Predictors of post-stroke aphasia recovery: a systematic review-informed individual participant data meta-analysis: A Systematic Review-Informed Individual Participant Data Meta-Analysis

Background: The factors associated with recovery of language domains affected by stroke remain uncertain. We described recovery of overall-language-ability, auditory comprehension, naming and functional-communication across participants’ age, sex and aphasia chronicity in a large, multilingual, international aphasia dataset. Methods: Individual participant data (IPD) meta-analysis of systematically-sourced aphasia datasets described overall-language-ability using the Western Aphasia Battery Aphasia-Quotient (WAB-AQ); auditory comprehension by Aachen Aphasia Test Token Test (AAT-TT); naming by Boston Naming Test (BNT) and functional-communication by AAT Spontaneous-Speech Communication subscale (AAT-SC). Multivariable analyses regressed absolute score-changes from baseline across language domains onto covariates identified a priori in RCTs and all study types. Change-from-baseline scores were presented as estimates of means and 95% confidence intervals (CI). Heterogeneity was described using relative variance. Risk of bias was considered at dataset and meta-analysis level. Results: Assessments at baseline (median=43.6 weeks post-stroke; interquartile range (IQR) [4-165.1]) and first-follow-up (median=10 weeks from baseline; IQR [3, 26]) were available for n=943 on overall-language-ability, n=1,056 on auditory comprehension, n=791 on naming and n=974 on functional-communication. Younger age (<55years, +15.4 WAB-AQ points [CI: 10.0-20.9], +6.1 correct on AAT-TT [CI: 3.2-8.9]; +9.3 BNT points [CI: 4.7-13.9]; +0.8 AAT-SC points [CI: 0.5-1.0]) and enrolment <1 month post-onset (+19.1 WAB-AQ points [CI: 13.9-24.4]; +5.3 correct on AAT-TT [CI: 1.7-8.8]; +11.1 BNT points [CI: 5.7-16.5]; and +1.1 AAT-SC point [CI: 0.7-1.4]) conferred the greatest absolute change-from-baseline across each language domain. Improvements in language scores from baseline diminished with increasing age and aphasia chronicity. Data exhibited no significant statistical heterogeneity. Risk-of-bias was low to moderate-low. Conclusion: Earlier intervention for post-stroke aphasia as crucial to maximise language recovery across a range of language domains though recovery continued to be observed to a lesser extent beyond 6 months post-stroke.