Trial Design

This 12-month study was a rater-blinded, single cohort design of patients with dysarthria due to progressive ataxia, using a single study arm—ClearSpeechTogether. Participants acted as their own controls by implementing a 2-week no treatment phase. No adjustments were made to the methodology following registration of this study in the ISRCTN clinical trial database [50].

Sample Size

The study was intended to function as a pilot study to establish suitability of the intervention approach for a larger RCT. For this purpose, it was decided to run two groups of five participants each, aiming for a total of ten recruits for the study.

Participants

Eligibility criteria for the study included a confirmed diagnosis of progressive ataxia, the presence of mild to moderate predominantly ataxic dysarthria, the absence of a functional voice disorder other than can be expected as part of the ataxia, the absence of visual, hearing or cognitive impairments that would have impact on participation in the assessment or treatment regime, age above 16 years, and availability and ability to use the technology necessary to complete assessment and treatment sessions online via Zoom.

Advertising took place via the funder website and social media campaigns. In addition, people with ataxia who had requested to stay informed about upcoming trials from a previous study were contacted directly via email. All participants self-selected and were provided with study information by email after contacting the research team for more information about the study. For those still interested, suitability to participate was established during a Zoom call with the first author, during which consent was also taken for those recruited to the study.

Study Design

Patient involvement in this study lasted for 16 weeks. This included a 2-week pre-therapy assessment period, 6 weeks of intervention and a further 8-week assessment period. Given the distance of study participants’ homes to the investigators and to each other, and the fact that the UK was undergoing various COVID-19 related lockdown measures at the time of the study, all assessments and individual and group therapy sessions were delivered remotely via Zoom. The feasibility of telehealth provision in this population had been established in our previous study using Skype [25].

Assessments required participants to record themselves at home. For this purpose, they were provided with information on how to use freely available recording software AudacityR (version 3.0.3). Two participants had iPads and used the inbuilt voice recorder instead. Each participant was provided with a headset microphone to ensure stable mouth-to-microphone distance and a low-cost speech intensity meter (Cadrim Digital Sound Level Meter). They were sent a OneDrive link to securely upload their recordings after each assessment session. Backup recordings were made using Zoom cloud recordings with participants’ consent.

Assessment Tasks

The study included multiple baseline assessments (sessions 1 & 2, administered 2 weeks apart prior to treatment), and two post-therapy assessments, one within 1 week of completing treatment, and another 8 weeks post-treatment (sessions 3 & 4). Assessments were conducted by the first author who was not involved in the treatment of participants.

In line with the ICF model, we assessed participant’s communication at impairment, activity and participation level. In addition, we collected information on fatigue and their medical history, as summarised in Table 1.

Table 1 Assessment summary

Fatigue was captured with the overall score of the Fatigue Impact Scale (FIS [51]). For speech, two repetitions of maximum phonation time were collected unless the participant clearly performed within the normal range with a duration of around 20 s or more on their first attempt. Where two attempts were collected, the better of the two was used for subsequent analysis. Connected speech samples were captured by a reading task and a spoken monologue. The reading passage comprised the first paragraph of the Caterpillar passage [52], which resulted in 20 to 30 s samples, and for the monologue participants were asked to talk about a topic of their choice, such as a holiday, a recent memorable event or a hobby for about one minute. In addition, we recorded ten repetitions of the sentence “Tony knew you were lying in bed” to measure the consistency of sentence production (utterance to utterance variability, UUV [53]). This measure had previously shown some promise of being sensitive to intelligibility levels and could thus potentially quantify any post-treatment improvements in this parameter.

During the post-treatment assessment sessions, all participants were explicitly requested to apply the speech strategies developed during the intervention phase.

Finally, we captured participation by asking participants to complete the Communication Participation Item Bank (CPIB [54]) and to score their level of communication confidence on a 10 point scale.

Analysis

The primary speech outcomes measures were duration in the MPT, and intelligibility of connected speech (reading and monologue samples). Secondary outcomes included consistency of sentence production, measures of communication participation and confidence, fatigue ratings and patient perceptions. All examiners were blinded to the time point of the samples they analysed.

Vowel Prolongation

Vowel prolongation was analysed in terms of MPT and voice quality. Oscillographic and wide-band spectrogram data viewed in Praat ([44], version 6.0.43) were used for duration measures. In addition, four experienced speech and language therapists (SLTs) used the GRBAS [55] to provide perceptual evaluations of voice quality. This tool provides scores for Grade (G—overall severity), roughness (R), breathiness (B), asthenia (A—weak voice) and strain (S).

Connected Speech

Intelligibility in the reading and monologue tasks were rated by four experienced SLTs. Due to the repetitive nature of the reading material, listeners scored the samples using the Direct Magnitude Estimation (DME) method [56] using the first recording of the reading sample (session 1) as the standard. For the evaluation of the monologue, listeners scored the samples on a percentage scale. The samples comprised of around 30-s continuous speech without interruptions from the examiner or extraneous noise. As for reading, all samples were presented in randomised order of assessment but grouped by speaker.

Communication Participation

We conducted semi-structured interviews in sessions 1, 3 and 4 to establish the form, severity and impact of speech problems experienced by participants, and how these were affected by the intervention. We also asked them to complete the short form of the Communication Participation Item Bank (CPIB) [54] on these occasions, and to provide a single score on a scale of 1–10 of their confidence when communicating with people outside their immediate social circle.

Acceptability of the Approach

The participant interviews also focused on the content and presentation of the treatment, discussing areas such as appropriateness of the exercises, treatment intensity, group dynamics, online nature of presentation and balance of individual versus group input.

Inter-rater Reliability and Statistical Analysis

To assess inter-rater reliability, we conducted an independent analysis of four participants for the various measures performed. Agreement was excellent with an intraclass correlation coefficient (ICC) of 0.999 for the MPT task. In addition, agreement between the four expert listeners for the perceptual analysis of the data was good with an ICC of 0.804 for reading intelligibility, and 0.884 for the monologue intelligibility, and 0.808 for voice quality.

Due to the small sample size and variability of speaker presentation non-parametric statistics were applied to avoid overinterpretation of results. The Friedman Test was performed to assess changes across time, using the Wilcoxon-signed rank test for the post hoc analyses. We chose not to employ Bonferroni corrections given the sample size and highly exploratory nature of the investigation, but considered these factors in the interpretation of the results. For correlational analyses we employed Spearman’s Rho. Listener agreement was calculated with the Intraclass correlation coefficient.

Treatment Schedule

Treatment was administered over a period of 6 weeks. This included an initial 2 weeks of individual therapy with two sessions of 45–60 min per week (4 individual sessions) and twice daily homework tasks. This was followed by 4 weeks of intensive peer supported group practice, consisting of daily 1 h virtual meetings with the group (20 group sessions). The group phase was supported by a weekly meeting with the SLT. There was the option to provide further individual input for participants if the clinician determined that they were not using the speech strategies effectively or showed adverse reactions. A non-specialist volunteer was present during the non SLT-led group sessions to support the participants with any technical issues. The SLT-led sessions were administered by two expert clinicians who were highly experienced in treating patients with ataxia.

Treatment Focus

In line with previous trials for people with progressive ataxia [25, 27, 29, 57], two global speech strategies were focused on in this study – LOUD and CLEAR. Principles of the Lee Silverman Voice Treatment (LSVT LOUD®) programme were adopted in terms of the focus on voice, however, unlike in Parkinson’s disease, volume is not necessarily in issue for people with ataxia. Even though we maintained the cue “LOUD” for participants, it represented effective voice use and clinicians also ensured that voice quality was not strained or effortful, and produced at the appropriate pitch. “CLEAR” speech production aimed to maximise intelligibility for a communication partner by encouraging participants to over-articulate. The individual sessions were used to introduce participants to the two therapy concepts and to establish their use at least at single word level. The group phase then involved participants working through a handbook of graded exercises in line with the LSVT LOUD® programme. These briefly involved further practice at the single word level and then moved quickly on to phrases, sentences and increasingly complex reading and free speech exercises by the end of week 4. Participants also practised ten daily phrases during the sessions, and completed prolonged vowel exercises as a warm up before the group meeting, to maximise time during the session for targeted speech activities.

Participants were provided with materials to practise, but also increasingly asked to prepare their own materials to build independence during the post-therapy phase for continued practice. The SLT met with the group at the end of each week to monitor each participant’s progress, suggest adjustments as necessary and to explain the upcoming tasks for the following week. The weekend was available for participants to prepare materials as necessary.

Participants were invited to reflect and comment on each other’s performance in a constructive way. This was intended to provide support but also developed participants’ ability to monitormselves and others. All exercises were designed to be executed in turns, ensuring active involvement of all participants throughout the session. In addition, participants rotated as “session chair”, which involved time management and ensuring all exercises were attempted. They were also responsible for contacting the research team if they were unclear about any exercise or if any other problems arose.

Speech exercises were designed to last 20 to 30 min though sessions tended to last 45 to 60 min depending on how much social chat was included at the start and end of the meeting..