Complex speech-language therapy interventions for stroke-related aphasia: the RELEASE study incorporating a systematic review and individual participant data network meta-analysis

Aphasia and psychosocial impacts

Aphasia directly influences a person’s perception of their own identity16,17 isolating the person with the communication impairment from their spouse, family and wider social networks. 18,19 Family members have also described feeling isolated. 20 Aphasia intensifies social problems more generally associated with a stroke, restricting or altering social activities. 21 This leads to fewer friendships and smaller social networks than before the stroke, and than both healthy peers and stroke survivors with preserved language abilities. 18,21 With restricted opportunities for social participation, people with aphasia become socially isolated, which severely affects their emotional well-being. 19 Spontaneous recovery of language and communication abilities has been thought to be limited beyond the first year after aphasia onset. 22 Emerging research has questioned this premise,23 and further evidence indicates that focused therapeutic interventions may provide benefit in later stages after stroke. 24–26

Aphasia recovery and participant demographics

Clinical guidelines recommend that people with aphasia are provided with ‘realistic recovery prospects’ following aphasia,27 but, based on current evidence, accurate prognosis is difficult. Although age has been associated with language recovery, the nature of that interaction remains unclear. 3,13,28,29 Patient sex has been reported by some to be linked to the incidence of aphasia13 and initial aphasia profile (but not recovery3,28), although several large, well-controlled studies found no evidence of this effect. 2,3,30 Conducting a large-scale investigation of aphasia after stroke has often required compromise, with language data based on a single item in a global stroke severity scale,2,3,29 and, in some cases, restricted to one aspect of language use. 3 Details on the ability to speak, understand, read, write and participate in everyday communication activities are often unavailable in these tools, thereby limiting our clinical interpretation of aphasia impact and recovery.

The contribution that socioeconomic status and educational background make to recovery may also be relevant, but the interconnectedness of these factors with cognition and performance on measures of language ability makes this a difficult concept to untangle. 3,31–33 The incidence of aphasia is similar among monolingual and bilingual stroke survivors, although it has been suggested that bilinguals experience a milder aphasia. 34 However, studies using imaging data found that bilinguals’ performances on a detailed language assessment protocol were worse than monolinguals’ performances. 35 Social support prior to and following a stroke may also positively influence outcome36,37 and, although opportunities to practise functional language use have been shown to decrease as a consequence of aphasia,18 they are also an important factor in aphasia recovery38,39 and research. 38

Aphasia recovery and stroke profile

Initial stroke severity13,28 and time since stroke may be important factors in language recovery, the latter specifically in relation to the acceptability of high-intensity therapy. 40,41 There is growing evidence that the site and size of stroke lesion (although perhaps not structural connectivity) are indicators of language recovery. 31,42 Stroke-related problems, including depression and cognition, vision and hearing impairments, are also thought to be important to language recovery and rehabilitation. 43–45 Others suggest that the initial aphasia profile (i.e. severity, domains involved, abilities retained) may predict the pattern of language recovery. 3,28,31 Concomitant stroke-related speech disorders, such as dysarthria (a motor speech disorder) and apraxia (disordered planning of movement to produce speech), may also affect initial functional outcomes and recovery. 46

Clinical guideline recommendations

UK and international clinical guidelines on aphasia after stroke highlight the paucity of guidance to support therapists’ provision of aphasia rehabilitation interventions. 58 Evidence-based stroke clinical guidelines rarely offer specific therapy approach, regimen or mode of delivery recommendations, in contrast to movement rehabilitation recommendations for strength and fitness training, orthoses, electric stimulation and more. 59–61

Early62,63 and high-intensity SLT64 may be effective, but these historic reviews of SLT for aphasia were limited in their design (time-bound, English language only, small number of studies included, summary data analysis only). The Cochrane review (57 RCTs, n = 3002 participants) found that the potential benefit of high-intensity SLT was confounded by a significantly higher dropout rate from intensive SLT groups. 41 Tailoring to the individual is considered essential,65 but is based on best practice recommendation. High-quality information from large-scale research projects does not currently exist to guide therapists’ choice of an effective therapeutic approach best suited to a specific patient with aphasia (and their families).

Data sets capturing a large number of people with aphasia after stroke are typically based on single-centre data with relatively small numbers of individual participant data (IPD). In studies for which larger numbers of IPD are available (n > 200), the quality of information on language outcomes is often restricted, informed by a language data item on a generic stroke scale or screening tool gathered by non-language specialists2,3,29 and lacking a full description across a range of language skills.

Availability of speech and language therapy for people with aphasia

Information on clinical services for people with aphasia after stroke is difficult to gather. A 2003 UK survey found that therapists providing adult services reported that only one-quarter of their time was spent on aphasia. Of available SLT time for aphasia, just 5% was spent on interventions with an intensity of > 3 hours weekly. 66 Half of SLT time was invested in interventions at an intensity of < 3 hours weekly. 66 An earlier survey reported in 2000 indicated that the average total duration of SLT interventions for people with aphasia ranged from one to five sessions in the UK and Australia. 67 North American SLT interventions ranged from 16 to 20 sessions. 67 Since then, hospital-based SLT in Australia may have increased, with a 2009 survey describing typical intervention intensity of 4 hours weekly (2 hours weekly in acute settings). 68 Information about the SLT received by 278 people with aphasia from across the UK prior to their involvement in the Big Clinical and cost-effectiveness of aphasia computer treatment versus usual stimulation or attention control long term post-stroke (Big CACTUS) trial69 showed that, in the first 3 months from the index stroke and pre recruitment, participants had received a total dosage of 11 hours of SLT, delivered at an intensity of approximately 1 hour weekly. People who were ≥ 5 years after the index stroke received a total dosage of 2.75 hours over the preceding 3-month period, delivered at an intensity of 45 minutes monthly,69 even lower than the 1 hour weekly reported for people with aphasia in the Australian community. 68

National audits, although capturing information on access to clinical services, rarely distinguish between time spent with a speech and language therapist for dysphagia (swallowing problems) and aphasia. 2 The Sentinel Stroke National Audit Programme (SSNAP) does, however, highlight the lack of resources for SLT services, a sentiment echoed by therapists. 70 The 2016–17 SSNAP data71 highlighted stroke specialist services with no speech and language therapist, whereas others had a very limited service (one or less full-time equivalent therapist), leaving service provision for aphasia vulnerable should a therapist be on leave, move post or retire.

There is some suggestion, however, that standard SLT services may be changing in some centres. Usual care provided to participants with aphasia in a recent Australian Phase III trial72 found that the usual care arm provided early SLT intervention to 81% of participants, compared with just 12% in the Phase II trial. 56 Similarly, the Phase III trial72 usual care participants received more frequent SLT, and a SLT dosage of 9.5 hours over the first 7 weeks (approximate intensity of 1.5 hours of SLT weekly), compared with an average intensity of 14 minutes of SLT weekly in the Phase II trial. 56 Although the triallists acknowledge that this could be related to the sites simply being part of a trial (the observer effect), it may also suggest a shift in usual SLT interventions over time in those centres. 56,72

Types of studies

Data sets with IPD on people with aphasia after stroke were sought. We particularly looked for IPD data sets from RCTs, recognising their rigour and robustness for our planned analyses, by using a RCT-optimised search strategy. However, other study designs that met our eligibility criteria were also included, for example non-RCT, cohort and case series studies,89,90 as well as clinical and research registries (Box 2). Non-RCT designs (including cohort and case series) were included in covariate-adjusted analyses of recovery profiles and predictors of prognosis after stroke. 89,91 We applied no restrictions to study design, data set age, country, minimum number of time points, duration of follow-up, language of data collection or publication.

Types of study designs included

The primary research studies included in the RELEASE study were categorised according to study design:

• RCTs – participants were randomly allocated to study groups that typically differed in the intervention received [or received a control (or no) intervention].

• Non-RCT studies – participants were recruited but random allocation was not applied to group selection by the primary research team.

• Cohort/case series – data sets from a single participant group that did not compare different SLT interventions.

• Registries – clinical or research databases holding information (e.g. demographic, stroke, aphasia and SLT intervention details) on people with aphasia after stroke, which was not directly gathered to address a single research question.

Interventional or non-interventional

Studies were also classified according to whether they were (1) interventional, whereby SLT was provided in the context of the study, or (2) non-interventional, whereby no SLT was provided in the context of the study.

Crossover data sets

We identified crossover data sets (crossover RCTs or non-RCTs) and carefully extracted all data up to the point of crossing over.

Social support and stimulation of communication

Social support and stimulation interventions provided informal support and functional language stimulation to the person with aphasia, but did not include targeted therapeutic interventions that aimed to resolve participants’ speech and language impairments. Historically they have been used as an attention control in trials of SLT for aphasia or as an adjunct to SLT providing an opportunity to apply new learning in functionally relevant communicative situations.

Conventional speech and language therapy

Interventions described as ‘conventional’, ‘traditional’, ‘standard’, ‘typical’, ‘as directed by the therapist’ or ‘usual’ SLT (and for which absence of detail prevented more specific categorisation) were referred to as ‘conventional SLT’. Such interventions were often delivered in a comparative group or clinical registry. We recognise that conventional therapy in one context may not be comparable with another.

Speech and language therapy co-interventions

We did not explore the impact of pharmacological or neurostimulation (e.g. magnetic or electrical stimulation such as transcranial direct-current stimulation) co-interventions, which are not typically within the remit of therapists in the UK. Examining the contribution of these co-interventions to language recovery among people with aphasia was beyond the scope and remit of the RELEASE study. We extracted information about the presence of such co-interventions and the timing of delivery and noted this in the data analysis.

Main outcome

The main outcome was change from baseline to follow-up in language performance (language recovery) on formal measures of language ability across six prespecified domains (overall language ability, auditory comprehension, spoken-language production, reading comprehension, writing and functional communication) (Box 3).

Given the multinational and multidisciplinary nature of this study, we did not predefine a list of all relevant language assessments and language versions of those measures. Instead, we established eligibility criteria that included any language assessment that (1) captured an outcome of relevance to this study, (2) was published and (3) met with the approval of the RELEASE study collaborators. Agreement was achieved on all measures included and their alignment with specific language domains. We excluded screening tools owing to their variable psychometric properties and lack of sensitivity (due to ceiling effects), and the subsequent impact that this would have on our analysis. 97

Data collection time points

We extracted data on participants at recruitment or entry to the primary research study and at all available time points following this. When follow-up data were available, the time of follow-up was recorded in a continuous fashion in days since primary study baseline. When time to follow-up was given in a category (e.g. 3–6 months), the mean of this time span was taken (i.e. 4.5 months in this example) and converted to days.

Electronic search methods

We searched the following databases from inception to September 2015 using a RCT-optimised search strategy: Cochrane Stroke Group Trials Register, Cochrane Central Register of Controlled Trials and other Cochrane Library databases (Cochrane Database of Systematic Reviews, Database of Abstracts of Reviews of Effects and Health Technology Assessment Database), MEDLINE, EMBASE, Cumulative Index to Nursing and Allied Health Literature, Allied and Complementary Medicine Database, Linguistic and Language Behavior Abstracts and SpeechBITE (see Appendix 1). We also reviewed all studies included in and excluded from a 2016 systematic review. 41 We imposed no language restrictions.

Study records

All study records identified in the searches were screened for eligibility for inclusion. When references described potentially eligible data sets, we invited the primary research teams to confirm eligibility and (where appropriate) to contribute their IPD data sets to our RELEASE study database. We also extended an invitation to contribute through our pre-existing networks, including the Collaboration of Aphasia Trialists [www.aphasiatrials.org (accessed 5 June 2020)].

Selection of studies

One researcher (LRW or KVB) screened titles and abstracts of the records identified using the methods described above and applied the inclusion criteria. Obviously irrelevant titles or reports were excluded. Full-text copies of all the remaining data set reports were reviewed by an independent researcher (KVB, LRW or MCB). Eligibility criteria were ascertained through review of published reports for each study. When published reports were unavailable (for recently completed studies, clinical registries or similar), we clarified eligibility through discussion with the primary research team. Any disagreements about eligibility were resolved through discussion with a third reviewer.

Data encryption and storage

All data contributed to the RELEASE study database were stored at the Nursing, Midwifery and Allied Health Professions (NMAHP) Research Unit at Glasgow Caledonian University (GCU), UK, according to the university’s information systems policy. Original primary research data sets and any additional study data were stored on a secure hard drive [a 1 terabyte IronKey™ Enterprise H300 password-protected and encrypted external hard drive (Kingston Technology Corporation, Fountain Valley, CA, USA)] with centralised management. The hard drive was kept in locked storage in a password-protected room. These data were available as data descriptors on the GCU hard drive, which restricted access to the GCU-based project management team only. The primary research data sets were preserved unmodified. Anonymised, password-protected working copies of these primary research data sets were created and stored on the NMAHP Research Unit’s designated portion of the university server. Adjustments to these working data sets did not affect the original primary research data.

Security and access

Primary research IPD data sets, data sets created for this project and analyses data sets were accessed only by the RELEASE study project management team at GCU. 100 Data were not accessible to those outside the project collaboration. The project management team and collaborators governed use of the research data, and participated in data analysis plans, interpretation of the findings and preparation and review of the manuscripts based on the planned analyses. All collaborators had an opportunity to contribute to the planned secondary analyses [which were conducted in SAS^® version 2.8 (SAS Institute Inc., Cary, NC, USA)].

Data preparation for analysis

We undertook a rigorous process of preparing the IPD data sets for analysis to ensure that the data were valid, as complete as possible, consistent and uniform.

Quality assurance and verification

Each data set was cross-checked against the primary source, such as published papers, protocols and other outputs. We checked within data sets for duplicate participant data sets. We also identified duplication of participant demographics whereby participants taking part in one study went on to participate in a subsequent study led by the same research team, or were reported across dual publications (e.g. as a study group in one study but a comparison group in another). Duplicate participants were identified through (1) careful review of the supporting literature and study materials across primary research teams, (2) checking demographic data across IPD data sets contributed by the same primary research teams and (3) when possible, correspondence with the primary research team. Duplicate entries and any risk of double-counting participant data were excluded from the analysis. This was achieved by removing duplicate participant data sets from one of the data sets, and removing duplicate demographic data prior to any overview in which the participant descriptors would be double-counted. Other data were considered on a variable-by-variable and analysis-by-analysis basis.

Variable labels for data sets collected in non-English languages were translated by the contributing primary research teams. We confirmed the accuracy and completeness of translations with contributors. Following data extraction, the representative from each primary research team was sent an overview of the data items and details relating to their contributed data set. They were asked to confirm that the information was accurate and complete.

Data cleaning and mapping to unit of analysis

Data were contributed in a range of different formats such as Microsoft Excel^® (Microsoft Corporation, Redmond, WA, USA), SAS and Statistical Product and Service Solutions (SPSS) (formerly Statistical Package for the Social Sciences) [IBM Corporation, Armonk, NY, USA (IBM SPSS Statistics from version 19 onwards) or SPSS Inc., Chicago, IL, USA (version 18 and below)]. In the first instance, we cleaned all data sets to reclassify duplicated variable names, removed blank lines and transposed data into a one row per participant format (short form). We converted all data sets into SAS. Prior to analysis, data relating to prespecified variables needed for the RELEASE study analyses (see Box 3) were cleaned. These variables were decided a priori by the research and analysis team and presented to the wider RELEASE study collaboration for review and comment. We ensured that the variable entry matched an agreed uniform format to inform our planned analyses. For example, time since stroke was recorded in days, months and years across RELEASE data sets; the RELEASE collaboration agreed to record ‘time since stroke’ in the format of ‘days since stroke’. When necessary, entries as years or months after stroke were converted to days. In some cases, the record format was a date entry from which days since stroke could be calculated when date of study entry was considered. For each variable of relevance, the primary research data sets and other data sources (publications, protocols, data dictionary) were searched for that variable or indicative surrogate measures. Decisions on how to synthesise these data into a common nomenclature were based on agreement of the wider collaboration. A decision tree was used to ensure standardisation of variable names and ranges when converting data from the original data sets to the RELEASE study analysis data sets.

For some variables, the relevant data were common to all participants across a data set and not included in the contributed electronic IPD data set. For example, if the primary research entry criteria specified right-handed participants, then all participants in that data set were right-handed, whether or not this was recorded in the submitted electronic data set. Similarly, when a specific intervention was delivered to one randomised group, then (unless otherwise specified) the intervention description was common to all participants in that group. In such cases, we imputed and translated those data using one-to-many merging and in keeping with the agreed format for the RELEASE study analyses.

The collaboration then agreed on standard definitions to describe therapeutic interventions in terms of the theoretical and target approach (see Box 3). These definitions were applied to all intervention studies. In some instances, those participants who were allocated to no SLT were permitted to receive usual care SLT after the study’s active intervention period had finished. In these cases, we identified the follow-up times that corresponded to the duration of no SLT and took those observations forward for analysis in the no-SLT group. We also identified cases of usual SLT care, study SLT intervention and the corresponding follow-up time points associated with each. We identified the presence of crossover in treatment and took all measurements up to the point of crossover. These data were identified, cleaned to a standardised format and entered into the analysis data set.

According to the needs of the analysis, data-cleaning processes included examining the data set for validity of contents, such as receipt of all relevant participants’ data; matching numbers of IPD transferred with those reported in publications; matching variable names to known published assessments and ascertaining any missing domains of assessment from the original data set; examining the content of fields to ensure that they matched prescribed minimums and maximums for known assessments; ascertaining any transformations of data that had occurred prior to transfer and resolving such queries [e.g. use of a mix of t-scores and raw scores in the Aachen Aphasia Test (AAT) across studies]; examining odd values in a given range (e.g. data sets containing a zero value for assessments when this indicated that no assessment had been conducted); and converted data into a numeric format.

Anchor and minority measures

We prespecified several language domains or outcomes for analysis in the RELEASE study:

• overall language ability

• auditory comprehension

• spoken-language production (including naming)

• reading comprehension

• writing

• functional communication.

For each of these outcomes, a range of assessments (sometimes in various language adaptations) was used. To synthesise the IPD from across these multiple measures, we first identified the most commonly used measurement for each outcome. We profiled all measures used to capture data on that outcome, the number of studies using that measure and the number of IPD available, and calculated the median score, interquartile range (IQR), minimum and maximum for each measurement tool. Language or version variations were treated as separate tools.

An ‘anchor measure’ was identified for each language domain (e.g. reading) as the measurement instrument used by the most data sets in the RELEASE study database. When this approach resulted in two possible anchor measures, the assessment that had the greater number of IPD in that language domain was identified as the anchor measure. All other assessment tools measuring the same domain, but which were informed by smaller numbers of data sets, were considered ‘minority measures’. Minority measures were then transformed to match the format and range of each anchor measure (one for each language domain) through an adapted version of the method previously used in large-scale data synthesis (see Appendix 2, Box 4).

Meta-analysis and network meta-analysis

Meta-analyses of aphasia intervention research already exist and have provided evidence-based review and data synthesis of the clinical effectiveness of SLT for aphasia after stroke (e.g. Brady et al. ,41 Robey63 and Bhogal et al. 64). Methodologically, these meta-analyses are usually based on the data synthesis of aggregate group-level summary data. Large group summary approaches taken by RCTs and meta-analysis of RCTs have limitations, particularly in relation to the risk of ecologic bias; aggregate data generated across a complex participant population, such as people with aphasia, risk concealing the possibility of varying individual response to a complex rehabilitation intervention. 76

Traditional meta-analysis is limited to direct between-treatment (or control) comparisons and group-level descriptors of interventions and aggregated group data on the treatment effects. 103 A more detailed examination of the effectiveness of a specific therapy component is not possible owing to the variability of complex SLT across data sets. Similarly, such approaches cannot accommodate or control for participant demographics or other factors relevant to recovery. 41 Network meta-analysis approaches, however, allow for an estimate of the difference between two (or more) treatments that are not directly compared. For example, studies comparing A versus B or A versus C (direct comparisons) can be combined to give an estimate of B versus C (indirect comparison). 104

Individual participant data network analysis approach

Network meta-analysis can be applied to aphasia intervention research using group-level summary statistics,38 but this approach does not address the limitations of ecological bias, highlighted in the previous section. Confounding is also a risk. We conducted a network meta-analysis based on a large IPD database, which allowed us to explore the subtleties of a highly heterogeneous group of people with aphasia after stroke, control for individual predictors and support the detailed exploration of the influence of participant-level covariates on SLT treatment effects across a range of language-specific measures. 104 Meta-analyses based on aggregated data are typically unable to examine such interactions. In addition, we were aware of some studies that had recorded valuable information on SLT intervention at IPD level.

One-stage versus two-stage approach

There are two main approaches to an IPD network meta-analysis: a one- or two-stage approach. 104 Two-stage approaches are very similar to the standard meta-analysis approach, working first with IPD to centrally generate aggregate data for each contributed data set (rather than using the primary research team’s reported summary statistics). These aggregate data could then be meta-analysed with summary statistics from other trials for which only aggregate data were available. This approach can, however, lead to bias in effects, greater heterogeneity and lower power to detect associations between language outcomes and continuous variables. 104 It is also very similar in methodology to existing reviews and meta-analyses of summary data.

In contrast, a one-stage approach would allow all available IPD from across the contributed data sets and IPD retrieved from the public domain, filtered for relevance to each analysis, to be combined into a single model (which considered the clustering of IPD within a study). One benefit of this methodological approach is that it can address confounding by enabling the examination of the impact of several participant and language variables on an intervention effect at the same time.

Network meta-analysis

We performed our one-stage network meta-analysis in SAS, using PROC MIXED, with fixed demographic and treatment effects, and study as a random effect. The variance structure was allowed to be unstructured. Study homogeneity was assessed by examining the variance parameters. Network diagrams for each treatment factor were drawn up. Each comparison within each domain was illustrated by a network graph, and supplemented by a table of the number of available IPD in each RCT and treatment node. We used the graphs to explore the balance of the networks and whether or not all levels in a treatment were compared across and within studies. Each comparison in each domain was illustrated by a network graph, and supplemented by a table of the number of available IPD in each RCT and treatment level. Feasibility was contingent on the number of trials and IPD available.

The populations in each data set were expected to differ, and so a base model of age, sex, time since stroke and baseline score was created for each language domain. We also planned to examine interactions between demographic characteristics and treatments, should sample sizes be sufficient. Given the number of tests likely to be undertaken, we were aware of the high risk of false positives. However, the aim of the RELEASE study exploratory analysis was not to generate definitive answers, but to use inference to suggest priority avenues for future research; therefore, we were more concerned with estimated effect sizes than statistical significance.

Primary analysis

We used RCT data as the gold standard for each planned analysis, with only participants from RCTs included in the primary analyses. Analyses that included data from non-RCTs, cohort/case series studies, observational studies, registries and other study design types (as relevant) followed the primary analysis, to judge whether or not they supported the RCT findings (see Appendix 3, Figure 26). When they did not agree, the data were further explored to quantify this difference.

To maintain the independence of subjects, we ensured that IPD did not appear in more than one data set for intervention type comparisons, although they could appear in more than one data set when grouped by language domains (e.g. overall language ability, auditory comprehension, reading) (see Appendix 3, Figure 26).

Natural history of language recovery

We defined the natural history population as a population for whom no active SLT intervention was present, and for whom there was no history of previous SLT. People who were allocated to the no-SLT group in a study were typically permitted to receive usual care SLT after the intervention period of the study had been completed. Any SLT intervention could have an impact on the post-intervention follow-up measures. Depending on the time since the onset of their aphasia, they may also have had historical SLT prior to entering the study as part of usual care. We therefore implemented the following restrictions when extracting eligible IPD to support the examination of the natural history of aphasia recovery:

1. Allocation to the no SLT group in an intervention study.

2. Enrolment to the study with baseline assessment commencing within 15 days of index stroke, to reduce the possibility of unreported initiation of SLT as part of usual care.

3. No history of prior stroke.

4. When longer-term follow-up measurements were available, we reported on the first post-intervention follow-up only, when applicable, to eliminate all instances of access to SLT as part of usual care SLT after the study intervention period, as identified in our data extraction.

We plotted baseline language domain scores along with corresponding language domain values that were available at first follow-up, to examine the trajectory of the natural history of language recovery over time in those participants who were allocated to no-SLT groups.

Language recovery in the context of speech and language therapy

We also examined the trajectory of language recovery among participants who had access to SLT; we identified those who had access to SLT either as a study participant (SLT group) or reported (or reasonable grounds to assume) historical access to SLT as part of standard health care prior to study enrolment (historical SLT group). We also identified those who were allocated to the historical SLT group (with enrolment beyond 15 days post stroke to exclude those who were part of the natural history analysis) and assessed whether or not there were sufficient data available at the first follow-up period. We sought to eliminate the impact of any SLT intervention (even usual care) in any follow-up data collection time points.

We described baseline and follow-up scores in all available groups for each language domain, when sufficient data were available. Data were presented as plots of language domain scores over time (relative to study entry) and, when applicable, stratified by SLT, no SLT or historical SLT intervention.

Language recovery: change in language scores from baseline by time since stroke

We examined the role that time since index stroke and study enrolment played on the proportion of recovery observed for each domain score. The study enrolment period in the primary data set was stratified into the following groupings:

• 0–1 month

• > 1–3 months

• > 3–6 months

• > 6 months following index stroke.

We ascertained whether or not participants were allocated to receive a SLT intervention; when possible, we described the absolute proportion of change observed for SLT versus historical SLT (excluding those who were assessed as part of the natural history analysis). We calculated the proportion of change (%) in each language domain score from baseline to first follow-up and plotted these proportion changes by study enrolment period (as categorised earlier). This was described for the RCT population, and then for all study types. We plotted each language domain separately to visualise the proportion of change observed in each domain at different study entry time points. We opted not to stratify data according to baseline severity stratum in combination with aphasia chronicity, as this would have resulted in very small strata and much wider confidence limits, thus precluding informative analyses. We also accounted for the skewness of data measuring language impairment by opting to present medians and IQRs, rather than means and standard errors of the means.

Distribution of language scores at study entry by time since stroke

All participants were selected from the total population of people with aphasia after stroke. We identified all data available at study entry for each language domain. This time point was selected as all participants were assessed prior to receiving any study intervention. We then ascertained the time of assessment relative to stroke onset. We plotted all transformed language domain observations available at study entry against time of study entry, relative to stroke onset, for each language domain. We then fitted a regression line to these observations and conducted linear regression analyses to describe whether there was a significant trend towards increased or decreased language domain scores with later time to study entry. We considered whether or not the baseline data reflected the trajectory findings based on participants’ language scores over time (see Chapter 4). We considered the distribution of language scores at clinically relevant time points (e.g. study entry between 0 and 1000 days since stroke). Finally, we considered whether or not there was any suggestion of selection bias in recruitment to aphasia studies based on participants’ language scores over time since stroke.

Factors associated with language score changes: age, sex and time since stroke

We performed regression, modelling a trend towards increasing proportion of change, and including age, sex and time since stroke (days) as covariates in this model, to ascertain whether or not there was a relationship between age, sex and time since index stroke on the degree of recovery observed.

Distribution of language scores at study entry: age, sex and living context

Building on the earlier analyses, we identified all participants’ data from across all data sets for the specified language domains at study entry in advance of any study-level intervention. Using the data on time since stroke, we plotted the language scores and explored their distribution across different age groups (≤ 55, 56–65, 66–75 and > 75 years), sex and living contexts.

Beneficial speech and language therapy approaches to language recovery for participant subgroups

We explored whether or not some interventions (or intervention components) may be more beneficial for some participant subgroups than others. When demographic/stroke variables were significant to the model, a series of one-stage subset analyses were carried out to explore whether or not certain participant groups responded better to certain therapy interventions (see Appendix 3, Figure 26d). We also examined participant clustering within trials, which would allow us to distinguish individual participant-based interactions from any data set-based interactions. 103

Selection bias

We incorporated a systematic review component as the first stage to building our IPD database and actively invited data set contributions from primary research teams that we had no prior collaboration with. We also identified and extracted IPD reported in the public domain, including the grey literature, from primary research teams with which we had no contact. We considered the possible impact of potentially eligible but unavailable IPD data sets. 106 When relevant, we tested the comparability of groups at baseline.

Publication bias

Through our rigorous approach to identification and selection of eligible data sets, we identified IPD that were conducted across a range of languages, unpublished or reported in the grey literature, thus reducing the risk of reporting bias. 107 Where the limited availability of suitable data sets rendered a funnel plot unfeasible, we examined risk of publication bias by looking at the distribution between sample size and those studies that reported significant findings. It was not possible to look at publication bias in the unpublished registry data sets.

Availability bias

We considered the impact of potentially eligible data sets that were unattainable, unavailable to our study or for which the researchers were uncontactable.

Other biases

We considered the possible risk from other sources of bias, including relevance of the primary research study’s objectives. We used the ROBIS tool to consider to what extent the RELEASE study meta-analysis and network meta-analysis were at risk of bias. 105

Missing completely at random

Data were examined for patterns of loss or whether or not the data were missing completely at random by recoding the variable of interest into ‘missing’ or ‘not missing’. This variable was compared with demographic variables that may have had an influence on recording that information (e.g. participants’ age, sex, stroke), and then with the other variables available. We used the t-test for comparisons of missing data with continuous variables such as age (or the Mann–Whitney U-test, depending on the normality of the distribution). When comparisons involved categorical variables, we used the chi-squared test to indicate whether or not the missing data were substantially different from the data reported (not missing). When there was no evidence of an influence, the data were treated as if assumed to be missing completely at random. Data missing at random were excluded from our modelling and analysis.

Choice of language measurement

We explored the impact of our prespecified choice of assessment tool in situations in which a data set had two or more assessment tools eligible for synthesis in a single language outcome. We considered the impact on findings had the alternative assessment been selected to go forward to data synthesis.

Exclusion of minority measures

We considered what impact the exclusion of the minority measures had on the findings. To examine the contribution that each language assessment tool made to the transformed language measures, we plotted the distribution of transformed language measures over time, stratified by the original assessment tool and presented by language domain.

Random or fixed effects

We examined the impact of including the data set identifier as the random effect or as a fixed effect using the Wu–Hausman test. 108 For simplicity, the comparison was based on a simple model of language change by baseline. The model was run twice, once with data set as a fixed effect and once with data set as a random effect. The differences between the estimates of the covariate and the standard error of the baseline score were used to calculate both the Wu–Hausman and the alternative Wu–Hausman for one parameter. If the W-score was < 3.841 (≈ χ² with 1 degree of freedom), or the alternative W was < –1.96 or > 1.96 [≈ N(0,1)], then we rejected the null hypothesis that random effects were more efficient.

The inclusion of historic data sets

We explored what bias there may have been in the inclusion of historic data sets (e.g. collected prior to the establishment of specialist inpatient stroke units), when approaches to the provision of stroke care differed. 109,110 Response to treatment may be influenced by overall health and improvements in stroke care, such as thrombolysis, which has been linked to better aphasia recovery. 29 Similarly, a response to SLT for aphasia depends on the therapy provided, and such approaches can change over time (as suggested by Godecke et al. 56,72). We conducted a sensitivity analysis (RCT data sets only) to consider whether or not the recruitment date had any impact on recovery patterns by language domain.

Excluding non-randomised controlled trial data sets

We explored the impact of excluding data sets graded as being at a high risk of bias, in particular the non-RCT data sets, from the analysis. In each analysis (except the planned subgroup analysis; see Subgroup analysis), we analysed and presented the IPD gathered in the context of RCT studies before then conducting the analysis on the wider data set inclusive of non-RCT designs.

Registry data sets and impact on findings

We explored the impact and contribution that registry data sets (clinical and other research archives contributed to the RELEASE study) made to the analysis and considered whether or not these data sets may have altered the findings.

Excluded data sets

We excluded 433, out of 1131, full texts that failed to meet our eligibility criteria; 77 data sets had fewer than 10 IPD; four data sets had qualitative data only; 272 data sets were not about aphasia, stroke or language impairment; and 52 were reviews, secondary data analysis or commentary pieces (see Figure 3).

Unavailable data sets

We were unable to confirm whether or not the 267 remaining data sets identified as potentially eligible (reported across 399 records) met the RELEASE study inclusion criteria. Following careful review, it is possible that 53 were RCTs and 214 were a non-RCT design (see Figure 3). Of the 53 RCTs, 46 included, but did not necessarily evaluate, a SLT intervention.

Source

The date of participant recruitment was available for 79 out of 174 (45.4%) primary research studies (Table 2). These data sets reflected participants recruited from 1973 to 2016, although most were recruited since 2000 [67/79 (84.8%)]. Recruitment dates were unavailable for 95 out of 174 data sets (54.6%), particularly those from the public domain. We considered date of publication as a surrogate indicator of data set age [161/174 data sets (92.5%); 4627/5928 IPD (78%)]. Thirteen data sets did not have an associated publication, including recently completed research or clinical registries. Of the 161 data sets that were published between 1973 and 2017, most [135/161 (77.6%)] were published since 2000.

Data set design

One-quarter of the primary research data sets included were RCTs [47/174 data sets (27.0%); 1778/5928 IPD (30.0%)]. The remainder were case series or cohort studies [104/174 data sets (59.8%); 2886/5928 IPD (48.7%)] and non-RCTs [18/174 data sets (10.3%); 411/5928 IPD (6.9%)]. Five data sets were derived from registries [5/174 data sets (2.9%); 853/5928 IPD (1.4%)]. Approximately half the data sets included a SLT intervention [91/174 data sets (52.3%); 2834/5928 IPD (47.8%)], whereas the remainder were non-interventional in nature.

In 20 data sets (11.5%; 311 IPD), participants crossed over from one intervention to another. An additional five data sets (2.9%; 240 IPD) delayed the start of treatment for a group. Seven data sets (4%; 128 IPD) allocated participants across three arms, and three data sets (1.7%; 84 IPD) allocated across four arms.

Some participants contributed to more than one data set, in some instances, participating in a follow-on study conducted by the same research group. In such cases, their demographic details were counted only once in our data set (e.g. Table 3). For other participants their language performance data had been duplicated across data sets. We identified these duplications and removed them from our database.

People with aphasia

Of the 5928 people with aphasia following stroke represented in the RELEASE study database, 61.4% were male (3407/5928 IPD) and 38.6% (2143/5928) were female. They had a median age of 63 years (IQR 53–72 years). Most were right-handed [3719/3879 IPD (95.9%)]; the remainder were left-handed [133/3879 (3.4%)] or ambidextrous [27/3879 (0.7%)]. The median number of years of formal education, for those participants who had such information recorded [84/174 data sets (48.3%); 3125/5928 IPD (52.7%)], was 12 (IQR 10–16 years). Ethnicity was recorded for one-quarter of participants [1475/5928 IPD (24.9%)] across 24 out of 174 data sets (13.8%). Three-quarters of participants were from a White ethnic background [1116/1475 IPD (75.7%)]; the remainder were recorded as being of Asian [248/1475 (16.8%)], Black [67/1475 (4.5%)], Latino/Hispanic [9/1475 (0.6%)], Mixed [2/1475 (0.1%)] or Other [33/1475 (2.2%)] ethnic backgrounds (see Table 3).

Intervention context

Most participants receiving an intervention lived at home with others [195/2330 IPD (8.4%)]. Interventions were typically delivered in a formal health-care setting [1866/2330 IPD (80.1%)] on a one-to-one [1613/2330 IPD (69.2%)], face-to-face [1957/2330 IPD (84.0%)] basis, where therapy was provided by a professional [1871/2330 IPD (80.3%)] (see Table 6).

Intervention description

Therapy targeted a range of language impairments including spoken language and/or naming [734/2330 IPD (31.5%)], auditory comprehension [68/2330 IPD (2.9%)], mixed spoken language and auditory comprehension [583/2330 IPD (25%)] and reading [10/2330 IPD (0.4%)]. No intervention specifically targeted the rehabilitation of writing. Most of the spoken-language interventions targeted improvements on naming (also known as word-finding) therapy [489/2330 IPD (21%)].

A range of theoretical approaches were identified, including semantic [34/2330 IPD (1.5%)], phonological [124/2330 IPD (5.3%)], semantic and phonological [260/2330 IPD (11.2%)], constraint-induced aphasia therapy [113/2330 IPD (4.8%)], melodic intonation therapy [61/2330 IPD (2.6%)] and functional–pragmatic therapy approaches [246/2330 IPD (10.6%)].

Intervention regime

We gathered data on the frequency [2057/2330 IPD (88.3%); 66/67 data sets (98.5%)], duration [1960/2330 IPD (84.1%); 64/67 data sets (95.5%)], intensity [1882/2330 IPD (80.8%); 60/67 data sets (89.6%)] and dosage [1978/2330 IPD (84.9%); 62/67 data sets (92.5%)] of the SLT interventions. Information on whether or not the SLT intervention had been tailored by difficulty of the tasks was available for 1145 out of 1373 IPD (83.4%) [34/67 data sets (50.7%)]; information on whether or not the SLT intervention had been tailored by functional relevance to the individual participant was available for 649 out of 928 (69.9%) IPD [22/67 data sets (32.8%)]. Whether or not participants had been prescribed home-based practice tasks was available for 1306 out of 2330 IPD (56.1%) [31/67 data sets (46.3%)] (see Table 6).

Risk of bias

The sample size of the primary research data sets ranged from 10 IPD of people with aphasia after stroke [25/174 data sets (14.4%)] to one data set with 420 IPD. An overview of each contributing data set can be found in Report Supplementary Material 1, but a summary is provided in Quality of primary research data sets.

Quality of primary research data sets

Transformations

A total of 64 assessments were used across data sets to capture participants’ language performance. Some assessments comprised several subtests, which also contributed to the data synthesis for other language domains [e.g. Western Aphasia Battery (WAB) data contributed to overall language ability, auditory comprehension, naming and reading outcomes]. Language-specific adaptations of an original assessment were profiled individually to accommodate differences between languages.

We collated and categorised each assessment and their relevant subtests by language domain. For each data set, we noted the scoring format. Data from some tools [e.g. the Token Test (TT)] were reported in raw scores, percentiles, t-scores and error scores and were either positively or negatively scored. All scores were converted to raw scores and positive scoring.

For each language domain, we identified the assessment used by the most data sets as the ‘anchor measure’ (see Appendix 7, Table 14). In most cases, this choice also reflected the assessment with the greatest availability of IPD. The availability of IPD on an assessment contributed to the choice of anchor measure only in situations in which the greatest number of data sets using an assessment was shared across an equal number of data sets.

We examined the contribution that each assessment made to the overall data for that language domain and plotted the distribution of each score at baseline across increasing time since stroke time points and stratified by the original assessment (see Appendix 6, Figure 27). We concluded that the data points generated by each of the minority and anchor measures were valid and within the expected ranges.

The maximum, minimum, median, lower quartile (LQs) and upper quartile (UQs) were calculated for each of the identified anchor measures (see Appendix 7, Table 14). The values for the maximum, minimum, median, LQ and UQ were calculated for each of the remaining assessment tools that captured that language domain (minority measures). These values were then transformed to the anchor measure scale, ensuring that the values remained within the minimum and maximum values of the anchor measure scale (see Appendix 7, Table 14).

For each analysis, we collated the data with other required data items (e.g. participant demographic data items or intervention data), which resulted in a reduction of overall data to inform our analysis. We profile this drop-off in data availability in Table 5.

Selection bias in data set recruitment

The systematic review-based approach to the identification of eligible data sets recruited an additional 122 data sets to those originally identified through our research network alone, yielding important demographic and language IPD. For example, our final database has 5928 IPD on time since stroke (our original recruitment estimate was 3181 IPD) and 2922 IPD on naming outcomes (originally estimated as 2007 IPD). Methodologically, employing a review-informed approach to building the database was essential. It reduced the risk of bias from an analysis based only on data sets contributed through pre-existing networks and increased the feasibility of analysis on a sparse database with limited overlap in outcomes and demographic data across data sets.

Publication bias in data set recruitment

Overall, the database included data sets collected across 28 countries and 23 international languages (see Tables 2 and 4). We included 13 unpublished data sets. We considered the risk of publication bias by examining the proportion of significantly reported results by decade or other suitable period grouping. Any suggested imbalance was tested with the chi-squared test (and is reported in Chapter 8).

Availability bias in data set recruitment

In the systematic review, we identified 399 out of 698 (57.2%) records reporting 267 potentially eligible data sets, which, despite best efforts, were not included (see Report Supplementary Material 2). Of these, 53 data sets appeared to be RCTs, with 46 out of the 53 RCTs including a SLT intervention that was part of the study protocol or standard care. Of the of 47 RCTs that contributed group summary level data to the Cochrane review meta-analysis of SLT for aphasia after stroke,41 we secured IPD for 17. We also secured IPD from five RCTs that were ‘ongoing’ at the time of the Cochrane review41 and from a further four RCTs that were identified but unavailable to the Cochrane review. 41

Trajectory of overall language ability over time

We examined the trajectory of overall language ability over time by looking at scores on aphasia test batteries transformed to the anchor measure [Western Aphasia Battery-Aphasia Quotient (WAB-AQ)], in which higher scores indicated better language performance. We had no data capturing overall language ability over time for participants who had received only historical SLT in the RCT population, meaning we could not conduct comparative analyses with the participants who had SLT during the primary studies on this population. We therefore present the trajectory of overall language ability scores over time for participants allocated to SLT in RCTs (601 IPD; 18 RCTs; Figure 4a) and across all study designs (2345 IPD; 74 data sets; see Figure 4b). In both cohorts, there was a trend towards increased scores on aphasia language batteries over time.

FIGURE 4.

Overall language ability scores over time. (a) RCTs; and (b) all study types.

Trajectory of auditory comprehension over time

In our database, all participants who were allocated to receive no study-based SLT went on to receive SLT after the study intervention period was over. Due to inadequate longer-term follow-up assessments beyond the study intervention period, we could not describe the trajectory of auditory comprehension scores over longer time periods in those allocated to SLT versus those who received only historical SLT. We examined the trajectory of auditory comprehension scores over time in both the RCT population (536 IPD; 17 data sets; Figure 5a) and using all study designs (2463 IPD; 62 data sets; see Figure 5b). The anchor measure was Token Test from the Aachen Aphasia Test (TT-AAT).

FIGURE 5.

Auditory comprehension scores over time. (a) RCTs; (b) all study types; (c) within earlier follow-up times in RCTs; and (d) earlier follow-up times in all study types.

Of 536 IPD from 17 RCTs, 521 IPD (17/17 RCTs) were allocated to receive a SLT intervention, whereas 15 IPD (2/17 RCTs) received historical SLT only (see Figure 5a). Those with historical SLT experienced a decline in their auditory comprehension scores over time. In contrast, those who had study access to SLT experienced an increase in auditory comprehension scores over time. We examined this trajectory more closely by plotting follow-up between the two groups up to 90 days (see Figure 5c).

Despite randomisation, we observed slightly higher initial auditory comprehension scores at baseline in the historical SLT group; this declined over time. Across all primary research study designs (2463 IPD; 62 data sets; see Figure 5b), participants allocated to receive SLT demonstrated an increase in auditory comprehension scores over time. Data from participants who received only historical SLT in this population were available at only three time points and were therefore too sparse to make strong inferences (see Figure 5d).

Trajectory of naming over time

We transformed all naming scores to the Boston Naming Test (BNT) distribution. There were inadequate naming data at follow-up time points immediately after study intervention comparison periods from historical SLT participants in the primary studies (29 IPD; 3 RCTs). We therefore could not compare the trajectory of naming scores over longer follow-up time periods between participants with study access to SLT and participants with historical access to SLT in the RCT population. Instead, we present the trajectory of naming scores over shorter follow-up times (up to 90 days) for participants in RCTs (623 IPD; 22 RCTs; Figure 6a) and across all study designs (2592 IPD; 68 data sets; see Figure 6b).

FIGURE 6.

Naming scores over time. (a) RCTs; (b) all study types; and (c) within earlier follow-up times in RCTs.

Historical SLT participants had higher naming scores at baseline than those allocated to receive SLT. Naming scores appeared to decline over time for the historical SLT participants, whereas scores for the SLT group experienced a steady upwards trajectory over time (see Figure 6a–c). Across all study types, there were limited data on naming scores among the historical SLT participants. For those who had study SLT access, naming scores increased with time (2592 IPD; 68 data sets; see Figure 6b).

Trajectory of other spoken language over time

There were no data available on other spoken-language outcomes for participants who received only historical SLT in RCTs. This meant that we could not describe the trajectory of spoken-language scores among this group or make any comparison with the trajectory of spoken-language scores over time among participants allocated to study SLT. We therefore present the trajectory of spoken-language scores over time [transformed to the verbal (1) subtest of the Porch Index of Communicative Ability (PICA)] for participants allocated to SLT in a RCT (172 IPD; 3 RCTs; Figure 7a) and across all study types (353 IPD; 8 data sets; see Figure 7b). The trajectory appeared to be stable over time in both populations.

FIGURE 7.

Other spoke language scores over time. (a) RCTs; and (b) all study types.

Trajectory of reading comprehension over time

We reviewed the transformed reading comprehension scores [transformed to fit the range of the Comprehensive Aphasia Test (CAT) reading subtest] for historical SLT participants and for those who had access to SLT in the primary studies. All historical SLT participants were permitted access to conventional SLT beyond the active intervention comparison period. Reading comprehension data beyond the immediate study intervention periods were inadequate for comparative meta-analysis. The stable trajectory of available reading comprehension scores over time among participants who had SLT across all primary studies is presented in Figure 8 (650 IPD; 9 data sets).

FIGURE 8.

Reading comprehension scores over time for all study types.

Trajectory of writing over time

Participants with available writing scores who received only historical SLT during the primary studies were typically given access to SLT immediately after the study intervention comparison period. Thus, writing data from later follow-up time points were insufficient to permit comparative meta-analysis by SLT status over time. We report writing data from SLT participants only. Data from three RCTs (41 IPD; Figure 9a) suggest a slight score decline across time. Across all study types (635 IPD; 10 data sets; see Figure 9b), following transformation to the anchor score (CAT writing subtest), we observed stable writing scores over time.

FIGURE 9.

Writing scores over time. (a) RCTs; and (b) all study types.

Trajectory of observer-rated functional communication (activity) over time

For functional communication (activity) over time, we examined 828 IPD from 17 RCTs (Figure 10a). The anchor measure for functional communication domain (activity) was the Aachen Aphasia Test-Spontaneous Speech Communication (AAT-SSC). There were inadequate data on participants who received only historical SLT, which did not permit comparison of trajectories with the SLT group; once again, there were few after-intervention data points and people who received only historical SLT during the study were all permitted to receive SLT after the intervention period was complete. In the SLT group, we observed better observer-rated activity scores over time. This was also evident when looking at all study types (1272 IPD; 28 data sets; see Figure 10b). Data on observer-rated participation measures showed the same patterns for better participation scores over time (see Report Supplementary Material 3, figure 1).

FIGURE 10.

Observer-rated activity scores over time. (a) RCTs; and (b) all study types.

Trajectory of self-rated functional communication (activity) over time

Data were available on self-rated activity scores for 11 IPD (one RCT), and for 43 IPD (two data sets of all study types). The anchor measure for self-rated activity scores was the Assessment for Living with Aphasia (ALA). We observed an increase in self-rated activity scores with time (Figure 11) when using all study designs. Similar observations were seen for self-rated participation scores (see Report Supplementary Material 3, figure 2).

FIGURE 11.

Self-rated activities over time in all study types.

Overall language ability at study entry

Following transformation of overall language ability scores from minority measures to match the range of the WAB-AQ (anchor measure), we plotted language ability scores at study entry for RCT participants (700 IPD; 19 RCTs), and for all participants across all study designs for which overall language ability data were available (2571 IPD; 78 data sets).

Regression analyses based on data from the RCT population revealed that time since stroke at baseline (in months) was positively associated with better overall language ability (OR 1.010, 95% CI 1.006 to 1.014; p < 0.0001) (see Report Supplementary Material 3, figure 3). We observed similar overall increasing language ability with increasing time since stroke among all participants across all study designs (OR 1.003, 95% CI 1.002 to 1.005; p < 0.0001) (see Report Supplementary Material 3, figure 3).

When examining the distribution of overall language ability scores within more clinically applicable study enrolment time frames since index stroke (0–1000 days; 1780 IPD; 78 data sets), we observed a similar trend towards greater overall language ability with increasing time (in months) since stroke at baseline (OR 1.026, 95% CI 1.017 to 1.035; p < 0.0001). A 1-year increase in time from stroke to study enrolment was associated with an improvement in overall language ability scores (OR 1.36, 95% CI 1.22 to 1.51; p < 0.0001).

Auditory comprehension at study entry

Auditory comprehension scores were transformed to fit the range of the anchor measure (the TT-AAT). We plotted these transformed scores at study entry for the RCT population (633 IPD; 19 RCTs; see Report Supplementary Material 3, figure 6) and across all study designs (2597 IPD; 67 data sets; see Report Supplementary Material 3, figure 6).

In the RCT population, auditory comprehension scores appeared to be stable with increased time to study entry. We examined whether or not the slope of the regression line was influenced by the small number of baseline data from later time points by examining the distribution within a more clinically applicable enrolment period of up to 1000 days since stroke in RCTs (506 IPD; 19 data sets; see Report Supplementary Material 3, figure 6). We observed a stable trend in auditory comprehension scores as time since stroke increased (see Report Supplementary Material 3, figure 6). This association was not significant when all baseline time to inclusion was considered (OR 1.001, 95% CI 0.996 to 1.005; p = 0.7312) nor when time to inclusion was limited to 1000 days after stroke (OR 0.999, 95% CI 0.980 to 1.019; p = 0.9571).

We then examined the distribution of auditory comprehension scores at increasing time to enrolment in all study types (see Report Supplementary Material 3, figure 6). In contrast to the RCT data set, we observed a slight increase in auditory comprehension scores at baseline with increasing time since stroke (2597 IPD; 67 data sets; see Report Supplementary Material 3, figure 6). Regression analysis showed no significant association between time to enrolment (in months) and auditory comprehension at baseline (OR 1.001, 95% CI 0.999 to 1.002; p = 0.3763). We examined the distribution of baseline auditory comprehension scores further, looking at clinically applicable enrolment periods of 0–1000 days since stroke for all study types (1879 IPD; 66 data sets; see Report Supplementary Material 3, figure 6). There was no association between time since stroke at baseline (months) and auditory comprehension scores for the 0–1000 days since stroke (OR 1.008, 95% CI 0.999 to 1.017; p = 0.0863).

Naming scores at study entry

We described the distribution of naming scores at study entry in RCTs (748 IPD; 25 RCTs; see Report Supplementary Material 3, figure 9) and all study designs (2760 IPD; 72 data sets; see Report Supplementary Material 3, figure 9). In the RCT population, naming scores appeared to be stable with increasing time from stroke to study entry (OR 1.00, 95% CI 0.996 to 1.003; p = 0.95). In contrast, for all study designs, an increase in baseline time (in months) was associated with increased naming scores (OR 1.005, 95% CI 1.004 to 1.007; p < 0.0001). A similar trend towards better naming scores with increased time from stroke to study enrolment (months) was observed within a clinically applicable period of 0–1000 days (1836 IPD; 71 data sets; see Report Supplementary Material 3, figures 9–11) (OR 1.031, 95% CI 1.022 to 1.040; p ≤ 0.0001).

Other spoken language at study entry

We described the distribution of other spoken-language scores at study entry for participants who were enrolled in a RCT (163 IPD; 2 RCTs; see Report Supplementary Material 3, figure 12) and across all study designs (344 IPD; 7 data sets; see Report Supplementary Material 3, figure 12). We observed relatively stable spoken-language scores among the RCT population as study entry time increased (OR 0.999, 95% CI 0.992 to 1.005; p = 0.69). There appeared to be relatively stable spoken-language scores with increasing time to study enrolment among participants from all study types (OR 1.002, 95% CI 0.998 to 1.007; p = 0.26) (see Report Supplementary Material 3, figure 12). Similar patterns were observed when examining a clinically applicable study enrolment period (0–1000 days; see Report Supplementary Material 3, figure 12).

Reading comprehension at study entry

We plotted the distribution of reading comprehension scores at study entry. These scores were transformed to fit the range of the CAT reading comprehension subtest. Data from the RCT population were available for only 11 IPD from one RCT, and therefore did not meet our prerequisites for analysis. We examined data at study entry from all study types (652 IPD; 9 data sets; see Report Supplementary Material 3, figure 13). We observed stable reading scores over time from stroke to study enrolment (OR 1.001, 95% CI 0.999 to 1.003; p = 0.46). We examined the distribution of scores during a more clinically applicable time span of 0–1000 days (see Report Supplementary Material 3, figure 13). We observed similarly stable reading comprehension scores among participants recruited later after stroke onset (OR 0.996, 95% CI 0.978 to 1.014; p = 0.68).

Writing at study entry

Randomised controlled trial data on writing scores at study entry were available for only 11 IPD from one RCT, and were therefore not taken forward for further analysis. Data on 673 IPD (11 data sets; see Report Supplementary Material 3, figure 14) were available across all study types. Writing scores appeared to increase slightly with increasing time to study enrolment, but this was not significant (OR 1.000, 95% CI 0.998 to 1.002; p = 0.96), nor was it significant within a more clinically applicable period of 0–1000 days (421 IPD; 11 data sets; see Report Supplementary Material 3, figure 14) (OR 1.003, 95% CI 0.986 to 1.021; p = 0.71).

Observer-rated functional communication (activity) at study entry

Data on observer-rated functional communication (activity) were converted to match the range of the AAT-SSC domain. Data were available for 998 IPD from 18 RCTs. We observed stable activity scores with increasing time to inclusion in RCTs (see Report Supplementary Material 3, figure 15) (OR 0.999, 95% CI 0.994 to 1.003; p = 0.51). We observed increased activity scores with increasing time to enrolment in RCTs when examining a more clinically applicable time period (0–1000 days after stroke; see Report Supplementary Material 3, figure 15) (OR 1.021; 95% CI 1.001 to 1.042; p = 0.036).

Data were available on activity measures for 1437 IPD from 29 data sets across all study types (see Report Supplementary Material 3, figure 15). We observed relatively stable activity scores with increasing time from stroke to study enrolment (OR 1.0; 95% CI 0.997 to 1.003; p = 0.94). Within a clinically meaningful enrolment period (0–1000 days after stroke), we observed similarly stable scores with increasing time to study enrolment (OR 1.005, 95% CI 0.992 to 1.019; p = 0.46) (see Report Supplementary Material 3, figure 15).

Observer-rated participation data were derived from this data set and also demonstrated similar observations across RCTs and all study types (see Report Supplementary Material 3, figure 16). A further breakdown of language ability scores for study entry at 0–15, 15–30, 30–90, 90–180, 180–365 and 365–1095 days can be found in Report Supplementary Material 3, figure 17 (RCTs) and 18 (all study types).

Observed-rated functional communication (participation) at study entry

Observer-rated participation data were also derived from this data set and also demonstrated similar observations across RCTs and all study types (see Report Supplementary Material 3, figure 18).

Self-rated functional communication at baseline

Data were available for only 11 IPD from one RCT on self-rated activity and participation. Data were therefore examined from all study types (see Report Supplementary Material 3, figure 19). Across all baseline times to study inclusion, self-rated activity scores appeared stable (OR 1.004, 95% CI 0.993 to 1.014; p = 0.48). This was also apparent when examining a more clinically meaningful enrolment period of 0–1000 days (OR 0.991, 95% CI 0.914 to 1.075; p = 0.83). Data on self-rated participation were also generated from these IPD and showed similar distributions (see Report Supplementary Material 3, figure 20).

Recovery in the randomised controlled trial population

Recovery in all study types

Age and overall language ability at baseline

We examined the distribution of overall language ability in RCTs (700 IPD; 19 RCTs; Figure 14a), stratified by four age groups: ≤ 55 years (209 IPD; 19 data sets), 56–65 years (204 IPD; 19 data sets), 66–75 years (149 IPD; 18 data sets) and aged 75+ years (138 IPD; 13 data sets). As time to inclusion increased, so too did baseline language ability scores across all age strata. When examining a clinically meaningful enrolment period (0–1000 days post stroke; see Figure 14c), the CIs for all age groups appeared to overlap, indicating that these differences were not significant.

FIGURE 14.

Overall language ability scores by age for all participants at baseline. (a) RCTs; (b) all study types; (c) for enrolment in RCTs within 0–1000 days of stroke; and (d) for enrolment in all study types within 0–1000 days of stroke.

We then examined overall language ability at baseline for participants in all study types (2571 IPD; 78 data sets; see Figure 14b). Data were available for the following age groups: ≤ 55 years (845 IPD; 76 data sets), 56–65 years (708 IPD; 75 data sets), 66–75 years (605 IPD; 71 data sets) and 75 + years (413 IPD; 48 data sets). All age groups appeared to have increased overall language ability scores as time since stroke at baseline increased. We examined overall severity scores up to 1000 days after stroke onset to reflect typical trial recruitment periods after index stroke (see Figure 14c and 14d). For participants recruited within 1000 days of index stroke, the confidence limits for each age group overlapped, indicating no difference in the distributions of overall language ability scores by age in all study types.

Age and auditory comprehension at baseline

We examined the distribution of auditory comprehension scores at different baseline time points, stratified by age (633 IPD; 19 RCTs; Figure 15a). Data were available on participants aged ≤ 55 years (208 IPD; 19 RCTs), 56–65 years (201 IPD; 19 RCTs), 66–75 years (137 IPD; 18 RCTs) and > 75 years (87 IPD; 14 RCTs). The confidence limits overlapped for each age stratum, indicating that there was no evidence of a difference in distributions of auditory comprehension scores across each age group (see Figure 15a), with similar findings for the data from all study designs (2597 IPD; 67 data sets; see Figure 15b). Across all study designs, data on auditory comprehension scores at baseline were available for participants aged ≤ 55 years (939 IPD; 65 data sets), 56–65 years (721 IPD; 67 data sets), 66–75 years (626 IPD; 64 data sets) and > 75 years (311 IPD; 44 data sets). When we examined data for enrolment within 1000 days of index stroke (see Figure 15c and d), we observed similar overlaps in auditory comprehension scores across each age stratum, indicating no significant difference in the scores across different age groups.

FIGURE 15.

Auditory comprehension scores by age for all participants at baseline. (a) RCTs; (b) all study types; (c) for enrolment in RCTs within 0–1000 days of stroke; and (d) for enrolment in all study types within 0–1000 days of stroke.

Age and naming at baseline

We examined the distribution of naming scores at baseline by time since stroke and stratified by age in RCTs (748 IPD; 25 RCTs) and in all study designs (2760 IPD; 72 data sets; Figure 16a and 16b).

FIGURE 16.

Naming scores by age for all participants at baseline. (a) RCTs; (b) all study types; (c) for enrolment in RCTs within 0–1000 days of stroke; and (d) for enrolment in all study types within 0–1000 days of stroke.

Participants’ naming scores in RCTs were stratified into the following age groups: ≤ 55 years (259 IPD; 25 RCTs), 56–65 years (233 IPD; 24 RCTs), 66–75 years (168 IPD; 24 RCTs) and > 75 years (88 IPD; 16 RCTs) (see Figure 16a). RCT participants aged > 75 years appeared to suggest better naming scores as time since stroke increased at baseline; however, this could be a function of the relatively short time to baseline period for which data on this stratum were available. When examining RCT enrolment within 0–1000 days of stroke, the distribution of naming scores did not appear to differ in those aged > 75 years (see Figure 16c).

We plotted the distribution of participants’ naming scores at baseline across all study types and stratified by age (2760 IPD; 72 data sets; see Figure 16b). Data were available for participants aged ≤ 55 years (1005 IPD; 72 data sets), 56–65 years (779 IPD; 71 data sets), 66–75 years (693 IPD; 69 data sets) and > 75 years (283 IPD; 42 data sets). Once again, among participants aged > 75 years, a trend towards better naming scores was apparent when time from stroke to enrolment increased, compared with the younger participant age categories, but this difference was not significant, nor was it evident within 0–1000 days of stroke (see Figure 16b and 16d).

Age and other spoken language at baseline

We considered the distribution of other spoken-language scores at increasing time from index stroke to study entry, stratified by age, looking first at 163 IPD from two RCTs (see Report Supplementary Material 4, figure 1). Data were available for participants aged ≤ 55 years (90 IPD; 2 data sets), 56–65 years (58 IPD; 2 data sets) and 66–75 years (15 IPD; 1 data set). There were no IPD for those aged > 75 years.

Other spoken-language scores at baseline remained stable across various times since stroke for the different age groups. There appeared to be no significant differences in the distributions of other spoken-language scores by age group. We found similar distributions in the 344 participants from across seven data sets of all study designs [aged ≤ 55 years (191 IPD; 7 data sets), 56–65 years (112 IPD; 7 data sets), 66–75 years (37 IPD; 4 data sets) and > 75 years (4 IPD; 2 data sets)]. Inadequate data on study entry within 0–1000 days of stroke precluded further meaningful analyses of other spoken language (98 IPD; 2 data sets).

Age and reading comprehension at baseline

We examined the distribution of baseline reading scores at different baseline time points, stratified by age. Data were available from only one RCT and did not meet the prespecified criteria for analysis.

Participants’ reading comprehension scores across all study types at baseline (652 IPD; 9 data sets) were distributed across several age groups: ≤ 55 years (240 IPD; 8 data sets), 56–65 years (185 IPD; 9 data sets), 66–75 years (160 IPD; 8 data sets) and > 75 years (67 IPD; 6 data sets) (see Report Supplementary Material 4, figure 2). Participants aged > 75 years appeared to suggest higher reading scores with greater time since stroke at baseline when compared with participants in the other age groups, but this difference was not significant. We examined whether or not sparse data points at later time points contributed to the slopes of any of the regression lines (see Report Supplementary Material 4, figure 2) and found that reading scores at baseline did not significantly differ between different age strata for enrolment within 1000 days of index stroke.

Age and written language at baseline

We examined the distribution of writing scores among participants across all study types at baseline, stratified by age. There were insufficient data to consider RCT participants’ written language scores in isolation: data were available on 47 IPD from three RCTs. For participants across all study types (673 IPD; 11 data sets; Figure 17a), data on written language scores were available across the following age groups: ≤ 55 years (265 IPD; 11 data sets), 56–65 years (189 IPD; 11 data sets), 66–75 years (156 IPD; 8 data sets) and > 75 years (63 IPD; 5 data sets). There was no evidence of a difference in the distributions of writing scores for different age strata across the whole baseline sample, or when study entry was within 0–1000 days of index stroke.

FIGURE 17.

Writing scores by age for all participants at baseline. (a) All study types; and (b) for enrolment in all study types within 0–1000 days of stroke.

Age and observer-rated functional communication (activity) at baseline

We examined the distribution of observer-rated functional communication (activity) in RCTs (998 IPD; 18 RCTs; Figure 18a), stratified by four age groups: ≤ 55 years (207 IPD; 18 data sets), 56–65 years (236 IPD; 17 data sets), 66–75 years (283 IPD; 18 data sets) and > 75 years (272 IPD; 15 data sets). Observer-rated activity scores overlapped across each age stratum. This pattern was also seen when examining all study types (1437 IPD; 29 data sets; see Figure 18b) and when examining a clinically meaningful enrolment period of 0–1000 days post stroke (see Figure 18c and 18d).

FIGURE 18.

Observer-rated functional communication (activity) scores by age for all participants at baseline. (a) RCTs; (b) all study types; (c) for enrolment in RCTs within 0–1000 days of stroke; and (d) for enrolment in all study types within 0–1000 days of stroke.

Observer-rated participation scores were based on the same populations, and, similarly, showed no significant differences across each age stratum, regardless of study type and within clinically meaningful enrolment periods.

Age and self-rated functional communication (activity) at baseline

Data on self-rated functional communication (activity) scores were available on only 11 IPD in one RCT, and therefore did not meet our prerequisites for analysis. For all study types, data were available on 62 IPD from three data sets across all baseline time points, and for 43 IPD from three data sets for which enrolment occurred within 1000 days of index stroke (see Report Supplementary Material 4, figure 3). We observed no significant differences in the self-rated functional communication (activity) scores across each age stratum. Data on self-rated functional communication (participation) were derived from the same data set and showed similar results (see Report Supplementary Material 4, figure 4).

Sex and overall language ability at baseline

When data were available, we considered overall language ability at baseline by sex for participants in RCTs (700 IPD; 19 RCTs; Figure 19). The confidence bands overlapped, indicating that this difference was not significant. We found similar patterns when we examined IPD across all study designs (2380 IPD; 68 data sets; see Figure 19b).

FIGURE 19.

Overall language ability scores by sex for all participants at baseline. (a) RCTs; (b) all study types; (c) for enrolment in RCTs within 0–1000 days of stroke; and (d) for enrolment in all study types within 0–1000 days of stroke.

When we examined overall language ability scores within a more clinically relevant time point of 0–1000 days since stroke (599 IPD; 19 data sets; see Figure 19c), the data suggested that women appeared to have higher scores at later baseline time points than men, but there was no significant difference in the distributions. The findings were similar when applied to the data gathered at baseline within 1000 days of index stroke from all study designs (1665 IPD; 68 data sets; see Figure 19d).

Sex and auditory comprehension at baseline

We stratified the distribution of auditory comprehension scores at baseline time points by sex. Data were available from RCTs (622 IPD; 19 RCTs; Figure 20a) and across all study designs (2428 IPD; 60 data sets; see Figure 20b). Sex differences were not significant (see Figure 20).

FIGURE 20.

Auditory comprehension scores by sex for all participants at baseline. (a) RCTs; (b) all study types; (c) for enrolment in RCTs within 0–1000 days of stroke; and (d) for enrolment in all study types within 0–1000 days of stroke.

We then examined whether or not the presence of a smaller number of data points at later baseline time periods influenced the slope of the regression line by plotting scores within a more clinically relevant recruitment period (1778 IPD; 59 data sets; see Figure 20c and d), but we found similar distributions and no evidence of a sex difference.

Sex and naming at baseline

We examined the available naming data at baseline for RCT data sets (745 IPD; 25 RCTs; Figure 21a) and across all study designs (2629 IPD; 65 data sets; see Figure 21b). With increasing time from stroke to study entry, naming scores declined in the RCT population. Women maintained marginally higher scores than men. However, the confidence limits overlapped, thus indicating that there was no significant sex difference between the distributions of naming scores. Overlapping confidence limits between men and women were also observed across all study designs (see Figure 21b and 21d). When examining these differences in a more clinically relevant enrolment period of 0–1000 days since stroke at baseline, we found a trend towards women having higher naming scores than men, as time to baseline assessment increased in RCTs (583 IPD; 24 RCTs; see Figure 21c) and across all data sets (1766 IPD; 64 data sets; see Figure 21d).

FIGURE 21.

Naming scores by sex for all participants at baseline. (a) RCTs; (b) all study types; (c) for enrolment in RCTs within 0–1000 days of stroke; and (d) for enrolment in all study types within 0–1000 days of stroke.

Sex and other spoken language at baseline

We considered the distribution of other spoken-language scores at baseline by time since stroke and stratified by sex for RCT data sets (163 IPD; 2 RCTs; Figure 22a) and across all study types (344 IPD; 7 data sets; Figure 22b). There did not appear to be any sex differences in the distributions of spoken-language scores at baseline assessment in RCT data sets or across all study designs.

FIGURE 22.

Other spoken-language scores by sex for all participants at baseline. (a) RCTs; (b) all study types; (c) for enrolment in RCTs within 0–1000 days of stroke; and (d) for enrolment in all study types within 0–1000 days of stroke.

Sex and reading comprehension at baseline

Data on reading comprehension in the RCT population were available for only 11 IPD in one data set and were therefore not taken forward for further analysis. When data were available, reading comprehension scores at baseline were plotted against time since stroke at baseline and stratified by sex across all study types (652 IPD; 9 data sets; Figure 23). Data suggested that women appeared to have better reading scores at baseline time points later after stroke than men at a similar stage since stroke, but this difference was non-significant.

FIGURE 23.

Reading comprehension scores by sex for all participants at baseline. (a) All study types; and (b) for enrolment in all study types within 0–1000 days of stroke.

Sex and writing at baseline

We also examined writing scores at baseline by time since stroke and stratified by sex. In the RCT population (47 IPD; 3 RCTs; Figure 24a) and all study types (673 IPD; 11 data sets; see Figure 24b), there were no significant differences in writing scores between men and women. Examination of a more clinically relevant time point (of 0–1000 days) also revealed no significant differences between women and men (see Figure 24c and 24d).

FIGURE 24.

Writing scores by sex for all participants at baseline. (a) RCTs; (b) all study types; (c) for enrolment in RCTs within 0–1000 days of stroke; and (d) for enrolment in all study types within 0–1000 days of stroke.

Sex and observer-rated functional communication (activity) at baseline

When examining observer-rated functional communication activity measures (Figure 25), men in the RCT population (995 IPD; 18 RCTs) appeared to have better observer-rated activity scores; however, this observation was not apparent when looking at a more clinically relevant enrolment period (see Figure 25a and 25c). In all study types (1434 IPD; 29 data sets; see Figure 25b), there was no significant difference in the scores observed between men and women. Observer-rated participation scores were also based on this data set and displayed the same pattern.

FIGURE 25.

Observer-rated functional communication (activity) scores by sex for all participants at baseline. (a) RCTs; (b) all study types; (c) for enrolment in RCTs within 0–1000 days of stroke; and (d) for enrolment in all study types within 0–1000 days of stroke.

Sex and self-rated functional communication (activity) at baseline

Randomised controlled trial data were available for only 11 IPD from one data set. Therefore, these data were not carried forward to further analyses. Data for self-rated participation scores were also derived from this data set and were also not carried forward to further analyses. Data were available for 62 IPD from three data sets (see Report Supplementary Material 4, figure 5). We observed overlaps in activity scores (and participation scores; see Report Supplementary Material 4, figure 6) in men and women across all baseline time points, and across all clinically relevant enrolment time points.

Living context and overall language ability

When data permitted, we examined the distribution of overall language ability scores at baseline by time since stroke and stratified by living context, in RCTs (158 IPD; 3 RCTs; see Report Supplementary Material 4, figure 7) and across all designs (282 IPD; 8 data sets; see Report Supplementary Material 4, figure 7). There were insufficient data points to compare overall language ability scores at baseline by living context in RCTs. Using data from all study designs, it appeared that people living in formal care environments trended towards better overall language ability at baseline with greater time since stroke than those who were living alone. However, as the confidence limits overlapped, and there were sparse data points, these differences were not significant.

Living context and auditory comprehension

We examined the distribution of auditory comprehension scores at baseline across different time points after stroke and stratified by a participant’s living context, in RCT data sets (121 IPD; 3 RCTs; see Report Supplementary Material 4, figure 8). There were too few data points with the RCT data to adequately establish a relationship between living context and auditory comprehension scores over time. When we included data from all study designs (143 IPD; 5 data sets), there continued to be inadequate data at common baseline time points to enable comparison of patterns of auditory comprehension scores.

Living context and naming

We examined the baseline naming scores at time points since stroke and stratified by living context since stroke among RCT participants (163 IPD; 6 RCTs) and across all study types (176 IPD; 7 data sets; see Report Supplementary Material 4, figure 9). There were insufficient data points to adequately examine the differences in participants’ living contexts and naming.

Living context and spoken language, reading comprehension and writing

We sought data on other spoken-language scores at baseline, by time since stroke and participants’ living context after stroke, but identified only 10 IPD (from one data set), preventing further examination. Similarly, we lacked sufficient data to examine reading comprehension (10 IPD; 1 data set) or writing (10 IPD; 1 data set) at baseline.

Living context and observer-rated functional communication (activity)

Randomised controlled trial data on observer-rated functional communication (activity) were available for 280 IPD from six data sets, and for 290 IPD from seven data sets for all study types (see Report Supplementary Material 4, figure 10). There were inadequate data points at longer times to enrolment to permit reliable analyses or description after stratification by living context. Functional communication (participation) measures were also derived from this data set and showed similar results.

Living context and self-rated functional communication (activity)

Only 11 IPD from one RCT were available, which did not meet the prerequisite minimum data set for analysis. There were no further IPD available for self-rated activity or participation by living context across all study types.

Following these exploratory analyses of the data and predictive factors identified in other stroke recovery research,114–117 we identified age, sex and time since stroke as potential key variables for further examination in relation to recovery and absolute change in scores across each language domain.

Factors associated with increased proportion of change in each language domain

We examined the potential factors that were associated with a trend towards a greater absolute percentage change from baseline to first follow-up in each of the language domains using regression and adjusting for common covariates (age, sex and time since stroke), when these data were available.

After excluding participants from the natural history group (who were allocated to no SLT and were enrolled within 15 days of stroke), there were only two RCTs (22 IPD) and three data sets in all study types (34 IPD) comprising participants who received only historical SLT, and for whom data were available at the first follow-up time point following study intervention. For most of this group, participants had been permitted to receive SLT following the study intervention period and had follow-up time points that were inclusive of that usual care intervention. We were thus unable to separate out the effects of only historical SLT from study SLT in these participants. Analyses were therefore based on participants who had access to SLT in all study types to maximise available data.

Factors associated with increased proportion of change using data from all study types

Predictors of recovery following aphasia after stroke

We sought to examine the predictors of aphasia recovery across seven language domains in relation to participant demographics and stroke profile. Generalised linear models were used to estimate the impact of the initial severity of language impairment on subsequent recovery outcomes (defined as the absolute changes in raw scores since baseline), and this was repeated for each language domain (overall language ability, auditory comprehension, naming, other spoken language, reading comprehension, writing and functional communication).

In these analyses, we quantified the association between aphasia and eight variables (baseline score, sex, hemisphere of stroke, type, handedness, language used, age and time since stroke) using univariable and multivariable analyses. We considered language measures at baseline and first post-intervention period follow-up.

There were insufficient data on bilingualism (12 IPD available). Data on ethnicity (1475 IPD available) were also limited, but were further minimised as the variable was distributed across seven possible categories. Socioeconomic status was reported in several formats and no single composite score could be derived to describe socioeconomic status across data sets. Educational history was reported for 3125 IPD, but the variable was recorded in multiple formats, including ‘years spent in education’, ‘highest qualification’ and ‘age on leaving full-time education, making synthesis of these variables uncertain. Alternatively, analysis of years of education alone resulted in an insufficient data set. Consequently, we did not consider monolingualism/multilingualism, ethnicity, socioeconomic status or educational history in our analysis. Similarly, when a variable at each level was available only in an isolated data set, the effect of that variable on the outcome could not be distinguished from the study effect. In such cases, it was not possible to calculate estimates of the effect for non-English languages.

We first retrieved the key demographic data and identified the availability of overlapping IPD with interventions data, before excluding duplicate records, IPD that had no baseline or no follow-up data points, and non-RCT data (see Table 5). We then looked at the availability of IPD on age, sex, handedness, language used, stroke type, hemisphere, time since stroke and aphasia baseline score which would inform our planned analyses (see Appendix 8, Table 15). As prespecified, we removed variables with missing data > 20% and adjusted for initial aphasia severity. Sex, time since stroke and age are key predictive factors in recovery from stroke;114–117 although they were not always significant in this model, we included them in the analysis of population characteristics to adjust for differences between studies and changes in magnitude of effect and to inform later stages of the analyses (Table 8).

Overall language ability

Eleven RCTs (518 IPD) had data on overall language ability at study baseline, but only 482 IPD had both baseline and follow-up data of an SLT intervention and were eligible for analysis. Missing data were observed for the following variables: education (56%), handedness (44%), ethnicity (95%) and hemisphere of stroke (30%).

Auditory comprehension

A total of 556 IPD from 16 RCTs had auditory comprehension data at baseline and follow-up, but only 550 IPD had both baseline and follow-up auditory comprehension analysis, relevant demographic data and a SLT intervention to support the planned analysis (see Appendix 8, Table 15).

Naming

We identified 13 RCT data sets involving 391 IPD with naming measures available at baseline and follow-up. Six participants who received a social support intervention were excluded; therefore, the naming analysis was based on 385 IPD (13 RCTs; see Appendix 8, Table 15).

Other spoken language

Three RCTs (182 IPD) and three non-RCT data sets (49 IPD) reported other spoken-language outcome measures. The data were transformed to the anchor measures (PICA with a scale of 0–16). There were too few comparisons using RCT data sets only, and so we report an analysis that was inclusive of all data sets (see Appendix 8, Table 15). Other spoken language scores by time and age group at baseline were significantly associated with change in scores at follow up in multivariable analyses age (group p = 0.015) and time since stroke (group p = 0.01) (see Appendix 8, Table 15).

Reading comprehension

Only one RCT (11 IPD) reported reading outcomes, and so the following analyses were performed on that RCT and three non-RCT data sets (77 IPD) with reading data on participants at baseline and follow-up (see Table 8). All data (88 IDP; 4 data sets) were transformed to the CAT reading subtest (scale of 0–74).

Writing

On review of the data, we identified three RCTs (47 IPD) that reported data on writing at baseline and at a subsequent follow-up time point (see Appendix 8, Table 15). No non-RCT data sets provided writing data at these time points. All writing data were transformed to the anchor measures (the CAT with a scale of 0–76).

Functional communication

Functional communication was captured using three distinct measurement approaches: (1) tools that captured data using an observer-rated approach, (2) tools that asked participants to self-rate their functional communication abilities and (3) discourse analysis approaches that captured and quantified the presence of particular language or communicative features in a defined sample of language. Measures captured using the three different approaches could not be meta-analysed across approaches, and are thus analysed and presented separately.

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.

Age

We investigated whether or not the younger (≤ 65 years; 277 IPD) and older participants’ (> 65 years; 205 IPD) language recoveries differed in the context of SLT variations. Relative variance was modest (< 50%) across comparisons.

Time since aphasia onset

Participants for whom it was ≤ 3 months since their aphasia onset (early group: 324 IPD) showed higher change scores across all domains than participants for whom it was > 3 months after stroke (late group: 158 IPD), potentially reflecting spontaneous recovery alongside SLT intervention benefits. Relative variance was modest (< 50%) across comparisons (see Appendix 30, Table 34, and Appendix 34, Table 39).

Aphasia severity

Based on their language abilities at baseline, participants were dichotomised into mild–moderate (253 IPD) and moderate–severe aphasia groups (280 IPD). For each domain, median transformed language scores on the anchor measures, together with all available IPD in the RELEASE study database, supported the creation of two aphasia severity groups (see Appendix 34, Table 40). Across all domains, participants with moderate–severe aphasia achieved higher gains from baseline than those with mild–moderate aphasia. It is possible that the mild–moderate group’s recovery on the anchor measures may have been masked by the assessments’ ceiling effects, whereas the moderate–severe group had more scope to improve (see Appendix 30, Table 34, and Appendix 34, Table 40).

Sex

Generally, female participants (211 IPD) had higher language gains across all domains than male participants (329 IPD), although these differences declined with increasing age (> 66 years for naming and > 75 years for functional communication). Interestingly, the best overall language gains for the oldest females [> 75 years: 16.6 points (95% CI 8.38 to 24.89 points) on the WAB-AQ (68 IPD; 6 RCTs)] were similar to those of the youngest females [≤ 55 years: 16.3 points, 95% CI 8.84 to 23.72 points (55 IPD; 10 RCTs)] (see Appendix 30, Table 34, and Appendix 34, Table 38). Relative variance was modest (< 50%) (see Appendix 30, Table 34, and Appendix 34, Table 41).

Specification of study eligibility criteria; selection bias

The systematic review yielded a large number of records, which we screened for eligibility (see Chapters 2 and 3). Research teams were invited to contribute and collaborate. We had had no prior collaboration with 55 data set contributors. The systematic, inclusive approach to recruitment supported data set contributions that may not otherwise have been volunteered. We identified and extracted IPD from the public domain and grey literature in a range of languages (including 183 IPD from four RCTs, and 1988 IPD from 96 non-RCTs). These methodological steps were important in reducing the risk of selection bias in the meta-analyses.

The systematic development of the RELEASE study database used a priori eligibility criteria; reflected the scope of the research questions; and incorporated participant, study design and outcome data and IPD availability. Some research questions required SLT intervention data (see Chapters 6 and 7). Non-intervention data sets were also important, contributing to the exploration of patterns and predictors of language recovery (see Chapters 4 and 5). The database eligibility criteria were deliberately inclusive and were adhered to throughout.

We accepted diagnosis of participants’ aphasia in the primary research, confirming that it followed stroke. Aphasia associated with other neurological aetiologies was excluded. Formal approaches to aphasia diagnosis rely on validated language assessments, data required for inclusion in the study database and which informed our analysis. Data collection that was not quantitative or based on informal measures of language, were excluded as they were irrelevant to the research questions and analysis plan. We had no exclusions by study date, language of data collection or source, which might have affected electronic database searching or the contribution of unpublished clinical data sets. We made no other exclusions.

Data set identification and selection; availability bias

We made substantial efforts to identify all eligible IPD data sets, employing a broad search strategy (see Appendix 1) across databases and grey literature, and systematically inviting contributions from identified (potentially eligible) records and members of the Collaboration of Aphasia Trialists (see Chapter 3). The RELEASE study database also reflects data sharing with large independent aphasia databases [e.g. AphasiaBank, the Cantonese Aphasia Bank, Predicting Language Outcome and Recovery After Stroke (PLORAS)119] and clinical registries.

The RCT-optimised systematic search applied no restrictions to date, language or publications. We independently screened titles and abstracts, followed by double-checking to minimise any errors in the selection of eligible texts. We listed the data sets identified as potentially eligible for inclusion but for which the IPD were unavailable, which suggests additional RCTs (46 inclusive of a standard care or experimental SLT intervention) may exist (see Report Supplementary Material 2). We were unable to confirm that these data sets (and associated IPD) typically described as aggregate data in full text reports met our eligibility criteria or that the IPD remained available. Confirmation of the eligibility of included datasets was only possible following communication with the primary researchers and careful review of the IPD (see Chapter 3). Yet, it is probable that some eligible IPD existed but were unavailable to our analysis.

From a research-waste perspective, the possible existence of eligible but unavailable data is disappointing given that several of the planned analyses were not possible or unstable owing to lack of data availability (see Chapters 4–7). Additional data would have strengthened the network meta-analyses, potentially providing more links between network nodes (see Chapters 6 and 7) and greater insights. The number of included data sets, however, was sufficient to address the main RELEASE research questions.

The RELEASE study data set was predominantly based on data from English-speaking, high-income countries with well-developed SLT and stroke services. Non-English-language aphasia assessment tools are lacking, and limited (or absent) SLT services are important constraints on data-gathering in some countries. Thus, it is unlikely that we omitted many large, robust language data sets from non-English speaking countries.

Data collection and study appraisal (ROBIS domain 3)

We sought to minimise data extraction and synthesis errors. Data extraction was rigorously checked by a second reviewer. Data extraction of the complex SLT intervention descriptions was independently conducted by two researchers and was reviewed by the primary researchers. Collaborators informed the categorisation of narrative descriptions of SLT interventions by (i) target of impairment and (ii) a theoretical approach that supported data synthesis and analysis. We provided an opportunity to examine the data sets that informed the analyses, presenting aggregate data set descriptions (see Chapter 3) and individual data set overviews (see Report Supplementary Material 1). We detailed the approach to data synthesis, data item descriptions and group labels used in the analysis (see Chapter 3). Formal approaches to risk-of-bias analysis at primary data set level were described and undertaken by an independent reviewer, with a second reviewer rigorously checking that analysis (see Chapter 3).

Synthesis and findings

Choice of language measurement

Some primary data sets had IPD on two or more assessment tools that were eligible contributors to data synthesis relating to a single language outcome. Inclusion of all relevant data was not possible as it would result in ‘double-counting’ participants in that outcome. We chose which language assessment would contribute to the data synthesis for that language domain (see Report Supplementary Material 18) using a structured, systematic approach to support decision-making:

• We reviewed the numbers of missing data (considering data at study entry and follow-up time points). The more complete data set (IPD) was prioritised.

• When there was equivocation (i.e. both assessments had equal IPD numbers, including at follow-up), we prioritised the inclusion of IPD collected on the anchor measure.

Most overlapping data situations were resolved using this approach. Only a few remained, involving a small number of IPD and data sets, and these decisions were unlikely to affect the study findings and conclusions. For example, one data set using the TOMs included data from both activity and participation subtests, available on equal numbers of IPD. We arbitrarily included the activity data, but conducted a sensitivity analysis by replacing the activity item with the participation data in the network analysis (see Appendix 9, Tables 16–18). There was no indication of a notable difference between the findings. Each case of overlapping data eligibility was recorded, along with the choices available and decisions reached (see Report Supplementary Material 18). The numbers of IPD and data sets were insufficient to support any formal sensitivity analysis or the production of any meaningful result.

Exclusion of minority measures

We plotted the distribution of all transformed measures over time and by language domain, and observed that the spread of data points gathered on all language measures as profiled by each language domain overlapped (see Appendix 6, Figure 27 and Appendix 7, Table 14). There did not appear to be a clustering of transformed language measures that were generated from any one contributing assessment. The transformed values generated by each of the language assessments were valid and within an expected range.

Random or fixed effects

We considered the impact of the choice of a random-effects model over a fixed-effects model, using both the Wu–Hausman and the alternative Wu–Hausman. We compared the difference between the estimates of the covariate and standard error of baseline score. We had inadequate data for the reading comprehension item (one RCT). For all other language domains, the W score was < 3.841. For the alternative W, most values were < –1.96 or > 1.96, except for those based on functional communication (–1.9680). We also noted a relative variability of 41.1% for the naming data set, although this dropped to 30% when other variables were included in the model (see Appendix 33, Table 37). We rejected the null hypothesis that fixed effects were more efficient.

The inclusion of historic data sets

We explored the availability of historic data sets and what bias there may be in the inclusion of data sets gathered in the 1970s, as treatment responses may reflect health-care interventions (SLT and stroke related) at that time. We found no evidence that date of recruitment contributed to the findings based on language outcomes (see Appendix 31).

Excluding non-randomised controlled trial data sets

Throughout the analysis relating to patterns of recovery, predictors of recovery and optimum approaches to SLT, we prioritised the data analysis on IPD from RCT data sets only, and then re-ran the analysis inclusive of IPD from non-RCT data sets. We observed only modest (non-significantly) higher values in the context of some analyses inclusive of non-RCT data set (often with wider CIs) (see Chapters 4–6). Inclusion of the non-RCT data sets augmented the IPD on language abilities of people with aphasia at a specified time point from onset and in analyses where RCT data were lacking or absent. The IPD derived from non-RCTs and RCTs contributed to the calculation of median aphasia severity and age at the baseline assessment for the prespecified subgroup analysis (see Chapter 7). However, given the challenges and risk of overstating subgroup analysis results, we restricted the subgroup analyses to RCT IPD only.

Registry data sets and impact on findings

The database included five registry data sets (853/5928 IPD; 14%) that provided valuable data on international, clinical populations with aphasia. Summary statistics on the registry population demographics indicated that the data sets were broadly similar to other non-RCT data sets in terms of age, sex, time since stroke, aphasia severity, handedness and years of education (see Appendix 5, Table 13). Registry data contributed to the analysis of patterns and predictors of recovery only (see Chapter 4). Reporting of therapy interventions in the registries was mostly absent or insufficient to support profiling of the intervention, and thus inclusion in that analysis.

Patterns of language recovery

We explored the patterns of language recovery among participants who had access to SLT (in the context of the primary research). We looked at when recovery occurred and in what language domain. We had insufficient data to consider whether or not the patterns of language recovery varied by different languages spoken.

Participants allocated to receive SLT demonstrated higher language scores over the study duration across all language domains, with modest gains observed on measures of other spoken language, reading comprehension and writing. Data for participants who had only historical SLT were limited, as many were permitted SLT access after the comparison group completed the SLT intervention period and prior to follow-up data collection points (see Table 11).

We had insufficient data from a pure natural history population (participants who, we could be reasonably confident, had no access to SLT after stroke) to plot language recovery by language domain. With the benefits of SLT for aphasia well established, it is ethically unlikely that such data will be available in the future, unless gathered in an international context in which SLT for aphasia is not part of usual care; however, in such contexts, other aspects of stroke care are also likely to differ.

When considering participant data at study entry time points only, time since stroke was inversely associated with the severity of language impairment across all language domains, suggesting spontaneous improvement or selection bias in data sets recruiting at later time points following stroke. Absolute proportion of change in participants’ language abilities over time was greatest on measures of functional communication, overall language ability and auditory comprehension.

Predictors of recovery

We explored the predictors of language recovery outcomes following aphasia by individual characteristics (age, education, cognition, monolingual or multilingual), rehabilitation environment (social support, socioeconomic demographics, ethnicity), stroke profile (severity, lesion type, size, location) and by language domain (overall language ability, auditory comprehension, spoken language, reading comprehension and writing). We were unable to take some analyses forward because of insufficient data (the role of monolingualism or multilingualism, ethnicity, cognition) or because the format of available data prevented meaningful synthesis (education, socioeconomic demographics). Lesion type and size are currently being investigated by a collaborating research group (e.g. PLORAS119); to avoid research waste, these were not taken forward in this study.

Across all language domains, when a significant change from baseline was evident, the greatest change in language performance from baseline was among participants enrolled within 1 month of stroke. Change in language performance scores from baseline appeared to lessen with increasing time from stroke to study enrolment.

Using regression, we examined the proportion of recovery within each language domain and found that increasing time from stroke to study entry was significantly associated with poorer recovery on all language domains. Increasing age was significantly associated with poorer recovery on overall language ability. Sex was also included owing to its known relationship with other stroke outcomes (see Table 7). Each of these covariates contributed to base models in subsequent analyses (see Table 11).

Optimum speech and language therapy interventions for aphasia after stroke

Using a one-stage network meta-analysis with direct and indirect comparisons, we explored language recovery in the context of SLT aphasia rehabilitation intervention regimen (timing, frequency, duration, intensity, dosage, home-based therapy), approach (theoretical approach and impairment target) and delivery model (functional relevance, one to one or group).

Controlling for age, time since stroke and sex, we found that SLT frequency, duration, intensity and dosage were significantly and selectively associated with language outcomes. Auditory comprehension gains were associated with the frequency, intensity, dosage (including home practice) and tailoring of SLT. Naming gains were associated with the frequency and dosage of SLT. Functional communication gains were associated with dosage (see Report Supplementary Materials 5–17).

We considered language gains made from baseline, and found that SLT regimen was associated with language recovery and varied by domain. Overall language ability and functional communication scores improved most in the context of receiving SLT for up to 2 hours weekly, delivered daily (> 4–5 days weekly). Auditory comprehension gains were greatest alongside SLT in excess of 9 hours weekly, over 3–4 days. The best overall language gains were associated with 20–50 hours of SLT. The greatest overall language and functional communication gains were seen in the context of mixed SLT approaches, whereas SLT aimed at improving word-finding was associated with the greatest gains in auditory comprehension. Language recovery on these measures was greatest in the context of SLT tailored by functional relevance to the participant and when home practice was prescribed. We found no indication that the expertise of the SLT provider, the mode of therapy (face to face, group or computer-based), the context of SLT delivery or the theoretical approach were related to participants’ language change from baseline, although naming, duration and the theoretical approach findings were based on a limited data set (see Table 11).

Speech and language therapy interventions: age, time since stroke, severity and sex subgroups

Subgroups were defined by individual (age, sex), stroke (time since onset) and aphasia characteristics (severity at baseline). The association between language scores and SLT intervention components was examined using a network meta-analysis approach. Participant numbers were low and the findings were treated with caution.

Language recovery was best for both younger and older participants in the context of 20–50 hours of SLT over several SLT days weekly. For both groups, auditory comprehension improved most when associated with highly intensive SLT (> 9 hours weekly). The best language gains for the older group on other domains occurred alongside less intensive approaches (up to 2 hours of SLT weekly).

When significant changes from baseline were observed, participants within 3 months of aphasia onset made most language gains in the context of 20–50 hours of SLT, delivered two to four times weekly for up to 2 hours each week. In contrast, participants starting SLT > 3 months after stroke made their best gains alongside SLT in excess of 50 hours in total, delivered four to five times weekly, for 3–4 hours weekly. The only variation to this profile was that, regardless of time since onset, both groups made their best gains in auditory comprehension with > 9 hours of SLT weekly.

Language gains from baseline were greater for participants with moderate–severe aphasia than for those with a milder aphasia across language domains. People with milder aphasia, already scoring highly on assessments, may have had less scope to demonstrate improvement on assessment tools unable to capture language change at or close to the highest possible score (the ceiling effect). The moderate–severe group’s best language gains were associated with frequent SLT (three to five times weekly), for between 2 and 4 hours weekly, and a total dosage of 14–50 hours, or more, depending on the language outcome.

Language gains across all domains were higher for females than for males. Females’ highest language gains occurred with up to 2 hours of SLT weekly, delivered over 2–4 days. Males’ highest language gains occurred with intensive, daily SLT (3–4, or even > 9, hours weekly, depending on language modality). Males’ and females’ best overall language gains were associated with 20–50 hours of SLT, although higher SLT dosage (> 50 hours) was associated with the highest auditory comprehension gains for female participants and the highest functional communication gains for male participants.

Diversity of data set

This international collaboration of multidisciplinary researchers created one of the largest IPD data sets on 5928 people with aphasia and on language after stroke, gathered across 174 clinical or research settings, incorporating pre-existing databases, RCTs, non-RCT designs and clinical registries. The diversity of data sets, languages, severities and contexts (clinical and research) supports generalisation of the results. The primary research data sets met all the eligibility criteria and were further filtered for each research question. The database was sparse, with each data set typically including just some of the variables relevant to the analysis. Thus, the high number of data sets included in the database was essential to ensure sufficient overlap in the availability of variables to support the planned analysis. In addition, we required enough data sets to ensure an adequate sample size to explore the characteristic diversity of this population group, a wide range of language outcomes and the complexity of SLT interventions for aphasia.

Quality of data set

The database included RCTs generally graded as being at a low risk of bias (random sequence generation, concealment of allocation, blinding and few with unexplained attrition). For non-RCT data sets, the lack of randomisation was a risk of bias. Blinding of outcome assessors and the comparability of the groups at baseline were reassuring. The analyses based on IPD derived from RCTs were typically similar to the analyses that included the non-RCT IPD data sets.

We included unpublished IPD and eligible data sets identified by the systematic review. Responses to invitations to contribute data sets and inclusion of extensive IPD from the public domain (including complete RCT data sets) contributed to a reduction in the risk of publication bias. The formal assessment of publication and historical bias found no evidence to suggest that the findings were influenced by such biases.

Data relevant to the analyses were carefully identified, extracted and subsequently checked by a second researcher and (when possible) the contributing primary researchers. Data were additionally checked by reviewing plots and distributions, network graphs and through constant reference to the primary research data set to ensure accuracy. We standardised language recovery outcome by language domain across data sets and categorisation of SLT interventions by impairment target and theoretical approach through consensus with collaborators.

Language data

The language data reflect participants’ performance across several language modalities, gathered using specialised language measurement tools rather than a brief language item on a stroke scale. 29 Only published language measures were included. Screening tool data were excluded. 97 We preserved a clinically relevant language measurement reference, transforming the data to a single anchor measure for each outcome. Thus, measures of language change could be interpreted in the context of a clinically relevant scale as used in aphasia rehabilitation. The international multilingual collaboration supported data extraction and analysis of non-English aphasia assessment tools.

Speech and language therapy data

Working across an international, multilingual context supported the SLT data extraction, analysis and interpretation. Classification of SLT approaches by treatment target and theoretical approach was undertaken by the collaboration, and, following consensus review and approval (with minor edits), was sustained throughout the project. The aphasia intervention classification framework is an important adjunct to the TIDieR checklist to encourage transparency of reporting in SLT intervention studies for aphasia after stroke. 86

Individual participant data one-stage approach to meta-analysis

Findings from IPD and group summary meta-analysis can be equivalent, but the IPD data synthesis approach allowed inclusion of different data in existing meta-analyses. We included unpublished, clinical data sets, IPD, outcomes and extended follow-up time points excluded from aggregated published reports addressing some of the risks of outcome reporting bias. Using participants as the unit of analysis, we also had more power to identify clinically relevant interactions, which can be more easily interpreted to specific clinical populations than data derived from aggregate data alone. Study variability was monitored; only findings for which study variability was low to modest were reported, and these variations were highlighted to the reader. The multiple analyses were presented in a balanced manner, with all non-significant findings available for reference.

Exploratory approach

We adopted an exploratory approach to the analysis, and results should be interpreted in this context. The ambition was to inform prioritisation, design, delivery and reporting of definitive research to confirm (or refute) these findings. It was impossible to anticipate the completeness of the final database. Network diagrams accompanied each comparison by language domain and every effort was made to ensure well-constructed networks. When it was not possible to make comparisons between all interventions, this was reported as an unreliable analysis. There were few direct comparisons or, when available, comparison was limited to one or two data sets. Network graphs were reasonably balanced, but only because of the sparsity of direct comparisons between each node. The availability of data and the network geometry highlighted comparisons that were unavailable; these evidence gaps should be addressed as a priority in future RCTs.

Diversity of data set

This database demonstrated a marked imbalance towards English-language data (more than half the data set), with other languages often represented by a single data set. We applied no language restrictions to the searching or inclusion criteria. The imbalance in language representation is reflective of the wider aphasia research literature. Recovery from aphasia in monolingual and bilingual stroke survivors may differ. Languages vary from each other in grammatical structure (e.g. English grammar, compared with agglutinating languages such as Finnish) and writing systems (Latin alphabetic, e.g. English, vs. Greek alphabet or Chinese logographic scripts). Non-English aphasia research has historically been hampered by the lack of rigorous, valid and reliable aphasia assessments, although some co-ordinated initiatives are under way that aim to change this landscape. 120

Reading comprehension and writing interventions and outcomes were also limited in their representation in the database, hindering the insight into impact on and recovery of these language domains.

Data collection limitations

Demographic, language and SLT intervention data, although sufficient for the purposes of the primary research activities, were often limited in their contribution to the secondary data analyses. For example, the analysis was often limited by a lack of outcome data captured at both baseline and follow-up time points. Historical SLT groups were rarely measured following the comparator intervention period. Participants allocated to receive SLT in the primary research had limited post-intervention follow-up, thereby limiting our insight into long-term maintenance of language recovery gains.

We also undertook a comprehensive literature search and systematically invited contributions and included IPD from the public domain to ensure that the breadth of included data sets would compensate for the anticipated sparsity of the data. Data extraction of IPD in the public domain was, as expected, often limited to baseline data, but also included four complete RCT IPD data sets. It was not possible to include all datasets identified as potentially eligible. In many cases we had no response to our inviation, while in others the IPD was no longer available.

Descriptions of speech and language therapy interventions

Speech and language therapy intervention descriptions typically lacked information on tailoring, adherence and home practice tasks. SLT intervention descriptions were rarely available at IPD level, and SLT intervention data were insufficient to support all the planned analyses, in some cases resulting in a network meta-analysis with too few connections to permit complete direct and indirect comparisons. For other networks, we used a reference group to close the analysis loop and strengthen the analysis, but some orphan groups were still evident (see Appendices 11–14). These networks gaps should be addressed in targeted RCT-based comparisons. Other SLT interventions are reported in the wider literature, but these were not represented in this database because of unavailability or issues with eligibility.

Language outcome limitations

We sought to adopt the ICF framework in the RELEASE study to describe the impact of aphasia and rehabilitation interventions. Language impairment was well represented across several language domains. Measures of communication participation or activity items, however, were typically represented by isolated data sets and available data items could not be meaningfully synthesised. The development of new assessments that capture such constructs121 and recent consensus on a core outcome set for aphasia research122 are important steps to ensuring that future aphasia research will reflect outcomes of importance to people with aphasia, their families and health-care professionals. 122

Stroke description limitations

The description of stroke (and related health data) was limited in the included aphasia data sets. Stroke severity and acute care stroke interventions, such as thrombolysis, are important rehabilitation and recovery factors. 29,123 Comorbidities and stroke risk factors are relevant to aphasia recovery29 and stroke rehabilitation. 124 Standardisation of data collection on these items across all stroke and aphasia recovery research would support further clarity on their relevance to aphasia recovery. 125,126

Inclusion limitations

The planned analysis focused on specific language outcomes to ensure feasibility; however, we acknowledge that many other important questions remain, including the role of repetition and word fluency in recovery, quality of life, and the impact of aphasia on (and interventions for) carers. Information on stroke lesion location and size were also excluded from the analysis. We included non-RCT and RCT data sets to support the research questions. Although non-RCT data sets pose a higher risk of bias (because of lack of randomisation), many of those we included were of high quality, reporting outcome assessor blinding and comparability of groups at baseline. The analysis prioritised IPD derived from RCT data sets, before we included IPD from non-RCT data sets. In analyses where RCT data was lacking, the non-RCT data provided an overview of the available data to date. Where RCR data was available to support the analysis, we found little indication of a difference in the findings (see Chapters 4–7).

Methodological limitations

Across the study, we undertook multiple subgroup analyses, and we acknowledge the risk of false-positive findings. The prespecified statistical analysis plan developed with the collaborators had a strong rationale for conducting the exploratory analyses in this way. We considered compensating for this risk statistically, but, given the exploratory approach, it was important to highlight all future research avenues. The findings should be interpreted in this context. The analysis approach could not consider or adjust for all possible confounders (e.g. dropout rates, interactions between age or severity and delivery of services). The examination of spontaneous recovery following aphasia in a natural history group was hampered by the lack of data availability. Ethically, it may be difficult to gather such data now in a clinically relevant population. Participants ranged from a few days to 38 years from stroke at study enrolment. We used the term ‘study entry’ to highlight this and controlled for time since stroke in the analysis. Despite these steps, it was not possible to distinguish spontaneous recovery of language from treatment-induced effects in the analysis. Aphasia researchers and therapists alike continue to work in this context.

Interpretation of the data in relation to other research

The results are in keeping with previous meta-analyses, although this IPD meta-analysis approach afforded a greater insight into the interactions between participant profile, therapy interventions and language outcomes than is possible in an aggregate data meta-analysis. Some meta-analyses were based on language-limited literature searching (e.g. English only),62–64 included RCTs only,41 included both RCT and non-RCT data sets,62 and highlighted data availability limitations. 41,62

Others have suggested the importance of early intervention after stroke,62,63 reporting it to be twice as effective as spontaneous recovery. 62 Gains made later after stroke were acknowledged, but were smaller. 62 Some meta-analyses have suggested that ≥ 2 hours of SLT weekly is the optimal intensity,63 whereas others indicated that this may be subtherapeutic and that a minimum of 9 hours of SLT weekly is required for gains. 64 Interventions that provided a total dosage in excess of 90 hours of therapy found benefits while those providing a dose of less than 44 hours did not. 64

One review based on 23 case studies (57 participants), concluded that time since stroke was not a predictor of rehabilitation effect. 127 A 2016 Cochrane review update41 found that intensive SLT benefited measures of overall language ability (group summary data from five RCTs) and functional language (two RCTs); however, these gains occurred alongside significantly higher numbers of dropouts from the intensive SLT compared with less intensive SLT approaches. They suggest that not all patients may be able to tolerate high-intensity SLT interventions for aphasia (4–15 hours per week). 41 A 2018 SLT RCT of an SLT intervention for aphasia did not experience the same level of dropout. 26

Further randomised controlled trials

Targeted RCTs of aphasia treatment interventions are required to address the many uncertainties identified in the research. We noted many high-quality non-RCT data sets, which could have greatly reduced the risk of possible bias and increased confidence in the study findings with the application of a randomisation process on recruitment of the participants.

We noted the limited availability of data sets describing interventions that targeted improvements in reading comprehension and writing or reporting of reading and writing language outcomes. Greater research focus on these language domains is warranted.

A co-ordinated series of RCTs is required to definitively address the questions raised by this exploratory research and the possibility of a SLT–language recovery dose–response relationship. Researchers should specifically consider the development of participant inclusion criteria, subgroups of interest and experimental SLT interventions in the context of these findings. Understanding of complex rehabilitation interventions delivered to a complex population can be enhanced and methodologically addressed through improvements in research design, data capture, description and reporting, and the continuation of collaborative initiatives such as the RELEASE study.

Time since stroke at enrolment and clinical relevance

We noted a trend towards better language scores with increasing time and across all domains among participants at study entry, which may suggest a risk of selection bias. Time since stroke was also a predictor of language recovery. Thus, time since aphasia onset should be considered in participant recruitment, interpretation of study findings and their clinical relevance.

Core data set

The aphasia research community recently achieved international consensus on a core outcome set, but we also require consensus on a core data set that supports adequate description of participant demographics, stroke profile and, potentially, biomarkers, alongside a comprehensive measurement of language.

Similarly, advances made in the transparent reporting of complex interventions using the TIDieR checklist could be further enhanced when reporting complex SLT interventions by including a description of the language impairment target and the theoretical approach. 132 In the light of the findings from this study, details on tailoring of the SLT interventions to participants by functional relevance or level of difficulty, and the prescription of home-based practice tasks (and adherence to those practice tasks) would enhance the reporting of therapy dosage.

Legacy database

This collaborative, international, multidisciplinary initiative has resulted in the creation of a large IPD database that (subject to approvals from the contributing research gatekeepers) will continue to support future secondary-analysis aphasia research activities.

Search strategy

1. exp aphasia/

2. language disorders/or speech disorders/or anomia/

3. (aphasi$ or dysphasi$ or anomia or anomic).tw.

4. ((speech or language$ or linguistic or communicat$) adj5 (disorder$ or impair$ or problem$ or dysfunction or difficult$)).tw.

5. 1 or 2 or 3 or 4

6. exp aphasia/rh, th or language disorders/rh, th or speech disorders/rh, th or anomia/rh, th

7. speech-language pathology/or exp “rehabilitation of speech and language disorders”/

8. ((speech or language$ or linguistic or aphasi$ or dysphasi$ or anomia or anomic) adj5 (therap$ or train$ or rehabilitat$ or treat$ or remediat$ or intervention$ or pathol$)).tw.

9. (SLT or SLP).tw.

10. (melodic intonation therap$ or MIT).tw.

11. 6 or 7 or 8 or 9 or 10

12. Randomized Controlled Trials as Topic/

13. random allocation/

14. Controlled Clinical Trials as Topic/

15. control groups/

16. clinical trials as topic/or clinical trials, phase i as topic/or clinical trials, phase ii as topic/or clinical trials, phase iii as topic/or clinical trials, phase iv as topic/

17. double-blind method/

18. single-blind method/

19. Placebos/

20. placebo effect/

21. cross-over studies/

22. randomized controlled trial.pt.

23. controlled clinical trial.pt.

24. (clinical trial or clinical trial phase i or clinical trial phase ii or clinical trial phase iii or clinical trial phase iv).pt.

25. (random$ or RCT or RCTs).tw.

26. (controlled adj5 (trial$ or stud$)).tw.

27. (clinical$ adj5 trial$).tw.

28. ((control or treatment or experiment$ or intervention) adj5 (group$ or subject$ or patient$)).tw.

29. (quasi-random$ or quasi random$ or pseudo-random$ or pseudo random$).tw.

30. ((control or experiment$ or conservative) adj5 (treatment or therapy or procedure or manage$)).tw.

31. ((singl$ or doubl$ or tripl$ or trebl$) adj5 (blind$ or mask$)).tw.

32. (cross-over or cross over or crossover).tw.

33. (placebo$ or sham).tw.

34. trial.ti.

35. (assign$ or allocat$).tw.

36. controls.tw.

37. or/12-36

38. 5 and 11 and 37

39. exp animals/not humans.sh.

40. 38 not 39

41. (pediatric or paediatric or infant or infants or child or children$ or childhood or neonat$ or juvenile$ or toddler$).ti.

42. (child/or child, preschool/or adult children/or adolescent/or exp infant/) not exp adult/

43. 41 or 42

44. 40 not 43

45. 36 not 39.

Search strategy

1. (MH “Aphasia+”)

2. (MH “Speech Disorders”) or (MH “Language Disorders”) or (MH “Anomia”)

3. TI (aphasi* or dysphasi* or anomia or anomic) OR AB (aphasi* or dysphasi* or anomia or anomic)

4. TI ((speech or language* or linguistic or communicat*) N5 (disorder* or impair* or problem* or dysfunction or difficult*)) or AB ((speech or language* or linguistic or communicat*) N5 (disorder* or impair* or problem* or dysfunction or difficult*))

5. 1 OR 2 OR 3 OR 4

6. (MH “Aphasia+/RH/TH”) or (MH “Speech Disorders/RH/TH ”) or (MH “Language Disorders/RH/TH”) or (MH “Anomia/RH/TH ”)

7. (MH “Rehabilitation, Speech and Language”) or (MH “Speech-Language Pathologists”) or (MH “Speech-Language Pathology”) or (MH “Speech Therapy+”) or (MH “Language Therapy”)

8. TI ((speech or language or linguistic or aphasi* or dysphasi* or anomia or anomic) N5 (therap* or train* or rehabilitat* or treat* or remediat* or intervention* or pathol*)) or AB ((speech or language or linguistic or aphasi* or dysphasi* or anomia or anomic) N5 (therap* or train* or rehabilitat* or treat* or remediat* or intervention* or pathol*))

9. TI (SLT or SLP) or AB (SLT or SLP)

10. TI (melodic intonation therap* or MIT) or AB (melodic intonation therap* or MIT)

11. 6 OR 7 OR 8 OR 9 OR 10

12. (MH “Randomized Controlled Trials”) or (MH “Random Assignment”) or (MH “Random Sample+”)

13. (MH “Clinical Trials”) or (MH “Intervention Trials”) or (MH “Therapeutic Trials”)

14. (MH “Double-Blind Studies”) or (MH “Single-Blind Studies”) or (MH “Triple-Blind Studies”)

15. (MH “Control (Research)”) or (MH “Control Group”) or (MH “Placebos”) or (MH “Placebo Effect”)

16. (MH “Crossover Design”) OR (MH “Quasi-Experimental Studies”)

17. PT (clinical trial or randomized controlled trial)

18. TI (random* or RCT or RCTs) or AB (random* or RCT or RCTs)

19. TI (controlled N5 (trial* or stud*)) or AB (controlled N5 (trial* or stud*))

20. TI (clinical* N5 trial*) or AB (clinical* N5 trial*)

21. TI ((control or treatment or experiment* or intervention) N5 (group* or subject* or patient*)) or AB ((control or treatment or experiment* or intervention) N5 (group* or subject* or patient*))

22. TI ((control or experiment* or conservative) N5 (treatment or therapy or procedure or manage*)) or AB ((control or experiment* or conservative) N5 (treatment or therapy or procedure or manage*))

23. TI ((singl* or doubl* or tripl* or trebl*) N5 (blind* or mask*)) or AB ((singl* or doubl* or tripl* or trebl*) N5 (blind* or mask*))

24. TI (cross-over or cross over or crossover) or AB (cross-over or cross over or crossover)

25. TI (placebo* or sham) or AB (placebo* or sham)

26. TI trial

27. TI (assign* or allocat*) or AB (assign* or allocat*)

28. TI controls or AB controls

29. TI (quasi-random* or quasi random* or pseudo-random* or pseudo random*) or AB (quasi-random* or quasi random* or pseudo-random* or pseudo random*)

30. 12 OR 13 OR 14 OR 15 OR 16 OR 17 OR 18 OR 19 OR 20 OR 21 OR 22 OR 23 OR 24 OR 25 OR 26 OR 27 OR 28 OR 29

31. 5 AND 11 AND 30

32. TI (pediatric or paediatric or infant or infants or child or children* or childhood or neonat* or juvenile* or toddler*)

33. ((MH “Adolescence+”) or (MH “Child+”) or (MH “Infant+”)) not (MH “Adult”)

34. 32 OR 33

35. 31 not 34