Blood Glucose Levels and Mortality in Patients With Sepsis: Dose-Response Analysis of Observational Studies

Background: We undertook a systematic review and meta-analysis to investigate the relationship between blood glucose levels and mortality in patients with sepsis. Methods: Medline and EMBASE were searched from inception to April 8, 2018. Cohort studies or case-control studies reported the association between blood glucose and mortality in patients with sepsis were selected. Study characteristics, baseline characteristics, definition of hyperglycemia, and outcomes of interest were extracted. We performed a dose-response meta-analysis to assess the effect of blood glucose level on mortality. We also conducted meta-analysis for patients with or without diabetes separately. Results: Ten cohort studies involving 26 429 patients were included, of which 5 were prospective studies and 5 retrospective studies. Dose-response analysis showed that the effect of blood glucose on mortality may differ in patients with versus without diabetes. There was a U-shaped relationship for patients with diabetes and a J-shaped relationship for patients without diabetes, with blood glucose at 145 to 155 mg/dL corresponding to lowest mortality both in patients with and without diabetes. Conclusions: Current evidence suggested U-shaped relationship between blood glucose and mortality in all patients irrespective of their diabetes status. Diabetic patients with blood glucose below 145 mg/dL may have poorer prognosis compared to patients without established diabetes.