Glycaemic variability is associated with all-cause mortality in COVID-19 patients with ARDS, a retrospective subcohort study

Abstract

There is high mortality among intensive care unit (ICU) patients with acute respiratory distress syndrome (ARDS) caused by coronavirus disease (COVID-19). Important factors for COVID-19 mortality are diabetes status and elevated fasting plasma glucose (FPG). However, the effect of glycaemic variability on survival has not been explored in patients with COVID-19 and ARDS. This single-centre cohort study compared several metrics of glycaemic variability for goodness-of-fit in patients requiring mechanical ventilation due to COVID-19 ARDS in the ICU at University Hospital Aachen, Germany. 106 patients had moderate to severe ARDS (P/F ratio median [IQR]: 112 [87-148] mmHg). Continuous HRs showed a proportional increase in mortality risk with daily glycaemic variability (DGV). Multivariable unadjusted and adjusted Cox-models showed a statistically significant difference in mortality for DGV (HR: 1.02, (P) < 0.001, LR(P) < 0.001; HR: 1.016, (P) = 0.001, LR(P) < 0.001, respectively). Kaplan-Meier estimators yielded a shorter median survival (25 vs. 87 days) and a higher likelihood of death (75% vs. 31%) in patients with DGV ≥ 25.5 mg/dl (P < 0.0001). High glycaemic variability during ICU admission is associated with significant increase in all-cause mortality for patients admitted with COVID-19 ARDS to the ICU. This effect persisted even after adjustment for clinically predetermined confounders, including diabetes, median procalcitonin and FPG.

Figure 1

Flowchart of patient enrolment. The register population represents all patients included in the COVAS cohort. A total of 106 out of 271 patients were enrolled in this subgroup analysis. ARDS acute respiratory distress syndrome; DGV daily glycaemic variability.

Figure 2

Comparison of HbA1c and variability metrics. Comparison of multivariable Cox-Proportional Hazard models, adjusted for age, sex and history of type 2 diabetes, for HbA1c and established variability metrics: standard deviation (SD), bias corrected Coefficient of Variation (CoefVar), Neuman’s (root) Mean Square of Successive Differences (rMSSD and MSSD) and median of the absolute successive difference between successive values (= DGV, daily glycaemic variability). Includes Hazard Ratio and 95% confidence intervals, p-value for the parameter and likelihood ratio test (LR), Akaike Information Criterion (AIC) and concordance index (C-Index).

Figure 3

Relation between fatality risk and glucose variability. Continuous hazard ratios for fatality or non-survival were calculated by Cox proportional hazard models of daily glycaemic variability (DGV) adjusted for age and sex, including 95% CI for endpoint. The dashed red line notates a hazard ratio of 1.0, the solid black line the dichotomisation threshold, which we determined using a regression tree. The threshold of 25.5 mg/dl for DGV was used for our summary statistics (Tables 1 and 2), Kaplan–Meier curves (Fig. 4) and the adjusted DGV Cox proportional hazard model (Fig. 5).

Figure 4

Kaplan–Meier Survival for ARDS Patients. Kaplan–Meier survival curves for patients with low (< 25.5 mg/dl) and high (≥ 25.5 mg/dl) daily glycaemic variability (DGV). Each cross on a curve represents a censored patient (survived to discharge), while each step in the curve represents a deceased patient. The x-axis represents the time in days to discharge. Survival curves were compared, and p-values calculated by log-rank test.

Figure 5

Adjusted hazard ratios for non-survival by DGV. Multivariable Cox proportional hazard model for median daily glycaemic variability (DGV) cut-off as a continuous covariate dichotomised at DGV ≥ 25.5 mg/dl, adjusted for clinically predetermined confounders: age, sex, type 2 diabetes status, dexamethasone treatment, median of procalcitonin, glucose and DGV during ICU admission.