*Article* **Does Route of Full Feeding Affect Outcome among Ventilated Critically Ill COVID-19 Patients: A Prospective Observational Study**

**Dimitrios Karayiannis 1,\*, Sotirios Kakavas 2, Aikaterini Sarri 3, Vassiliki Giannopoulou 3, Christina Liakopoulou 3, Edison Jahaj 3, Aggeliki Kanavou 3, Thodoris Pitsolis 3, Sotirios Malachias 3, George Adamos 3, Athina Mantelou 3, Avra Almperti 1, Konstantina Morogianni 1, Olga Kampouropoulou 3, Anastasia Kotanidou <sup>3</sup> and Zafeiria Mastora <sup>3</sup>**


**Abstract:** The outbreak of the new coronavirus strain SARS-CoV-2 (COVID-19) highlighted the need for appropriate feeding practices among critically ill patients admitted to the intensive care unit (ICU). This study aimed to describe feeding practices of intubated COVID-19 patients during their second week of hospitalization in the First Department of Critical Care Medicine, Evaggelismos General Hospital, and evaluate potential associations with all cause 30-day mortality, length of hospital stay, and duration of mechanical ventilation. We enrolled adult intubated COVID-19 patients admitted to the ICU between September 2020 and July 2021 and prospectively monitored until their hospital discharge. Of the 162 patients analyzed (52.8% men, 51.6% overweight/obese, mean age 63.2 ± 1.9 years), 27.2% of patients used parenteral nutrition, while the rest were fed enterally. By 30 days, 34.2% of the patients in the parenteral group had died compared to 32.7% of the patients in the enteral group (relative risk (RR) for the group receiving enteral nutrition = 0.97, 95% confidence interval = 0.88–1.06, *p* = 0.120). Those in the enteral group demonstrated a lower duration of hospital stay (RR = 0.91, 95% CI = 0.85-0.97, *p* = 0.036) as well as mechanical ventilation support (RR = 0.94, 95% CI = 0.89–0.99, *p* = 0.043). Enteral feeding during second week of ICU hospitalization may be associated with a shorter duration of hospitalization and stay in mechanical ventilation support among critically ill intubated patients with COVID-19.

**Keywords:** SARS-CoV-2 virus; energy target; enteral nutrition; parenteral nutrition; critical illness

#### **1. Introduction**

The recent COVID-19 pandemic, which resulted from SARS-CoV-2 coronavirus infection, contributed to a rapid increment in hospital and intensive care unit (ICU) admissions [1]. Most patients exhibit mild to moderate symptoms, such as fever, cough, fatigue, respiratory failure, and multiple effects on the gastrointestinal tract, which appear especially in patients who are over 60 years old and those with concomitant diseases [2]. Approximately 5% of patients develop severe disease and require ICU admission, where the provision of adequate nutritional support is a challenge given the complex and fluctuating metabolic changes that accompany their nutritional status over time [3]. Critical illness, in general, is characterized by an inflammatory response that elicits a catabolic response and

**Citation:** Karayiannis, D.; Kakavas, S.; Sarri, A.; Giannopoulou, V.; Liakopoulou, C.; Jahaj, E.; Kanavou, A.; Pitsolis, T.; Malachias, S.; Adamos, G.; et al. Does Route of Full Feeding Affect Outcome among Ventilated Critically Ill COVID-19 Patients: A Prospective Observational Study. *Nutrients* **2022**, *14*, 153. https:// doi.org/10.3390/nu14010153

Academic Editor: Arved Weimann

Received: 24 November 2021 Accepted: 26 December 2021 Published: 29 December 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

can be divided into different metabolic diseases: the early acute phase, which occurs in the first 48 h after entering the ICU, and the next, called the late acute phase, which usually lasts days after admission [4]. In severely ill patients with COVID-19, the acute phase presents with a prolonged severe inflammatory response, which is associated with a significant increase in resting energy expenditure [5,6], and current strategies for nutritional therapy seem to be unsuccessful in covering energy/protein needs [7]. In addition, COVID-19 patients spend extended periods in the ICU compared to the usual severely ill [8], and thus, metabolic phases of the disease are likely to be different and last longer.

Nutritional support is essential for all critically ill COVID-19 patients because acute illness carries an increased risk of malnutrition, muscle wasting, and increased mortality and requires a complex calculation of feeding route, time of onset, and amount and type of nutrients [4,9]. All the above may significantly affect patient's outcomes, such as length of hospital stay or duration of mechanical ventilation [10]. Recently, various guidelines from international societies were developed and published early in the pandemic without providing convincing answers to many critical questions [3,11,12]. One important question is route of nutritional support administration (enteral vs. parenteral), the effects of which on outcome remain unclear [13,14]. Up to the present moment, administration of energy needs via the enteral route is the preferred method of choice according to published guidelines [15,16] although it is affiliated with higher rates of gastrointestinal intolerance [17]. Parenteral feeding is more invasive, provides higher amounts of calories, and is associated with a higher risk of complications [18]. There are several published papers comparing different types of interventions regarding the route of feeding and ICU outcome but mainly among non COVID-19 patients during their first days of ICU stay [19].

Despite it being quoted in various guidelines, there is lack of data comparing enteral vs. parenteral feeding among critically ill patients with COVID-19 disease as well as the effect of full feeding (after 5–7 days of ICU stay) on clinical outcomes [20–23]. Therefore, the purpose of this study was to describe full feeding practices and highlight their long-term consequences among intubated patients with COVID-19 disease, who were treated within ICU at a reference hospital in Athens.

#### **2. Methods**

#### *2.1. Participants*

We conducted this single-center observational study from September 2020 to July 2021 in the first intensive care unit of Evaggelismos Athens General Hospital (tertiary hospital). Adult patients (aged > 18 years), requiring mechanical ventilation for >48 h, projected to receive nutritional support for at least 5 days, and with laboratory-confirmed COVID-19 by a positive reverse-transcriptase polymerase-chain-reaction (RT-PCR) assay of a nasopharyngeal swab specimen were eligible. The Institutional Review Board at Athens Evangelismos Hospital: 116/31-03-2021 approved data collection and waived the need for informed consent. Exclusion criteria were existence of metabolic diseases requiring a specific diet (for example phenylketonuria), history of gastrectomy or esophagectomy, initiation of nutrition support 5–7 days before ICU admission, ICU length of stay (LOS) shorter than 7 days, pregnancy, or patients that were expected to require no more than 48 h of invasive mechanical ventilation.

#### *2.2. Study Design*

Data were retrieved from patient electronic file and daily hospital sheet on the admission, 7th day, 14th day, and until discharge from hospital. During ICU hospitalization, anthropometric data, such as height (in cm) and weight (in kg), were collected, while body mass index (in kg/m2) was determined. Besides patients' basic characteristics, the Acute Physiology and Chronic Health Evaluation (Apache) score was calculated 24 h after ICU admission. After the initiation of nutrition support, various data, such as pre-existing illness, initiation/duration of mechanical ventilation, resource to organ support (vasopressors and

kidney replacement therapy), use of prone position, and laboratory markers, such as serum albumin and CRP levels, were also computed.

#### *2.3. Feeding Practices*

Nutritional support was initiated as soon as possible after admission to the ICU and no later than 48 h after intubation. Data on feeding practices and tolerance to feeding were collected every day by the Nutrition Support Team (NST), while nutrient delivery was based on ESPEN guidelines by the hospital NST [4]. Feeding practices included enteral and parenteral feeding, while nutritional parameters assessed included calorie (from nutrient and non-nutrient calorie sources such as propofol and dextrose) and protein intake Regarding caloric requirements, calculations were based on previously published indirect calorimetry measurements in a subset of patients [6]. In our ICU, we had established protocols for nutrition support from the first day of the COVID-19 pandemic, based on the ESPEN guidelines or local consensus. The NST made the decision on the nutrition regimen as well as the route of nutrition support in collaboration with the treating physician.

Patients in the enteral nutrition group were fed through a nasogastric or nasoduodenal catheter depending on tolerance and comorbidities. The aim was to meet the caloric requirements by the 7th day of hospitalization using isosmotic, isocaloric, high-protein, polymeric preparations, after which the decision was at the discretion of the bedside physician [9]. In all cases, nutrition support was administered continuously (mL per h). Malnourished patients (BMI < 17 or clinical diagnosis) were assigned to a parenteral nutrition protocol designed to reduce the risk of refeeding syndrome [11,16]. In the parenteral group (presence of uncontrolled shock, uncontrolled life-threatening hypoxemia, hypercapnia or acidosis, hemodynamic instability, or when energy intake was consistently <50% of targets over 5–7 days), all patients received only parenteral nutrition via a central venous catheter depending on the results of daily haemodynamic assessments and nonfunctional digestive tract. We excluded from the analysis patients using supplemental parenteral nutrition to achieve predefined calorie targets after day 8. Parenteral nutrition was stopped and replaced with enteral nutrition at the flow rate needed to achieve the pre-defined calorie target once the patient met pre-defined criteria for haemodynamic stability. Bedside physician provided extra water, electrolytes, vitamins, and trace elements using standard preparations.

#### *2.4. Outcome*

The primary outcome was death by day 30. Secondary outcomes included mortality up to day 60 of hospitalization, ICU mortality, length of hospital and ICU stay, days of mechanical ventilation support, days of renal replacement, and percentage of patients who received mechanical ventilation, vasopressors and renal replacement therapy.

#### *2.5. Statistical Analysis*

The categorical variables are expressed as frequencies while the continuous variables as means (with standard deviations) or medians (with interquartile ranges (IQRs)), according to the rejection/not rejection of null hypothesis by the Shapiro–Wilk test, and then compared using Students *t*-test or Mann–Whitney U test. Due to the lack of information regarding effect of the feeding route on COVID-19 patients at the time of study design, it was not possible to perform a sample size calculation prior to the beginning of the current study when we performed hypothesis testing. The comparison between the groups was made using the Wilcoxon rank sum tests for continuous variables, while a chi-square test of independence (*χ*2) (Fisher exact tests where appropriate) was performed for categorical variables. We used generalized linear models to test associations between feeding groups (enteral vs. parenteral nutrition) and study final outcomes. A Poisson distribution with log-link function was used to test association of number of count data, while a Binomial distribution with logit link function was used for clinical endpoints. The results are presented as relative risk (RR) and 95% confidence intervals (CIs). Statistical tests were considered significant if *p* < 0.05. All statistical analyzes were conducted using SPSS version 21.0.

#### **3. Results**

#### *3.1. Patients*

From September 2020 to July 2021, 192 patients with a confirmed SARS-CoV-2 infection during the second and third wave of the COVID-19 pandemic were admitted to the Evangelismos intensive care unit (ICU). Of these, 16 patients were excluded from the study because they either died within the first 48 h after admission or were under nutritional support, while four had mechanical ventilation started more than 24 h earlier, resulting in 176 patients (Figure 1). Participants were intubated within 24 h of admission, and their mean age was 62.1 ± 10.9 years, while 51.2% were male. The demographic, baseline, and nutritional characteristics of the patients analyzed are presented in Table 1. There was no clear categorization of patients as malnourished or not at admission, as it was not possible to collect information on dietary intake, weight loss, and the usual body weight. Almost half of the participants had Nutric Score values >5, indicating a high nutritional risk upon admission to the ICU.

**Figure 1.** Patient recruitment flow chart.

**Table 1.** Baseline characteristics of the participants (*n* = 162).



**Table 1.** *Cont.*

Values represent median (IQR) or means (+SD) or number of subjects (*n*, %). \* Denotes statistically significant different between groups at <0.05 level, *p* = *p* value for Students *t*-test or Mann-Whitney *U* test or Chi square test. Abbreviations: APACHE, Acute Physiology and Chronic Health Evaluation; COPD, chronic obstructive pulmonary disease; BMI, body mass index; PEEP, positive end expiratory pressure; Nutric Score, Nutrition Risk in the Critically Ill; Fi02, fraction of inspired oxygen; NMBAs, neuromuscular blocking agents.

#### *3.2. Outcomes*

Data on feeding practices are presented in Table 1. A total of 117 individuals were fed enterally, while the rest were fed through the parenteral route. Of those belonging to the parenteral nutrition group, 5.6% experienced side effects, such as electrolyte disturbances, pneumothorax, hypoglycemia, and cholestasis, while in those receiving enteral nutrition, vomiting (32.3%) was the most common side effect.

Fifteen of 45 patients (33.3%) in the parenteral group and 38 of 117 patients (34.2%) in the enteral group died, with no significant between-group difference by day 30 even after adjusting for various baseline factors, such as age, sex, and race, APACHE score, Nutric Score (as a dichotomous variable), BMI, diabetes, and chronic kidney disease (relative risk in the enteral group, 0.97; 95% confidence interval (CI), 0.88 to 1.06, Table 2). There were significant reductions in the enteral group as compared with the parenteral group for in-hospital length of stay (RR = 0.92; 95% CI, 0.86 to 0.98; *p* = 0.039), ventilator days (RR = 0.94; 95% CI, 0.89 to 0.99; *p* = 0.043), and elevated liver enzymes ((RR = 0.91; 95% CI, 0.85 to 0.97; *p* = 0.039); *p* = 0.022). However, there were also significant differences regarding the rates of adverse gastrointestinal events between the parenteral group and the enteral group (32.3% vs. 22.1% for vomiting and 37.2% vs. 29.2% for diarrhea, *p* < 0.05). There was no significant difference in the duration of survival up to 60 days. Initiation of feeding was delayed in 37 patients in the parenteral nutrition group and in 41 patients in the enteral nutrition group, while caloric and protein intake are shown in Table 1. The energy target of 25–30 calories per kilogram of body weight per day was reached in the majority of patients in both groups, and the average caloric intake was almost identical in both groups.

**Table 2.** Patients primary and secondary outcomes (*n* = 162).


Values represent median (IQR) or means (+SD) or number of subjects (*n*, %). \* Denotes statistically significant different between groups at <0.05 level **#** Covariates were selected a priori, incorporating demographic information (age, sex, and race). APACHE score, Nutric Score (dichotomous variable), BMI, diabetes, and chronic kidney disease. Abbreviations: ICU, intensive care unit; RRT, renal replacement therapy.

#### **4. Discussion**

This single-center prospective study evaluated the effect of the nutritional support administration route after the seventh day of hospitalization on outcome among critically ill intubated COVID-19 patients. Our data suggest that enteral feeding is superior to parenteral feeding in terms of length of hospital stay and mechanical ventilation, but these findings were not accompanied by a corresponding improvement in mortality both in-hospital and out-of-hospital at 60 days. These two groups also did not exhibit any differences regarding infections incidence, days on RRT, and septic shock prevalence.

Are there other data available to date on feeding practices and their effect on outcome among COVID-19 critically ill patients? Only a few studies so far have provided limited data about nutrition support. A USA cohort [24] revealed that more than half of the participants (56%) presented intolerance to enteral nutrition, which was associated with higher ICU stay and in-hospital mortality, whereas a similar study among intubated patients from Mexico revealed a lower prevalence of intolerance to enteral nutrition—about 32% [25]. A series of 176 critically ill patients with COVID-19 disease [26] managed to reach their energy and protein requirements during the first week of admission especially through the use of supplemental parenteral nutrition. Parenteral nutrition use was comparable to our study, as approximately 35% of patients were fed through the parenteral route compared with 27.7% of our patients. Reports from another study—the ISIIC point prevalence study also suggest that there is a growing interest in the role of nutrition support, which resulted in providing COVID-19 patients with higher amounts of energy and protein compared to non-COVID-19 patients [27].

There is also lack of available data regarding the effect of feeding administration after the first week of ICU hospitalization when energy and protein targets are theoretically achieved [26]. All previous major studies tried to investigate the effect of either early administration (during the first 72 h following admission) of nutritional support [28] either up to the fifth day of hospitalization [19,29] or the first week of hospitalization [14] on outcome. The two largest multicenter clinical trials among critically ill patients assessing the effects of nutritional support route during first week of ICU stay suggest nonsignificant difference in all-cause mortality, frequency of infectious complications, ICU, and hospital length of stay (LOS) [14,19]. In our cohort, feeding practices were investigated during the second week of hospitalization and not the first. This was settled for two reasons: so far, most studies that explore the relationship between feeding practices during the first week of ICU and outcome have not provided any clear association [10,14,19,21]. The other reason is that in critically ill patients, energy metabolic demands usually peak within five to seven days before returning to normal (ebb/flow phases) [30]. Regarding COVID-19 disease, there is a stable hypermetabolic condition probably because of hyperinflammatory response, which persists during the second week of hospitalization and appears to stabilize after the 10th day of ICU stay, as recent studies have demonstrated [5,6].

The total length of hospital stay was lower in those patients receiving enteral nutrition although ICU length of stay (LOS) was not different. A systematic review and meta-analysis show that early enteral feeding is associated with a reduced postoperative length of stay among patients undergoing lower gastrointestinal surgery [31], while among critically ill patients, route of nutrition support during first week of hospitalization was not related to LOS although there was a trend for reduced ICU stay [14]. Harvey et al. revealed that there was no difference in both length of hospital and ICU stay among critically ill patients according to feeding rout e [19]. Another main finding of this study was that those receiving enteral feeding presented a shorter duration of mechanical ventilation support. As compared to earlier times, enteral feeding is no longer contraindicated in patients under mechanical ventilation and is the preferred route of feeding administration [32], while some data do not suggest that enteral feeding is superior [33]. On the other hand, parenteral feeding is well recognized for its higher caloric intake, leading to hyperglycemia and increased ventilator stay [34].

In addition to these findings, parenteral nutrition group was more likely to experience an increase in liver enzymes during hospitalization. Research on the effects of parenteral nutrition has focused on the possibility of liver dysfunction as well as an increase in liver enzymes [35]. This phenomenon is a potential result of cholestasis, which is caused by biliary obstruction or impaired secretion of bile [36]. On appearance, γ-glutamyl transpeptidase, alkaline phosphatase, and conjugated bilirubin levels are elevated. This is a clinical manifestation accompanying long-term parenteral nutrition therapy among pediatric patients or adults. Long-term PN therapy-induced cholestasis is a significant consequence that can progress to cirrhosis and liver failure [37].

The present study was not without limitations. First, our study had a relatively small sample size, which limits the generalizability to other ICUs and countries. Second, the prospective nature of this study does not define a causal relationship and reflects only associations between the implemented nutrition protocols and study outcomes. Moreover, we did not obtain data regarding energy expenditure by using indirect calorimetry in the whole sample because it was very time consuming to get such a large quantity of data. Our study was not a randomized evaluation of feeding practices; even if the indication for parenteral and enteral nutrition followed guidelines and local treatment protocols, a selection bias cannot be excluded. We should also point out that a higher proportion of patients belonging to the parenteral nutrition group exhibited high Nutric Score values, but we have controlled for this confounding factor in the multivariate models. Furthermore, it was not possible to assess all possible covariates among our patients, and future studies might lead to different results regarding the provision of nutrition therapy in critically ill patients.

#### **5. Conclusions**

There are scarce data on feeding practices among critically ill COVID-19 patients and their effects on outcome. Our study demonstrated that the majority of COVID-19 intubated patients were fed through the enteral route during their second week of ICU hospitalization. Enteral nutrition was not superior compared to parenteral feeding in relation to the main study outcome, which was 30-day mortality, but it was associated with reduced length of hospital stay, less demand for mechanical ventilation support, and a more favorable profile for liver enzyme levels.

**Author Contributions:** Conceptualization, D.K., Z.M. and A.K. (Anastasia Kotanidou); formal analysis, D.K. and A.A.; methodology, Z.M., A.A., K.M., D.K., E.J. and T.P.; project administration, D.K.; supervision, D.K. and Z.M.; writing—original draft, D.K.; writing—review and editing, A.K. (Aggeliki Kanavou), A.M., V.G., D.K., A.S., C.L., S.K., O.K., S.M., A.S. and G.A. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Institutional Review Board Statement:** The study was conducted in accordance with the Declaration of Helsinki, and approved by the Institutional Review Board at Athens Evangelismos Hospital: 116/31-03-2021.

**Informed Consent Statement:** The Institutional Review Board at Athens Evangelismos Hospital: 116/31-03-2021 approved data collection and waived the need for informed consent.

**Acknowledgments:** We are grateful to Athina Kardara and Emy Valavani for their assistance in data collection.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **Abbreviations**


#### **References**


## *Article* **Stress Hyperglycemia and Osteocalcin in COVID-19 Critically Ill Patients on Artificial Nutrition**

**Francisco Arrieta 1,2, Victoria Martinez-Vaello 1, Nuria Bengoa 1, Marta Rosillo 3, Angélica de Pablo 4, Cristina Voguel 4, Rosario Pintor 5, Amaya Belanger-Quintana 6, Raquel Mateo-Lobo 1, Angel Candela <sup>4</sup> and José I. Botella-Carretero 1,2,\***


**Abstract:** We aimed to study the possible association of stress hyperglycemia in COVID-19 critically ill patients with prognosis, artificial nutrition, circulating osteocalcin, and other serum markers of inflammation and compare them with non-COVID-19 patients. Fifty-two critical patients at the intensive care unit (ICU), 26 with COVID-19 and 26 non-COVID-19, were included. Glycemic control, delivery of artificial nutrition, serum osteocalcin, total and ICU stays, and mortality were recorded. Patients with COVID-19 had higher ICU stays, were on artificial nutrition for longer (*p* = 0.004), and needed more frequently insulin infusion therapy (*p* = 0.022) to control stress hyperglycemia. The need for insulin infusion therapy was associated with higher energy (*p* = 0.001) and glucose delivered through artificial nutrition (*p* = 0.040). Those patients with stress hyperglycemia showed higher ICU stays (23 ± 17 vs. 11 ± 13 days, *p* = 0.007). Serum osteocalcin was a good marker for hyperglycemia, as it inversely correlated with glycemia at admission in the ICU (*r* = −0.476, *p* = 0.001) and at days 2 (*r* = −0.409, *p* = 0.007) and 3 (*r* = −0.351, *p* = 0.049). In conclusion, hyperglycemia in critically ill COVID-19 patients was associated with longer ICU stays. Low circulating osteocalcin was a good marker for stress hyperglycemia.

**Keywords:** COVID-19; hyperglycemia; parenteral nutrition; enteral nutrition; osteocalcin

### **1. Introduction**

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is the cause of the coronavirus disease 2019 (COVID-19) pandemic, which has affected more than two hundred million individuals and is the cause of more than four million deaths worldwide as of this writing (https://coronavirus.jhu.edu/map.html, last accessed on 24 August 2021). Hyperglycemia is a risk factor for a more severe course of the disease, as hospitalized COVID-19 patients with diabetes show longer hospital stays than patients without diabetes [1]. They suffer disproportionately from acute COVID-19, with higher rates of serious complications and death [2]. Chronic inflammation, increased coagulation activity, immune response impairment, and potential direct pancreatic damage by SARS-CoV-2 might be among the underlying mechanisms for this association [3]. Furthermore, an

**Citation:** Arrieta, F.; Martinez-Vaello, V.; Bengoa, N.; Rosillo, M.; de Pablo, A.; Voguel, C.; Pintor, R.; Belanger-Quintana, A.; Mateo-Lobo, R.; Candela, A.; et al. Stress Hyperglycemia and Osteocalcin in COVID-19 Critically Ill Patients on Artificial Nutrition. *Nutrients* **2021**, *13*, 3010. https://doi.org/10.3390/ nu13093010

Academic Editors: Dimitrios T. Karayiannis and Zafeiria Mastora

Received: 19 July 2021 Accepted: 25 August 2021 Published: 28 August 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

important proportion of COVID-19 patients need admission to an intensive care unit (ICU) and may develop stress hyperglycemia—whether or not having a previous history of diabetes mellitus—which has been shown to be a prognostic factor [4,5].

Among the factors associated with stress hyperglycemia, the need for parenteral nutrition (PN)—which is normally part of the nutritional therapy together with enteral nutrition (EN) in critically ill patients [6]—might be one of them, especially in older individuals [7]. Besides, several interactions between some counter-regulatory hormones, adipokines, and inflammatory cytokines produce excessive production of glucose by the liver and insulin resistance at the peripheral tissues [8]. The resultant hyperglycemia further exacerbates the inflammatory and oxidative stress response, potentially setting up a vicious cycle whereby hyperglycemia leads to further hyperglycemia [9].

In the last few years, circulating osteocalcin, an osteoblast-specific protein, has shown extraskeletal metabolic activity, such as promoting insulin secretion and increasing peripheral insulin sensitivity [10]. Reduced circulating osteocalcin was found in patients with type 2 diabetes mellitus and also associated with insulin sensitivity [11,12].

Therefore, given this previous knowledge, we aimed to study the possible association of stress hyperglycemia in COVID-19 critically ill patients with a worse prognosis compared to non-COVID-19 patients. We also aimed to explore the possible associations of hyperglycemia with circulating osteocalcin concentrations, the composition of artificial nutrition, and serum markers of inflammation.

#### **2. Materials and Methods**

#### *2.1. Subjects and Measurements*

Fifty-two patients were included in this study: 26 consecutive patients with severe COVID-19 requiring admission to the ICU and 26 non-COVID-19 critically ill postsurgical patients. The latter were historical controls before the COVID-19 pandemic occurred and had the following procedures: 7 cardiovascular surgery, 12 renal/urological surgery, 6 pelvic surgery, and 1 surgery for traumatic hemothorax. Patients with COVID-19 were diagnosed by the presence of SARS-CoV-2 in respiratory specimens by real-time polymerase chain reaction (RT-PCR) in pharyngeal swabs. All COVID-19 patients were critically-ill and on mechanical ventilation. They received our standard treatment protocol, including low molecular-weight heparin, glucocorticoids, and tocilizumab. Non-COVID-19 patients were all complicated postsurgical ones in need of mechanical ventilation and vasoactive drugs. Patients under 18 y were excluded as well as those with active cancer, pregnancy, or with end-stage renal or liver disease before the diagnosis of COVID-19.

Patients in both groups were categorized as with or without diabetes mellitus based on a previous diagnosis. Stress hyperglycemia was defined as a plasma glucose level of ≥140 mg/dl. Continuous insulin infusion (50 IU of Actrapid in 50 mL of 0.9% saline using an IV pump) was started when blood glucose was ≥180 mg/dl and adjusted for a glycemic range of 140–179 mg/dl [13]. Capillary glycemia was measured every one or two hours to adjust the infusion rate. When blood glucose fell to <140 mg/dl, insulin infusion was stopped and subcutaneous insulin started. Capillary glycemia was then measured every six hours. ICU stay and total hospital stay were recorded as well as mortality.

The type of composition of artificial nutrition administered to the included patients were recorded. PN was delivered through a central line as soon as the patient was hemodynamically stable. Individualized formulae or standardized commercial bags were prepared by the hospital pharmacy. We aimed at 20–25 kcal/Kg/day, with a proportion of 3–6 g/Kg/day for glucose, 1.0 g/Kg/day for amino-acids, and less than 1 g/Kg/day for lipids, with 7–10 g/day of essential fatty acids. Vitamins and trace elements were also added by the hospital pharmacy. For those patients with ICU stays above 7 days, and whenever possible, EN was started with a standard, fiber-free formulation and PN gradually tapered as EN was tolerated and increased.

#### *2.2. Ethics*

The study protocol was approved by the Institutional Ethics Committee of our center (study code 147/20) and performed according to the Declaration of Helsinki. Written or verbal informed consent was obtained from all patients.

#### *2.3. Analytical Assays*

Serum concentrations of glucose and other biochemical variables were measured with an Architect c16000/i2000-analyzer (Abbot Diagnostics, Maidenhead, UK) and HbA1c by high-performance liquid chromatography (A. Menarini Diagnostics, Bagno a Ripoli, Italy). Immunoanalysis was employed for the measurement of C-reactive protein (CRP), procalcitonin (Abbott, Illinois, IL, USA) and D-dimer (Siemens, Münich, Germany), and interleukins 6 (IL-6) and 12 (IL-12) by enzyme-linked immunosorbent assay (Invitrogen, Waltham, MA, USA) at ICU admission. Serum osteocalcin was measured by electrochemiluminescence Cobas-e601 (Roche Diagnostics, Rotkreuz ZG, Switzerland), with normal range of 15.0–45.0 μg/L also at ICU admission. The intra- and inter-assay coefficients of variation were below 10%.

#### *2.4. Statistical Analysis*

We used GRANMO 7.12 [14] for sample size analysis. The primary outcome was to study the possible association of stress hyperglycemia in COVID-19 critically ill patients with a worse prognosis compared to non-COVID-19 patients. In order to find a mean difference of 10 days in ICU stay with a SD of 10, we needed a sample size of at least 16 individuals in each group for a two-tail estimates setting α at 0.05 and β at 0.20. Secondary outcomes were the possible associations of hyperglycemia with circulating osteocalcin concentrations, the composition of artificial nutrition, and serum markers of inflammation. Based on our previous results [12], in order to find a mean difference of 7.6 μg/L in osteocalcin concentrations with a SD of 8.8, we needed a sample size of at least 21 individuals in each group for a two-tail estimates setting α at 0.05 and β at 0.20. Sample size analyses for the composition of artificial nutrition and inflammatory markers were not performed.

Results are expressed as means ± SD unless otherwise stated. The Kolmogorov– Smirnov statistic was applied to continuous variables. Logarithmic or square root transformations were used as needed to ensure normal distribution of the variables. To compare discontinuous variables, we used the χ<sup>2</sup> test and Fisher's exact test as appropriate. Unpaired *t*-tests or Mann–Whitney U tests were used to compare the central tendencies of the different groups as appropriate. Bivariate correlation was employed to study the association between two continuous variables using Pearson or Spearman's tests as appropriate. Two-way ANOVA was employed to analyze the effect of both COVID-19 and stress hyperglycemia on the studied continuous prognostic variables and corrected χ<sup>2</sup> test for categorical ones. Finally a multivariate linear regression analysis was also performed with a backwards strategy. Analyses were performed using SPSS 17 (SPSS Inc, Chicago, IL, USA). *p* < 0.05 was considered statistically significant.

#### **3. Results**

#### *3.1. Baseline Characteristics of Patients*

Of the 52 initially included patients, the results of 49 patients were finally analyzed, as three patients with COVID-19 had undetectable levels of osteocalcin (Table 1). Patients with COVID-19 were younger, with a higher proportion of males. Previous diagnosis of diabetes mellitus was similar in both groups as well as their HbA1c.


**Table 1.** Characteristics of the critically ill included patients.

Data are means ± SD unless otherwise stated. TPN, total parenteral nutrition; EN, enteral nutrition; CRP, C-reactive protein; IL, interleukin; ICU, intensive care unit. \* Stress hyperglycemia was defined as glycemia ≥ 140 mg/dl.

We did not find a difference in glycemia at ICU admission in the proportion of patients with stress hyperglycemia or in mean glycemia at ICU. However, patients with COVID-19 more frequently needed insulin infusion therapy for glycemic control and showed lower osteocalcin concentrations. Patients with COVID-19 were also on artificial nutrition for longer (both PN and EN) and received higher energy per day than non-COVID-19 ones but with similar amounts of delivered glucose per day. Serum CRP and procalcitonin at ICU admission were similar in both groups (Table 1).

Mortality did not differ between groups. Non-COVID-19 patients died as a result of cardiac arrest (*n* = 3), pulmonary embolism (*n* = 1), and multiorgan failure after septic shock (*n* = 3). COVID-19 patients died as a result of respiratory distress syndrome (*n* = 6) and pulmonary embolism (*n* = 2). The possibility of stroke and or any other CNS complications as the cause of death could not be accurately assessed due to profound sedation in these patients (Table 1).

#### *3.2. Impact of Stress Hyperglycemia*

When both patients with and without COVID-19 were considered together and classified according to the presence of stress hyperglycemia, the latter group showed a higher proportion of previous diagnosis of type 2 diabetes mellitus and higher HbA1c levels

(Table 2). They were on PN for longer, received higher amounts of energy and glucose from artificial nutrition, and showed higher ICU stays.


**Table 2.** Characteristics of patients with and without stress hyperglycemia \*.

Data are means ± SD unless otherwise stated. TPN, total parenteral nutrition; EN, enteral nutrition; CRP, C-reactive protein; ICU, intensive care unit. \* Stress hyperglycemia was defined as glycemia ≥ 140 mg/dl.

Total hospital stay did not correlate with glucose control, but ICU stay correlated with glycemia at admission in ICU (*r* = 0.337, *p* = 0.018) and at day 2 (*r* = 0.427, *p* = 0.015). Those patients with stress hyperglycemia showed higher ICU stays. As expected, those patients with longer ICU stays needed more days on artificial nutrition for both PN (*r* = 0.741, *p* < 0.001) and EN (*r* = 0.829, *p* < 0.001).

#### *3.3. Circulating Osteocalcin as a Marker for Hyperglycemia and Prognosis*

Circulating osteocalcin was lower in patients with COVID-19 (Table 1). It was lower in patients with stress hyperglycemia (Table 2) and in those in need for insulin infusion therapy (7.0 ± 4.7 vs. 12.3 ± 7.0 μg/L for subcutaneous insulin, *t* = 3.672, *p* = 0.001).

Circulating osteocalcin concentrations inversely correlated with glycemia at the day of admission in the ICU (*r* = −0.476, *p* = 0.001) and at days 2 (*r* = −0.409, *p* = 0.007) and 3 (*r* = −0.351, *p* = 0.049) (Figure 1). Osteocalcin did not correlate with HbA1c (*r* = −0.207, *p* = 0.225) or age (*r* = 0.145, *p* = 0.319), and their levels were similar in men and women (*t* = 0.482, *p* = 0.632).

**Figure 1.** Correlations of circulating osteocalcin with glycemia and HbA1c.

#### *3.4. Impact of the Composition of Artificial Nutrition*

The amount of energy delivered through PN did not correlate with glycemia at the day of admission in the ICU (*r* = 0.245, *p* = 0.097), at day 2 (*r* = 0.211, *p* = 0.245), or 3 (*r* = 0.069, *p* = 0.750). Conversely, the amount of glucose delivered through PN was positively correlated with glycemia at the day of admission in the ICU (*r* = 0.395, *p* = 0.006). In addition, mean glucose and energy delivered through artificial nutrition correlated with mean glycemia during ICU stay (*r* = 0.399, *p* = 0.007 and *r* = 0.291, *p* = 0.048, respectively) (Figure 2).

**Figure 2.** Correlations of energy and glucose delivered through artificial nutrition with glycemia.

Those patients in need of insulin infusion therapy compared with those on subcutaneous insulin received higher energy (1221 ± 256 vs. 923 ± 294 kcal/day, respectively, *t* = 3.709, *p* = 0.001) and glucose delivered through artificial nutrition (146 ± 22 vs. 133 ± 22 g/day, respectively, *t* = 2.112, *p* = 0.040).

#### *3.5. Circulating Inflammatory Markers*

Glycemia at the day of admission in the ICU was positively correlated with IL-12 (*r* = 0.454, *p* = 0.038) but not with CRP (*r* = 0.065, *p* = 0.722), procalcitonin (*r* = 0.145, *p* = 0.446), IL-6 (*r* = 0.138, *p* = 0.551), or D-dimer (*r* = 0.232, *p* = 0.312) (Figure 3).

**Figure 3.** Correlations of serum inflammatory markers with glycemia.

Patients with stress hyperglycemia had more similar inflammatory markers at ICU admission than those without it (Table 2). The same occurred between those patients in need for insulin infusion therapy compared with those on subcutaneous insulin (*p* > 0.05 for IL-12, CRP, procalcitonin, D-dimer, and IL-6).

#### *3.6. Ancillary Analyses*

In order to correct for the effects of both the presence of COVID-19 and stress hyperglycemia in prognostic parameters and osteocalcin, we performed a two-way analysis of variance for continuous variables and corrected χ<sup>2</sup> test for categorical ones. Separated data in four groups are shown in Table 3.


**Table 3.** Prognosis and serum markers of patients with and without COVID-19 and with or without stress hyperglycemia \*,†.

Data are means ± SD unless otherwise stated. ICU, intensive care unit; CRP, C-reactive protein. \* Stress hyperglycemia was defined as glycemia ≥ 140 mg/dl. †, *<sup>p</sup>* show the results of two-way ANOVA except for categorical variables that were analyzed by corrected <sup>χ</sup><sup>2</sup> tests.

> ICU stay was higher with both COVID-19 diagnosis and the presence of stress hyperglycemia, indicating these two were independent prognostic factors. Conversely total hospital stay and mortality were not associated with either COVID-19 diagnosis or the presence of stress hyperglycemia. Circulating osteocalcin was lower in both COVID-19 patients and with the presence of stress hyperglycemia, but CRP and procalcitonin showed no associations (Table 3).

> Insulin infusion was more frequent for both COVID-19 diagnosis and the presence of stress hyperglycemia. The time on artificial nutrition, both on TPN and EN, was longer in those patients with COVID-19 and with stress hyperglycemia. The amount of energy and glucose delivered by artificial nutrition were also associated with both the presence of COVID-19 and stress hyperglycemia and with their interaction (Table 3).

> Finally, a multivariate linear regression analysis was also performed with a backwards strategy, introducing COVID-19 diagnosis, the presence of stress hyperglycemia, the need for insulin infusion, age, and sex as independent variables and ICU stay as the dependent variable. Both COVID-19 diagnosis (β = 0.488, *p* < 0.001) and the presence of stress hyperglycemia (β = 0.394, *p* = 0.001) were retained by the model (R2 = 0.383, F = 14.297, *p* < 0.001). When osteocalcin was introduced as the dependent variable, both COVID-19 diagnosis (β = −0.375, *p* = 0.005) and the presence of stress hyperglycemia (β = −0.344, *p* = 0.009) were retained by the model (R2 = 0.347, F = 12.224, *p* < 0.001).

> Another multivariate regression analysis model was performed as to take into account the artificial nutrition characteristics on ICU stay, which was introduced as the dependent

variable. The independent variables in this model were the days on artificial nutrition, both TPN and EN, the amount of energy and glucose delivered by artificial nutrition, and the need for insulin infusion. The presence of COVID-19 and stress hyperglycemia could not be introduced in the model, as they showed collinearity with the variables of artificial nutrition. The variables retained by the model (R2 = 0.965, F = 438.5, *p* < 0.001) were the days on both TPN (β = 0.535, *p* < 0.001) and EN (β = 0.652, *p* < 0.001) and the need for insulin infusion (β = 0.066, *p* = 0.031).

#### **4. Discussion**

In the present study, we have shown that hyperglycemia in critically ill COVID-19 patients was associated with longer ICU stays and higher amounts of glucose delivered through artificial nutrition. Low circulating osteocalcin was lower in COVID-19 patients, in those with stress hyperglycemia, and in those in need for insulin infusion therapy. Therefore, osteocalcin could be considered a useful marker for stress hyperglycemia and prognosis at ICU.

Osteocalcin, while playing important roles in bone remodeling, also contributes to glucose metabolism by affecting both insulin secretion and insulin sensitivity [15]. In vitro, co-cultures of pancreatic islets and wild-type osteoblasts stimulated insulin secretion, whereas knockout osteocalcin osteoblasts did not [15]. Furthermore, the link between bone and glucose metabolism is supported by clinical observations indicating that patients with diabetes show an increased risk of fractures because of osteopenia or osteoporosis [16,17] and similarly in animal models [10].

Circulating osteocalcin was reduced in patients with severe COVID-19, in accordance with a recent report in which 40 patients were compared with 57 non-COVID-19 controls in a cross-sectional design [18]. We further analyzed the serum osteocalcin association with glycemia in COVID-19 patients, as we and other authors reported in the past this relationship in other type of patients [12,19]. We found that circulating osteocalcin was inversely correlated with glycemia: it was lower in those with stress hyperglycemia, those in need for insulin infusion therapy, and also associated with longer ICU stays. To our knowledge, this is the first time that this association is reported in critically ill patients.

Diabetes mellitus has shown to be a risk factor for a worse prognosis in patients with COVID-19 [20,21] as well as those who develop stress hyperglycemia [4]. Furthermore, the severity of COVID-19 illness increases progressively in relation to glucose abnormalities at admission [22], and this has also been shown to happen in patients without a previous diagnosis of diabetes [23]. Conversely, a recent report has shown no difference in mortality based on the diabetes status, previous control, or complications [1]. Therefore, glycemic control may be important to all COVID-19 patients even if they have no pre-existing diabetes, as most COVID-19 patients are prone to glucose metabolic disorders as a result of stress hyperglycemia and probably the adverse effects of several treatments.

Among the factors that may influence the appearance of stress hyperglycemia in critically ill patients, the release of inflammatory mediators might be one of the physiopathological pathways. Inflammatory cytokines excessively produce glucose by the liver and insulin resistance at the peripheral tissues [8], and the resultant hyperglycemia further exacerbates the inflammatory and oxidative stress response [9]. The modulation of the immune response in patients receiving insulin treatment may partially explain a reduction in morbidity and mortality [24]. In agreement, we have shown that circulating IL-12 concentrations correlated with hyperglycemia at ICU admission and that those patients with stress hyperglycemia showed higher ICU stays.

The amount or glucose delivered through artificial nutrition might be another associated factor with stress hyperglycemia. It has been shown that among the metabolic complications of parenteral nutrition, hyperglycemia is one of them, and this may be especially important in older individuals [7]. Therefore, nutrition-support regimens need to minimize stress hyperglycemia and assist glucose management [25]. Recent recommendations from the European Society of Clinical Nutrition and Metabolism (ESPEN) state

that in critically COVID-19 patients who do not tolerate full-dose EN during the first week in the ICU, initiating PN should be weighed on a case-by-case basis [14]. Conversely, in the last two decades, evidence-based recommendations suggest PN use in patients in whom EN cannot be initiated within 24 h of ICU admission or injury [26–28], as it produces similar outcomes as EN alone. Even a combination of PN and EN in critically ill patients has been recently recommended as a better approach [29].

However, there is still a paucity of data in critically ill COVID-19 patients regarding the recommendation of the type of artificial nutrition and also its composition. Our results showed that the amount of glucose delivered by artificial nutrition did not differ between COVID-19 and non-COVID-19 patients. It is true that the former needed higher rates of insulin infusion, but this could be related to the use of glucocorticoids as part of their treatment. In the present study, we have also shown that the amount of glucose delivered through parenteral nutrition was associated with higher glycemia at ICU independently of COVID-19 diagnosis.

An important limitation of our study is that it was enabled to find differences in ICU stay and serum osteocalcin but not in the composition of artificial nutrition. In this regard, we showed after performing a multivariate analysis that both ICU stay and osteocalcin were independently associated with COVID-19 diagnosis and the presence of stress hyperglycemia. However, the latter were also associated with longer artificial nutrition support and higher energy needs, so the design of our study precludes us from reaching valid, nonbiased associations regarding artificial nutrition. Therefore, future studies are needed to address the role of the type and composition of artificial nutrition in critically ill COVID-19 patients.

#### **5. Conclusions**

Hyperglycemia in critically ill COVID-19 patients was associated with longer ICU stays and with higher amounts of glucose delivered through artificial nutrition in both COVID-19 and non-COVID-19 patients. Low circulating osteocalcin was a good marker for stress hyperglycemia.

**Author Contributions:** J.I.B.-C. and F.A. equally contributed to the conception and design of the research; A.C. contributed to the design of the research; N.B., V.M.-V., M.R., A.d.P., C.V., R.P., A.B.-Q. and R.M.-L. contributed to the acquisition and analysis of the data; J.I.B.-C., A.C. and F.A. contributed to the interpretation of the data; and J.I.B.-C. and F.A. drafted the manuscript. All authors critically revised the manuscript, agree to be fully accountable for ensuring the integrity and accuracy of the work, and read and approved the final manuscript. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by Vegenat Healthcare, grant for partly support study 147/20.

**Institutional Review Board Statement:** The study was conducted according to the guidelines of the Declaration of Helsinki and approved by the Ethics Committee of Hospital Universitario Ramón y Cajal (study code 147/20).

**Informed Consent Statement:** Written or verbal informed consent was obtained from all patients.

**Data Availability Statement:** Restrictions apply to the availability of data generated or analyzed during this study to preserve patient confidentiality or because they were used under license. The corresponding author will on request detail the restrictions and any conditions under which access to some data may be provided.

**Acknowledgments:** CIBERobn is an initiative of Instituto de Salud Carlos III (ISCIII).

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**


## *Article* **Bad Prognosis in Critical Ill Patients with COVID-19 during Short-Term ICU Stay regarding Vitamin D Levels**

**Lourdes Herrera-Quintana 1, Yenifer Gamarra-Morales 1, Héctor Vázquez-Lorente 1, Jorge Molina-López 2,\*, José Castaño-Pérez 3, Juan Francisco Machado-Casas 3, Ramón Coca-Zúñiga 3, José Miguel Pérez-Villares <sup>3</sup> and Elena Planells 1,\***


**Abstract:** Background and aims: Vitamin D inadequacy may be involved in the mechanisms of SARS-CoV-2 infection and in potential risk factors for disease propagation or control of coronavirus disease 2019 (COVID-19). This study assessed a short-term evolution of vitamin D status and its influence upon different clinical parameters in critically ill patients with COVID-19. Methods: A prospective analytical study in which 37 critically ill volunteers between 41 and 71 years of age with COVID-19 were evaluated at baseline and three days of intensive care unit (ICU) stay. 25-OH-D3 and 25-OH-D2 were analyzed by liquid chromatography–tandem mass spectrometry and total 25-OH-D levels were calculated as the sum of both. Results: All patients presented low 25-OH-D levels at baseline, decreasing total 25-OH-D (*p* = 0.011) mainly through 25-OH-D2 (*p* = 0.006) levels during ICU stay. 25-OH-D2 levels decreased a mean of 41.6% ± 89.6% versus 7.0% ± 23.4% for the 25-OH-D3 form during the ICU stay. Patients who did not need invasive mechanical ventilation presented higher levels of 25-OH-D2 at baseline and follow-up. Lower 25-OH-D and 25-OH-D3 levels were associated with higher D-dimer at baseline (*p* = 0.003; *p* = 0.001) and at follow up (*p* = 0.029), higher procalcitonin levels (*p* = 0.002; *p* = 0.018) at follow up, and lower percentage lymphocyte counts (*p* = 0.044; *p* = 0.040) during ICU stay. Conclusions: Deficient vitamin D status in critical patients was established at the admission and further worsened after three days of stay. Lower vitamin D levels were related to key altered clinical and biochemical parameters on patients with SARS-CoV-2 infection. Given the different response of the 25-OH-D3 and 25-OH-D2 forms, it would be useful to monitor them on the evolution of the critically ill patient.

**Keywords:** coronavirus disease 2019; SARS-CoV-2; Vitamin D; critical care; intensive care patient

#### **1. Introduction**

The global public health crisis caused by the SARS-CoV-2 virus has created the need for urgent actions in order to reduce the risk of infection, progression, and the severity of coronavirus disease 2019 (COVID-19) [1], which triggers an acute inflammatory process and uncontrolled oxidative stress [2]. This in turn results in severe acute respiratory distress syndrome (ARDS) characterized by a cytokine storm, mainly in critical cases [3], which may lead to multiple organ damage [4] and further complicate the patient's critical condition previously described during their ICU stay [5,6].

There is currently great concern regarding the clinical management and intensive care of patients with critical stages of the disease and who are at a higher risk of death [7]. The implementation of prompt and appropriate nutritional assessment in COVID-19 must be

**Citation:** Herrera-Quintana, L.; Gamarra-Morales, Y.; Vázquez-Lorente, H.; Molina-López, J.; Castaño-Pérez, J.; Machado-Casas, J.F.; Coca-Zúñiga, R.; Pérez-Villares, J.M.; Planells, E. Bad Prognosis in Critical Ill Patients with COVID-19 during Short-Term ICU Stay regarding Vitamin D Levels. *Nutrients* **2021**, *13*, 1988. https://doi.org/10.3390/nu13061988

Academic Editors: Dimitrios T. Karayiannis and Zafeiria Mastora

Received: 4 May 2021 Accepted: 5 June 2021 Published: 9 June 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

considered, [8,9] because possible modulation of the status of key micronutrients appears to be a relevant factor influencing the development of this disease [10]. No information about nutritional monitoring in critical patients with COVID-19 is available to date [11], and this lack of data precludes the definition of firm micronutrient recommendations in this particular risk population [12].

Certain micronutrients are essential for adequate immunocompetence and antioxidant defense, which are related to inflammatory response, such as vitamin D [13]. 25-Hydroxyvitamin D (25-OH-D) is the metabolite used to assess vitamin D status, due to its long half-life in plasma or serum (one month) [14], and is characterized by synergic action of its two main forms: 25-hydroxyvitamin D2 (25-OH-D2), which is obtained from plant sources, and 25-hydroxyvitamin D3 (25-OH-D3), which comes from animal products and endogenous synthesis in skin through exposure to sunlight [15], both of them can be supplemented with commercial products [16]. Recently, vitamin D has generated particular interest because of its role in reducing the risk of pneumonia and viral upper respiratory tract infections at a physical barrier and cellular natural and adaptive immunity level [17,18]. The underlying mechanisms can be grouped into two main actions: anti-inflammatory and anti-infective [19]. Vitamin D is associated with a decrease in proinflammatory cytokines, reducing the cytokine storm induced by the innate immune system, which is exacerbated in COVID-19 [17,20]. Moreover, it must be noted that serum 25-OH-D is considered as a negative acute phase reactant [21] and low vitamin D status in critical ill patients may be related to a decrease of binding protein concentration [22]. On the other hand, vitamin D is able to reduce viral infection and replication rates by inducing transcription of proteins with antimicrobial functions, enhancing autophagic encapsulation of viral particles, favoring lung epithelial cell barrier integrity, and ultimately regulating both innate and adaptive immunity [17–20,23,24].

Thus, vitamin D inadequacy has emerged as a factor that may be involved in the mechanisms of virus infection and in potential risk factors for disease propagation or control [24,25]. In fact, low vitamin D status in patients with COVID-19 has been reported [26–28], being associated with a poorer prognosis, infection risk, or unregulated inflammation. Thus, due to the lack of evidence on the importance of monitoring vitamin D status in critical patients with COVID-19, the present study was designed to assess the short-term evolution of the status of vitamin D and its influence upon different clinical parameters in critically ill patients with COVID-19 in the province of Granada, Spain.

#### **2. Materials and Methods**

#### *2.1. Study Design and Subjects*

A prospective analytical study was carried out of patients monitored from the first day of admission to the intensive care unit (ICU) (baseline) until day three of stay (follow-up). Of a total of 43 initially recruited patients, 37 participants from the province of Granada (Spain), aged 41–74 years, were included in the period from 1 March to 1 June 2020, after been informed about the study protocol. Six patients died during the study and were excluded. All eligible participants enrolled in the study were critical patients aged 18 years or older and hospitalized for more than 48 h, who agreed to participate in the study or for whom approval of participation was obtained from the family. All patients had a diagnosis of critical active SARS-CoV-2 infection according to the Chinese Clinical Guideline for the classification of COVID-19 [29] (analyzed by real-time reverse transcriptase–PCR (RT-PCR)) testing of nasal and pharyngeal swab samples) and had an ICU stay of at least three days and did not receive vitamin D support. The present study was conducted in accordance with the principles of the Declaration of Helsinki, following the International Conference on Harmonization/Good Clinical Practice standards, and was approved by the Ethics Committee of the University of Granada (Ref. 149/CEIH/2016).

#### *2.2. Data Collection*

Data including age, sex, body mass index (BMI), smoking habits, comorbidities, respiratory and clinical parameters, ICU length of stay, length of hospitalization, and 28-day mortality were retrieved from the hospital electronic database system and recorded for each study participant at ICU admission (baseline) and after three days (follow-up). The Acute Physiology and Chronic Health Evaluation II (APACHE-II) score and Sequential Organ Failure Assessment (SOFA) score were obtained by intensivists at baseline and follow-up.

Patient clinical outcomes were recorded both at admission and during the ICU stay: heart rate (beats per minute); respiratory rate (breaths per minute); mean blood pressure (mmHg); positive end-expiratory pressure (PEEP); fraction of inspired oxygen (FiO2); partial oxygen arterial pressure/fraction of inspired oxygen (PaO2/FiO2); ARDS; invasive mechanical ventilation (IMV).

#### *2.3. Blood Sampling and Biochemical Parameters*

Two measurements were performed (baseline and follow-up). Blood sampling was carried out in the morning under fasting conditions, followed by centrifugation (4 ◦C for 15 min at 3500 rpm) to separate the plasma. The samples were frozen at −80 ◦C until analysis of the different parameters. All samples were measured in one run, in the same assay batch, and blinded quality control samples were included in the same assay batches to determine laboratory error in the measurements.

The recorded biochemical parameters were total proteins, albumin, prealbumin, ferritin, transferrin, glucose, total cholesterol, glutamic oxaloacetic transaminase or aspartate transaminase (GOT or AST), glutamic pyruvic transaminase or alanine transaminase (GPT or ALT), lactate dehydrogenase (LDH), C-reactive protein (CRP), procalcitonin (PCT), hemoglobin, leukocytes, neutrophils, lymphocytes, platelets, D-dimer, fibrinogen, calcium (Ca), phosphorous (P), and magnesium (Mg), using routine hospital analytical assays (ECLIA, Elecsys 2010 and Modular Analytics E170, Roche Diagnostics, Mannheim, Germany).

#### *2.4. Analytical Determination of Vitamin D*

Vitamin D was measured in plasma samples by liquid chromatography–tandem mass spectrometry (LC-MS/MS). Plasma sample treatment involved protein precipitation adding 500 μL of acetonitrile in an Eppendorf flask with 200 μL of plasma and 20 μL of 25-hydroxyvitamin D3 and 25-hydroxyvitamin D2 deuterated solutions as Internal Standard (IS) (Sigma Aldrich, St. Louis, MO, USA) (0.5 μg/mL). The samples were slightly shaken for 1 min on a plate shaker and centrifuged at 10,000 rpm for 15 min at 4 ◦C. The supernatant was collected in another Eppendorf flask and dried with N2. The dry residue was vortexed for 30 s after the addition of 200 μL of ethyl acetate and 100 μL of deionized water and centrifuged at 3000 rpm for 5 min. The supernatant was again collected in another Eppendorf flask, and the previous steps were repeated with the remaining liquid phase, subsequently pooling the second supernatant with the first. The total supernatant was dried with N2.

For samples derivatization, we prepared a solution of 4-phenyl-3H-1,2,4-triazole-3,5(4H)-dione (PTAD) (Sigma Aldrich, St. Louis, MO, USA) in acetonitrile (0.5 mg/mL), using 50 μL of this solution in standards and in each sample, with vortexing. All samples were placed on the plate shaker for 1 h at room temperature and covered with aluminum foil. Lastly, the samples were transferred to vials, diluted with 50 μL of deionized water and stored in freezer at −20 ◦C covered with aluminum foil until injection into the chromatograph. For the calibration line, increasing concentrations of 1, 2, 5, 10, 25, 50, and 100 ppb of the standards 25-OH-D3 and 25-OH-D2 (Sigma Aldrich, St. Louis, MO, USA) with 20 μL of IS were used and dried with N2 and derivatized at the same time as the samples. For sample measurements, use was made of a Waters Acquity UHPLC I-Class System chromatograph (Waters, London, UK), with the Acquity UHPLC BEH C18 column 2.1, 50 mm, 1.7 m at room temperature. The mobile phase of channel A was water

with 50 mM of ammonium formate, while that of channel B was methanol. The injection volume of the sample was 10 μL and the flow rate was 0.4 mL/min. The detector was a Waters XEVO-TQ-XS Triple Quadrupole Low Resolution Spectrometer. Total 25-OH-D was calculated as the sum of the 25-OH-D3 and 25-OH-D2 forms. According to the Endocrine Society Practice Guidelines on Vitamin D, the threshold for biochemical 25-OH-D sufficiency values was considered to be >30 ng/mL, with deficiency being defined as 20–29 ng/mL and insufficiency as <20 ng/mL [28].

#### *2.5. Statistical Analysis*

Qualitative variables were presented as frequencies and percentages. Quantitative variables with normal distribution were expressed as the arithmetic mean and standard deviation (SD), and variables with non-normal distribution were expressed as the median and the interquartile range. Normal data distribution for continuous variables was tested using the Shapiro–Wilk test. The Wilcoxon signed-rank test for nonparametric samples was used for the comparative analyses at baseline and follow-up. The unpaired Student *t*-test for parametric samples was used for the comparative analysis based on clinical outcomes. The effect size (ES) was estimated and interpreted as follows: small = 0.01, moderate = 0.06, and large = 0.14 [30]. Correlation analyses and partial correlation coefficients were performed using the Spearman test. Statistical significance was considered for *p* < 0.05. The SPSS version 22.0 statistical package (IBM SPSS, Armonk, NY, USA) was used throughout.

#### **3. Results**

The demographic and clinical characteristics of the 37 patients enrolled in the study are shown in Table 1. The median age of the study population was 60 years, and the gender distribution of the sample was 26 males and 11 females. With regard to the anthropometric parameters, over a third of the patients were overweight, and more than half of them were obese. Most of the patients had one or more underlying diseases. With regard to the severity parameters, the mean APACHE-II and SOFA scores were 12.3 and 6.54, respectively, upon admission. A total of 26 of 37 patients (70.2%) had at least 1 infection, and 5 of them had 3 or more infections during their admission in ICU. The most frequent infection was bacteremia in 25 out of 48 total infections (52%), followed by respiratory infections 15 out of 48 total infections (31%), and finally urinary tract infections associated with urethral catheterization in 7 out of 48 total infections (14.5%). The most frequent germs causing these infections were Gram-positive (39%), followed by Gram-negative bacteria (33%) and fungi (20%). The mean length of hospitalization was 39.5 days. More than two-thirds of the patients presented ARDS on admission, requiring invasive mechanical ventilation with a mean duration of over 20 days. The mortality rate after 28 days of ICU stay was over two-thirds of the total study population.

The clinical and biochemical parameters of the study population at baseline and follow-up are shown in Table 2. Respiratory parameters were altered with significant changes in FiO2 and PEEP after three days. Total proteins (*p* = 0.012), albumin (*p* = 0.035), prealbumin (*p* = 0.017), LDH (*p* = 0.002), CRP (*p* = 0.001), hemoglobin (*p* = 0.001), and fibrinogen (*p* = 0.001) were outside the reference values and decreased significantly after three days of ICU stay. Parameters such as ferritin, transaminases, or D-dimer were also outside the reference values although no changes were observed in their evolution during the ICU stay. The results showed the 25-OH-D, 25-OH-D3, and 25-OH-D2 levels to be lower at follow-up versus baseline—with statistical significance being reached for 25-OH-D (*p* = 0.011) and 25-OH-D2 (*p* = 0.006).

Figure 1 shows the distribution of patient vitamin D status upon ICU admission and after three days of stay. In no case were the 25-OH-D levels > 25 ng/mL. Only 16.7% (6/37) of the patients had 25-OH-D > 20 ng/mL at baseline, versus 3.2% (1/37) at follow-up. Furthermore, 22.2% (8/37) presented 25-OH-D < 10 ng/mL—this percentage reaching 25.8% (10/37) after three days of stay.


**Table 1.** Demographic and clinical characteristics of the patients.

*N* = 37. Data expressed as mean ± standard deviation. Abbreviations: SD = standard deviation; Min–Max = minimum–maximum; CI = confidence interval; M/F = male/female; BMI = body mass index; COPD = chronic obstructive pulmonary disease; APACHE-II = Acute Physiology and Chronic Health Evaluation II; SOFA = Sequential Organ Failure Assessment; PaO2/FiO2 = partial oxygen arterial pressure/fraction of inspired oxygen; ARDS = acute respiratory distress syndrome; IMV = invasive mechanical ventilation; ICU = intensive care unit.

**Figure 1.** Vitamin D status in the study population at ICU admission (baseline) and after three days (follow-up).


**Table 2.** Clinical and biochemical parameters of the critical patients with COVID-19 at baseline and follow-up.

*N* = 37. Data expressed as median (interquartile range, IQR). Abbreviations: SD = standard deviation; ES = effect size; bpm = beats per minute; brpm = breaths per minute; PEEP = positive end-expiratory pressure; PaO2/FiO2 = partial oxygen arterial pressure/fraction of inspired oxygen; GOT or AST = glutamic oxaloacetic transaminase or aspartate transaminase; GPT or ALT = glutamic pyruvic transaminase or alanine transaminase; LDH = lactate dehydrogenase; CRP = C-reactive protein; PCT = procalcitonin; Ca = calcium, P = phosphorous, Mg = magnesium, \*10<sup>3</sup> = multiplied by 1000. The sixth column reports statistical significance after the Wilcoxon signed-rank test; evolution is shown after 3 days. ES effect size calculations were also made to determine the effect of ICU stay (ES: small ≤ 0.01, moderate = 0.06, and large ≤ 0.14) [28]. Statistical significance = *p* < 0.05.

> Figure 2 corresponds to the comparative analysis of clinical parameters in relation to 25-OH-D, 25-OH-D3, and 25-OH-D2 levels upon ICU admission and after three days of stay. Based on the sepsis (Figure 2A) and infectious processes (Figure 2B), there was a trend toward statistical significance in the follow-up observing lower levels of 25-OH-D and 25-OH-D3 for septic patients and lower levels of 25-OH-D2 for infected patients. Patients who did not need invasive mechanical ventilation presented higher levels of 25-OH-D2 at baseline and 25-OH-D and 25-OH-D2 at follow-up (Figure 2C).

> Table 3 shows the Spearman bivariate correlations between the 25-OH-D, 25-OH-D3, and 25-OH-D2 levels at baseline and follow-up and the clinical and biochemical parameters analyzed in our study. At baseline, the 25-OH-D levels were correlated to albumin (*p* = 0.021), hemoglobin (*p* = 0.028), D-dimer (*p* = 0.003), and fibrinogen (*p* = 0.020) with albumin also being correlated to 25-OH-D2 (*p* = 0.037); and D-dimer (*p* = 0.001) and fibrinogen to 25-OH-D3 (*p* = 0.029). Respiratory rate was negatively correlated to 25-OH-D2 (*p* = 0.025). At follow-up, 25-OH-D and 25-OH-D3 were significantly correlated

to PCT (*p* = 0.002; *p* = 0.018) and lymphocytes (*p* = 0.044; *p* = 0.040). In the case of Ddimer and Ca, an inverse correlation to 25-OH-D3 was observed (*p* = 0.029 and *p* = 0.006, respectively). Finally, the 25-OH-D2 levels showed a significant correlation to both the fibrinogen (*p* = 0.003) and Ca levels (*p* = 0.030). We did not find correlation between mortality rate after 28 days of ICU stay and 25-OH-D3, 25-OH-D2, and 25-OH-D levels.

**Table 3.** Matrix for correlation coefficients (rho) showing the simple linear relationship between clinical and biochemical parameters with 25-OH-D, 25-OH-D2, and 25-OH-D3 levels.


Matrix correlations are presented as correlation coefficients (rho). Abbreviations: BMI = body mass index; APACHE-II = Acute Physiology and Chronic Health Evaluation II; SOFA = Sequential Organ Failure Assessment; FiO2 = fraction of inspired oxygen; PCT = procalcitonin; Ca = calcium. Statistical significance <sup>a</sup> = *p* < 0.05; <sup>b</sup> = *p* < 0.01.

#### **4. Discussion**

The main finding of the present study was the low 25-OH-D levels in the patients upon admission (baseline), followed by a significant decrease after three days of ICU stay. The entire population was below the sufficiency reference values for 25-OH-D, and most of them presented insufficient 25-OH-D status. We also analyzed the 25-OH-D3 and 25-OH-D2 levels, both of which were seen to decrease after three days (though statistical significance was only reached in the case of 25-OH-D2), thus influencing upon 25-OH-D decreased levels and presenting a worsening during their stay at 3 days. Moreover, vitamin D was associated with clinical parameters such as the need for mechanical ventilation or respiratory frequency and with biochemical parameters also associated with the severity of the critically ill patient such as albumin, hemoglobin, D-dimer, fibrinogen, PCT, and lymphocytes.

Previous evidence points to poorer COVID-19 outcomes associated with factors such as the male gender, older age, BMI > 35 kg/m2, and the presence of certain comorbidities [31]. The demographic and clinical characteristics of our patients (Table 1) are consistent with this evidence. In effect, the population was fundamentally elderly, two-thirds were males, and there was a high prevalence of comorbidities and obesity. It should be noted that metabolically ill patients with obesity may have a high risk of suffering inflammatory processes [32], which could contribute to a greater probability of poorer outcomes.

Many of the clinical and biochemical parameters in our patients were altered (Table 2). In the case of PEEP and FiO2, levels were above reference values and, they even decreased significantly in three days and remained altered in the most cases. It should be noted that ferritin, CRP, D-dimer, and fibrinogen were well above the reference values. Parameters related to inflammation (such as CRP) or coagulation (such as D-dimer) have been correlated to a poor prognosis and have been described as possible predictive biomarkers of COVID-19 [33].

Our results with reference to vitamin D status showed all patients to have insufficient levels (<30 ng/mL) both upon admission and during the study period—with the vitamin D status being seen to worsen in only three days. The great majority of patients presented vitamin D deficiency (<20 ng/mL), while extreme deficient values of <10 ng/mL were recorded in a quarter of the study sample. It is known that, compared to the general population, the prevalence of hypovitaminosis D is greater in the critically ill and may constitute a risk factor for adverse outcomes [34]. Moreover, these levels could be influenced by seasonality. In an observational study carried out in critical ill patients from Austria, significant differences were noted in the prevalence of vitamin D deficiency and in the mean 25-OH-D values between the winter and summer months [35]. Our study covered the period from March to June; we, therefore, could not demonstrate the influence of seasonality in our patients.

On the other hand, there is concern about the high prevalence of hypovitaminosis D in the general population—being regarded as a global health issue with important consequences [36]. Moreover, SARS-CoV-2 positivity has been strongly and inversely correlated to circulating 25-OH-D levels—a relationship that persists across latitudes, races/ethnicities, both genders, and age ranges [37], thereby evidencing that the COVID-19 fatality rates parallel the vitamin D deficiency rates [38]. These negative correlations between vitamin D deficiency and the number of COVID-19 cases and mortality have also been reported in another 20 European countries [39]. It could be expected that countries such as Spain have a better vitamin D status and therefore less severe consequences than other countries in northern Europe. However, our results reinforced the evidence of a possible widespread vitamin D deficiency in the Spanish population. Indeed, vitamin D deficiency in Italy and Spain (the countries presenting the highest age-specific case fatality ratio) [40] is more severe than elsewhere in Europe [41], particularly in the aging population [39].

On the other hand, Maghbooli et al. found 25-OH-D levels >30 ng/mL to reduce the risk of adverse clinical outcomes in patients with COVID-19 [42]. None of our critical

patients with COVID-19 presented 25-OH-D levels >25 ng/mL, which is consistent with the findings of Maghbooli. Low 25-OH-D levels have been reported by many authors in hospitalized patients with COVID-19 [43], with such deficiency being associated with a greater mortality risk [26]. In fact, Vassiliou et al. found that low 25-OH-D levels in patients with COVID-19 at ICU admission could predispose to an increased 28-day mortality risk [44]. In the present study, although we did not observe an association between 25-OH-D levels with 28-day mortality risk, we found an association with clinical outcomes, reporting that higher 25-OH-D and 25-OH-D2 levels were associated with those patients who did not require invasive mechanical ventilation. Moreover, lower 25-OH-D values, as a result of decreased 25-OH-D3 and 25-OH-D2 levels were observed with the presence of infection (bacterial/fungal) and sepsis (Figure 2). It should be noted that 25-OH-D2 levels were in agreement with previous studies in the Spanish population [45], being also similar to those found in other studies performed in Belgian and Chinese populations [46,47]. Likewise, observational studies have also shown that higher 25-OH-D levels would be associated with better clinical outcomes in respiratory diseases [48]. Nevertheless, reference values are needed to have more contrastable evidence on 25-OH-D2 levels.

In relation to infection, a high percentage of our patients (70.3%) showed bacteriological or fungal infection, which would support the idea that the daily risk rate of infection in COVID-critically-ill patients is increased during ICU stay [49]. Recent studies [50], consistent with the high prevalence of hypovitaminosis D observed in our study, suggest a possible role of low vitamin D in the increased risk of SARS-CoV-2 infection and subsequent hospitalization. Likewise, 25-OH-D levels were inversely associated with coagulation and sepsis, in addition to major comorbidities. Both, 25-OH-D3 and 25-OH-D2 forms tended to respond differently in patients with bacteriological or fungal infection and in patients presenting sepsis. On the other hand, it was observed that 25-OH-D2 levels decreased a mean of 41.6% ± 89.6% versus 7.0% ± 23.4% for the 25-OH-D3 form during the ICU stay, which would suggest that the lack of vitamin D support during ICU stay could have allowed this more pronounced decrease in case of the 25-OH-D2 form, since it would depend on the intake through diet (before the admission) or its supplementation (during ICU stay). A greater decrease in 25-OH-D2 relative to 25-OH-D3 could be also related to the lower affinity of 25-OH-D2 for vitamin D–binding protein, leading to a shorter half-life and a higher rate of clearance from the circulation [51], and in some cases, it even caused a decline, thereby, precipitating in vitamin D deficiency [52]. This, together with the fact that there is literature that already reports a possible different role of 25-OH-D3 and 25-OH-D2 forms [51,53], although not in critical patients, could evidence the different correlations obtained with bacterial/fungal infection or sepsis and invasive mechanical ventilation previously described. Therefore, it would suggest that a lower vitamin D status at admission and worsening during three days of ICU stay may be a modifiable risk factor and an early predictive marker of adverse outcomes in hospitalized patients with COVID-19.

On associating the concentrations of both 25-OH-D and 25-OH-D3 with other biochemical severity parameters, significantly lower vitamin D levels were correlated to higher D-dimer and PCT levels and a lower percentage of lymphocytes (Table 3). A recent metaanalysis has demonstrated that patients with severe COVID-19 tend to present increased leukocyte and neutrophil counts, neutrophil–lymphocyte ratio, PCT and CRP levels, and a decreased number of total lymphocytes, among parameters, compared to nonsevere individuals [54]. Furthermore, blood hypercoagulability is common among hospitalized patients with COVID-19. Elevated D-dimer levels are consistently reported in this scenario, and a gradual increase of this parameter in the course of the disease is particularly associated with patient worsening. Similarly, lower fibrinogen levels were found in nonsurvivor patients with COVID-19 [55]. In this line, the relationship between fibrinogen and 25-OH-D levels is often reported in the literature in noncritical patients [56,57]. In our study, 25-OH-D and 25-OH-D3 levels were inversely correlated with D-dimer levels at baseline. Furthermore, 25-OH-D3 levels have been correlated with fibrinogen levels at baseline. Our results may reflect a better vitamin D status (mainly due to vitamin D3) in

patients with a more appropriate hematological profile. We observed a positive correlation at baseline between albumin (which was below the reference values) and the levels of both 25-OH-D and 25-OH-D2. Recently, low albumin levels have been regarded as more of a disease severity marker than as a marker of malnutrition, when such low levels are detected upon admission to hospital [58].

The present study has limitations and strengths. As limitations, the present study enrolled fewer patients than desired due to the difficulty in obtaining the sample and the patient's own clinical situation and severity both at admission and during the ICU stay. Therefore, the data should be treated with caution in order to generalize the findings of the study. Thus, on a comparative level, the effect size was shown for a better understanding. We had no reliable data on exposure to sunlight, dietary factors, or vitamin D supplementation—all of which affect vitamin D status. The overall negative results may be related to the heterogeneity of the subjects and their underlying disease conditions or severity, which may all influence the plasma 25-OH-D levels. Our findings cannot be generalized to other populations or ethnic groups, especially considering the wide range of COVID-19 prevalence. Replicating this study in a larger, prospective, and heterogeneous population and taking into account a control group, would allow for other stratified analyses based on demographic and biochemical characteristics, taking seasonality into account, and could further corroborate our findings. Additional research is therefore needed to validate our findings. As strengths, the present study used LC-MS/MS, which is the gold standard for assessing the levels of 25-OH-D [59], affording greater sensitivity, flexibility, and specificity than the enzymoimmunoassay techniques commonly used in clinical practice, which tend to overestimate the 25-OH-D values in cases of deficiency [60,61].

Recent studies have reported encouraging results after vitamin D intervention [62,63]. However, further evidence is needed to confirm that improving vitamin D status is of benefit in reducing disease severity and mortality and the probability of developing a critical clinical condition. It is essential to ensure close patient monitoring before establishing intervention guidelines [64]. This study is one of the few that have been conducted in this context, assessing the short-term evolution of the 25-OH-D levels (through 25-OH-D2 and 25-OH-D3 levels) and its impact in critical patients with COVID-19.

#### **5. Conclusions**

Our data reflect a high prevalence of hypovitaminosis D in all the critical patients at ICU admission, which increased after only three days of ICU stay. On the other hand, the associations observed between 25-OH-D levels, through 25-OH-D3 and 25-OH-D2 values and key clinical outcomes and biochemical altered parameters, suggests that it might be helpful to assess vitamin D status in patients with SARS-CoV-2 infection. Given the different response of the 25-OH-D3 and 25-OH-D2 forms, it would be useful to analyze them to elucidate the role of each form on the evolution of the critically ill patient. Further investigations are needed to define underlying mechanisms in vitamin D deficiency and useful strategies based on vitamin D interventions aimed at preserving vitamin D status and enhancing the clinical and biochemical profile of critical patients with COVID-19.

**Author Contributions:** Conceptualization, E.P., J.C.-P. and L.H.-Q.; methodology, J.M.-L. and J.M.P.- V.; software, R.C.-Z., J.M.-L. and E.P.; validation, Y.G.-M. and J.F.M.-C.; formal analysis, L.H.-Q. and H.V.-L.; investigation, H.V.-L. and L.H.-Q.; resources, E.P. and H.V.-L.; data curation, J.M.-L.; writing original draft preparation, L.H.-Q., Y.G.-M. and E.P.; writing—review and editing, H.V.-L. and J.M.-L.; visualization, L.H.-Q., Y.G.-M. and H.V.-L.; supervision, J.M.-L. and E.P.; project administration, E.P. and J.M.-L.; funding acquisition, E.P. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received external funding by FIS Projects from Carlos III Health Institute [REF. PI10/1993]. L.H.-Q. and H.V.-L. are under a FPU fellowship from the Spanish Ministry of Education. **Institutional Review Board Statement:** The study was approved by the Ethical Committee of the hospital, and informed consent was obtained from the patients or their family, who agreed to participate in the study.

**Informed Consent Statement:** Informed consent was obtained from all subjects involved in the study.

**Data Availability Statement:** The datasets generated and analyzed during the current study are not publicly available because the database is very extensive and includes data from other studies complementary to this but are available from the corresponding author on reasonable request.

**Acknowledgments:** Thanks are due to all the patients who participated in our study and the personnel from the Hospital Virgen de las Nieves, and FIBAO foundation (Fundación Pública Andaluza para la Investigación Biosanitaria de Andalucía Oriental Alejandro Otero) from Granada. We also acknowledge the expertise of the translator who provided English editing.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results.

#### **References**

