Late recovery after a first ischemic stroke is highly variable and its predictors are unknown. The present study aims at determining whether MRI data obtained one to four months after a first ischemic stroke help to predict clinical evolution up to 2 years.

Patients included in the PERFORM MRI study, an ancillary study of the PERFORM randomized control trial of terutroban against aspirin in secondary prevention of vascular ischemic events were selected. Mixed-effect regression modelling was used to test whether MRI data obtained one to four months after a first ischemic stroke ameliorate the prediction of further recovery, up to 2 years, compared to clinical data alone. Outcomes to predict were disability (modified Rankin Scale [mRS] and NIH Stroke Scale [NIHSS]) and cognition (MMSE, Isaac's Set Test [IST] and Zazzo's Cancellation Test [ZCT]). MRI markers were designed as the total lesion load on FLAIR (FLAIR_vol) and brain volume (brain parenchymal fraction [BPF]) on T1, both normalized to intracranial cavity volume. Age, gender, level of education and initial value of the outcome to predict were systematically entered as covariates in predictive models based on clinical data alone. FLAIR_Vol, BPF and microbleed number were added to those variables in predictive models based on clinical and imaging data. Predictive ability of both types of models were compared.

Five hundred and ninety-two patients of mean age 67.3±7.8 years were included. Models based on clinical and MRI data were significantly better predictors of mRS, MMSE, IST and ZCT, compared to models based on clinical data alone.

MRI data that can be easily extracted from routine sequences help to predict further recovery even months after a first ischemic stroke. The use of MRI in this context may help to select patients for which rehabilitation will be the most beneficial.