Nonlinear association between albumin-to-globulin ratio and in-hospital mortality after primary liver cancer resection: A retrospective cohort study

Background This study aimed to investigate the nonlinear association between preoperative albumin-to-globulin ratio (AGR) and in-hospital mortality following primary liver cancer resection. Methods A retrospective cohort analysis was conducted using data from 1,241 patients undergoing curative hepatectomy at Beijing You’an Hospital (2008–2025). Restricted cubic spline (RCS) models and multivariate logistic regression were employed to evaluate threshold effects and adjusted odds ratios (OR). Subgroup analyses explored interactions across sex, comorbidities, and tumor characteristics. Results In-hospital mortality occurred in 3.4% (42/1241) of patients. AGR exhibited a nonlinear relationship with mortality ( P for nonlinearity = 0.02), with an inflection point at 1.58. Below this threshold, each unit increase in AGR reduced mortality risk by 92% (adjusted OR = 0.079, 95% CI: 0.018–0.349, P  < 0.001), while no significant association was observed above 1.58. The multivariable regression analysis demonstrated that the AGR independently predicted reduced in-hospital mortality following primary liver cancer resection.Tertile-based analysis confirmed higher AGR (T3 vs. T1) correlated with a 60% risk reduction (OR = 0.40, 95% CI: 0.18–0.89). The subgroup analysis did not yield statistically significant results for sex, biliary tract diseases, hepatitis, hypertension, diabetes mellitus (DM), pneumonia, hypersplenism, degree of differentiation and microvascular invasion (MVI) grade when testing for interactions. No interactions were observed in the subgroups except for sex (all P  > 0.05). Conclusions Preoperative AGR demonstrates a nonlinear, threshold-dependent association with post-resection mortality, identifying high-risk patients at AGR ≤ 1.58. These findings highlight AGR’s utility in perioperative risk stratification and personalized interventions.