2.1

Participants

We included 32 individuals with aphasia after a first stroke who were followed 2 years post-stroke. Inclusion criteria were aphasia after a first ischemic or haemorrhagic stroke of the left hemisphere, usable auditory comprehension (see below) that permitted understanding basic questions from questionnaires, French-speaking and right-handed. Exclusion criteria were impairment of consciousness or coma, illiteracy, pre-stroke dementia, severe dysarthria, and psychiatric disorders. Participants were followed by a physician specialised in physical medicine and rehabilitation and received multidisciplinary rehabilitation (e.g., physiotherapy and/or occupational therapy) according to the French and International guidelines and depending on their initial deficiencies. Hence, if needed, patients had physiotherapy 4 to 5 times a week the first months and occupational therapy to train upper-limb motor impairment or limitations in daily life activities. To treat aphasia, all patients received conventional speech therapy at least for the first 6 months after the stroke, as much as possible depending on their tolerance and at least 3 sessions a week. Fig. 1 presents the overlap of the lesions.

Overlap of brain lesions of 30/32 patients with aphasia after stroke. Two patients presented contraindications for MRI.

2.2

Measures

In addition to socio-demographic information, the following instruments were administered to each participant at the 2 time points (i.e., as soon as possible in the first 14 days post-stroke and 2 years post-stroke [±3 months]), during a face-to-face interview by a trained physician and a speech therapist at baseline and by 2 speech therapists at follow-up, who were blinded to baseline results.

QoL was assessed at follow-up by Sickness Impact Profile (SIP-65), the French short version of the SIP-136, a specific scale for patients with aphasia . The SIP-65 is feasible and recommended to assess health-related functional status and QoL of French people with aphasia, excluding those without reliable comprehension (i.e., patients with Z-score less than the mean in comprehension subtests). This survey can be completed by patients themselves, with the speech therapist to assist comprehension impairment, and consists of 65 items divided into 8 categories exploring physical, psychological and social components of QoL.

The severity of aphasia was determined by using the French version of the Goodglass and Kaplan’s Boston Diagnostic Aphasia Examination (BDAE) and its severity subscore, the Aphasia Severity Rating Scale (ASRS) . The ASRS is rated on a 6-point Likert scale from 0, no usable speech or auditory comprehension, to 5, very slight language impairment. This measure was performed in the acute phase (i.e., as soon as possible in the first 14 days post-stroke) and at follow-up. Aphasia severity was considered mild with ASRS score 4-5, moderate with ASRS score 3, and severe with ASRS score < 3. Good outcome at month 6 was defined as ASRS score 4 or 5. Hence, to explore the impact of good recovery on QoL, we were able to distinguish between 3 categories based on the initial severity:

• •
  

  patients with mild or moderate aphasia initially, known to have good outcomes and who all recovered;

  
  
• •
  

  patients with severe aphasia who did not recover and:

  
  
• •
  

  patients with severe aphasia who recovered.

Comprehension subscores allowed for screening impairments in comprehension for reliable assessment at follow-up. On the basis of the Z-score of the BDAE, we excluded patients with scores < 56/72 for the subscore picture recognition “words discrimination” (mean 68.4 [SD 5]), < 10/15 for the item order execution “commands” (mean 13.2 [SD 2.4]) or < 6/12 for the item complex material (mean 9.5 [SD 1.9]).

Communication activity limitation was assessed by the Verbal Communication Scale of Bordeaux (VCSB) , exploring current communication behaviours in daily life. It also includes questions about motivation for communication and strategies implemented by the patient to cope with their difficulty. Several aspects of communication were explored: basic communication, conversation, phone use, shopping, social communication, reading and writing. This measure was performed only at follow-up because of the lack of relevancy in the acute phase.

Fatigue was assessed by the French version of the Fatigue Severity Scale (FSS) , a self-administered questionnaire consisting of 9 items that identifies features of fatigue. Each item is graded from 1, strongly disagree, to 7, strongly agree. The scale was validated in stroke . It was performed only at follow-up because of its lack of relevancy in the acute phase.

Limitations on activities of daily life were assessed at baseline and at follow-up by the Barthel index (scored from 0, no functional abilities, to 100, no limitation). This measure was performed in the acute phase and at follow-up. Initial stroke severity was assessed by the US National Institutes of Health Stroke Score (NIHSS) at baseline.

Anxiety and depression were assessed by the French version of the Hospital Anxiety and Depression Scale (HADS) , with 2 subscores assessing depression (HADS-D) and anxiety (HADS-A). Each subscore includes 7 items scored from 0 to 3; a score > 7 indicates anxiety or depression . This measure was performed only at follow-up because of its lack of relevancy in the acute phase.

2.3

Statistical analysis

First, we explored the association between QoL and age, sex, completed speech therapy, level of education, antidepressant treatment, ASRS at baseline and 2 years post-stroke, the category of patients taking into account initial severity, VCSB, FSS, Barthel Index, NIHSS, and HADS-A and HADS-D. Pearson and Spearman correlations, one-way ANOVA for multiple groups and independent t tests for binary variables (normal distribution) were used. We then used multivariable analysis (linear regression, backward method) with SIP-65 as the dependent variable and significant factors assessed at follow-up. Regarding aphasia, the category of patients taking into account initial severity or VCSB were added to the model. This method first placed all variables in the model and calculated the contribution of each of them. Their contribution was then compared against a removal criterion. The variable with the least contribution to the model was subsequently removed and the reduced model re-estimated for the remaining variables. The contribution of the remaining variables was reassessed in an iterative way until the model reached statistical significance. Moreover, for each predictor, the relative importance and contribution percentages of the predictors were calculated ; p-values were adjusted for multiplicity to control the overall type I error rate . Finally, all subscores of VCSB were added in a model with SIP-65 as a dependent variable by using the same approach to explore which subscore was the most predictive. All statistical analyses were performed with R3.5.0 ( www.R-project.org ). P < 0.05 was considered statistically significant.