Background: Predicting aphasia recovery after stroke has been difficult due to substantial variability in outcomes. Few studies have characterised the nature and extent of recovery, beginning with baselines at 24–72 hours after stroke onset.
Aim: To characterise the course of language recovery after first-time stroke.
Methods: Using our Performance and Recovery in Stroke Study (PARIS) database, we evaluated consecutive first-time stroke patients with aphasia and diffusion-weighted-image-positive lesions on admission and at 90 days.
Results: Twenty-two of 91 patients had language disorders. Initial syndrome scores were positively correlated with 90-day scores (r = 0.60) and negatively correlated with the change in score from baseline to follow-up (r = −0.66). Neither lesion size, age nor education correlated with initial syndrome severity or with performance at 90 days. Level of education was not associated with degree of recovery. A multiple regression model that combined lesion size, age and initial syndrome was significant (p = 0.03) but only explained 29% of the variance. Patients with severe deficits at baseline in individual language domains could recover, improve to a less severe deficit or not improve at all.
Conclusion: There was significant variability in language recovery after first-time stroke, even in more severe, initial syndromes. Traditional predictors of post-stroke language outcomes did not reliably predict function at 90 days. These data suggest that other factors that account for functional stroke recovery have not yet been identified.
Statistics from Altmetric.com
Most patients who have suffered aphasia and related language disorders after stroke recover at least some degree of function.1 Although long-term follow-up has shown recovery after many years,2 3 the greatest degree of spontaneous recovery appears to occur within the first 3 months after a stroke.4–7 Among the factors purported to determine language recovery after stroke, initial syndrome severity and lesion size have been reported to be important predictors.8–10
Many prior studies of aphasia recovery, however, did not collect baseline data until several days to weeks after admission, did not exclude patients with prior stroke, and reported data as mean values for patient groups based on initial aphasia diagnosis, obscuring potential variability across individual patients. The importance of understanding and predicting recovery in the acute period lies in the recent data that show brain re-organisation occurring earlier than previously thought11 12 and that earlier post-stroke therapy may improve patient outcomes.13 We had the opportunity to follow the evolution of language recovery in the Columbia Performance and Recovery in Stroke (PARIS) study, a prospective database of first-time stroke patients. Our interest in restricting our sample to first-time events arises from the difficulty in interpreting stroke-induced disorders of higher cerebral function if there has been prior clinical stroke.14 The major aim of this study was to characterise the nature and extent of recovery of aphasic deficits from 24–72 hours after stroke onset to 90-day follow-up.
SUBJECTS AND METHODS
The Columbia PARIS database is comprised of English- and Spanish-speaking patients with first-time stroke, at least 18 years of age, presenting on admission with a new deficit in motor, language and/or visual–spatial dysfunction, and a positive MRI diffusion-weighted image (DWI) for stroke. All were examined for baseline function within 24–72 hours of stroke onset. The follow-up examination reported here took place at 90 days after the qualifying stroke because it is felt that most spontaneous recovery occurs by this point.5 6 Patients with prior clinical stroke or other known neurological disease or condition were ineligible, as well as those with first-time stroke who were somnolent. Regardless of the qualifying syndrome for entry into PARIS, each patient was administered the comprehensive battery of tests evaluating aphasia, hemineglect and hemiparesis. All patients signed institutional-approved informed consent and those with severe aphasia were eligible from an Institutional Review Board exemption.
The aphasia battery consisted of three test domains: Comprehension, repetition and naming. For comprehension, patients were given either the dictated commands subtest from the Boston Diagnostic Aphasia Examination15 (BDAE) or the Sequential Commands subtest from the Western Aphasia Battery16 (WAB). Naming was assessed via the Object Naming subtest from the WAB. Repetition was tested with either the low and high probability phrases from the BDAE, or with the Repetition subtest from the WAB. Follow-up testing for individual patients always involved the same measures as those used at baseline. The published, translated version was used for Spanish-speaking patients. Handedness was established by self-report, or by the family when patients were too aphasic to do so. Patients were not considered to have a defect in repetition or naming as a result of dysarthria, although disorders of articulation can co-occur with aphasia.
To establish comparability across the three assessment modalities, all tests were converted to a 10-point scale, with impairment defined as a score worse than 8.0. This cut-off was chosen based on the published normative values for each assessment instrument that would be between 2 and 3 standard deviations below the mean. Impaired scores were dichotomised further into less (4–7.9) or more severe (<4). To be considered aphasic in PARIS, patients had to have at least one abnormal language domain on the aphasia baseline examination. A function was considered to be “recovered” if the follow-up score was 8 or above. To provide a single, overall measure of language function, the scores of the three test domains were combined into a single variable, with a range of 0–30. Education was graded on a continuous 6-point scale: 1 = Less than a high school degree; 2 = High school degree; 3 = Some post-secondary education; 4 = College/University Degree; 5 = Some post-graduate courses; 6 = Graduate degree.
Relative lesion volumes on MRI were calculated by measuring the largest X and Y diameters of the lesion on axial diffusion imaging, and multiplying by the number of visualised slices times the slice thickness (5 mm) divided by 2, similar to methods used to calculate haematoma volume.17
There were 91 patients in the complete PARIS database, accrued from 15 April 2001 to 18 July 2005, of whom 22 were considered to have a focal language disorder at the time of stroke admission because each had at least one impaired language domain (see below). Table 1 displays demographic characteristics, total scores at baseline and follow-up, and lesion locations. Twenty patients in this aphasic cohort were reported to be right-handed. There were 17 males and 5 females, with a mean age of 63.8 years (SD 5.6). Seventeen patients had some form of speech-language therapy after stroke; we could not ascertain whether or not language intervention occurred for the remaining five patients. Figure 1 depicts the total language function (total possible = 30) at baseline and at the 90-day follow-up. The mean combined baseline language function score for these 22 patients was 17.13 (SD 0.97) and the mean combined 90-day score was 23.60 (SD = 6.08). The five patients for whom we had no information regarding speech-language therapy had a mean baseline score of 21.84 (SD 0.95) and a follow-up score of 25.92 (SD 0.2). As a group, there was significant overall improvement in language functioning 90 days after stroke onset (p<0.001). Initial syndrome scores were positively correlated (Pearson) with 90-day scores (r = 0.60, p = 0.001) and negatively correlated with the change in score from baseline to follow-up (r = −0.66, p = 0.001).
Table 2 shows that, as expected, the most common disorder at baseline was a naming disorder (15 patients). Among this aphasic cohort, 5 participants had only one language abnormality, 13 had two spheres of language dysfunction and 4 had impairment in all three aphasic spheres, and so there is no assumption that these measurements represent independent observations. For example, the correlation of baseline naming scores with baseline comprehension scores was 0.53 (p = 0.007). By 90 days, 23 functional deficits had resolved, with only 4 deficits that were still more severe and 15 that were less severe. Comparing less severe (score = 4–7.9) with more severe deficits (score = 0–3.9) at baseline, Chi-square analysis showed that patients were as likely to recover (score ⩾8) from a more severe deficit as from a less severe one (naming: p = 0.49; repetition: p = 1.0; comprehension: p = 0.37).
Regression analysis demonstrated that baseline language functioning was predictive of 90-day language functioning, accounting for 33% of the variance (p<0.01). Baseline language functioning was also predictive of 90-day language recovery, accounting for 41% of the variance (p<0.01). Although the group data show improvement, visual inspection of figure 1, however, does show that there could be wide variability in 90-day scores among patients with similar baselines. For example, two patients had baseline scores of 1.5 or lower, one improved to 6.5, and the other had no recovery at all. Multivariate regression analysis was used to determine whether addition of the variables age, education and lesion size would account for a greater portion of the variance in 90-day language functioning or recovery. Addition of these variables to the equation did not reliably improve r2.
Figure 2 depicts individual function scores at the time of stroke onset and at the 90-day follow-up for comprehension (left), naming (middle) and repetition (right). Curves represent the performances for individual patients on each task. This graph demonstrates the wide range of variability in outcomes that could occur with the same degree of deficit on stroke admission. For example, five patients had scores less than 1 when tested for comprehension at baseline. At 90 days, however, three patients had nearly a complete recovery of comprehension, one patient improved to 3/10 and one made no improvement at all.
Calculation of Pearson correlation coefficients also revealed that lesion size was not associated with baseline language functioning (r = −0.23; p = 0.30), 90-day language functioning (r = −0.19; p = 0.40), nor with the total recovery (ie, change in total score) in language functioning. Other non-statistically significant correlations were between the baseline global score and age (r = −0.31; p = 0.14) and education (r = 0.03; p = 0.89); global change score and age (r = 0.08; p = 0.70) and education (r = −0.11; p = 0.59); and 90-day global score and age (r = −0.28; p = 0.17) and education (r = −0.08; p = 0.70). t-tests comparing the change in total score across the four lesion locations displayed in table 1 showed no statistical differences. Similarly, there was no correlation between education and 90-day language functioning (r = 0.02, p = 0.91) nor recovery of language functioning (r = −0.05; p = .82). Age did not correlate with 90-day language score (r = −0.30; p = 0.18) nor the change in score from baseline to follow-up (r<0.01; p = 0.98).
Among the aphasic patients studied prospectively in the Columbia PARIS database, most functional deficits improved to some extent from acute stroke admission to the 90-day follow-up, which is consistent with prior literature.18 Our goal was to demonstrate the nature and extent of the recovery that can occur in a first-time stroke cohort. As such, we did not intend to characterise the evolution of global aphasia syndromes following acute stroke, per se, to establish the features of vascular aphasias, nor to further study the relationship between lesion locations and traditional aphasia/language syndromes.
The results showed that it was difficult to predict how individual patients would fare on the basis of their initial aphasic presentation. Although baseline language functioning was predictive of outcome, only 32–41% of the variance could be accounted for. The interpretation of the statistical relationship between initial and final language functioning, however, is confounded by two artifacts. First, a positive correlation between the initial and final scores is likely to occur when there are mild deficits at syndrome onset. Second, a negative correlation between initial total score and the change in score from baseline to follow-up can come about from initial greater deficits (lower scores) having more room to change than the initial high scores.19 Neither age nor education was predictive of recovery. A multiple regression analysis that combined a number of these independent factors showed a statistically significant relationship to aphasia outcomes but the model did not account for a greater proportion of the variance than baseline language function, indicating that there are other, as yet unidentified, factors governing recovery. Indeed, when we looked at the recovery patterns of individual language domains, we found that it was as likely that a patient would recover as not recover from a more severe initial deficit, although a larger population might have revealed differences.
Most prior studies of aphasia recovery have not gathered their baseline language data during the acute stroke period,8 9 20 although it is recognised that evaluating patients too soon after stroke onset runs the methodological risk of misconstruing the focal impact of reversible ischaemia as part of the permanent syndrome.21 We therefore evaluated aphasia characteristics early in their course (<72 hours), but waited at least 24 hours after the onset of stroke symptoms to minimise the influence of transient, perfusion-related deficits. Nevertheless, we did find that the two patients who had the most severe baseline syndromes were seen at 24 hours, so that it is possible that there was residual cortical hypoperfusion.22 A direct effect of thrombolysis at the time of study entry was unlikely, as our baseline evaluations were at least 24 hours after its possible administration. Some studies have established language profiles during the acute admission period, but failed to exclude a significant number of patients with previous stroke.1
The importance of establishing the features of acute aphasia, specifically, and other post-stroke syndromes, in general, lies in the increasing evidence for early brain re-organisation and perhaps early opportunities for behavioural therapy. Non-human animal models, for example, have shown enhanced cortical activity shortly after induced ischaemic lesions.23 24 Such acute physiological changes may account for our own observation of fMRI changes in human motor activation within 24 hours after stroke25 and by other demonstrations of MRI language activation within 72 hours after an ischaemic event.26 These early signs of brain plasticity have led to new studies suggesting that pairing behavioural and/or pharmacological therapy with these acute alterations in brain organisation may improve patient function.11 13
The advantage of evaluating individual deficits rather than global characterisations (eg, Wernicke’s and Broca’s aphasias) is that within each of these syndromes is a wide range of profiles, and that disturbances can be highly idiosyncratic, especially when the lesions are smaller.27 Moreover, it has been shown in large studies of acute post-stroke aphasia that only slightly more than half of patients have the classic syndromes that comprise aphasiology.28 We chose comprehension, naming and repetition because of their sensitivity to deficits that can arise from anterior or posterior lesions, and because of their objectivity in measurement.
Every effort was made to accrue consecutive patients to ensure an unbiased sample, with an enrollment rate greater than 95% of the first-time stroke cases admitted to our service, and we were able to capture, at study entry, levels of language dysfunction ranging from severe global to anomic aphasia. The age range of our cohort was typical of a stroke population,29 and although not described in this report, the stroke risk factors of diabetes, hypertension, elevated lipids, obesity and lack of exercise were representative of patients admitted to an large urban stroke unit.30 None of these patients had any severe systemic illnesses such as infection or hyperglycaemia, seizures, psychiatric disorders or known dementia, all of which were exclusion criteria for study entry. None had any known language disorders prior to their index stroke.
Among the limits of this study was the small sample size; a larger cohort might have increased the extent to which our regression model would explain variance, and perhaps altered the outcomes of univariate analyses. We did not look at the potential roles of silent infarction, lacunes, white matter disease or atrophy, although the contribution of these factors to language recovery is not well established. It should be noted that seven of our patients did not have a naming disorder, which was once thought a necessary element for the diagnosis of aphasia. More recently, however, aphasia has been regarded more generically as a disorder of linguistic processing.31 It is not clear in our non-anomic cases whether there had been a rapid amelioration of their language disorders, including naming, but our data did show that 6/7 patients did fully recover by 90 days. Because they were part of our database and had least one abnormal language domain at baseline, we included them in this analysis. Unfortunately, we did not have information regarding the nature or intensity of speech-language therapy among the 17 patients who received such intervention, nor a comparison group of those who did not receive therapy. As a result, we could not assess the impact of therapy on the recovery in our cohort. As we enrolled a relative young group of patients (mean age, 63.8 years), we were not able to address the nature of recovery among very old patients, who are at greater risk for aphasia after first-ever stroke.32
The major implication of these findings is that aphasia recovery cannot be predicted in the earliest stages after stroke onset, based solely on age, education, syndrome severity and lesion size. Our findings suggest that there are probably other factors affecting language recovery that have yet to be fully explored, and which may predict those individuals who would probably benefit from therapy. For example, we recently demonstrated, via fMRI activation within 48 hours after stroke onset, that there may be an independent, global pattern of brain activity that predicts motor recovery at 90 days.33 The extent to which a common neural mechanism regulates language recovery has not yet been established. Another regulatory factor might be the degree of neuroinflammation at the time of stroke, as suggested by the enhancement of neurogenesis in animal stroke models following the administration of anti-inflammatory agents.34 35 In any case, it is likely that stroke recovery, in general, and post-stroke language recovery, specifically, is multiply-determined.
This research was funded in part by NIH grant 5R01-HD043249 (RML), the Richard and Jenny Levine Foundation, and the Doris and Stanley Tananbaum Foundation.
Competing interests: None.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.