Optimising speech and language therapy interventions for aphasia after stroke

Overview

We examined the impact of the modifiable components of SLT interventions for aphasia after stroke (e.g. regimen, intervention approach and delivery; see Table 1 and Box 3). We included age, time since aphasia onset and sex to account for population differences between data sets, in advance of the planned network meta-analysis considering the impact of intervention variables. These factors are considered important predictors of general stroke recovery and were important predictors for language recovery in the preceding analysis across selected language domains for absolute changes in raw scores from baseline to first follow-up (see Chapters 4 and 5).

We considered language measures at first follow-up after SLT intervention periods. Data sets and IPD filtering in preparation for analysis were reported in Table 5. We explored the impact of SLT intervention by regimen, therapy approach, delivery model and tailoring. When a SLT intervention variable was reported for an isolated data set only, then the effect of that variable on language could not be distinguished from the study effect, and findings were treated with caution.

Network meta-analysis

Each SLT intervention variable was considered simultaneously (with continuous dose variables such as frequency grouped) in the context of language recovery (overall language ability, auditory comprehension, naming and functional communication). There were insufficient data to consider the impact of SLT variables on other spoken language, reading comprehension, writing or functional communication (self-rated) domains (see Table 5). The following analysis presents the overall language ability, auditory comprehension, naming and functional communication (observer-rated) findings (see Appendices 11–14).

The functional communication domain includes data gathered on the TOMs, when subtest data on communication activity or participation were available. We arbitrarily included the TOMs activity data set in the main analysis, but conducted a sensitivity analysis, repeating the analysis inclusive of the participation data to consider whether or not our decision had affected the findings (see Appendix 9, Tables 16–18, and Chapter 8).

Within-study clustering was examined for the four language domains and the four treatment regimen categories: frequency, duration, intensity and dosage (see Appendix 15, Table 19). This form of structuring of the data, which allows for correlation between patients within studies, did not significantly improve the fit of our predictive model to the data. The additional modelling of within study clustering was either non-significant or caused the failure of the analysis due to the G matrix not being positive definite (see Appendix 15, Table 19).

We examined the contribution to the model (which included baseline score, sex, age group and time since stroke) of each SLT intervention by regimen (frequency, duration, intensity, dosage, home practice), intervention approach (impairment target, theoretical approach) and delivery (provider expertise, context, tailoring by functional relevance and difficulty). We explored the magnitude of differences seen with different degrees of the SLT component of interest (e.g. providing 5 or 14 or 20 hours of SLT; see overview in Appendix 16, Table 20).

Frequency of speech and language therapy for aphasia

We started with a base model that included the language domain absolute raw score as the outcome, and age, sex and time since stroke as covariates. We added frequency of SLT (i.e. the number of days per week that SLT was delivered) to this model and assessed its impact on language domain scores. Frequency, in addition to age, sex and time since stroke, significantly contributed to auditory comprehension (p = 0.0091) and naming (p = 0.03) outcomes, but not to overall language ability or functional communication outcomes (see Report Supplementary Material 5). We saw significant improvements in participants’ scores from baseline to first follow-up across overall language ability, auditory comprehension, naming and functional communication language domains. Data on reading outcomes were absent from included RCTs (see Appendix 17, Table 21). Based on RCT IPD, the greatest overall language ability [14.95 points (95% CI 8.67 to 21.23 points) on the WAB-AQ (194 IPD; 6 RCTs)] and functional communication gains [0.78 points (95% CI 0.48 to 1.09 points) on the AAT-SSC (155 IPD; 8 RCTs)] occurred alongside delivery of SLT 4 to 5 days weekly. The highest auditory comprehension gains [5.86 points (95% CI 1.64 to 10.01 points) on the TT-AAT (114 IPD; 5 RCTs)] were associated with receiving SLT 3 to 4 days weekly. The greatest naming gains [12.05 points (95% CI 6.52 to 17.59 points) on the BNT (42 IPD; 3 RCTs)] occurred alongside receiving SLT up to 2 days weekly, but this was based on small numbers of IPD and RCTs, and a network geometry with limited connections; therefore, caution should be used when interpreting this finding (see Appendix 13, Figure 31). The non-RCTs IPD supported these findings (see Appendix 17, Table 21).

Duration of speech and language therapy for aphasia

Speech and language therapy duration reflected the total number of weeks that the intervention lasted. The addition of the duration variable to the base model was significantly associated with auditory comprehension outcomes (p = 0.0013; see Report Supplementary Material 6), but not with other domains. Importantly, the SLT duration network meta-analysis geometry was unstable, with few studies comparing therapy durations and little variation on maximum gains in the context of SLT of different durations (see Appendices 11–14 and 18); thus, the SLT duration findings should be interpreted with caution. The greatest gains on overall language ability [17.27 points (95% CI 9.71 to 24.82 points) on the WAB-AQ (45 IPD; 3 RCTs)] and auditory comprehension [6.79 points (95% CI 1.69 to 11.88 points) on the TT-AAT (93 IPD; 5 RCTs)] were associated with SLT lasting 10–20 weeks. Naming gains from baseline were greatest [8.38 points (95% CI 1.24 to 15.51 points) on the BNT (134 IPD; 3 RCT)] alongside > 20 weeks of SLT. The best gains in functional communication [0.8 points (95% CI 0.29 to 1.31 points) on the AAT-SSC (57 IPD; 2 RCTs)] occurred with 4–10 weeks of SLT (see Appendix 18, Table 22). Non-RCT data were generally confirmatory (see Appendix 18, Table 22).

Intensity of speech and language therapy for aphasia

Intensity of SLT was measured by the number of SLT hours delivered weekly. Including the intensity variable in the base model (with age, sex and time since stroke) significantly contributed to auditory comprehension outcomes (p = 0.0004), but not to any other domains (see Report Supplementary Material 7). Although clinically similar gains occurred alongside different SLT intensities, peak gains on overall language ability [15.85 points (95% CI 8.06 to 23.64 points) on the WAB-AQ (74 IPD; 3 RCTs)], naming [13.23 points (95% CI 5.83 to 20.64 points) on the BNT (18 IPD; 2 RCTs)] and functional communication [0.77 points (95% CI 0.36 to 1.19 points) on the AAT-SSC (83 IPD; 4 RCTs)] were all associated with receiving SLT for up to 2 hours weekly. In contrast, peak auditory comprehension gains [7.3 points (95% CI 4.09 to 10.52 points) on the TT-AAT (141 IPD; 6 RCTs)] occurred when delivery of SLT was in excess of 9 hours weekly. The non-RCT data set provided similar findings (see Appendix 19, Table 23).

Dosage of speech and language therapy for aphasia

Including the SLT dosage variable (total number of SLT hours) to the base model of age, sex and time since stroke significantly contributed to naming (p = 0.040) and functional communication (p = 0.018) outcomes, but not to other domains (see Report Supplementary Material 8).

The greatest gains in overall language ability [18.37 points (95% CI 10.58 to 26.16 points) on the WAB-AQ (31 IPD; 4 RCTs)] and auditory comprehension [5.23 points (95% CI 1.51 to 8.95 points) on the TT-AAT (90 IPD; 7 RCTs)] were associated with 20–50 hours of SLT. Functional communication gains [0.94 points (95% CI 0.34 to 1.55 points) on the AAT-SSC (11 IPD; 3 RCTs)] were highest in the context of a slightly lower SLT dosage of 14–20 hours. Naming gains were highest alongside up to 5 hours of SLT [12.48 points (95% CI 1.34 to 23.63 points) on the BNT (28 IPD; 2 RCTs)] (see Appendix 20, Table 24). For functional communication and naming gains, IPD and RCT numbers were low and the naming network analysis geometry was unstable (see Appendix 13, Figure 31). It is noteworthy that, for both functional communication and naming, the second-highest gains occurred in the context of 20–50 hours of SLT [naming: 9.23 points (95% CI 1.75 to 16.7 points) on the BNT (81 IPD; 5 RCTs); functional communication: 0.77 points (95% CI 0.43 to 1.1 points) on the AAT-SSC (96 IPD; 9 RCTs)] (see Appendix 20, Table 24). Addition of the non-RCT IPD was generally confirmatory, although the greatest (and only significant) naming gains from baseline [11.32 points (95% CI 5.92 to 16.72 points) on the BNT (135 IPD; 8 data sets)] were associated with 20–50 hours of SLT (see Appendix 20, Table 24).

Home practice

Prescribed SLT home practice augments face-to-face therapy dosage. Included data sets rarely described home practice dosage, so we considered whether or not data sets reported the prescription of home practice (yes/no). When home practice was not reported, we assumed that it was not present. Adding the home practice variable to the base model (sex, age, time since stroke) significantly contributed to auditory comprehension outcomes (p = 0.019; see Report Supplementary Material 9), but not to other language domains. The best gains from baseline on overall language ability [16.69 points (95% CI 10.01 to 23.37 points) on the WAB-AQ (87 IPD; 2 RCTs)] and auditory comprehension [5.28 points (95% CI 2.19 to 8.37 points) on the TT-AAT (278 IPD; 7 RCTs)] were associated with prescribed home practice. In contrast, naming gains [7.08 points (95% CI 1.81 to 12.35 points) on the BNT (166 IPD; 9 RCTs)] and functional communication gains [0.74 points (95% CI 0.39 to 1.09 points) on the AAT-SSC (267 IPD; 10 RCTs)] were slightly higher in RCTs that did not report home practice than in those that did. Including the non-RCT data generally confirmed these findings, except for functional communication, when greater gains were achieved by the group for which home practice was reported [0.69 points (95% CI 0.34 to 1.04 points) on the AAT-SSC (187 IPD; 5 RCTs) than by the group for which home practice was not reported (see Appendix 21, Table 25).

Speech and language therapy regimen

We considered the contribution that unit increases in the SLT intervention regimen (frequency, duration, dosage and intensity) made to specific language outcomes using simple linear regression models and using the IPD RCT data set only (Table 9). For example, for every unit increase in SLT intensity (i.e. every additional SLT hour weekly), the average overall language ability score change from baseline was an increase of 0.23 points on the WAB-AQ. These average changes by regimen, language domain and anchor measure are presented in Table 9. Examination of the categorical data suggests that excess treatment may have a negative effect on participants in some domains. Naming scores appear to be negatively affected by increased SLT frequency, intensity and dosage; thus, further examination of SLT regimen in the context of network meta-analyses is required. Changes from baseline may have been reduced in the higher-dose bandings.

Target of speech and language therapy intervention

Including the SLT target variable in the base model did not significantly contribute to the language outcomes (see Report Supplementary Material 10). Few significant gains from baseline were observed. The greatest gains in overall language ability [15.62 points (95% CI 8.82 to 22.43 points) on the WAB-AQ (245 IPD; 8 RCTs)] and functional communication [1.05 points (95% CI 0.52 to 1.58 points) on the AAT-SSC (72 IPD; 3 RCTs)] were associated with mixed SLT approaches (see Appendix 22, Table 26). The greatest gains in auditory comprehension [4.46 points (95% CI 0.31 to 8.62 points) on the TT-AAT (136 IPD; 7 RCTs)] and naming [8.82 points (95% CI 3.15 to 14.49 points) on the BNT (174 IPD; 7 RCTs)] were associated with word-finding SLT. The inclusion of non-RCT data was generally confirmatory (see Appendix 22, Table 27).

Theoretical approach of speech and language therapy intervention

Including the SLT theoretical approach variable in the base model of age, sex and time since stroke did not significantly contribute to the model on any language domain (see Report Supplementary Material 11). The numbers of IPD, data sets and network connections were generally low (see Appendices 11–14); therefore, the data were unreliable, and findings should be considered cautiously. Few significant gains from baseline were observed. When the greatest gains were identified, overall language ability [20.39 points (95% CI 1.90 to 38.88 points) on the WAB-AQ (60 IPD; 2 RCTs)] and auditory comprehension gains [11.93 points (95% CI 1.44 to 22.43 points) on the TT-AAT (35 IPD; 1 RCT)] were associated with semantic–phonological SLT approaches. Functional/pragmatic approaches were associated with the greatest functional communication gains [1.82 points (95% CI 0.36 to 3.28 points) on the AAT-SSC (8 IPD; 1 RCT)] (see Appendix 23, Table 27).

Adding multimodal approaches to the base model together with age, sex and time since stroke significantly contributed to functional communication outcomes, but not to any other outcome (see Report Supplementary Material 12). Functional communication gains (as measured by the AAT-SSC) were observed in the context of multimodal approaches [verbal: 1.96 points, 95% CI 0.85 to 3.07 points (38 IPD; 2 RCTs); total: 1.90 points, 95% CI 0.86 to 2.93 points (66 IPD; 3 RCTs)] and verbal SLT approaches [1.95 points, 95% CI 0.92 to 2.99 points (273 IPD; 10 RCTs)] (see Appendix 24, Table 28). No significant gains from baseline were observed in the context of multimodal SLT (verbal or total) on auditory comprehension or naming (see Appendix 24, Table 28).

Expertise

We categorised the expertise of the person delivering SLT as either professional (therapist or researcher) or non-professional (trained volunteer or family member). Adding the expertise variable to the base model of age, sex and time since stroke did not significantly contribute to any outcomes (see Report Supplementary Material 13). The availability of IPD based on non-professional-delivered SLT interventions was poor (< 15 IPD across the language domains), and findings should be interpreted with caution (see Appendix 25, Table 29). Inclusion of the non-RCT data did not alter these findings.

Context of delivery

The SLT context (whether participants were inpatients or outpatients during their therapy sessions) was extracted. Adding this variable to the model of age, sex and time since stroke did not make a significant contribution to language outcomes when based on IPD from RCTs or non-RCT data (see Report Supplementary Material 14). The SLT context made little impact on the gains made from baseline (see Appendix 26, Table 30).

Mode of delivery

We considered the SLT delivery mode as either face to face, computer supported or self-managed. Adding the delivery mode variable to the base model did not significantly contribute to the language outcomes (see Report Supplementary Material 15).

Overall, we observed some slight, but clinically insignificant, difference in gains between face-to-face SLT and non-face-to-face approaches, for example overall language ability (as measured on the WAB-AQ) gains associated with face-to-face SLT (14.13 points, 95% CI 8.34 to 19.91 points) versus non-face-to-face SLT (11.06 points, 95% CI 2.63 to 19.49 points) (351 IPD; 11 RCTs) (see Appendix 27, Table 31).

Tailoring

The content of a SLT intervention may be tailored to an individual patient’s needs in two ways: tailoring by functional relevance and/or tailoring by level of language difficulty. When tailoring was not reported, we assumed that it had not taken place. There were no within-study comparisons of tailored and untailored approaches.

Tailoring by functional relevance

The addition of the tailoring by functional relevance variable to the base model significantly contributed to auditory comprehension outcomes (p = 0.029), but not to other domains (see Report Supplementary Material 16).

When significant changes were observed, higher gains occurred in the context of functionally relevant SLT for overall language ability [16.47 points (95% CI 10.95 to 21.99 points) on the WAB-AQ (232 IPD; 6 RCTs)], naming [8.79 points (95% CI 1.95 to 15.63 points) on the BNT (113 IPD; 5 RCTs)] and functional communication [0.74 points (95% CI 0.38 to 1.10 points) on the AAT-SSC (249 IPD; 6 RCTs)] than with non-tailored SLT interventions. Similarly, significant gains from baseline for auditory comprehension were only evident in the context of SLT tailored by functional relevance [5.26 points (95% CI 2.05 to 8.47 points) on the TT-AAT (194 IPD; 7 RCTs)] (see Appendix 28, Table 32). The analysis inclusive of IPD from non-RCTs confirmed these findings.

Tailoring by difficulty

Including the SLT tailoring by difficulty variable in the base model significantly contributed to auditory comprehension outcomes (p = 0.043), but not to any other language outcome. Inclusion of the non-RCT data supported this finding (see Report Supplementary Material 17).

Tailoring of SLT by difficulty was associated with higher gains in overall language ability [14.46 points (95% CI 8.82 to 20.09 points) on the WAB-AQ (210 IPD; 7 RCTs)] than was untailored SLT [13.24 points (95% CI 7.50 to 18.99 points) on the WAB-AQ (207 IPD; 6 RCTs)], but this difference is not clinically relevant. In contrast, auditory comprehension gains from baseline were only significant in the context of SLT tailored for level of difficulty [4.57 points, 95% CI 1.55 to 7.60 points (331 IPD; 10 RCTs)] as participants receiving SLT untailored by level of language difficulty made no significant gains (see Appendix 29, Table 33).

In contrast, SLT untailored to participants’ levels of language difficulty was associated with higher gains in naming [10.21 points (95% CI 2.75 to 17.67 points) on the BNT (79 IPD; 4 RCTs)] and functional communication [0.81 points (95% CI 0.34 to 1.27 points) on the AAT-SSC (141 IPD; 5 RCTs)] than tailored SLT (see Appendix 29, Table 33). The non-RCT data supported these findings.

Heterogeneity

We considered heterogeneity with reference to risk of selection bias, study variability and methodological quality.

We examined the similarity of participant groups in the primary data sets (RCTs and non-RCTs) at baseline, comparing participants’ age, sex, time since stroke, aphasia severity, handedness and years of education (when available), and they were broadly similar (see Appendices 4 and 5).

Speech and language therapy interventions included in the analysis differed between intervention groups and across studies, but there was consensus that all were SLT interventions. Variations in SLT approaches (by delivery, intervention approach and regimen) were explored in the analysis.

The early pilot analysis resulted in unacceptably high relative study variability, in excess of 25%. Following careful examination of the contributing data sets and groupings in this analysis (e.g. the dose thresholds), data set checks and a minor adjustment to the algorithm (see Appendix 2, Box 4), the variance subsequently lay between 10% and 25% and all remained under 30%. Higher variance values observed were considered to be related to small IPD in some cells. No analysis exceeded the prespecified threshold of 50% relative variability due to study differences.

Summary

Having established a base model of the language domain absolute raw scores as outcomes, and age, time since stroke and sex as covariates, we included SLT intervention parameters in succession. We found that several SLT intervention parameters significantly contributed to selective language domain outcomes: auditory comprehension and SLT frequency, intensity, home practice and tailoring (by functional relevance and difficulty); naming and SLT frequency and dosage; and functional communication and SLT dosage. The addition of SLT impairment target, provider’s expertise, the mode of SLT and therapy context to the base model made no significant contribution to language outcomes.

The greatest gains across language outcomes were observed in the context of SLT regimens that varied by frequency, intensity and dosage; these are summarised in Chapter 8. Duration and naming data were treated with caution across all language domains because of unstable network geometry and, in many cases, small numbers of IPD and data sets.

Overall language ability gains were highest in the context of receiving SLT 4 to 5 days weekly, for up to 2 hours weekly and over 20–50 hours in total, using mixed expressive and receptive language impairment approaches, tailored to the participant (by functional relevance and level of difficulty) and with prescribed home practice tasks.

The highest auditory comprehension gains were associated with receiving SLT 3 days weekly, for > 9 hours weekly, for a total of 20–50 hours, tailored to the participant (by functional relevance and level of difficulty) word-finding SLT approaches augmented with prescribed home practice tasks.

Participants’ functional communication gains were best in the context of receiving SLT 4 to 5 days weekly, for up to 2 hours weekly, for 14–50 hours in total, using functionally relevant approaches, targeting expressive and receptive language impairment (mixed SLT). The mode of delivery (face to face, computer supported or self-managed) and the context of SLT (inpatient vs. outpatient settings) made little clinical difference to the language gains observed from baseline. As planned, variability in language outcome gains and the SLT context in which they occur was further explored by participant subgroups (age, sex, time since stroke and aphasia severity), and is reported in Chapter 7.