Recovery from aphasia remains difficult to predict. Among the language features predicting recovery from aphasia, production scores such as repetition or phonology seem to be more relevant predictors than only severity. As motor cortex is strongly involved in language processes, both production and perception, the present study aimed to determine whether the integration of an electrophysiological measure of the motor networks using motor-evoked potentials MEP in the acute phase of stroke can improve the prediction of recovery from post-stroke aphasia.

Hundred and twelve patients with aphasia within 14 days after a first ischemic or haemorrhagic stroke completed the study. The severity of aphasia was assessed using the Aphasia Severity Rating Scale (ASRS) at baseline and 6 month (M6) after the stroke. The ASRS is a 6-point Likert scale from the lowest score 0 to 5. A severe aphasia initially was defined as an ASRS score<3.

MEPs were recorded from the abductor pollicis brevis muscle and orbicularis oris. The resting motor threshold (rMT) was defined for each muscle and a ratio between left and right rMT (rMTr) was calculated for upper-limbs and for lips.

A first-level model, including only clinical variables (i.e. initial severity) predicted severity at six months for severe patients. When the rMTr of upper limbs was added in a second-level model, the predictive power significantly increased from 41 to 51%, as well as adding in a third-level model rMTr of upper-limbs and initial fibres number ratio of the corticospinal tracts (41 to 55%). With the changes of severity as the dependent variable, the same factors made a significant contribution and the predictive power of a second-level model increased from 4 to 20% to the same extent as in a third-level model (4% to 27%).

Added with usual clinical factors, MEPs of upper-limbs and lips contribute to the prediction of aphasia recovery, particularly for patients with severe aphasia initially.