Vitamin D insufficiency in COVID-19 and influenza A, and critical illness survivors: a cross-sectional study

Objectives The steroid hormone vitamin D has roles in immunomodulation and bone health. Insufficiency is associated with susceptibility to respiratory infections. We report 25-hydroxy vitamin D (25(OH)D) measurements in hospitalised people with COVID-19 and influenza A and in survivors of critical illness to test the hypotheses that vitamin D insufficiency scales with illness severity and persists in survivors. Design Cross-sectional study. Setting and participants Plasma was obtained from 295 hospitalised people with COVID-19 (International Severe Acute Respiratory and emerging Infections Consortium (ISARIC)/WHO Clinical Characterization Protocol for Severe Emerging Infections UK study), 93 with influenza A (Mechanisms of Severe Acute Influenza Consortium (MOSAIC) study, during the 2009–2010 H1N1 pandemic) and 139 survivors of non-selected critical illness (prior to the COVID-19 pandemic). Total 25(OH)D was measured by liquid chromatography-tandem mass spectrometry. Free 25(OH)D was measured by ELISA in COVID-19 samples. Outcome measures Receipt of invasive mechanical ventilation (IMV) and in-hospital mortality. Results Vitamin D insufficiency (total 25(OH)D 25–50 nmol/L) and deficiency (<25 nmol/L) were prevalent in COVID-19 (29.3% and 44.4%, respectively), influenza A (47.3% and 37.6%) and critical illness survivors (30.2% and 56.8%). In COVID-19 and influenza A, total 25(OH)D measured early in illness was lower in patients who received IMV (19.6 vs 31.9 nmol/L (p<0.0001) and 22.9 vs 31.1 nmol/L (p=0.0009), respectively). In COVID-19, biologically active free 25(OH)D correlated with total 25(OH)D and was lower in patients who received IMV, but was not associated with selected circulating inflammatory mediators. Conclusions Vitamin D deficiency/insufficiency was present in majority of hospitalised patients with COVID-19 or influenza A and correlated with severity and persisted in critical illness survivors at concentrations expected to disrupt bone metabolism. These findings support early supplementation trials to determine if insufficiency is causal in progression to severe disease, and investigation of longer-term bone health outcomes.


INTRODUCTION
Vitamin D metabolites contribute to bone metabolism, calcium homeostasis and immunomodulation. Vitamin D is a steroid pre-pro-hormone which is converted to the main circulating form 25-hydroxy vitamin D (25(OH)D), and subsequently to the active hormone 1,25 dihydroxy vitamin D. This second activation step occurs in the kidney, modulated by parathyroid hormone (PTH), for 'endocrine' calciotropic effects, and also under local control within extra-renal tissues, including immune cells, for direct action. These 'intracrine' actions on immune cells mediate antimicrobial and anti-inflammatory effects. 1 The majority of 25(OH)D circulates Strengths and limitations of this study ► Liquid chromatography-tandem mass spectrometry was used to quantify 25-hydroxy vitamin D (25(OH) D) in plasma samples from well-characterised hospitalised people with COVID-19 and influenza A and survivors of non-selected critical illness. ► Biologically active free 25(OH)D was measured by ELISA in COVID-19 plasma samples for the first time. ► Samples from people with COVID-19 and influenza A were obtained early in the course of the disease. ► Binary logistic regression multivariable models were used to assess the association of plasma 25(OH) D concentration with outcomes in COVID-19 and influenza A, correcting for other known relevant covariates. ► The observational nature of the study means it is not known whether vitamin D status led to poor clinical outcome or was a consequence of illness severity.
Open access bound to proteins, principally vitamin D binding protein (85%-90%), and the relatively small unbound ('free') fraction is available to immune cells. 2 In the context of infectious diseases, vitamin D insufficiency (routinely determined by total 25(OH)D measurement) is associated with increased incidence and severity of respiratory tract infections, [3][4][5] including COVID-19. 6 7 A geographical association between vitamin D deficiency prevalence and COVID-19 incidence and mortality has been reported. 8 Free 25(OH)D has not yet been investigated in COVID-19, but this is required to fully understand vitamin D status during acute illness and any associations with systemic inflammation. 9 Clinical trials of vitamin D supplementation in respiratory diseases have returned mixed results. [10][11][12][13] Potential beneficial effects of vitamin D supplementation may be pathogen-specific and dependent on timing and route of administration. In addition to an interest in modifying acute illness outcomes, longerterm effects on bone health warrant consideration as critical illness is associated with loss of bone mineral density after recovery. 14 In this cross-sectional study, we report measurements of total and free 25(OH)D in hospitalised people with COVID-19 and total 25(OH)D in hospitalised people with influenza A and survivors of critical illness. We use these three data sets to test the hypotheses that vitamin D insufficiency in severe respiratory virus infections scales with severity and persists in survivors of critical illness.

Non-selected critical illness survivors
We include a post-hoc analysis of the RECOVER (Evaluation of a Rehabilitation Complex Intervention for Patients Following Intensive Care Discharge) trial of intensive rehabilitation after critical illness. 17 Full eligibility criteria have been published previously; briefly, adults who had received invasive mechanical ventilation (IMV) for at least 48 hours and were considered well enough for discharge from the intensive care unit (ICU) were recruited. Patients gave additional consent for participation in a biomarker substudy and blood samples were collected at ICU discharge. 18 All participants gave informed consent.

Patient and public involvement
There was no patient or public involvement in this study.
LC-MS/MS methods for total 25(OH)D analysis EDTA plasma concentrations (on samples obtained on the day of enrolment to the study) of 25(OH)D 2 and 25(OH) D 3 isoforms were measured by liquid chromatographytandem mass spectrometry (LC-MS/MS) and summed to derive the total 25(OH)D concentrations presented in the results. For patients with COVID-19 and critical illness survivors, analysis was performed by the Vitamin D Animal Laboratory using an assay which has been certified as proficient by the international Vitamin D External Quality Assessment Scheme and described in detail in an earlier manuscript, using 200 µL plasma. 19 The interassay precision (coefficient of variation, CV) of this method was <11.5% for both 25

Definition of vitamin D status
In addition to the absolute total 25(OH)D concentration, the relationship between vitamin D status and outcomes is often explored using a total 25(OH)D cutoff of 50 nmol/L to define populations that are vitamin D sufficient. 20 In this study, total 25(OH)D >50 nmol/L is reported as 'sufficient', 25-50 nmol/L as 'insufficient' and <25 nmol/L as 'deficient' (see online supplemental methods).

Free 25(OH)D ELISA
Free 25(OH)D was measured using the Free 25OH Vitamin D ELISA (DIAsource ImmunoAssays SA, Belgium), following the manufacturer's instructions, using 10 µL of plasma. Absorbance was measured at 450 nm against a reference filter set at 630 nm using the Tecan Sunrise Microplate Reader (Tecan). GraphPad Prism (V.7.0e for

Open access
MacOS X) was used to perform a four-parameter logistic function to create the calibration curve in order to read the mean concentration of duplicate samples. The lower limit of detection (LLOD) of the assay was 2.4 pg/mL. The intra-assay repeatability (CV) was ≤5.5% across three concentrations (low, mid and high concentrations on the standard curve) and the interassay precision (CV) was <6.5% across the three concentrations, calculated based on Clinical and Laboratory Standards Institute EP05-A3 and reported in the manufacturer's guidelines. Two control samples (low and high concentrations) were analysed in each batch in duplicate and data were only reported for the batch if the results of the controls were within the acceptance range outlined on each control sample vial. Each calibrator, control and patient sample were assessed in duplicate and the results only reported if the CV of the replicates was <10%.

Statistical analysis
For univariable analyses, the Shapiro-Wilk test was used to test for normal data distribution, and then appropriate tests, specified in the text, were used for comparisons. Associations between covariates and outcomes in COVID-19 and influenza A were assessed with binary logistic regression multivariable models. Sex, age, illness duration at time of sampling and comorbidity count were chosen as covariates. The comorbidity count was derived from the same comorbidities (table 1) from the two cohorts. To allow for potential non-linear relationship between predictors and the probability of an outcome, the models included smoothed thin plate regression spline terms for age, illness duration at time of sampling, comorbidity count and 25(OH)D concentrations. Multivariable models were estimated using the gam() function Open access of the R mgcv package using the default, thin plate regression smoothers. 21 22 The upper limit of smoother dimensionality was set to 9 for all variables, excluding the comorbidity count, where it was set to 7 as this variable was discrete with seven levels. Smoother parameters were estimated with restricted maximum likelihood. 25(OH)D concentrations were below the LLOD for 92 patients (free) and 2 patients (total) in the COVID-19 cohort. For the regression models, 25(OH)D values for these patients were imputed as the LLOD for the relevant analyte divided by the square root of 2. 23 As this is a commonly used but arbitrary method the regression analysis was repeated using 0 and the LLOD as imputed values to assess the sensitivity of the result to this assumption. Effects for categorical covariates are reported as OR; smoothed continuous covariates are reported graphically. Statistical analyses were conducted in R using the mgcv, tidyverse and gratia packages.  Obesity is a risk factor for severity and mortality in COVID-19 and can be associated with vitamin D deficiency. 15 However, there was no difference in total 25(OH) D concentration between patients with/without cliniciandefined obesity (online supplemental figure 5). Inflammatory mediator measurements had previously been performed on plasma samples from 66 patients included in this study. 24 Correlation matrix analysis demonstrated that total 25(OH)D was not significantly associated with circulating markers of systemic inflammation demonstrated to be involved in COVID-19 pathogenesis (online supplemental figure 6).

Patient characteristics
Multivariable analyses confirmed that total 25(OH) D concentration and vitamin D status (not sufficient) were both independently and negatively associated with receipt of IMV (table 2, figure 2A and online supplemental table 2). Two patients had total 25(OH)D concentrations below the LLOD; using 0 and LLOD, instead of LLOD divided by the square root of 2, had no substantive effect on significance of covariates or their effect sizes. Vitamin D status was also independently associated with in-hospital mortality, but total 25(OH)D concentration was not (online supplemental table 2 and online supplemental figure 7A).

Total 25(OH)D correlates with severity in influenza A
We then extended these observations to total 25(OH) D concentrations measured in people hospitalised with influenza A. Total 25(OH)D was not associated with age (p=0.1) or sex (p=0.  , table 2, online supplemental table 3 and online supplemental figure 7B).   -D and table 2). Free 25(OH)D was not associated with plasma inflammatory mediator concentrations (online supplemental figure 6).

DISCUSSION
Vitamin D insufficiency was prevalent and scaled with severity in patients with COVID-19 and influenza A and insufficiency persisted in survivors of critical illness. As determined by total 25(OH)D measurement, 73% of patients with COVID-19, 84% of patients with influenza A and 87% of critical illness survivors were vitamin D insufficient/deficient. We demonstrate evidence of a strong association between vitamin D status (insufficiency/ deficiency) during illness and both COVID-19 severity (receipt of IMV) and in-hospital mortality, with relevant confounders such as sex, age, comorbidities and day of illness adjusted for. This observation was replicated in influenza A, but the smaller sample size (n=93 compared with n=259) limited multivariable analyses. For the first time, we demonstrate a similar strong negative association between free 25(OH)D and COVID-19 disease severity and mortality. The results from this study extend earlier findings from other observational studies reporting associations between vitamin D status and SARS-CoV-2 infection and COVID-19 outcome. 6 7 25-27 Vitamin D may beneficially modulate the host response against SARS-CoV-2 via intracrine immune signalling. Vitamin D enhances intracellular pathogen clearance, primarily via the induction of autophagy. 28 Importantly, the ability of macrophages to produce cathelicidin, which has antiviral activity against influenza virus and respiratory syncytial virus, correlates with circulating 25(OH)D concentrations. 29 Although antiviral effects of vitamin D have not yet been demonstrated in vitro for SARS-CoV-2, they have been demonstrated for other bacterial and viral pathogens. 30 31 Consistent with vitamin D having a role in local immunomodulation, neither free nor total 25(OH) D correlated with circulating markers of systemic inflammation involved in COVID-19 pathogenesis (including C-reactive protein, interleukin 6 and granulocytemacrophage colony-stimulating factor 24 ).
Evidence for the importance of free versus total 25(OH) D in relation to the mechanisms by which vitamin D exerts antimicrobial and anti-inflammatory functions has been demonstrated. 32 33 We now demonstrate that free 25(OH) D was negatively associated with COVID-19 severity and in-hospital mortality. Studies directly measuring free 25(OH)D and immune responses to infection or during critical illness are limited. In a study of 30 critically ill patients, supplementation with high-dose vitamin D increased free 25(OH)D and plasma cathelicidin concentrations. 34 Another study of 30 patients with sepsis reported similar results when they examined the effects of vitamin D supplementation on bioavailable (combined albumin-bound and free fraction) 25(OH)D and cathelicidin concentrations. 35 Together, these findings suggest that low concentrations of free 25(OH)D may reduce vitamin D-induced antimicrobial and anti-inflammatory response, compromising immune defences.
We found that 30.2% of patients surviving critical illness and requiring IMV (prior to the COVID-19 pandemic) were vitamin D insufficient and 56.8% were deficient. Open access Vitamin D deficiency is common in critical illness, with a reported prevalence of between 40% and 70% in observational studies of both adults and children worldwide. 36 37 Although some patients may enter ICU in a deficient state due to pre-existing disease and malnutrition, vitamin D metabolism is dysregulated in critical illness 38 and concentrations fall rapidly after ICU admission. 39 The mechanistic link between acute illness and vitamin D deficiency is likely to be multifactorial, including reduced dietary intake/absorption, reduced cutaneous synthesis due to lack of sunlight and wastage due to reductions in vitamin D binding protein. 40 Furthermore, vitamin D insufficiency/deficiency has been associated with a range of poor outcomes in critical illness. 37 41-43 Vitamin D insufficiency leads to secondary hyperparathyroidism and a concentration of 50 nmol/L total 25(OH)D is required for optimum PTH concentrations. 44 Of the critical illness survivors, 87% had total 25(OH)D <50 nmol/L, which would be associated with secondary hyperparathyroidism and the potential for associated loss of bone mineral density. Critical illness survivors suffer accelerated loss of bone mineral density in the year after ICU discharge (compared with matched controls) and increased 10-year fracture risk. 14 Our findings implicate vitamin D insufficiency in this process. There is evidence that vitamin D supplementation can improve circulating total 25(OH)D concentrations in critically ill patients, 34 35 45 but evidence of a beneficial effect on outcomes is less clear. High-dose vitamin D supplementation in COVID-19 13 and critical illness 45 has been shown to increase plasma 25(OH)D concentrations 7 days post supplementation, but no significant reduction in the length of hospital stay or acute outcomes, including in-hospital mortality, admission to ICU or requirement for IMV, was demonstrated. 13 45 46 Longer-term outcomes such as bone health have not been evaluated. Conversely, two randomised trials reported high-dose 25(OH)D 3 (instead of vitamin D 3 as in the above-mentioned studies) on admission and then subsequent doses on either days 3 and 7, then weekly (n=76), 47 or days 3, 7, 15 and 30 (n=838), 48 were less likely to require ICU admission. We identified that vitamin D insufficiency was present early in the course of COVID-19 and influenza A (10 and 7 days after symptom onset, respectively), indicating that timing of supplementation may be an important factor when designing future supplementation studies. We propose that future studies examining effects on disease progression should investigate the effects of vitamin D supplementation given earlier in the course of the disease, closer to symptom onset rather than after hospitalisation. Evidence from an observational study of vitamin D supplementation usage supports this approach. 49 In a cohort of 8297 people with SARS-CoV-2 test results available, habitual vitamin D supplement intake prior to the pandemic was associated with a reduced risk of a positive test result after correction for known confounders including demographics and comorbidities. Furthermore, despite a decline in vitamin D following cardiothoracic surgery, postoperative outcomes (including organ dysfunction and mortality) are still associated with preoperative vitamin D status. 50 This suggests that supplementation prior to illness onset can still be expected to improve outcomes despite the fall in vitamin D concentration during acute illness. The longer-term effects of persistent vitamin D insufficiency/ deficiency in survivors of critical illness also require further investigation especially in the context of bone health, which could be independently evaluated using sequential measurement of bone turnover markers and serum PTH.
The current study has several important limitations. The observational design prevents any conclusions about a causal role of vitamin D status in poor clinical outcome being drawn. We cannot exclude the alternative explanation that the differences in vitamin D status were a consequence of illness severity. Although blood samples were obtained from people with COVID-19 and influenza A as soon after hospital admission as feasible, even these early measurements will still be subject to acute illness-related changes in vitamin D homeostasis. Healthy control data are presented (online supplemental figure 4), but these samples were obtained between the months of June and September, whereas samples from people with COVID-19 and influenza A were obtained between November and June (although no intermonth variation was observed). Data on prehospital or in-hospital vitamin D supplementation were not available. Finally, longer-term follow-up samples to assess vitamin D status in survivors of COVID-19 will be informative.
In conclusion, vitamin D deficiency/insufficiency was present in the majority of hospitalised patients with COVID-19 or influenza A and scaled with severity, highlighting that reduced concentrations of vitamin D are common to these disease states and distinct patient cohorts. For the first time, free and total 25(OH)D were studied in COVID-19, demonstrating consistent results. It is not clear whether vitamin D status led to poor clinical outcome or was a consequence of illness severity. Randomised trials will be necessary to determine whether a causal relationship exists between vitamin D early in the course of the disease and development of critical illness. Since vitamin D deficiency/insufficiency persisted at concentrations expected to disrupt bone metabolism in critical illness survivors, investigation of longer-term bone health outcomes is also warranted.