Incremental predictive value of multimodal over unimodal assessment for upper limb motor recovery after stroke: a systematic review and meta-analysis

Background Clinical scales predict post-stroke motor recovery but are insensitive to neural substrates. Neuroimaging and neurophysiological biomarkers capture corticospinal integrity, yet their incremental value over clinical assessment remains unquantified. Objective To quantify the pooled incremental predictive value (ΔAUC) of adding biomarkers to clinical models for upper limb motor recovery at 3–6 months post-stroke. Methods We systematically searched PubMed, Embase, and Web of Science from inception to January 31, 2026, for prospective cohort studies evaluating ≥2 assessment modalities. Random-effects meta-analysis pooled ΔAUC between multimodal and clinical-only models. Sensitivity analyses used alternative between-study variance estimators and the Hartung-Knapp adjustment. Meta-regression examined sources of heterogeneity. Risk of bias was assessed using QUIPS. Results Twenty-three studies ( N = 1,341) were included. Multimodal assessment significantly improved prognostic accuracy over clinical models alone (pooled ΔAUC = +0.112, 95% CI 0.061 to 0.163, p < 0.001). MEP status showed the largest incremental gain (ΔAUC = +0.150, 95% CI 0.023 to 0.277; I 2 = 71.0%). Ipsilesional CST FA was reported in one study (ΔAUC = +0.220, 95% CI 0.110 to 0.330). Other DTI-derived metrics did not reach statistical significance (ΔAUC = +0.074, 95% CI −0.006 to 0.155). Task-based fMRI showed promise in one study (ΔAUC = +0.190, 95% CI 0.088 to 0.292). EEG, kinematic parameters, and resting-state fMRI could not be pooled. Stroke phase was a significant moderator ( β = −0.127, 95% CI −0.247 to −0.007, p = 0.039), with larger gains in subacute versus acute assessments. Conclusion In this systematic review and meta-analysis, adding selected biomarkers to clinical assessment was associated with improved discrimination for upper limb motor recovery after stroke. MEP status showed the largest incremental gain. Stroke phase significantly moderated the incremental value, with larger gains in the subacute phase. However, given the small number of studies and the prediction interval crossing zero, these findings should be considered statistically promising but not yet robust enough to guide routine clinical practice. Further validation in larger prospective cohorts is needed before MEP-guided stratification can be applied clinically. Systematic review registration https://www.crd.york.ac.uk/PROSPERO/view/CRD420261288506 , Identifier: CRD420261288506.