Development and Validation of a Differential Diagnostic Models for Guillain–Barré Syndrome Based on Clinical and Laboratory Indicators: A Retrospective Study

Abstract

Objective: This study is aimed at developing a differential diagnostic model for Guillain–Barré syndrome (GBS) from other central nervous system diseases based on clinical and laboratory indicators.

Materials and Methods: A retrospective approach was conducted for the GBS patients and patients with other neurological diseases (non-GBS group, including viral encephalitis, peripheral neuropathy, multiple sclerosis, transverse myelitis, neuromyelitis optica spectrum disorders, and myasthenia gravis). The least absolute shrinkage and selection operator (LASSO) technique was integrated with multivariable logistic regression to perform predictor selection. The logistic regression model was established as the predictive framework, followed by the application of the Shapley additive explanation (SHAP) framework to quantify contributions of selected variables within the model. After that, patient data were collected for model validation.

Results: A total of 161 patients with GBS and 644 patients with non-GBS diseases were enrolled. Upper limb weakness, visual impairment, areflexia, hyperreflexia, total bilirubin (TBIL), mean corpusular hemoglobin (MCH), platelet large cell ratio (P-LCR), cerebral spinal fluid–protein (CSF-protein), dyslipidemia index, and oligoclonal band-serum/cerebral spinal fluid (SOB-CSF) emerged as independent predictors of GBS development. The logistic regression classifier demonstrated robust predictive performance, achieving an area under the curve (AUC) of 0.915 in the testing set, with an accuracy of 0.876, sensitivity of 0.823, and specificity of 0.889.

Conclusion: We developed and validated a logistic regression model incorporating multiple clinical indicators to differentiate GBS from other inflammatory neurological disorders (including MS, NMOSD, MG, TM, VE, and PN). The model demonstrated high diagnostic accuracy (AUC 0.92), supporting its potential as a supplementary tool for clinical decision-making.