Abstract
BACKGROUND: Glucose-to-potassium ratio (GPR) has demonstrated significant clinical value in various acute conditions. However, its prognostic utility in critically ill individuals with sepsis complicated by malignancy has not been unraveled. METHODS: The studied population with sepsis and malignancy from the MIMIC-IV database was split into four cohorts by baseline GPR. Survival probabilities across GPR cohorts were compared through Kaplan-Meier survival curves. The link of GPR to clinical outcomes was unraveled employing the Cox proportional hazards model. The potential nonlinear relation was examined using restricted cubic spline (RCS) models. Result robustness was rated via subgroup analyses. Additionally, we utilized the Integrated Discrimination Improvement (IDI) and Net Reclassification Improvement (NRI) to assess the incremental predictive value of adding GPR to the existing scoring systems. RESULTS: 4,802 critically ill individuals were encompassed. Their median age was 69.78 (IQR: 61.11–78.83), and 2,883 (60.04%) were male. KM analysis showed that patients with lower GPR values exhibited significantly higher ACM at 30, 90, and 360 days (log-rank P < 0.01). After the Cox models were fully adjusted, higher GPR levels were significantly associated with lowered likelihood of 30-day ACM (HR = 0.934, 95%CI 0.880–0.992, P = 0.026), 90-day ACM (HR = 0.923, 95%CI: 0.876–0.973, P = 0.003), and one-year ACM (HR = 0.954, 95%CI: 0.913–0.997, P = 0.038). RCS analysis revealed a nonlinear relation of GPR to ACM, with an inflection point at a GPR value of 1.725. Compared to patients with GPR levels above this threshold, those with lower GPR values had a 26.8% decreased 30-day ACM risk (HR = 0.732; 95%CI: 0.593–0.903). Furthermore, incorporating GPR into the SOFA and APS III scoring systems improved the predictive accuracy of the models. CONCLUSION: GPR is significantly related to ACM among critically ill people with sepsis and malignancy. GPR possibly serves as a potential prognostic indicator for both short-term and long-term death risk in this vulnerable population. CLINICAL TRIAL NUMBER: Not applicable.