Disease‐Specific Optimal Blood Glucose Control Strategies in Critically Ill Patients: Retrospective Cohort Study and Network Meta‐analysis

ABSTRACT Background Optimal glucose targets for critically ill patients are likely to vary by underlying condition, yet a ‘one‐size‐fits‐all’ approach persists. Aim Determining disease‐specific glucose thresholds may help achieve the lowest mortality rates. Study Design This study combined a retrospective cohort analysis of the Medical Information Mart for Intensive Care‐IV (MIMIC‐IV) database with a frequentist network meta‐analysis (NMA) of randomized controlled trials (RCTs). The glucose control targets, measured by time‐weighted average glucose (TWAg), were categorized into four strata: strict, intermediate strict, liberal and very liberal control. Results The retrospective analysis included 21 119 sepsis, 16 957 cardiac and 4708 neurological patients. In both sepsis and cardiac patients, compared with the intermediate strict control group (110–150 mg/dL, reference), both the strict (80–110 mg/dL: HR 1.10/1.26, both p < 0.05) and liberal (150–180 mg/dL: HR 1.38/1.78, both p < 0.05) control groups showed significantly higher 28‐day mortality. In contrast, among neurological patients, strict control conferred a survival benefit (HR 0.76, p = 0.013). The NMA of 35 RCTs corroborated these findings for sepsis and neurological disorders, with intermediate strict control ranking highest for sepsis (P‐score 74.9%) and strict control highest for neurological disorders (P‐score 71.2%), whereas liberal control ranked highest for cardiac disease (P‐score 83.5%); no pairwise differences were statistically significant. Conclusions This study provides evidence that optimal glucose targets in the ICU are disease‐specific. In the retrospective analysis, intermediate strict control was associated with the lowest mortality in sepsis and cardiac disease, while strict control may benefit neurological patients without excess hypoglycaemia. However, for cardiac disease, the NMA ranked liberal control highest, leaving the optimal target uncertain. Relevance to Clinical Practice These findings may support a pathophysiology‐guided, individualized approach to glycaemic management in critically ill patients.