Traces of systemic inflammation in stroke: Are NLR and NPAR strong predictors of mortality?

Introduction: The prognostic value of systemic inflammation in patients with acute ischemic stroke is gaining importance. This retrospective study aimed to examine the associations of the neutrophil-to-lymphocyte ratio (NLR) and the neutrophil percentage-to-albumin ratio (NPAR) with malnutrition, in-hospital infection, and mortality, and to identify the independent variables that effectively predict mortality. It also aimed to compare the performance of logistic regression with that of a machine learning–based random forest model.Methods: We retrospectively reviewed 290 patients followed in our clinic between November 2024 and May 2025. Demographic characteristics, laboratory findings, and mortality data were recorded. NLR and NPAR values were measured at admission. Mortality at day 90 was taken as the dependent variable. A logistic regression model was used for multivariable analysis. In addition, variable importance ranking and model discrimination were calculated with the random forest algorithm. ROC–AUC and model fit criteria were evaluated.Results: The mean NLR was 3.25 and the mean NPAR was 16.7. The mortality rate was 10.7%, the infection rate 19.0%, and the malnutrition rate 9.3%. In patients with malnutrition, both NLR (7.50 vs 2.81; p<0.001) and NPAR (24.7 vs 15.8; p<0.001) were significantly higher. NLR and NPAR were markedly elevated in the mortality group (p<0.001). In the logistic regression analysis, age, NLR, and NPAR were identified as independent factors that increased mortality risk (p<0.01), and the model showed high discrimination (AUC: 0.910). The random forest model similarly identified age, NLR, NPAR, and sedimentation rate as predictors of mortality (AUC: 0.814). In patients with an mRS score of 0–3, NPAR and NLR predicted mortality well, whereas in patients with mRS >4, only NPAR predicted mortality well.Discussion/Conclusion: In stroke patients, NLR and NPAR showed strong associations with malnutrition, infection, and mortality. The findings suggest that NPAR in particular may be useful in clinical decision support systems; however, they need to be confirmed in prospective, multicenter studies.Keywords: