associated with a positive COVID-19 test in the subgroup
with low vitamin D levels (Table 2: Model 3). Vitamin D levels were
associated with a positive SARS-CoV-2 test in the IMD quintile 5
subgroup (Table 2: Model 5) but not in the IMD quintiles 1-4 subgroup
(Table 2: Model 6).
The above analyses (Table 2: Models 3 and 5) suggests the possibility of
an interaction between IMD quintile 5 and low vitamin D levels, both
associated with positive SARS-CoV-2 test (Table 2: Model 2). Hence, the
total cohort was stratified into 4 groups by IMD quintiles (quintile 1-4
and 5) and vitamin D status (≤ 34.4, >34.4 nmol/L). A
Chi-square analysis did not show an association between these 4 groups
(p=0.41). Model 7 (Table 2) shows that individuals belonging to IMD
quintile 5 with vitamin D levels ≤ 34.4 nmol/L were more likely to have
a positive SARS-CoV-2 test (reference: vitamin D >
34.4nmol/L + IMD quintiles 1-4). Interestingly, the association with a
diagnosis of COVID-19 was clearly not evident in individuals with only
one risk factor (vitamin D ≤ 34.4 nmol/L or IMD quintile 5). The
analysis with the combined IMD and vitamin D subgroups was repeated in
patients aged ≤ 72 years (Model 8) and >72 years (Table 2:
Model 9). The association between the ‘at risk’ group (IMD quintile 5 +
vitamin D ≤ 34.4 nmol/L) appeared greater in patients aged
>72 years (Table 2: Model 9). Although we did not have
sufficient patient numbers to study the above associations in black and
ethnic minorities, we were able to analyse the data in the known white
Caucasian subgroup (n=59):– individuals belonging to IMD quintile 5
with vitamin D levels ≤ 34.4 nmol/L being the only patients
significantly associated (odds ratio: 19.00, 95% confidence interval:
2.73–132.27; p=0.003) with a diagnosis of COVID-19 (reference: vitamin
D >34.4 nmol/L + IMD quintiles 1–4).
Discussion There is accumulating evidence on the association between vitamin D
levels and COVID-19 incidence and outcome following infection [6–8,
15]. Ours is a relatively small study cohort fulfilling the PCR based
testing criteria for COVID-19. Univariate analyses showed that low
vitamin D levels and IMD scores were associated with a positive test for
SARS-CoV-2. It was clear that these associations appeared strengthened
in some subgroups; males (IMD scores), age >72 years
(vitamin D). Further, the association between low vitamin D and a
positive SARS-CoV-2 result was seen in the IMD quintile 5 subgroup only.
Logistic regression with age and gender added as confounding variables
confirmed the results of the above mentioned univariate association
between low vitamin D, deprivation (IMD quintile 5) and a positive
SARS-CoV-2 test. Interestingly, low vitamin D and IMD quintile 5 were
associated with a positive test in patients belonging to IMD quintile 5
(Table 2: Model 5) and low vitamin D (Table 2: Model 3), respectively,
and not the complementary subgroups.
These results are interesting as only a combination of low vitamin D
levels and deprivation (IMD quintile 5) appear to be associated with
COVID-19 (table 2: Model 7), this association perhaps more evident in
the older age group (Table 2: Model 9). Patients characterised by one of
these factors were not at any higher risk of a diagnosis of COVID-19.
Thus, associations between vitamin D levels, social deprivation and
COVID-19 infection will be dependent on cohort characteristics.
Consequently, we highlight the importance of acknowledging heterogeneity
and studying subgroups. Our results were influenced by the fact that
57.1% with IMD scores belonged to IMD quintile 5, the most deprived
group; each quintile should consist of ~20%, but
accordingly this shows the increased social deprivation present in this
cohort. The 2019 Index of Multiple Deprivation ranks Walsall as the 25th
most deprived English local authority (out of 317), placing Walsall
within the most deprived 10% of districts in the UK. High deprivation
is most likely a surrogate for either one or more risk factors. It is
established that rates of obesity and potentially of insulin resistance
are higher in more socially disadvantaged individuals [16]. Low
vitamin D could be a causative or a surrogate factor in COVID-19
infection. Only normalising vitamin D levels with supplements would
provide data on the role of the hormone in COVID-19 infection.
Interestingly, whilst addressing social deprivation is a long term
programme, normalising serum vitamin D concentrations can be easily
achieved in a relatively short period within existing clinical
guidelines. In our clinical practice, we have for >10 years
been using vitamin D supplements in patients with hormonal deficiency
(<30 nmol/L) or insufficiency (30–50 nmol/L), these
thresholds in synchrony with laboratory distribution data and at slight
odds with the UK National Institute for Health and Care Excellence
(NICE) recommendations
(https://cks.nice.org.uk/vitamin-d-deficiency-in-adults-treatment-and-prevention#!scenario
– last accessed on 10/05/2020). Depending on symptoms and
concentrations of serum vitamin D, a clinical decision based on the NICE
guidelines would be taken as to the dose of the hormone supplements
(https://cks.nice.org.uk/vitamin-d-deficiency-in-adults-treatment-and-prevention#!scenario
– last accessed on 10/05/2020). This should be followed by a repeat
test of vitamin D and a bone profile in 3–6 months to ensure adequate
treatment efficacy. Adverse effects are uncommon [17]; hence, the
importance of our findings.
In usual circumstances, ours would be considered an interesting
observation meriting further investigation. Even when a significant
association was observed, the relatively small numbers led to the odds
ratio having wide confidence intervals. We acknowledge that false
negative SARS-CoV-2 RT-PCR test results could have a significant bearing
on our study outcome. The relative clinical sensitivity and specificity
of the targets in the Roche assay are not yet fully known, but the false
negative rate for is estimated to be ~30% [18].
However, the current situation perhaps requires the usual scientific
method processes to be accelerated, especially if any associated
therapeutic intervention is perceived as safe. Our data adds to results
from other studies suggesting a possible role for vitamin D in COVID-19
infection [15]. As described previously there is mechanistic
plausibility in view of vitamin D modulating immune function and viral
replication [2,3]. A further positive aspect of our study is the
inclusion of SARS-CoV-2 test negative patients which were suspected of a
COVID-19 diagnosis for comparison (control group). A larger, more robust
study is required to validate our findings, with the cohort size
allowing for subgroup analyses of different populations (e.g. different
ethnic groups) at varying levels of risk [19]. It would also be
interesting to compare with similar global data to see if vitamin D
levels could, at least partially account for differences in incidence /
prevalence that is evident between nations
(https://www.who.int/docs/default-source/coronaviruse/situation-reports/20200304-sitrep-44-covid-19.pdf?sfvrsn=783b4c9d_2
– last accessed on 10/05/2020). This should be followed by large well
designed outcome studies (either randomised controlled trials or
longitudinal cohort studies) to see if vitamin D supplementation,
especially in individuals subject to social deprivation, alters outcome
trajectory at varying stages of the infection. However, at this time of
necessary urgent action, our data adds support to the recommendations
for normalising vitamin D levels in those with deficient / insufficient
levels and in groups at high-risk for deficiency.
In conclusion, combined low vitamin D levels and higher social
deprivation were most associated with COVID-19 infection. In older age,
this combination was even more significant. The importance of studying
subgroups and acknowledging heterogeneity is highlighted, with results
being dependant on the cohort studied.
Funding: None to declare.
Conflicts of Interest: None to declare.