Association of the glucose-to-lymphocyte ratio with stroke-associated pneumonia among acute ischaemic stroke patients with dysphagia: a secondary analysis of hospital-based observational data

Abstract Background Stroke-associated pneumonia (SAP) is a frequent complication after acute ischaemic stroke, particularly in patients with dysphagia. The glucose-to-lymphocyte ratio (GLR) is a simple biomarker measured at admission reflecting stress hyperglycaemia and stroke-related immune suppression, but its value in dysphagic stroke populations remains uncertain. In this secondary analysis of hospital-based observational data, we investigated the association between admission GLR and SAP in 681 Vietnamese patients with acute ischaemic stroke and dysphagia treated at a tertiary referral hospital in Hanoi. SAP was defined as a lower respiratory tract infection occurring within 7 days after stroke onset according to the modified Pneumonia in stroke consensus group criteria. GLR was measured in the first admission blood sample and analysed per 1-standard-deviation increase and by quartiles. Logistic regression models were adjusted for demographic and clinical covariates. The incremental value of GLR was assessed by comparing a clinical model with and without GLR. Results SAP occurred in 70 of 681 patients (10.3%; 95% CI 8.2 to 12.8). Patients with SAP were older, more often had previous stroke, more severe stroke and dysphagia, lower lymphocyte counts, higher GLR values and higher in-hospital mortality. In crude analysis, each 1-standard-deviation increase in GLR was associated with higher odds of SAP (odds ratio [OR] 1.24, 95% confidence interval [CI] 1.001 to 1.53), an association of borderline statistical significance that was attenuated after adjustment (adjusted OR 1.10, 95% CI 0.86 to 1.41). No significant trend was observed across GLR quartiles. GLR alone discriminated poorly (area under the curve 0.58), and adding GLR to a model containing established clinical predictors did not improve discrimination (area under the curve 0.820 to 0.821; difference 0.001, 95% CI − 0.002 to 0.012). Conclusions Admission GLR was not independently associated with SAP after multivariable adjustment and showed poor stand-alone discrimination. Adding GLR to a baseline clinical model did not improve discrimination or overall model performance. GLR may reflect overall acute illness burden rather than independent pneumonia risk. These findings do not support adding admission GLR to routine SAP risk assessment in this cohort.