Association of glycemic variability with 28-day mortality and adverse outcomes among non-diabetes patients after cardiac surgery with cardiopulmonary bypass: a retrospective analysis on the MIMIC-IV database

BMC Endocr Disord. 2026 Jun 29. doi: 10.1186/s12902-026-02373-0. Online ahead of print.

ABSTRACT

OBJECTIVES: To analyze postoperative glycemic variability (GV) patterns in non-diabetic patients undergoing cardiac surgery with cardiopulmonary bypass (CPB), and determine the association between various GV metrics and adverse outcomes, including 28-day mortality.

METHODS: A retrospective cohort study utilized data from the Medical Information Mart for Intensive Care IV (MIMIC-IV) database, including 2,940 non-diabetic patients who underwent cardiac surgery with CPB. GV was quantified by mean blood glucose (MBG), largest amplitude of glycemic excursion (LAGE), standard deviation of blood glucose (SDBG), coefficient of variation (CV), and mean amplitude of glycemic excursions (MAGE). Cox proportional hazards models were used to determine the relationship between GV indices and clinical outcomes. The restricted cubic spline (RCS) analysis examined non-linear associations, while receiver operating characteristic (ROC) curves were used to assess predictive accuracy for 28-day mortality.

RESULTS: Abnormal MBG levels (hypoglycemia and hyperglycemia) and elevated GV metrics (LAGE, SDBG, CV, MAGE) were significantly linked to adverse clinical outcomes (P < 0.05). Results of Cox regression analysis demonstrated that LAGE and CV could strongly predict 28-day mortality compared to MBG, SDBG, and MAGE. RCS analysis revealed non-linear associations, with MBG exhibiting a U-shaped relationship. ROC analysis confirmed a superior predictive accuracy for LAGE (AUC = 0.845) and CV (AUC = 0.817) compared with other metrics (P < 0.05).

CONCLUSIONS: Postoperative GV is an independent risk factor for 28-day mortality in non-diabetic CPB patients. LAGE and CV outperformed other GV indices in mortality prediction, highlighting the prognostic value of GV monitoring.

PMID:42374330 | DOI:10.1186/s12902-026-02373-0