Inflammation, Hyperglycemia, and Adverse Outcomes in Individuals With Diabetes Mellitus Hospitalized for COVID-19

Abstract

Objective: Diabetes mellitus (DM) is a major risk factor for severe coronavirus disease 2019 (COVID-19) for reasons that are unclear.

Research design and methods: We leveraged the International Study of Inflammation in COVID-19 (ISIC), a multicenter observational study of 2,044 patients hospitalized with COVID-19, to characterize the impact of DM on in-hospital outcomes and assess the contribution of inflammation and hyperglycemia to the risk attributed to DM. We measured biomarkers of inflammation collected at hospital admission and collected glucose levels and insulin data throughout hospitalization. The primary outcome was the composite of in-hospital death, need for mechanical ventilation, and need for renal replacement therapy.

Results: Among participants (mean age 60 years, 58.2% males), those with DM (n = 686, 33.5%) had a significantly higher cumulative incidence of the primary outcome (37.8% vs. 28.6%) and higher levels of inflammatory biomarkers than those without DM. Among biomarkers, DM was only associated with higher soluble urokinase plasminogen activator receptor (suPAR) levels in multivariable analysis. Adjusting for suPAR levels abrogated the association between DM and the primary outcome (adjusted odds ratio 1.23 [95% CI 0.78, 1.37]). In mediation analysis, we estimated the proportion of the effect of DM on the primary outcome mediated by suPAR at 84.2%. Hyperglycemia and higher insulin doses were independent predictors of the primary outcome, with effect sizes unaffected by adjusting for suPAR levels.

Conclusions: Our findings suggest that the association between DM and outcomes in COVID-19 is largely mediated by hyperinflammation as assessed by suPAR levels, while the impact of hyperglycemia is independent of inflammation.

Trial registration: ClinicalTrials.gov NCT04818866.

Figure 1

Risk of in-hospital outcomes in individuals with COVID-19 and with and without DM. The bar graphs depict the ORs comparing individuals with DM with individuals without DM (reference) and 95% CIs for the composite outcome (A) and the individual outcomes of in-hospital death (B), need for mechanical ventilation (C), and need for dialysis or continuous renal replacement therapy (D). Four different models were used: model 0 (unadjusted); model 1 (demographics) adjusted for age, sex, and race; model 2 (clinical characteristics) additionally adjusted for BMI and history of hypertension, coronary artery disease, and congestive heart failure (clinical characteristics); and model 3 (inflammation) further adjusted for suPAR level. *P < 0.05.

Figure 2

Variable importance plot to predict composite outcome in individuals with DM and COVID-19. The variable importance plot is based on the Gini index using a random forest approach. Shown are data from model 3 (adjusted for age, sex, race, BMI, admission suPAR, and history of preexisting coronary artery disease, hypertension, and heart failure) for predicting the composite outcome of in-hospital death, need for mechanical ventilation, and need for renal replacement therapy.

Figure 3

Associations among glucose, insulin, and combined outcome in individuals with DM in the M²C² subset. The forest plot depicts the ORs and 95% CIs for the association among glucose, insulin, and the composite outcome of in-hospital death, need for mechanical ventilation, and need for renal replacement therapy stratified by DM among individuals with COVID-19 in the M²C² subset (n = 1,608). All ORs are compared using the following reference categories for each variable: 0–1.25 for glucose coefficient of variance, 100% for glucose in range, 0% for high glucose, and 0 units/kg/day for insulin. The glucose coefficient of variation is calculated as the SD divided by the mean of all glucose measurements taken during hospitalization and then multiplied by 10. Percent in glucose range and high glucose are expressed as the percentage of all glucose measurements within each category during hospitalization. Insulin is calculated as the total amount of insulin (units) received during hospitalization divided by the patient’s weight (kg) multiplied by the number of days in the hospital. Models were adjusted for age, sex, race, BMI, and history of hypertension, coronary artery disease, and congestive heart failure.