Abstract
The Glucose-to-Albumin Ratio (GAR) has been identified as a significant predictor among sepsis and nonalcoholic fatty liver disease. However, its utility in predicting all-cause mortality among patients with acute pancreatitis (AP) has not been studied. Consequently, this study aimed to evaluate the predictive value of GAR for 28-day mortality among AP patients. This retrospective cohort study used data from the Medical Information Mart for Intensive Care (MIMIC-IV) database. According to the median GAR value, participants were divided into two cohorts. The primary endpoint was all-cause mortality within 28 days. Kaplan–Meier survival analysis and the log-rank test were employed to compare survival between the two cohorts. Cox proportional hazards models were then applied to estimate the hazard ratio of GAR, with and without adjustment for other significant clinical factors. Subgroup analyses were conducted to evaluate the relationship between GAR and mortality within strata of other significant factors. Finally, receiver operating characteristic (ROC) curves based on a logistic regression model were constructed to assess the predictive capability of GAR using the area under the curve (AUC), which was then compared with other predictors. A total of 459 patients with AP were included in the study. Kaplan–Meier curves and the log-rank test demonstrated that patients with a high GAR (> 48.85) had significantly higher 28-day all-cause mortality than those with a low GAR (p = 0.026). Cox proportional hazards models further indicated that high binary GAR was associated with an increased unadjusted hazard ratio (HR = 2.006, p = 0.029). When treated as a continuous variable, higher GAR remained significantly associated with an elevated hazard ratio after adjusting for age, gender, race, and BMI (HR = 1.010, p = 0.007). This association persisted in the final adjusted model derived through backward selection, where higher GAR continued to predict higher 28-day mortality risk (HR = 1.010, p = 0.008) after controlling for age, gender, race, BMI, and prothrombin time (PT). Subgroup analyses revealed no significant interaction between binary GAR and other covariates in predicting 28-day mortality. Compared with glucose, albumin, the Systemic Inflammatory Response Syndrome score, and the Glasgow Coma Scale score, GAR demonstrated superior predictive power, with a higher AUC (0.602) for predicting 28-day mortality. Our study originally identified that elevated GAR was correlated with increased 28-day all-cause mortality among AP patients, and GAR demonstrated strong predictive value for short-term mortality risk. SUPPLEMENTARY INFORMATION: The online version contains supplementary material available at 10.1038/s41598-025-34571-6.