Oral Presentation Australian and New Zealand Stroke Organisation Conference 2026

A systematic review of early neuroimaging and neurophysiological biomarkers for post-stroke upper limb prognostication (139275)

Cristina Levy 1 2 , Emily Dalton 1 3 , Jennifer Ferris 4 , Bruce Campbell 5 , Amy Brodtmann 5 6 , Sandra Brauer 7 , Leonid Churilov 5 , Kate Hayward 1 5
  1. Department of Physiotherapy, University of Melbourne, Parkville, Australia
  2. Department of Physiotherapy, Royal Melbourne Hospital, Parkville, Australia
  3. Department of Occupational Therapy, Royal Melbourne Hospital, Parkville, Australia
  4. Gerontology Research Centre, Simon Fraser University, Vancouver, Canada
  5. Department of Medicine and Neurology, Royal Melbourne Hospital, University of Melbourne, Parkville, Australia
  6. Department of Neurosciences, Monash University, Melbourne, Australia
  7. School of Health Sciences, University of Queensland, St Lucia, Australia

Background/Aim

It is unclear whether early neuroimaging or neurophysiological biomarkers can be used for post-stroke upper limb recovery prognostication. We performed a systematic review to assess the prognostic capacity of early neuroimaging and neurophysiological biomarkers for post-stroke upper limb outcomes.

Methods

MEDLINE/EMBASE were searched to identify cohort studies reporting biomarkers measured ≤14-days post-stroke and upper limb outcomes assessed 14-days to 24-months post-stroke. Analyses were classified as association, discrimination/classification, or validation. Measures of magnitude indicated prognostic capacity. Risk of bias was rated using the Quality in Prognostic Studies tool. Heterogeneity prevented meta-analysis. PROSPERO:CRD42022350771.

Results

From 26,486 title/abstracts, 72 studies (n=4,728 participants) were included. Biomarkers were measured median 5-days post-stroke, and outcomes from 1- to 24-months. Of 351 analyses (253 neuroimaging/86 neurophysiological), most were associative (81%). Lesion location, lesion size and corticospinal tract lesion load demonstrated limited prognostic capacity. Corticospinal tract fractional anisotropy and motor evoked potential status showed some promising associations, but classification performance was inconsistent and unvalidated. Most prognostic models were at the development stage (84%; 115/135 models) and interpretable performance measures were rarely reported. The PREP2 tool was validated and demonstrated excellent sensitivity but fair specificity. Methodological limitations across all studies included small sample sizes, moderate-high risk of bias, and heterogeneous outcome measures and time-points.

Conclusion

Evidence for post-stroke upper limb prognostic neuroimaging and neurophysiological biomarkers is limited. The absence of promising validated predictive performance prevents clinical translation. Multisite collaboration, integration of clinical and biomarker metrics, and standardised reporting of interpretable performance measures are critical to advance the field.