Introduction: Glycemic control is a major concern during critical illness. Several prospective studies have yielded conflicting results regarding its mortality effect. Current recommendations are to initiate insulin therapy for all patients when glucose levels are higher than 180 mg/dL. Some suggest decreasing this threshold for non-diabetic patients to 140 mg/dL. These thresholds haven't been compared to each other or to other glucose thresholds. This study aimed to find out whether different glucose levels are associated with 90-d mortality.
Methods: A retrospective cohort study. Critically ill patients who were admitted from 2019 to 2022 to a mixed medical-surgical intensive care unit for more than 48 h were included. Collected data included baseline characteristics, and all glucose levels recorded (time-indexed to the admission time). Glucose levels were considered constant until the following glucose level. The percentage of time above several chosen glucose cutoff levels was calculated and analyzed for mortality adjusted to other baseline covariates.
Results: 45,512 glucose measurements of 1429 patients were included in the study; 21.76 % of the patients had diabetes. Mean glucose level and glucose variability were higher in diabetic patients (165.86 mg/dL vs 135.47 mg/dL, p < 0.0001, and 30.81 % vs 20.86 %, p < 0.0001, respectively), along with a higher incidence of hypoglycemia (40.84 % vs 24.89 %, p < 0.001). 90-d mortality was higher in diabetic patietns (42.12 % vs 32.41 %, p = 0.0014) and was found associated with age, acute physiology and chronic health evaluation 2 score, medical or surgical admission reasons. Percentage of time above cutoffs ≥150 mg/dL was associated with 90-d mortality only in non-diabetic patients.
Conclusions: In non-diabetic patients, hyperglycemia greater than 150 mg/dL, was associated with increased 90-day mortality.
Keywords: Glucose control; Hyperglycemia; Hyperglycemia tolerance; Insulin therapy.
Copyright © 2024 European Society for Clinical Nutrition and Metabolism. Published by Elsevier Ltd. All rights reserved.