Fifty-five studies (75 estimates of treatment effects) were included. These comprised 1,491 participants in within-treatment effects, and 866 participants in between-treatment effects. Sixty of the effects were within-treatment effects, and the remaining 15 were between-treatment effects.
Within-treatment effects (n=60).
Untreated recovery: the average ES with a pre-test measure taken in the acute stage was 0.63 (95% confidence interval, CI: 0.31, 0.95), corresponding to the magnitude of spontaneous recovery. When the pre-test measure was taken between the third and twelfth month post-onset (i.e. the postacute stage), the average ES was 0.34 (95% CI: 0.04, 0.64). The average ES for chronic effects (those having the reference measure taken on or after the twelfth month post-onset) was 0.05 (the CI could not be calculated), reflecting the asymptote of the natural recovery function.
Recovery with treatment: when the reference measure was taken in the acute period, the average ES for recovery was 1.15 (95% CI: 0.5, 1.8), nearly twice that for untreated recovery (0.63). The average ES was 0.57 (95 CI: 0.06, 1.08) for treatment begun in the postacute period, and 0.66 (95% CI: -0.31, 1.63) for that begun in the chronic period.
Amount of treatment (n=12): the average ES for three high-intensity treatments provided by speech-language pathologists in the acute period was 1.39 (95% CI: 0.45, 2.33). The ES for moderate dosage in the acute period was 1.78 (the CI could not be calculated); however, this value represented a single effect. Overall, the more intense the treatment, the greater the change.
The ESs that were coded for amount of treatment showed a positive correlation (r) with the duration of treatment in weeks (r=0.76, p=0.004), and the total number of treatment hours at discharge (r=0.64, p=0.26).
Type of treatment: only two types of treatment were reported. The most frequently reported named treatment (5 effects), the Schuell-Wepman-Darley Multimodal Stimulation Treatment, gave an ES of 1.39 (95% CI: 0.18, 2.60). The most frequently reported treatment was that coded as 'not specified' (7 effects), giving an ES of 0.81 (95% CI: 0.28, 1.34).
Severity of aphasia: there were no studies looking at the effect on mild aphasia. The ES for treatment begun in the acute period of recovery was quite large for moderately and severely aphasic individuals: the ESs were 1.87 (95% CI: 0.91, 2.83, n=12) and 2.76 (no CI reported, n=6), respectively. Nineteen studies coded as heterogeneous (i.e. a mixture of severities among the patients in the primary study) gave an ES of 1.05 (95% CI: 0.22, 1.88). The ESs in the postacute period were considerably smaller than those obtained in the acute period.
Type of aphasia (n=6): four effects were obtained from nonfluent aphasic individuals and two from globally aphasic persons. The average ESs were 0.63 (95% CI: -0.76, 2.02) and 0.83 (95% CI: -0.04, 1.70) for nonfluent aphasic persons and variously diagnosed participants, respectively.
Between-treatment effects (n=15).
Treatment versus no treatment: two acute stage treatment versus control effects, resulting from active treatment provided by a speech-language pathologist in competition with a no treatment control,were pooled in a synthesis of effects; this gave an average ES of 0.61 (95% CI: 0.16, 1.06). In the postacute stage, the size of the treatment versus no treatment effect was 0.31 (this was a single effect, so it was possible to calculate a CI). A single mixed-stage effect of 0.23 was reported.
The validity of the meta-analysis was assessed by calculating the number of null findings that would be necessary, if they existed, to diminish the value of the ES to a critically low value. For the within-treatment effects, the critical value of ES for treated individuals was the corresponding value of ES for treated individuals. The results indicated that the findings of the meta-analysis were fairly robust.