Metabolic Syndrome and Acute Respiratory Distress Syndrome in Hospitalized Patients With COVID-19

Key Points Question What is the risk of acute respiratory distress syndrome (ARDS) and death in patients with COVID-19 with metabolic syndrome? Findings In this cohort study including 46 441 patients hospitalized for COVID-19, metabolic syndrome was associated with significantly increased odds of ARDS and death. With each metabolic syndrome criterion added from 1 to 4 criteria, the risk of ARDS significantly increased in an additive fashion. Meaning These findings suggest that metabolic syndrome and its associated comorbidities were critical risk factors associated with COVID-19–related ARDS and death.

Inclusion criteria included hospitalized, adult (age Ն18 years) patients with COVID-19 confirmed by reverse transcription-polymerase chain reaction testing. Exclusion criteria excluded any patient without prior research authorization, any non-COVID-19-related admissions, any patient without a completed discharge status, any patient without documented age, and any patient entered from a hospital site with a case volume of fewer than 10 participants. Patients were retrospectively divided into 2 groups: those with metabolic syndrome and control patients, and outcomes of hospital discharge or death were evaluated. Patients with metabolic syndrome were identified from admission data using modified World Health Organization criteria 17 as previously described 7 and as having at least 3 of the following: prediabetes (hemoglobin A 1c Ն5.7% [to convert to proportion of total hemoglobin, multiply by 0.01]), history of diabetes, or diabetes medication use; obesity (body mass index [BMI; calculated as weight in kilograms divided by height in meters squared] Ն30); history of hypertension or antihypertensive medication use; and serum triglyceride (TG) level of 150 mg/dL or greater (to convert to millimoles per liter, multiply by 0.0113), serum high-density lipoprotein (HDL) level less than 50 mg/dL for women and less than 40 mg/dL for men (to convert to millimoles per liter, multiply by 0.0259), or cholesterol-lowering medication use with history of dyslipidemia. Since TG and HDL laboratory measurements are frequently unavailable during hospital admission, these 2 criteria were combined into a single dyslipidemia criterion 18 to improve applicability in real-world practice and facilitate reproducibility by other investigators. 7,19 The control cohort included any patient in the population not meeting the definition for metabolic syndrome at admission in the same study period as patients with metabolic syndrome.
Data elements for this study included demographics, medications, admission BMI, comorbid conditions to estimate the Elixhauser Comorbidity Index, 20 worst level of oxygen support for each patient's hospitalization, last recorded hemoglobin A 1c level, serum TG level, and serum HDL level. Serum TG and HDL levels were obtained from laboratory values most recent to the index hospitalization within 1 year prior to admission, when available, as these levels may change with acute inflammatory episodes. 21 If measurements prior to hospitalization were not recorded, admission serum TG or HDL values were used when available; however, TG levels measured during or after administration of propofol were omitted, since propofol may lead to hypertriglyceridemia. 22

Outcomes
The primary outcome was hospital mortality. Secondary outcomes included a diagnosis of ARDS determined by site investigators using the Berlin Criteria, 23 intensive care unit (ICU) admission, invasive mechanical ventilation, need for noninvasive ventilation, supplemental oxygen use, hospital and ICU length of stay (LOS), and the worst level of oxygen support needed during hospitalization, defined by the lowest score on a 5-category ordinal scale adapted from the World Health Organization scale, 24 with 5 indicating death; 4, receiving invasive mechanical ventilation or extracorporeal membrane oxygenation (ECMO); 3, receiving noninvasive ventilation or high-flow oxygen devices; 2, requiring supplemental oxygen low flow (<15 L/min); and 1, not requiring supplemental oxygen. The primary outcome and secondary outcomes were compared between metabolic syndrome and control groups. Given the high prevalence of metabolic syndrome in the US, we compared rates of metabolic syndrome and each individual criterion (obesity, prediabetes or diabetes, hypertension, and dyslipidemia) in US and non-US hospital sites. Lastly, the primary outcome was then compared between patients hospitalized at US or non-US sites and stratified by metabolic syndrome.
Additionally, subgroup comparisons were performed to study the association of COVID-19 severity among patients with each metabolic syndrome criterion. First, to determine the risk increase associated with each cumulative metabolic syndrome criterion added, we defined subgroups according to the number of metabolic syndrome criteria present (ie, 0, 1, 2, 3, or 4 criteria) and compared mortality, ARDS, and LOS among them. The second subgroup analysis compared patients with each individual metabolic syndrome-associated criterion alone (eg, patients with obesity but without hypertension, dyslipidemia, or prediabetes or diabetes) with patients with no metabolic syndrome criteria (0 of 4 criteria) to explore which conditions, if any, may be associated with the greatest risk.

Statistical Analysis
The main objective was to determine whether metabolic syndrome was associated with greater mortality compared with patients without metabolic syndrome. We used t test to compare means of numerical variables in different groups, and Mood median test was used for comparison of medians, with an α of 0.05 to determine statistical significance. Pearson χ 2 test was used to compare the difference of distribution in groups for categorical variables. For comparison of the primary outcome and key secondary outcomes among metabolic syndrome and control groups, a clustered multivariable logistic regression model (or clustered multivariable linear regression model when appropriate) was constructed with the cluster of study site that patients came from and also including the following covariates selected using background knowledge of the factors connecting exposure to outcome: age, self-identified sex or gender (abstracted from health records and reported according to each site's policy), self-identified race and ethnicity, hospital case volume (as a measure of hospital site variation), and Elixhauser Comorbidity Index. 20 Quantile regression was used to test the difference of median of LOS, and a cumulative logistic model was used for the ordinal scale.
After inclusion and exclusion criteria, missing BMI data were substantial. Given that BMI is criterion standard to determine obesity and is one of the criteria to determine metabolic syndrome, incomplete data could result in possible misclassification bias. As such, additional steps were taken to account for this missingness. Since these data were assumed to be missing completely at random, a multiple imputation technique was used to create 20 imputed data sets to account for missing BMI values. Imputed data sets were analyzed separately and were reported as a pooled outcome across imputations. Multiple imputation and statistical analyses were performed using SAS Enterprise Guide

Study Population
Among 46 441 patients with COVID-19 who were hospitalized at 181 hospitals during the study period, 135 hospitals (74.6%) and 29 040 patients (62.5%) met eligibility criteria and were included in analyses (Figure 1). The mean (SD) age was 61.

Outcomes in Metabolic Syndrome vs Control Groups
Patients with metabolic syndrome, compared with control patients, were more likely to be women Furthermore, patients with metabolic syndrome consistently required higher levels of oxygen support, as measured by the worst 5-category ordinal scale score during their hospitalization, including supplemental oxygen, noninvasive ventilation or high-flow oxygen devices, and invasive mechanical ventilation or ECMO compared with control patients (Figure 2)

Additive Association of Metabolic Syndrome Criteria With Severe Outcomes
The outcomes of patients with 0, 1, 2, 3, and 4 of 4 metabolic syndrome criteria are portrayed in

Metabolic Syndrome Individual Conditions
In additional subgroup analyses, patients with only 1 individual metabolic syndrome-associated comorbid condition were compared with a separate cohort of patients who did not have any of the metabolic syndrome-associated conditions (eFigure 1 in Supplement 1). Compared with patients without these metabolic syndrome risk factors, rates of ARDS were significantly increased for patients with prediabetes or diabetes, hypertension, and dyslipidemia, but the opposite association was seen with obesity. Hospital mortality was also significantly increased for patients with prediabetes or diabetes and hypertension, but not dyslipidemia, compared with patients without metabolic syndrome risk factors. Again, obesity was associated with a significant improvement in mortality (eFigure 1 in Supplement 1).

Global and Hospital Case Volume Comparisons
Metabolic syndrome was significantly more common among patients with COVID-19 admitted to US hospitals compared with those admitted to non-US hospitals (4785 of 25 520 patients [18.8%] in US hospitals vs 284 of 3520 patients [8.1%] in non-US hospitals; P < .001). Furthermore, US hospitals, compared with non-US hospitals, admitted significantly higher proportions of patients with obesity

Discussion
In this international, multicenter, prospective cohort study including 29 040 adults hospitalized with COVID-19, the presence of metabolic syndrome was associated with significantly increased odds of death and ARDS irrespective of age, sex, race, ethnicity, hospital case volume, and comorbid conditions. The 4% absolute mortality rate difference between patients with metabolic syndrome and control patients was prominent, with similarly high rates of ARDS and invasive mechanical ventilation in patients with metabolic syndrome compared with control patients. Additionally, the proportion of patients who developed ARDS was significantly increased in a stepwise and additive fashion with each metabolic syndrome criterion. Greater resource utilization with more frequent ICU admission and higher disease severity among patients with metabolic syndrome patients were also associated with prolonged hospital and ICU LOS compared with control patients.
Several studies have investigated the association between COVID-19 and individual metabolic diseases, such as obesity, 5 diabetes, 25 hypertension, 26 and dyslipidemia 27 ; yet how these different conditions are mechanistically associated with COVID-19 risk and illness severity remain to be fully elucidated. 28 In smaller series, metabolic syndrome and its chronic low-grade inflammatory state 29 have been postulated as instrumental in predisposing patients to ARDS and subsequently mortality. 7,19 Our findings support this hypothesis, as patients with metabolic syndrome were not only at higher risk of ARDS and severe outcomes overall, but each additional metabolic syndrome criteria added was associated with greater risk of ARDS in an additive fashion among patients with 1, 2, 3, or 4 metabolic syndrome criteria.
Furthermore, although ARDS and death were increased among patients with metabolic syndrome, we also found that at every level of respiratory support, patients with metabolic syndrome experienced worse outcomes than control patients, with 43% increased risk of reaching a could be the high prevalence of metabolic syndrome in this population. Although our data show significantly increased rates of metabolic syndrome for patients in US hospitals compared with non-US hospitals, there were no significant differences in adjusted analyses comparing mortality between US and non-US sites. One explanation could be the association between hospital case volume and mortality, which demonstrated a significant decline when the higher-volume hospitals were compared with lower-volume hospitals, but these findings require further exploration.
Nonetheless, these results highlight a potential reason why the COVID-19 pandemic may have disproportionately affected the US in terms of critical illness burden.

Limitations
There are several important limitations that require consideration. Although VIRUS is one of the largest observational registries collecting data on patients with COVID-19, the conclusions drawn cannot imply causation. This study used multivariable regression analyses to address potential confounding, but unmeasured confounders, such as health care access, health system factors, or socioeconomic factors, may have contributed. Also, selection and information bias were considered important limitations owing to the unavailability of data from the electronic medical record.
However, to mitigate this potential bias, several analyses were performed to support the statistical approach in addressing incomplete data. Additionally, complete patient phenotyping was not available, so concurrent bacterial infections and other concomitant causes of mortality could not be explored. Furthermore, we were not able to assess consistency of adherence with lung protective ventilation strategies, an established high-quality intervention associated with improved survival.
This has been demonstrated to be challenging in patients who are obese and may have acted as an unmeasured confounder.

Conclusions
This cohort study found that metabolic syndrome, diagnosed by the clustering of obesity, prediabetes or diabetes, hypertension, and dyslipidemia, was associated with significantly increased mortality and ARDS in a global population of hospitalized patients with COVID-19. This increased risk was cumulative, with the proportion of ARDS increasing with each added metabolic syndrome criteria.