The verbal, non-verbal and structural bases of functional communication abilities in aphasia

Abstract The ability to communicate, functionally, after stroke or other types of acquired brain injury is crucial for the person involved and the people around them. Accordingly, assessment of functional communication is increasingly used in large-scale randomized controlled trials as the primary outcome measure. Despite the importance of functional communication abilities to everyday life and their centrality to the measured efficacy of aphasia interventions, there is little knowledge about how commonly used measures of functional communication relate to each other, whether they capture and grade the full range of patients’ remaining communication skills and how these abilities relate to the patients’ verbal and non-verbal impairments as well as the underpinning lesions. Going beyond language-only factors is essential given that non-verbal abilities can play a crucial role in an individual’s ability to communicate effectively. This study, based on a large sample of patients covering the full range and types of post-stroke aphasia, addressed these important, open questions. The investigation combined data from three established measures of functional communication with a thorough assessment of verbal and non-verbal cognition as well as structural neuroimaging. The key findings included: (i) due to floor or ceiling effects, the full range of patients’ functional communication abilities was not captured by a single assessment alone, limiting the utility of adopting individual tests as outcome measures in randomized controlled trials; (ii) phonological abilities were most strongly related to all measures of functional communication and (iii) non-verbal cognition was particularly crucial when language production was relatively impaired and other modes of communication were allowed, when patients rated their own communication abilities, and when carers rated patients’ basic communication abilities. Finally, in addition to lesion load being significantly related to all measures of functional communication, lesion analyses showed partially overlapping clusters in language regions for the functional communication tests. Moreover, mirroring the findings from the regression analyses, additional regions previously associated with non-verbal cognition emerged for the Scenario Test and for the Patient Communication Outcome after Stroke rating scale. In conclusion, our findings elucidated the cognitive and neural bases of functional communication abilities, which may inform future clinical practice regarding assessments and therapy. In particular, it is necessary to use more than one measure to capture the full range and multifaceted nature of patients’ functional communication abilities and a therapeutic focus on non-verbal cognition might have positive effects on this important aspect of activity and participation.


Materials and methods Participants
Thirty-seven participants were recruited for the present study (11 female, 26 male; mean age 64 ± 12 years, range 45-88 years; see Supplementary material for more details). All participants had a single left-hemispheric stroke (ischaemic or haemorrhagic) at least one year before assessment and imaging and had no additional significant neurological conditions and no contraindications for MRI. They were pre-morbidly right-handed native English speakers with normal or corrected-to-normal vision. Recruitment took place consecutively from local community clinics and from local NHS referrals. All participants had been diagnosed with aphasia, but no restrictions were applied regarding the type of aphasia or the severity, therefore the sample includes cases ranging from severely to mildly aphasic. Informed consent was obtained from all participants prior to participation, in line with the Declaration of Helsinki and as approved by the local NHS ethics committee.

Assessments
The data analysed here were collected in two phases, each comprising several sessions within a time frame of around two months. Due to the consecutive recruitment of patients, time intervals between testing phases ranged between 5 and 86 months. The first phase included a variety of language production and comprehension tasks as well as some other neuropsychological tests (detailed in Butler et al., 2014;Halai et al., 2017). In the second phase, the dataset was enriched with a broad range of standardized neuropsychological tests of attention and executive functions (detailed in Schumacher et al., 2019). Performance on all these tests was used to compute the percentage of impaired scores per patient, serving as an indicator of the severity of their verbal and nonverbal impairment, respectively. Moreover, a principal component analysis of these data revealed six orthogonal components, which were previously interpreted as Phonology, Semantics, Speech Quanta, Shift-Update, Inhibit-Generate, and Speed (Schumacher et al., 2019). More detailed information, also regarding the time interval between test phases, is provided in the Supplementary material.
The second phase also comprised assessments of functional communication abilities. Two tests and two versions of a rating scale were administered by the first author. The ANELT (Blomert et al., 1994) assesses verbal functional communication and comprises 10 descriptions of everyday situations that are read to the participant, followed by the prompt to say what they would say in that situation. Answers are scored regarding meaningfulness and intelligibility https://mc.manuscriptcentral.com/braincom with a maximum total score of 50 each. In the current study, only the meaningfulness was analysed. The Scenario Test (van der Meulen et al., 2010;Hilari et al., 2018) similarly uses everyday situations but in contrast to the ANELT, the situations are described and also depicted, and participants are allowed to use any mode of communication they want. Moreover, in case of unsuccessful conveying of the requested information, they are prompted to use a different way of communication. Failing this, closed questions are asked to assess comprehension of the Scenario Test. The test contains six scenarios with three questions each and the maximum score is 54. The COAST is a rating scale assessing perceived communicative effectiveness and quality of life (Long et al., 2008;Long et al., 2009). The patients' and the carers' version each contain 20 questions to be answered on a 5-point scale. The last five questions relate to the respondent's quality of life.

Statistical analysis
To elucidate the commonalities and differences between the measures of functional communication, correlations between their total scores were computed. For the COAST, sum scores were obtained including as well as excluding the five items on quality of life, as these were the items where patients and carers rated themselves. Analyses that did not include these five items are referred to as Patient/Carer COAST 1-15. Differences between patients' and carers' ratings were analysed by means of paired t-tests. The COAST items tap into various aspects of functional communication (Long et al., 2008;Long et al., 2009). Therefore, subscores were derived by applying principal component analyses with varimax rotation (extraction criterion of Eigenvalue > 1) on the Patient and Carer COAST separately. The individual factor scores on each component were then taken as sub-scores and used in further analyses.

The relationship between impairment-level measures and functional communication abilities
was elucidated by means of regression analyses. Three approaches including variables of differing specificity were used to relate the (sub)-scores obtained by the tests and rating scales assessing functional communication. The first and broadest approach included the patients' overall verbal and nonverbal impairment (in percentage of impaired scores in language and nonverbal tests, respectively), similar to previous research using severity measures or very general composite scores. The second, intermediate approach used the more fine-grained factor scores on three verbal components (Phonology, Semantics, Speech Quanta) and three nonverbal components (Shift-Update, Inhibit-Generate, Speed), derived from a principal https://mc.manuscriptcentral.com/braincom component analysis (see Schumacher et al., 2019) based on the 32 patients with no missing data. The main advantage of this intermediate approach, in addition to allowing insights beyond the verbal-nonverbal dichotomy, is that the components are orthogonal. Both approaches were carried out in a hierarchical fashion, with lesion volume (being the only patient characteristic significantly correlated with all functional communication measures) entered first, followed by language measures, and lastly by nonverbal measures, in order to assess the strength of their correlation with functional communication beyond the language impairment itself. Given the split in patients' performance on the Scenario Test (see below), we completed an additional exploratory regression analysis (forward selection of intermediate factor scores as independent variables) to elucidate the most important verbal and nonverbal abilities for patients performing either poorly or well on the Scenario Test (splitting the group at a score above or below 50).

Imaging data acquisition and analysis
High resolution structural T1-weighted Magnetic Resonance Imaging (MRI) scans were acquired on a 3.0 Tesla Philips Achieva scanner (Philips Healthcare, Best, The Netherlands) using an 8-element SENSE head coil. A T1-weighted inversion recovery sequence with 3D acquisition was employed, with the following parameters: TR (repetition time) = 9.0 ms, TE Structural MRI scans were pre-processed with Statistical Parametric Mapping software (SPM8: Wellcome Trust Centre for Neuroimaging, http://www.fil.ion.ucl.ac.uk/spm/). The images were normalised into standard Montreal Neurological Institute (MNI) space using a modified https://mc.manuscriptcentral.com/braincom unified segmentation-normalisation procedure optimised for focal lesioned brains (Seghier et al., 2008). Data from all participants with stroke aphasia and all healthy controls were entered into the segmentation-normalisation. Images were then smoothed with an 8 mm full widthhalf-maximum (FWHM) Gaussian kernel and used in the lesion analyses described below. An age and education matched healthy control group was used to determine the extent of abnormality per voxel. This was achieved using a fuzzy clustering fixed prototypes (FCP) approach, which measures the similarity between a voxel in the patient data with the mean of the same voxel in the control data (note: this method does not discriminate what caused the abnormality, but simply reflects how deviant the signal in the patient scan is from a healthy group). One can apply a threshold to the FCP to determine membership to abnormal/normal voxel. The default parameters were used apart from the lesion definition 'U-threshold', which was set to 0.5 to create a binary lesion image. We modified the U-threshold from 0.3 to 0. greater abnormality leads to poorer performance/lower ratings). A separate linear regression model (not including any covariates of no interest) was built for the four functional communication assessments. A voxel-level threshold of p < 0.001 and a family-wise error correction (FWEc) at cluster-level p < 0.001 were applied, unless noted otherwise. The anatomical labels for the clusters were determined using the Harvard-Oxford atlas for grey matter and on the John Hopkins white matter atlas for white matter tracts.

Data availability
Behavioural data are available in the Supplementary material. Further data are available by request to the last author. https://mc.manuscriptcentral.com/braincom

Results
Descriptive statistics of all functional communication measures are given in Table 1 and more details are available in the Supplementary material.

Relationship between different functional communication assessments
Correlations were computed to elucidate the commonalities and differences between the measures of functional communication. Statistically significant correlations were found between the ANELT, Scenario Test, ratings in the Patient and Carer COAST (with and without items on quality of life), as well as the factor scores from the first component of the Patient COAST (verbal communication), as shown in Table 2.
The strongest correlation was observed between the ANELT and Scenario Test. However, as clearly depicted in Figure 1, Scenario Test scores were highly variable in patients obtaining low ANELT scores but tended to be at ceiling for patients with higher ANELT scores. The latter, in turn, allowed for a finer grading of patients with high scores in the Scenario test. COAST scores (excluding the items on quality of life) revealed a significant difference (t(26) https://mc.manuscriptcentral.com/braincom = 2.29, p = 0.03), with patients rating their communicative abilities higher (mean 61.5 ± 14 SD) than carers (55.7 ± 13.3).
----- Table 2  analyses were carried out in a hierarchical fashion. Lesion volume was entered first, followed by language measures, and lastly by nonverbal measures. Table 3 shows, from left to right, the statistics of all the models computed in this way. For each functional communication (sub)score (rows), the first three columns of Table 3 show the Adjusted R 2 and the respective F-and p-values when only lesion load is entered as an independent variable. A p-value below 0.05 indicates that the independent variable(s) account for a significant amount of variance in the functional communication measure of interest. The next three columns show the same statistical information but after adding the language variable(s) to the regression model. A further column (sig. F change) indicates whether this model is significantly better than the previous one (thus significantly increasing the adjusted R 2 ). In the last four columns the same is shown for the full models (including the nonverbal variable(s)). For the specific approach, analyses were carried out using a stepwise forward selection approach and the relevant statistics are given further below in the text.
For all three approaches, the contribution of each variable included in the models are shown in

Brain-behaviour mapping of functional communication abilities
To elucidate whether there are associations between a patient's lesion and functional communication abilities, we performed separate VBCM analyses for each functional communication measure. Significant clusters emerged for all measures, as depicted in Figure   4 and detailed in Table 4. The cluster where tissue abnormality was associated with performance in the ANELT was mainly in the temporal lobe (including the temporal pole) but comprised also frontal (including inferior frontal gyrus) and parietal structures. The Scenario Test cluster overlapped with the ANELT cluster in the temporal lobe but extended more posteriorly into the lateral occipito-temporal cortex and into the parietal cortex. In addition, the Scenario Test cluster included right hemisphere structures, and both clusters contained subcortical regions (mainly parts of the left thalamus). In line with the results from the regression analyses, the ANELT cluster covers more classical language regions while a posterior part of the Scenario Test cluster has also been associated with performance in the nonverbal Shift-Update component (Schumacher et al., 2019). and is generally thought to play an important role in demanding visuo-spatial processing (Fedorenko et al., 2013;Humphreys and Lambon Ralph, 2017). The Patient COAST was associated with a cluster in the

How do functional communication measures relate to each other?
Administering two different objective measures of functional communication, the Scenario Test and the ANELT, to a sample of patients covering the whole range of aphasia severity made apparent the relative strengths and weaknesses of these tests. The overall high correlation between the two measures hides the fact that there are considerable floor and ceiling effects in both tests. Being limited to verbal communication, the ANELT can for instance not capture the occasionally remarkable functional communication skills in severely aphasic patients, while patients with relatively intact verbal abilities will obtain an undifferentiated high score on the Scenario Test. Thus, if one wants to grade patients and capture changes in functional communication, for instance within a randomized controlled trial, it is not sufficient to use only one of these measures, unless the sample is restricted in terms of aphasia severity or the scoring of the test is adapted. The German version of the Scenario Test (Nobis-Bosch et al., in press) for instance contains an extension of the scoring scheme to better account for high (verbal) performers.
The association between the two subjective measures, the patients' and carers' version of COAST was still relatively high, whereas the association between the objective and subjective measures was only moderate, in line with previous research (Hilari et al., 2018;Olsson et al., 2019). The lowest correlation emerged between patient's ratings of their communicative abilities and their performance in the Scenario Test. It seems that some patients tended to underestimate their abilities if they are using nonverbal modes to communicate.

What are the relationships to underlying cognitive and structural bases?
Given the close but not synonymous relationship between language and communication, it is perhaps not surprising that verbal impairment in general, and phonological abilities in particular, were strongly related to all measures of functional communication. This relationship https://mc.manuscriptcentral.com/braincom was most obvious for the ANELT where the regression analyses show that almost all of the variance can be accounted for by direct measures of the patients' verbal impairment -either in the form of the composite verbal factor score or individual tests of verbal short term memory such as digit span. Indeed, the variance explained by these regression models is such that they are equivalent to test-retest reliability of the ANELT itself. This would suggest that the timeconsuming ANELT assessment could potentially be replaced by much more efficient tests of language ability.
Communication can go beyond language alone, however, and consistent with this fact, we found that nonverbal abilities were critically important, beyond the lesion volume and the verbal measures, when language production was relatively impaired and other modes of communication were allowed, as in the Scenario Test. In these situations, the patients' nonverbal abilities move to the foreground and retained cognitive skills enable them to use and switch between nonverbal communication strategies. Similarly, a very recent study including only individuals with severe aphasia showed that the relationship between executive abilities (captured by screening tests) and Scenario Test performance was strongest in individuals with hardly any verbal output (Olsson et al., 2019).
The importance of nonverbal cognition became also apparent when patients and carers rated the basic communication abilities. Regression models for this aspect only became significant once overall nonverbal impairment or intermediate nonverbal factor scores, respectively, were included. Beyond this similarity, the bases influencing the carer's ratings differed in various ways from those of the patient's ratings and from the tests assessing functional communication.
Carers' ratings of patients' verbal communication ability for instance, was related to speech quanta (fluency and amount of speech production), which has been previously reported (Fridriksson et al., 2006). Carers' overall ratings, however, were only significantly related to lesion volume. One interpretation might be that carers' ratings reflect a more holistic judgment, integrating additional difficulties and resources a patient may have.
The available lesion and impairment-level data were overall considerably less useful for explaining variance in the subjective ratings than in the objective measures. One explanation intermediate verbal and nonverbal factor scores, specific tests) as well as of the outcome measures (functional communication (sub)-scores) critically influence which relationships, if any, are found between the measures of interest. In order to optimally capture potential changes in functional communication following an intervention, it is thus paramount to consider these differences and choose an appropriate type and level (or rather levels) of measurement.
Examples of approaches to evaluate more specific aspects of functional communication entail for instance the adaptation of scoring systems (Purdy and Koch, 2006;Nobis-Bosch et al., in press) or the creation of more specific assessments (Spitzer et al., 2019).
Secondly, a considerable proportion of the individuals in our sample had sufficient language production abilities to be able to solve the functional communication tests verbally. In everyday life, spoken language is the mode of choice for communication, which is also reflected in the fact that patient's ratings of communicative ability were heavily based on their language production abilities, and that being able to speak is the most important goal for the majority of patients. Only if the mode of choice is not (sufficiently) available, other abilities, including nonverbal cognition, gain relevance in solving a communicative task. However, it is important to bear in mind that a high score in the Scenario Test or ANELT does not necessarily mean that the individual's communicative abilities are equal to normal controls or to their pre-morbid abilities. Both tests assess common everyday situations, in which the relevant message can be conveyed with very limited output. The tests may not be very demanding for participants without severe language production deficits (see also Olsson et al., 2019). Moreover, other important aspects of communication, such as pragmatics (Irwin et al., 2002) or discourse (Barker et al., 2017), both more often considered in patients with right-hemispheric brain lesions (Bosco et al., 2017) or traumatic brain injuries (Galski et al., 1998), are not assessed.
Thus, further research will be needed to elucidate not only more specific but also more complex aspects of communication (see for instance MacDonald and Johnson, 2005) and their relation to (nonverbal) cognition in stroke aphasia.
To conclude, functional communication is very multifaceted and depends not only on verbal but also on nonverbal abilities. The latter gain importance when a functional communication assessment allows for nonverbal communication (as per its definition) and when verbal abilities are comparably low. Based on our findings, it seems advisable to use more than one measure to assess functional communication, particularly in the context of randomized controlled trials.
Moreover, a therapeutic focus on nonverbal cognition might have positive effects on this important aspect of activity and participation. Our thorough approach thus yielded findings that https://mc.manuscriptcentral.com/braincom        Note: L/R = Left or Right side of the brain, ant = anterior, fas = fasciculus, inf = inferior, p tri = pars triangularis, pos = posterior, temocc= temporo-occipital, coordinates in MNI space