The effect of two speech and language approaches on speech problems in people with Parkinson's disease: the PD COMM RCT

Parkinson’s disease

Speech and voice problems experienced by people with Parkinson's disease and their families

Rationale for trial

From 2009, several National Institute for Health and Care Research Health Technology Assessment (NIHR HTA) commissioned calls were issued to assess methods of SLT for improving speech intelligibility and effectiveness of communication in people with Parkinson’s disease.

The call from which the PD COMM trial was successfully funded, called for a pragmatic RCT in people with PD to assess the clinical and cost-effectiveness of NHS SLT compared with no treatment. The specifications included a follow-up of at least 1 year and the trial should follow a pilot phase. This recommendation was based on several factors:

• dysarthria is common in PD and increases in severity and prevalence as the disease progresses;

• quality of life is negatively affected by speech problems;

• speech and language therapy being offered by the NHS was variable;

• the lack of definitive studies investigating SLT for PD-related dysarthria.

Trial design

Driven by the HTA commissioned calls, the feasibility and acceptability of a large-scale SLT trial in people with PD were assessed and found to be acceptable in the PD COMM pilot trial36 funded by The Dunhill Medical Trust (grant R192/0511). The data from this pilot trial informed the design of the definitive RCT (PD COMM) to assess the clinical and cost-effectiveness of these SLT interventions. The pilot trial assessed eligibility, recruitment and retention, participant acceptability and treatment compliance. It also provided data to help inform the sample size calculations and to refine the choice of outcome measures including those used for the economic evaluation.

The 2017 NICE guideline10 for PD states that SLT should be offered to people with PD who have communication problems. The optimum treatment regimen including delivery method and theoretical basis were not and continue to remain unclear. 35 To progress the understanding of how SLT interventions are delivered and the mechanism of action, the PD COMM trial included a process evaluation.

Rationale for choice of interventions

Before the trial, guidance on SLT best practice for people with PD-related dysarthria as a result of PD could be found in two historical resources from the Royal College of Speech and language therapists. 37,38 Two Cochrane systematic reviews29,30 of the evidence published in 2012 were also available but could not provide clear, evidence-based recommendations for practice. These sources of evidence were not robust enough to be clear about the optimum treatment approach for PD-related dysarthria. The choice of treatments for the PD COMM trial was taken based on testing two commonly chosen treatment options in the UK in direct response to the HTA commissioned call.

Lee Silverman Voice Treatment (LSVT LOUD)

The focus of LSVT LOUD is to ‘think loud’, improving phonation and vocal loudness through better vocal fold adduction. 40 The intervention aimed to replicate the dose and content recommended by the originators and delivered in clinical practice and previous ‘standard’ LSVT trials. The LSVT LOUD intervention consisted of four face-to-face or remote 50-minute sessions per week delivered over 4 weeks. 40 Remote sessions were provided through the LSVT Companion software and could support remote intervention delivery for up to 8 of 16 sessions. Each session follows a similar structure which can be personalised for each patient: 25 minutes of repeated and intensive maximum effort drills, and 25 minutes of high-effort speech production tasks. 40 Participants were also prescribed 5 to 10 minutes of home-based practice tasks on treatment days, and up to 30 minutes of home-based practice tasks on non-treatment days. 43 The content of the intervention consisted of repeated repetitions of sustained ‘ah’ phonation, maximum fundamental frequency range high and low pitch glides, and functional sentence repetition for the first half of each session, and exercises using speech production hierarchy that progresses throughout the duration of the treatment programme (single word, phrases, sentences, paragraph reading, conversation) during the second half of the sessions. 43 Throughout all of the sessions, the focus of the intervention was to ‘think loud’, maintaining the vocal loudness produced during vowel phonation throughout all other tasks during the treatment. 40 Delivery of LSVT LOUD in the trial was as typical within an NHS setting.

National Health Service speech and language therapy

The content, intensity and dose of NHS SLT for people with PD-related voice or speech problems are poorly defined within the published literature. For this reason, the NHS SLT group took a pragmatic approach to the intervention and encompassed all local standard NHS SLT practices and techniques with the exception of LSVT (as per the LSVT LOUD Protocol). Therapists were free to tailor therapy to individual participant’s needs. NHS SLT could potentially include interventions targeted at rehabilitation of the underlying movement disorder functions that resulted in dysarthria, behavioural or compensatory strategies and AAC strategies to improve communicative function and participation. 44 The therapist involved participant’s family members or carer(s) as appropriate.

Treatments targeted at impairments of the underlying speech production movements may include exercises focused on improving capacity, control and co-ordination of respiration, techniques for improving phonation intensity and co-ordination with respiration (but not LSVT), and exercises to improve the range, strength and speed of the articulatory muscles. 45,46 Behavioural therapy includes interventions that target the reduction of prosodic abnormality47,48 such as exercises targeting pitch, intonation, stress patterns and volume variation,45–49 and techniques to address the overall rate of speech45,46 including the use of therapeutic devices such as pacing boards. 50,51 AAC strategies such as topic and alphabet supplementation through communication books and boards may be employed,44 along with AAC devices such as voice amplifiers, delayed auditory feedback systems and masking devices. 52–54

Pitch limiting voice treatment55 may also be utilised within the NHS SLT intervention. The above NHS SLT approaches may include techniques that are also used in LSVT LOUD (e.g. vocal intensity exercises) but will be distinct from that intervention as it was anticipated, based on the available literature, that they would be delivered in combination with other SLT strategies and at a different intensity and dose. Though dose and frequency were to be determined by the local therapist in response to participants’ individual needs, it was anticipated that it would reflect the median dose as reported in a survey of current UK SLT practice for PD by Miller et al. ,22 of six sessions delivered over 42 days. The PD COMM Pilot trial36 also found the median NHS SLT intervention dose to be six sessions (range 1–14) over an average of 9.6 weeks (standard deviation 6.1 weeks).

Control no speech and language therapy group

Participants were randomly allocated to the no SLT (control) arm for PD-related speech or voice problems for 12 months participation in the PD COMM trial. Participants may still receive SLT for swallowing problems (dysphagia). Since there is insufficient evidence to prove or disprove the benefit of SLT in PD, equipoise still exists. Therefore, it is ethical to randomise between SLT and no SLT. Investigators, however, remained vigilant throughout the 12-month trial follow-up period for participants randomised to the no SLT (control) group, who deteriorated to the point of needing therapy urgently for their speech or voice problems received SLT without delay via the usual local NHS services. At the end of the trial after their 12 months of patient and clinical assessments, participants in the control arm could be referred for SLT for their dysarthria by their usual care specialist through local NHS referral pathways.

Adherence to the interventions

As this was a pragmatic trial of two existing interventions in the NHS, our approach to fidelity was multidimensional. 56 The PD COMM pilot established that dosage was a key differentiating factor. For the practical purposes of statistical analysis, LSVT LOUD treatment adherence was defined as participants randomised to the LSVT LOUD arm receiving at least 14 out of the 16 prescribed LSVT LOUD sessions and having completed these sessions within 3 months of randomisation. A session was considered an LSVT LOUD session if at least 30 minutes of time was attributed to LSVT LOUD on the SLT treatment log (supervised or unsupervised). As a secondary assessment of LSVT adherence, we only included those who received all 16 sessions of LSVT within 3 months of randomisation. The first measure of adherence used a pragmatic approach, for example, a participant may have missed one session due to taking a holiday, whereas the second measure of adherence is strictly looking at those participants who received the intervention as prescribed.

As NHS SLT is not a prescriptive intervention, participants were considered adherent if they completed their SLT sessions within 3 months of randomisation that is they had their last session within 3 months of randomisation. These sessions should also not have any time attributed to LSVT LOUD. The NHS SLT was deemed to have been completed for a participant when a returned SLT log has been answered as ‘Yes’ to this being the last treatment session of the therapy course.

Adherence to the no SLT (control) arm was monitored through the resource usage form where participants could record whether they had received any SLT. If a participant in the no SLT (control) arm reported receiving SLT over the course of the 12-month follow-up, then they were considered non-adherent. The only exception to this was if SLT had been prescribed for the treatment of dysphagia only.

Training requirements

Prior to commencing recruitment, all site staff involved with the trial had to complete good clinical practice training. Key members of the site research team were required to attend either a meeting or a teleconference covering aspects of the trial design, protocol procedures, adverse event reporting, collection and reporting of data and record keeping. All therapists who took part were registered with UK regulatory body, the Health and Care Professions Council (HCPC), which sets standards for education, training and practice.

Only speech and language therapists or therapist assistants trained in LSVT LOUD, which includes a detailed manual and 2-year updates, could deliver the LSVT LOUD intervention. Therapists were provided with LSVT LOUD training through the trial by LSVT Global20 for free if they had not done this training or needed it to reregister as an LSVT therapist.

Ideally every therapist delivering LSVT therapy should have conducted at least three treatment sessions before delivering the intervention to trial participants; however, this was at the discretion of the trial sites.

Three workshops were held early into the trial, which brought together site speech and language therapists, research nurses and the research team to explore the PD COMM interventions. The workshop participants explored what was considered ‘core’ and ‘peripheral’ for each intervention and the barriers and challenges of delivering the interventions. The workshops helped to ensure that the trial ran smoothly.

Primary outcome

The primary outcome measure was the patient-reported Voice Handicap Index (VHI)57 total score at 3 months. The VHI comprises of 30 questions divided into emotional, functional and physical subscales. 57 It aims to assess the psychosocial consequences of voice disorders, and can be used to gain an overall perception of effectiveness of voice-related communication. The VHI total score ranges from 0 to 120 (with 0 being the best score and 120 the worst score) and the subscales range from 0 to 40.

Many previous trials have used vocal loudness as their primary outcome. It has been used as an outcome measure in an extended LSVT trial for PD16 and was also collected in the PD COMM Pilot trial. 36 The VHI was chosen as it was a patient-reported outcome that took little time to complete, was well-completed in our pilot trial36 and better reflected the focus of the PD COMM trial objectives.

Secondary outcomes

Patient-reported measures were used to assess the participant’s perception of how their voice impacted on daily activities and their quality of life, and to complement the primary outcome. The secondary outcome measures were:

• subscales of the VHI;57

• Parkinson’s Disease Questionnaire-39 (PDQ-39). 58

Quality of life was measured using the PDQ-39. 59 This is a validated, health-related quality-of-life measure specific to PD,59 and was the most widely used disease-specific quality-of-life rating scale for PD completed by the participant. It comprises of 39 questions divided into the following dimensions: mobility, activities of daily living, emotional well-being, stigma, social support, cognition, communication and bodily discomfort. The PDQ-39 summary index and each of the individual dimensions provide a score that can be converted into a 0–100 metric where 0 = no problem at all and 100 = worst or maximum level of problem.

• Questionnaire on Acquired Speech Disorders (QASD)60

Participation restriction related to speech and communication were assessed using the self-reported QASD. 60 The QASD questionnaire comprises of 30 questions which are scored 0–3 giving a total score that ranges from 0 to 90, where lower scores were better.

• EuroQol5D (5-level version). 61

The EQ-5D-5L61,62 is a well-established standardised instrument for use as a measure of health outcome. It provides a simple descriptive profile and a single index value for health status. It is completed by the participant, and comprises of the following five dimensions: mobility, self-care, usual activities, pain/discomfort and anxiety/depression. Each dimension can take one of five responses: no problems, slight problems, moderate problems, severe problems or extreme problems. There is also a 100-point visual analogue scale. It is often used together with resource utilisation questionnaires (see below) to provide data to inform the cost-effectiveness analysis.

• ICEpop Capabilities Measure for Older Adults (ICECAP-O).

The ICECAP-O63 is a measure of capability in older people for use in economic evaluations. Unlike most profile measures used in economic evaluations, the ICECAP-O focuses on well-being defined in a broader sense, rather than health. The measure covers attributes of well-being that were found to be important to older people in the UK and is completed by the participant. It comprises of five attributes: attachment (love and friendship); security (thinking about the future without concern); role (doing things that make you feel valued); enjoyment (enjoyment and pleasure) and control (independence).

• Adverse events (AEs) (see Chapter 2, Safety reporting).

• Hoehn and Yahr stage.

The Hoehn and Yahr stage64 is a clinician-rated measure of disease severity in PD. It is a standard staging scale for PD that is required to document the severity of PD in the participant population.

• Carer quality of life (Parkinson’s Disease Questionnaire – Carers).

Carer quality of life will be measured using the Parkinson’s Disease Questionnaire – Carer. 65 This is the first disease-specific measure of quality of life for carers of people with PD and is a validated and reliable tool. It is completed by the carer and comprises of 29 questions with 5 responses (Never/Occasionally/Sometimes/Often/Always). It is made up of four discrete scales: social and personal activities (12 items), anxiety and depression (6 items), self-care (5 items) and stress (6 items). The raw score of each scale can be calculated and converted to a 0–100 metric where 0 = no problem at all and 100 = worst or maximum level of problem. The sum of the scale scores can provide a single figure used to assess the overall quality of life of the individual questioned.

• Resource utilisation (collected for the Health Economic Evaluation).

Developed for use in the PD COMM Pilot, a disease-specific Resource Usage questionnaire was used to collect information on participant resource usage data. The questionnaire included items on primary care and secondary care healthcare utilisation, including the use of therapy services, and outpatient appointments. Further questions related to use of social services, including provision of meals and formal care. Finally, information was collected on time off work, participants’ out-of-pocket costs (e.g. travel, medication) and costs incurred by informal carers, to inform analysis from a societal perspective.

Global rating scale (transition item)

The transition item is a single question asked at the 3-month time point to the participant and carer: ‘Compared to 3 months ago (when you joined the trial), has your ability to communicate using speech changed?’ with seven levels of response ranging from ‘much worse’ to ‘much better’ or ‘Compared to 3 months ago (when your partner/family member/friend joined the trial), has his/her ability to communicate using speech changed’ is used for the participant and carer respectively. It measures whether the participant or carer has noticed any change in communication by voice or speech since the participant (with PD) entered the trial. The transition item was used to calculate the MCID for the VHI.

Sequence generation and allocation concealment

Following informed consent and completion of all baseline data collection, the participant could be randomised into the trial. Participants were randomised at the level of the individual via a central, secure, web-based computer-generated randomisation system developed and controlled by the BCTU, thus ensuring concealment of next treatment allocation. To randomise a patient into the trial, staff delegated the task of randomising patients into the trial either logged onto the trial database or rang the BCTU randomisation telephone line. The randomisation process used a minimisation procedure. The following minimisation variables were used:

• age (≤ 59, 60–70, > 70 years);

• disease severity measured using the Hoehn and Yahr staging64 (1.0–2.5, 3.0–5.0); and

• severity of speech measured using the VHI57 total score (≤ 33, mild 34–44, moderate 45–61, severe > 61).

To avoid any possibility of the treatment allocation becoming predictable, a random element was included within the randomisation process. Once the participant was randomised into the trial, they were given a unique trial identifier and the treatment allocation was confirmed by e-mail to the site.

Data completeness

Data returns and data completeness were good throughout the trial (see Appendix 1, Tables 21 and 22). The participant completed VHI, PDQ-39 and QASD had response rates of over 85% at each time point. As did the quality-of-life questionnaire completed by the carer (PDQ-Carer). The clinician-completed forms had a response rate of over 90% (99% at baseline; 94% at 12 months).

LSVT LOUD

Lee Silverman Voice Treatment LOUD was delivered over a median of 16 sessions and a mean of 6.6 (SD 6.5) weeks. The dose of the treatment was 1216 (SD 454) minutes, which consisted of 963 (SD 330) minutes of SLT content [mean SLT content per session: 63 (SD 10) minutes], with 752 (SD 287) minutes dedicated to LSVT LOUD [mean LSVT LOUD content per session: 54 (SD 7) minutes]. Participants were encouraged to continue with home-based practice and completed homework diaries. Ninety-six participants completed at least 1 week of the home-practice diary. Diary completion was high with a median completion of 4 weeks.

National Health Service speech and language therapy

National Health Service SLT was delivered over a median of five sessions over a mean of 11.4 (SD 11.4) weeks. The duration of NHS SLT was shorter than LSVT LOUD, totalling a mean of 404 (SD 234) minutes, with 298 (SD 171) minutes dedicated to SLT content. Mean active therapy time for NHS SLT was 149 (SD 113) minutes. The mean individual session length was similar to that of LSVT LOUD at a mean of 55 (SD 16) minutes. Therapy assistants provided a small number of treatment sessions across interventions and a few participants received the course of therapy in a group setting.

Primary analysis

The impact of the speech or voice problems at baseline across the three randomised groups was similar: a mean of 44.6 (SD 21.9, n = 130) in the LSVT LOUD group, 46.2 (SD 24.8, n = 129) in the NHS SLT group and 44.3 (SD 22.3, n = 129) in the no SLT (control) group. At 3 months all three group’s voice impact scores were lower compared with baseline, indicating less impact from speech or voice problems: the LSVT LOUD group had a mean score of 35.0 points (SD 20.1, n = 106), the NHS SLT group had a mean score of 44.4 points (SD 24.8, n = 102) and the no SLT (control) group had a mean score of 40.5 points (SD 21.5, n = 98) (Table 5).

The interventions were compared against each other for relative impact of speech or voice problems at 3 months by calculating the mean difference in VHI total score between groups. At 3 months, the VHI total score for the LSVT LOUD group was 8 points lower than for the no SLT (control) group [−8.0, 99% CI (−13.3 to −2.6); p = 0.0001]. For NHS SLT, at 3 months, the VHI total score was 1.7 points higher than the no SLT (control) group [1.7, 99% CI (−3.8 to 7.1); p = 0.4]. In the third comparison, the LSVT LOUD group was 9.6 points lower than the NHS SLT group [−9.6, 99% CI (−14.9 to −4.4); p < 0.0001] (see Table 5).

Per-protocol analyses

The main per-protocol analysis population included only those participants who both adhered to treatment and completed the 3-month VHI outcome assessment within the 1-month time window. This analysis gave similar results to the main ITT analysis: LSVT LOUD versus no SLT: −9.7, 99% CI (−16.0 to −3.4); NHS SLT versus no SLT: 1.1, 99% CI (−5.6 to 7.8); and LSVT LOUD versus NHS SLT: −10.8, 99% CI (−17.8 to −3.8). Similar results were also seen for the two other per-protocol analyses (see Appendix 1, Table 23).

Sensitivity analyses to assess impact of missing data

Various sensitivity analyses were undertaken to assess the impact of missing data. Different assumptions were made about the reasons for missing data to investigate the impact, if any, to our analysis of the primary outcome. All these analyses gave results that were in agreement with the primary ITT analysis (see Appendix 1, Table 24). For example, the sensitivity analysis using multiple imputation gave the following results: LSVT LOUD versus no SLT: −6.9, 99% CI (−10.9 to −2.9); NHS SLT versus no SLT: 1.8, 99% CI (−2.3 to 5.9); and LSVT LOUD versus NHS SLT: −8.7, 99% CI (−12.7 to −4.7).

Emotional subscale

At 3 months, the LSVT LOUD group had a lower score (i.e. better) than the no SLT (control) group [−3.0 (−5.1, −0.9); p = 0.0003], while there was no difference for standard NHS SLT compared to no SLT (control) [0.2 (−1.9, 2.4); p = 0.8]. LSVT LOUD resulted in a lower (i.e. better) emotional subscale score compared to standard NHS SLT [−3.2 (−5.3, −1.1); p < 0.0001] at 3 months. Similar results were seen for the comparisons at 6 and 12 months, and over the whole 12-month trial follow-up period (see Figure 4 and Appendix 1, Table 28).

FIGURE 4.

Mean VHI emotional subscale score profile plot with error bars over 12 months by intervention.

Functional subscale

At 3 months, the LSVT LOUD group had a lower score (i.e. better) than the no SLT (control) group [−2.9 (−4.8, −1.1); p < 0.0001], while there was no difference for NHS SLT compared to no SLT (control) [−0.0 (−1.9, 1.9); p = 1.0]. LSVT LOUD resulted in a lower (i.e. better) functional subscale score compared to NHS SLT [−2.9 (−4.7, −1.1); p < 0.0001] at 3 months. Similar results were seen for the comparisons at 6 and 12 months, and over the whole 12-month trial follow-up period (see Figure 5 and Appendix 1, Table 28).

FIGURE 5.

Mean VHI functional subscale score profile plot with error bars over 12 months by intervention.

Physical subscale

Only LSVT LOUD resulted in a lower (i.e. better) physical subscale score compared to NHS SLT [−2.2 (−4.1, −0.3); p = 0.003] at 3 months. Similar results were seen for the comparisons at 6 and 12 months, and over the whole 12-month trial follow-up period (see Figure 6 and Appendix 1, Table 28).

FIGURE 6.

Mean VHI physical subscale score profile plot with error bars over 12 months by intervention.

Design

The process evaluation was conducted as an integral part of the PD COMM trial (Figure 17) and it applied NPT to examine normalisation from the perspectives of both staff delivering the intervention and patient participants who needed to embed the trial interventions into their daily routine. All details of the process evaluation design and methods were published as a protocol for the study. 75

FIGURE 17.

PD COMM process evaluation flow chart.

This process evaluation used a variety of data collection methods. 75

Therapist questionnaire (TQ): All speech and language therapists and SLT assistants involved in the trial were asked to complete an online questionnaire at two time points: prior to the start of intervention delivery and after they had treated their last PD COMM participant. The TQ consisted of three sections A, B and C. Section A collected data on therapists’ roles and years of experience; section B consisted of an adapted version of the NPT NoMAD tool76 to evaluate attitudes towards the LSVT and NHS SLT interventions across the four NPT domains and the therapists’ self-efficacy as measured by the general self efficacy (GSE) questionnaire. 77 Section C comprised the LSVT skill set questionnaire78 to evaluate therapists’ confidence with completing LSVT in practice settings.

Semistructured in-depth interviews with study participants. Purposeful sampling ensured engagement of trial participants with different ages, PD severity (Hoehn and Yahr stage) and the presence of a family carer.

Semistructured in-depth interviews with study therapists responsible for delivering the two active treatment arms. Interviews were carried out at two time points: estimated midway through therapists’ involvement in the trial, and at least 1 year after they joined the trial.

All interviews were digitally recorded and fully transcribed and then checked by the interviewer, keeping all names of participants anonymous. All participants were asked to give consent to the recording prior to the start of the interview.

Quantitative and qualitative sets of data were analysed independently before being combined.

Quantitative data: therapist questionnaires

The online TQ was completed by 71 therapists prior to start of intervention delivery (time point 1). Of those therapists, 26 (36.6%) went on to complete the questionnaire approximately 1 year into the trial (time point 2). Therapists who completed the questionnaire were spread across all trial sites. Descriptive statistics were carried out to indicate the frequency of responses. Due to the limited sample size for time point 2 no within-group and between-group analyses were possible.

Of those therapists who completed the TQ 8% were Agenda for Change Band80 5, 39% were band 6 and 53% were band 7 or above. Therapists had different levels of experience of working as SLTs (Figure 18).

FIGURE 18.

Distribution of PD COMM speech and language therapists according to years of experience.

Sixty-nine per cent of therapists delivered LSVT LOUD as part of their current role in their Trust. Among different years of experience there was a similar spread across those who normally delivered LSVT as part of their role and those who did not (Figure 19).

FIGURE 19.

Distribution of SLTs according to years of experience and delivering LSVT as part of their current role.

Eighty-four per cent of the therapists had a GSE Total Score of 30 or higher. All therapists had high scores regardless of their years of experiences practising as a speech and language therapist (Figures 20 and 21).

FIGURE 20.

Distribution of GSE scores across all PD COMM therapists with different levels of experience grouped depending on whether SLTs deliver LSVT as part of their current role.

FIGURE 21.

Distribution of degree of familiarity of SLTs with LSVT and NHS SLT interventions across different levels of professional experience.

The answers to questions included in the adapted version of the NoMAD tool (Part B of the TQ) that addressed issues around familiarity with PD COMM interventions, the results show that therapists were overall familiar with both NHS and LSVT interventions and saw them as a normal part of their role. Therapists showed equipoise when it came to both interventions regardless of whether or not they delivered LSVT as part of their normal everyday job. Therapists were asked to rank from 1 to 10 how familiar both interventions felt to them. Survey data show a similar spread of responses across both interventions (see Figure 21).

All questions/items in section B of the TQ were grouped according to four of the NPT constructs that they represented (Coherence, Cognitive participation, Collective action and Reflexive monitoring). Data were analysed to describe the spread of responses (1 – Strongly disagree up to 5 – Strongly agree) and the mean value for each of the constructs for each of the interventions. Results show further evidence that confirms balance across both interventions (Figure 22).

FIGURE 22.

Mean score across four NPT constructs for each of the PD COMM interventions.

The last section of the questionnaire, section C (LSVT Skills level questionnaire), which explored therapists’ level of confidence in their ability to deliver all components of LSVT, showed that overall, all therapists who completed the questionnaire felt they were able to implement all aspects of the LSVT intervention (3 – ‘Got it’). This was true even for those who had recently completed their pre-registration training (Figure 23).

FIGURE 23.

Mean scores (from 1 – ‘I don’t get it’ to 3 – ‘Got it’) for each of the eight sections of the LSVT skills questionnaire, across all therapists with different levels of experience.

In summary, data collected in this questionnaire provided clear evidence of equivalence across PD COMM therapists with regard to their familiarity and preparedness to deliver both PD COMM SLT interventions.

Qualitative semistructured interviews

PD COMM therapists – identified themes

PD COMM participants – identified themes

Mapping of identified themes on to the normalisation process theory constructs

Once all themes from participants’ and therapists’ interviews were described in detail the research team mapped them on to the NPT framework74 and its five constructs (Table 10). This exercise revealed that for both, therapists and participants:

1. the trial and its aims ‘made sense’ in its own right

2. that they were ready to ‘drive it’ and ‘make it happen’

3. that they considered it the ‘right thing to do’ in order to bring about a positive outcome for those people affected by PD, their therapists and their SLT services

4. that therapists and participants were able to identify their own and the trial’s limitations and how these could play a role in the final outcome of the study.

An explanatory framework and model of ‘equipoise’ for the PD COMM trial interventions

The analysis and final synthesis of all data from all sources show that, via carrying out this process evaluation, the research team has been able to understand the process by which therapists have been able to deliver and implement and PD COMM intervention and participants have received such interventions. Based on the data, we are confident that equipoise between both PD COMM interventions was reached; in other words, there are no data to suggest that such equipoise has been compromised (Figure 24).

FIGURE 24.

Model of equipoise between PD COMM interventions.

Throughout their involvement in PD COMM, speech and language therapists have successfully balanced the demands placed on them as both, therapists and researchers. They have also been able to reach agreement balancing their clinical decision-making versus their need to remain true to the PD COMM intervention protocol. This balancing act has been a direct result of the learning that has taken place over time.

Speech and language therapy treatment record forms

Speech and language therapists or assistants completed one TRF for each therapy session. They documented minutes spent undertaking broad categories of therapy activities.

Broad categories for intervention components (Table 11) were agreed and piloted36 prior to the development of the TRF. Categories reflected clinical dysarthria management guidelines from the Royal College of Speech and Language Therapists which recommend a combination of educational, physiological, compensatory and augmentative approaches within a framework of body function and structures (impairment), activity, participation and contextual factors.

Therapy notes

Speech and language therapists (or assistants) routinely completed TNs as part of their usual practice after each therapy session. Data extraction from TNs was conducted under the following categories:

1. why the therapy was being done (rationale, goals and delivery plans)

2. what materials and equipment were reported as being used (technologies, biofeedback, numerical measurement, named approaches)

3. what therapy processes were reported (targets, cues, vocal activities, practice structure)

4. tailoring (personal and interpersonal).

Meanwhile, TRFs recorded the number of sessions (with/without direct contact), nature of sessions delivered (assessment/intervention), the provider (SLT/assistant/supervised student), duration of therapy (first to last session), therapy location, mode of delivery, time allocation assessment, goal setting, information provision and therapy (impairment, compensatory, AAC, generalisation, training, LSVT), use of LSVT companion, other therapist activities relating to patient treatment (liaison, other).

Technologies

Use of exercise materials was reported for almost all participant intervention sessions for both the NHS SLT and the LSVT LOUD; sheets, lists, pictures for description, reading passages and participant’s own materials such as magazines or books. LSVT-specific materials such as LOUD card prompts and sheets from the manual are featured. 26 Reference to using telephones or video calls as part of intervention (whether in the session or as a home-based practice task) was made in 69% of LSVT LOUD but only 13% of NHS LOUD^® notes.

The reported use of apps differed between the two therapy interventions (LSVT LOUD 37%; NHS SLT 17%), but the range of apps reported was more varied in NHS LOUD^® (which uniquely featured Voice Meter Pro,82 DAF pro,83 Bla Bla Bla84 and a trial of ClaroCom85 as AAC). In contrast, 10/19 reported apps used in the LSVT LOUD interventions specified Voice Analyst,86 and one used an app to encourage vocal effort through creating background noise. Parkinson’s UK apps87 and Decibel 1088 were reported in both interventions.

Biofeedback

Where biofeedback was mentioned in therapy activities, it was difficult to extract where visual or auditory biofeedback was being used. Visual biofeedback was described in 86% (44/51) LSVT LOUD and 57% (31/54) NHS LOUD^® interventions. Auditory biofeedback was reported in 41% (21/51) LSVT LOUD and 37% (20/51) NHS SLT interventions. Use of both visual and auditory biofeedback was reported in 37% (19/51) LSVT LOUD and 27% (14/54) NHS SLT interventions.

Numerical measurement

Numerical measurement was common in LSVT LOUD, whether from an object (such as a sound meter; 78%, n = 40/51) or Companion Software (16%; n = 8/51). This is compared with 52% (n = 28/54) of participants receiving NHS SLT using numerical measurement and two participants (4%) that used Companion Software. Use of self-rating scales was reported in 53% (n = 27/51) of LSVT LOUD and 39% (n = 21/54) of NHS SLT notes. Self-rating of effort is an expected component of LSVT interventions.

Intervention targets (speech subsystems)

A high proportion of LSVT LOUD interventions had one speech impairment target. Few had two. In contrast, half of NHS SLT interventions described a focus on three or more speech subsystems followed by two subsystems and the remaining interventions focused on one speech subsystem impairment target. The most common speech subsystems focused on within NHS SLT were phonation and breath support. Effectively all LSVT LOUD targeted one impairment target – phonation; one was ‘unreported’ due to limited participant therapy (Table 12).

Other intervention targets

Therapists described activities to increase participants’ insight into what was different in their communication as a consequence of PD, why this would make it harder for other people to understand them, and what they need to do to address it. In LSVT interventions this re-training of sensory perceptions is referred to as calibration. Both NHS SLT and LSVT LOUD interventions had high numbers of therapy reports targeting improved insight, with LSVT LOUD higher than the NHS SLT intervention. Both interventions reported a similarly low-level use of strategies to modify the environment (e.g. by reducing background noise) or deliberate use of background noise to encourage participants to use their loud voice.

Voice and speech cues

‘Loud’ or ‘Think Loud’ was reported by almost all LSVT LOUD records and in most cases, was the only recorded cue. In comparison, although a majority of NHS SLT cues reported were also ‘Loud’ or ‘Think Loud’ there was more variation with another cue referencing volume such as ‘strong’ voice described. Other cues documented in NHS SLT included ‘clear’ referencing articulation; those that referenced exaggerated movements; and those that cued ‘slow’ or ‘chunking’ of speech into smaller groups of words with pauses.

Vocal activities

Vocal activities were reported and there were differences in the use of most of the activities mentioned between the SLT treatment groups (Table 13).

Trial interventions

All resource use involved in the delivery of the trial interventions was obtained from the SLT TRFs, where the number and duration of NHS SLT and LSVT LOUD sessions attended by each participant were recorded. Details on who conducted the session, location and type of the session (one to one/group) were also collected. Additional sessions that were recorded in the SLT Initial Interview Log and not in the SLT treatment log were also added.

Sessions were costed individually based on the duration and location assuming that all sessions were conducted one to one. Both PD COMM SLT interventions were mainly delivered by a therapist, therefore, a cost of a therapist time in minutes (average of band 6 and 7) was used for all sessions. The cost of time of a therapist was multiplied by duration of contact (in minutes) to determine the total cost of the intervention per patient for LSVT LOUD and NHS SLT. The duration of contact included the time spent for data management, preparation and delivering the intervention and these were considered in the cost of the intervention. However, the duration of PD COMM SLT content was also estimated to calculate the cost of SLT content and presented as a subcategory of cost. Where a session was conducted in the participant’s home, 60 minutes of travel time was added to the total session time. For LSVT LOUD remote sessions, the costs of a computer loan from the NHS and companion software were added, and NHS SLT remote sessions were assumed to have delivered over the phone. The cost of LSVT training for therapists per patient was estimated using national data from SLT service providers with assumptions (see Appendix 2, Table 36) and was also included in the cost of the intervention. Experts were consulted to confirm that the assumptions were appropriate.

Additional SLT sessions for speech and/or dysphagia (outside the trial interventions) reported by participants were recorded and costed as 60 minutes standard sessions by a therapist. Reported SLT sessions in the treatment log by participants randomised to the control arm who had NHS SLT were considered as additional sessions outside the trial intervention. To avoid double counting, therapist reported SLT sessions and SLT services from the follow-up questionnaire completed by participants were compared and trial sessions were deducted from total additional SLT services (primary or secondary).

Missing data

Participants who returned the follow-up questionnaire (resource use and effectiveness) at any follow-up time point were included in the base-case analysis. A multiple imputation approach was applied to replace missing values for costs and health and capability outcomes, under the assumption that the variables with missing values were Missing At Random (MAR). This approach replaces missing observations with a set of plausible predicted values, derived from a posterior predictive distribution of the missing data given the available data. 107 Twenty-five data sets were imputed (reflecting the percentage of missing data), which were then combined using Rubin’s rules. 108 Age, sex, severity of PD (Hoehn and Yahr) and severity of speech impairment (VHI) were included in the imputation models. Total costs over 12 months were imputed for different cost components, whereas missing values for the EQ-5D-5L and ICECAP-O were imputed at each time point.

For participants who were known to have died after the first follow-up visit, a value of zero at the next follow-up point was assigned to EQ-5D-5L and ICECAP-O scores based on the date of death. For participants who died, resource use and costs were given a value of zero for all subsequent follow-up time points. EQ-5D-5L and ICECAP-O overall scores were considered as missing if the participant failed to respond to all questionnaire items. For all participants, if they returned the resource usage questionnaire and a response to an individual item was not completed, it was considered that the service was not utilised.

To explore the impact of missing data on cost-effectiveness, a complete case analysis was carried out where only patients with complete data on outcome measures and costs were included.

Statistical analysis

Outcome measures, resource use and costs were presented as means and standard deviations (SDs) and mean bootstrapped differences between interventions with 95% CIs, where applicable. Due to cost data being positively skewed, 95% CIs around the mean difference in costs were calculated using the bias-corrected and accelerated (BCa) bootstrap method using the analyses of 1000 resamples. 109 If the CIs of the difference in mean costs or outcomes between treatment arms did not cross zero, this indicated a significant difference between treatment arms. 109

Differences between treatments for all outcomes and cost components were adjusted for age at randomisation, outcome and sex as well as severity of PD (Hoehn and Yahr) and severity of dysarthria (VHI). The baseline score (EQ-5D-5L and ICECAP-O) was also included in the adjustment for outcome measures and calculation of total QALYs, YFC and YSC. 110

All analyses were performed using Stata^® version 16 (StataCorp LP, College Stations, TX, USA).

Economic evaluation

The first stage of the analysis involved reporting resource use, costs and health outcomes in a disaggregated manner by treatment arm. 103

The base-case analysis took the form of a cost–utility analysis (CUA) from the perspective of the UK NHS using total costs and QALYs for each trial arm of the trial. In line with the statistical analysis, economic analyses were conducted on an ITT basis. Differences between treatment arms in total costs were divided by differences in the outcome measure (QALYs) to calculate the incremental cost-effectiveness ratio (ICER), expressed as the additional cost per QALY gained. ICERs were generated for three comparisons: LSVT LOUD versus no SLT (control) NHS SLT versus no SLT (control); and LSVT LOUD versus NHS SLT. The NICE upper and lower thresholds of £30,000 and £20,000 per QALY were used to determine if an intervention was cost-effective. 89 Secondary analyses presenting the incremental cost per YFC and YSC calculated from ICECAP-O were also conducted. Given that trial involvement was for only 12 months, discounting was not applied to either the costs or outcomes. Further analyses were undertaken from the NHS/PSS perspective to include social care costs and a wider societal perspective to include the costs incurred by the participants and productivity losses.

Analysis of uncertainty

Bootstrapping was used to account for the overall uncertainty that occurs because of variations in sampling, by jointly bootstrapping mean cost and QALY differences. 111 The technique generated 5000 paired values of incremental costs and QALYs for the three treatment comparisons and were presented on a cost-effectiveness plane as a scatterplot to aid interpretation. 111,112 A cost-effectiveness acceptability curve (CEAC) was then constructed to reflect the probability that an intervention is cost-effective at different willingness-to-pay values per QALY gained. 113

Deterministic sensitivity analyses were conducted to explore the uncertainty in assumptions related to the delivery of the intervention and resource use. A series of different scenarios for LSVT LOUD delivery were explored by varying the location of session delivery (online or face to face), the staff grade of the session supervisor and supervised or unsupervised sessions. These scenarios were discussed with experts to agree on a typical scenario and sensible alternative options to reflect potential modes of delivery. Session delivery settings (outpatient or home delivery), level of session supervisors (therapist or therapy assistant), the method of delivery (face to face, online) and supervised or unsupervised sessions were varied using a combination of scenarios.

The following scenarios were explored:

1. All sessions are delivered by a therapist.

   1. Scenario 1: 100% of session in an outpatient setting.

   2. Scenario 2: 50% of the sessions are face to face in an outpatient setting and 50% as online sessions.

   3. Scenario 3: 50% of the sessions are in an outpatient setting and 50% at participant’s home.

   4. Scenario 4: 50% of the sessions are face to face at participant’s home and 50% as online sessions.

   5. Scenario 5: 100% of sessions at participant’s home.

2. 50% of the sessions are delivered by a therapist and 50% by a trained therapy assistant.

   1. Scenario 6: 100% in an outpatient setting.

   2. Scenario 7: 50% of the sessions are face to face in an outpatient setting and 50% as online sessions.

   3. Scenario 8: 50% of the sessions are in an outpatient setting and 50% at participant’s home.

   4. Scenario 9: All sessions delivered by a therapist where 50% of the sessions are face to face at participant’s home and 50% as online sessions.

   5. Scenario 10: 100% at participant’s home.

3. 50% of sessions are supervised and 50% are unsupervised.

   1. Scenario 11: 50% of the sessions are in an outpatient setting and the remaining 50% unsupervised at participant’s home.

   2. Scenario 12: 50% of the sessions are supervised at participant’s home and the remaining 50% unsupervised at participant’s home.

A single scenario for NHS SLT was used assuming that sessions were conducted in an outpatient setting. Online LSVT LOUD sessions and a mix of face-to-face and unsupervised LSVT LOUD sessions were assumed to be as effective as face-to-face sessions in terms of quality of life and speech outcomes. 114 Therefore, the QALY values for intervention groups from the trial data were held constant and used to calculate cost per QALY for each scenario. Further analyses were carried out. The first was related to adherence where only participants who were adherent to SLT interventions were included. The second assuming a vocal folds examination by an ENT consultant was undertaken for participants in the LVST LOUD arm before their first SLT initial session.

Data completion

Complete EQ-5D-5L data at baseline, 3, 6 and 12 months were available for 273 participants (92 LSVT LOUD, 89 NHS SLT and 92 no SLT (control)) out of 388 participants recruited, and 270 participants [91 LSVT LOUD, 89 NHS SLT and 90 no SLT (control)] had completed ICECAP-O at all time points.

The SLT TRF was completed by therapists or therapy assistants for 107 (82%) participants in the LSVT LOUD group and 117 (91%) in the NHS SLT group. Resource use data at all time points were available for 93 participants in the LSVT LOUD arm, 92 in the NHS SLT arm and 92 in the no SLT (control) arm. A total of 116 (88% of recruited) participants in the LSVT LOUD arm returned follow-up questionnaires (on effectiveness and resource use) at any time point, 113 (88%) in the NHS SLT and 112 (87%) in the no SLT (control) groups and were included in the base-case analysis where missing values were imputed.

Of the 388 participants randomised in the trial, 253 (63%) had complete data on resource use, treatment, EQ-5D-5L and ICECAP-O. Details on data completion and proportion of returned forms/questionnaires are presented in Appendix 2, Table 38.

Trial interventions

Table 15 presents the mean resource use and costs for trial interventions per participant including therapist time and travel time with additional costs including phone calls, LSVT training, companion software and laptop loan. The average cost of LSVT LOUD was £1315.55 (SD, 570.47) and £412.06 (SD, 275.01) for NHS SLT. Total session time was higher in LSVT LOUD group (1232.70 minutes compared with 412.31 minutes in the NHS SLT arm) and contributed to almost 82%–88% of the total cost of the SLT interventions. Very few participants used the LSVT companion software. The average number of SLT sessions was 16 for LSVT LOUD and 6 for NHS SLT.

The data from the treatment log showed that the majority of therapy sessions took place in an outpatient setting (73% in LSVT LOUD group vs. 77% in NHS SLT group). Almost 94% of LSVT LOUD sessions and 92% of NHS SLT sessions were supervised by a therapist. Only 26% of LSVT LOUD sessions took place in participants’ homes (6% of these sessions were unsupervised) compared with 16% of NHS SLT sessions. Remote sessions contributed to 1% of LSVT LOUD sessions and 6% of NHS SLT sessions.

Participants in both LSVT LOUD and NHS SLT arms required non-trial additional SLT sessions (an average of 2.53 per participant for LSVT and 2.01 for NHS SLT) with a small number of sessions reported by those in the no SLT (control) arm (average 0.90). These additional sessions were in part due to outliers, where 7 LSVT LOUD participants had attended at least 10 additional sessions. The clinical reason for these additional sessions (e.g. speech or voice problems, dysphagia) was not recorded.

National Health Service and social services resource use

Disaggregated mean resource use and costs per participant for those with complete resource use data over 12 months are presented by treatment group in Table 16. The level of resource utilisation was similar across treatment arms. The main cost drivers of primary care and therapy services were GP practice visits and PD nurse and physiotherapy appointments. PD-related outpatient appointments and consultations were comparatively more frequent in the NHS SLT arm compared to LSVT LOUD and no SLT (control) arms.

Social services resource use was reported by 20 (17%) LSVT LOUD, 17 (15%) NHS SLT and 16 (14%) no SLT (control) participants. Use of social services resources was much greater per person in the LSVT LOUD arm compared to NHS SLT and no SLT (control) arms and home help was the main cost component due to the presence of a small number of outliers. An observation was considered to be an outlier where the patient had over 120 visits over the 12-month period. Given that SLT interventions are highly unlikely to have any impact on physical functioning, normal daily activities or the requirement for institutional care, a decision was made to deviate from the original protocol and to exclude social services costs for the base-case analysis. Therefore, the base case was presented for the NHS perspective.

The results indicated that all arms had large standard deviations for the costs, particularly in the LSVT LOUD arm, driven by the skewed cost data caused by outliers.

Total costs

The mean total costs per participant for different cost categories across treatment arms are presented in Table 17. The greatest costs were PD medications, the cost of the SLT interventions and visits to primary care services and therapists. The cost of medication was highest for participants in the LSVT LOUD arm, who also had a higher mean baseline severity of PD. The aggregated cost of primary care, therapist and secondary care services were similar across treatment arms.

In terms of mean cost differences by category of resource use (see Table 17), the greatest differences were for the LSVT LOUD and NHS SLT interventions, both compared with each other and the no SLT (control) arm, and differences were statistically significant. Differences in the cost of additional SLT sessions were also statistically significant for both SLT interventions versus no SLT (control) but not when LSVT LOUD and NHS SLT were compared.

The mean total NHS cost differences indicated that LSVT was associated with additional costs of £1582.18 (95% CI, £1325.21 to £1892.23) and £652.76 (95% CI, £381.69 to £934.26) compared with no SLT (control) and NHS SLT, respectively. NHS SLT was £924.20 (95% CI, £641.08 to £1219.78) more expensive than no SLT (control). Differences in total costs were statistically significant. There were only small differences between trial arms for primary care services, therapists, secondary care services and medication. The NHS/PSS cost difference between LSVT LOUD and the other trial arms was much greater due to high social services costs reported by a small number of participants in the LSVT LOUD arm.

When considering the wider societal perspective with PSS costs, the difference in costs between LSVT LOUD and no SLT (control) was lower, and LSVT LOUD was associated with an additional cost of £922.56 (95% CI, £−26.65 to £1788.62). Compared with NHS SLT, LSVT LOUD had slightly higher costs of £118.67 (95% CI, £−834.29 to £1022.71). The difference in costs between NHS SLT and no SLT (control) was £754.03 (95% CI, £−251.52 to £1757.57). All differences were not statistically significant except for the difference between LSVT LOUD and no SLT (control). The reduced cost difference when the societal perspective was adopted was mainly related to the inclusion of productivity costs. These cost differences increased when social services costs were also included.

Base-case analysis

Results of the base-case analyses from the perspective of the UK NHS are presented in Table 18. The results suggest that LSVT LOUD was associated with an ICER of £197,772 per QALY gained and £77,017 per QALY gained compared to no SLT (control) and NHS SLT, respectively. Using the capability outcome measure, the cost of achieving an additional YFC for the LSVT LOUD arm was £87,899 compared with no SLT (control) and £46,210 compared with NHS SLT. The ICERs using the YSC as the measure of outcome were £121,706 and £51,344 per YSC gained for LSVT LOUD versus no SLT (control) and NHS SLT, respectively. For all three outcomes, standard NHS SLT was less effective and more costly than no SLT (control) and is therefore by definition ‘dominated’ by no SLT (control) and was not cost-effective.

The distribution of bootstrapped incremental costs and incremental effects (QALYs) presented on the cost-effectiveness planes supported the base-case results. The replications for LSVT LOUD versus no SLT (control) were in the north-west and north-east quadrants (Figure 25) showing that LSVT LOUD was more costly, with great uncertainty around the effectiveness (either more effective or marginally less effective). Similarly, compared with NHS SLT, LSVT LOUD was also more expensive and associated with both marginally more and fewer QALYs as illustrated in Figure 26. Estimates of incremental costs and QALYs between NHS SLT and no SLT (control) were also located in the northeast and northwest quadrants as demonstrated in Figure 27, with more points showing fewer QALYs. Overall, the wide spread of the replications with outliers reveals the degree of uncertainty around the effectiveness results.

FIGURE 25.

Scatterplot on the cost-effectiveness plane representing 5000 estimates of incremental costs and QALYs for the base-case analysis (NHS perspective) comparing LSVT vs. no SLT (control).

FIGURE 26.

Scatterplot on the cost-effectiveness plane representing 5000 estimates of incremental costs and QALYs for the base-case analysis (NHS perspective) comparing LSVT vs. NHS SLT.

FIGURE 27.

Scatterplot on the cost-effectiveness plane representing 5000 estimates of incremental costs and QALYs for the base-case analysis (NHS perspective) comparing NHS SLT vs. no SLT (control).

At NICE’s threshold of £20,000 per QALY, the CEACs show that the probability of LSVT LOUD being cost-effective was 0% and 3% compared to no SLT (control) NHS SLT, respectively (Figure 28). The probability of NHS SLT being cost-effective versus no SLT (control) was 2% (Figure 29). Considering the higher willingness-to-pay threshold of £30,000 per QALY, LSVT still had a low probability being cost-effective compared with NHS SLT and no SLT (control), with a probability of being cost-effective of 14% compared when compared with NHS SLT (Figure 30).

FIGURE 28.

Cost-effectiveness acceptability curve (CEAC) showing the probability of LSVT LOUD being cost-effective compared to no SLT (control) for the base-case analysis (NHS perspective).

FIGURE 29.

Cost-effectiveness acceptability curve (CEAC) showing the probability of NHS SLT being cost-effective compared to no SLT (control) for the base-case analysis (NHS perspective).

FIGURE 30.

Cost-effectiveness acceptability curve (CEAC) showing the probability of LSVT LOUD being cost-effective compared to NHS SLT for the base-case analysis (NHS perspective).

Sensitivity analysis

A number of sensitivity analyses were conducted to explore the impact of including social care costs, considering the wider societal perspective (with and without social care costs), missing data, adding the cost of ENT examination at the beginning of the treatment and participant adherence with the SLT protocol. Almost all the findings supported the results from the results of base-case analysis (see Appendix 2, Table 41). For the societal perspective (with and without social care costs) LSVT LOUD was also not cost-effective when compared to no SLT (control) if the society’s willingness to pay was £20,000 per QALY. However, when excluding social care costs, for LSVT LOUD versus NHS SLT the ICER was £9899 per QALY, suggesting that LSVT LOUD may be cost-effective. This result was due to the lower productivity costs in the LSVT LOUD arm. For both productivity costs and social services costs, the results were influenced by outliers from a very small number of participants. These costs are unlikely to be related to the intervention and therefore these estimates should be interpreted with caution.

The inclusion of only patients with complete data on all NHS costs and outcomes reduced the ICER to £78,281 per QALY for LSVT LOUD versus no SLT (control) £65,290 per QALY versus NHS SLT. Participant adherence resulted in greater QALY gains for the LSVT LOUD arm and a reduced ICER for LSVT LOUD versus NHS SLT of £32,291 per QALY.

Scenario analyses

Alternative scenarios were explored for the delivery of LSVT LOUD sessions. The cost of LSVT LOUD from each scenario was used to replace the trial LSVT LOUD cost and added to the other healthcare costs to generate the total cost per participant from the perspective of the NHS. The results of the scenario analysis are presented in Table 19 and show the new cost of delivering LSVT LOUD, the mean differences in costs and QALYs and the ICERs for each scenario. The analysis demonstrated that the least expensive scenario for LSVT LOUD delivery was a combination of supervised and unsupervised sessions. This scenario reduced the ICER to £27,808 per QALY compared with NHS SLT. If all sessions were supervised, conducting all sessions in an outpatient setting was the least costly method to deliver LSVT LOUD where sessions delivered by a mix of therapists and therapy assistants less costly than all sessions conducted by a therapist.

Clinical outcomes

LSVT LOUD was more effective at reducing participant-assessed impact of voice and speech problems than NHS SLT and no SLT (control) after 3 months. These results remain robust when the potential effects of non-adherence to treatment were investigated. The continued benefit at 12 months of LSVT LOUD on the impact of voice problems compared to NHS SLT and no SLT (control) is encouraging, but it remains possible that re-intervention may be required should the treatment effect wear off beyond 12 months. There was a benefit in communication-related quality of life for participants shortly after LSVT LOUD which exceeded the MCIC. In terms of AEs, the higher rate of vocal strain with LSVT LOUD treatment is generally a minor issue and is at an acceptable rate in relation to the level of benefit.

National Health Service SLT, which reflected current non-LSVT practice, was not effective when compared to both no SLT (control) or LSVT LOUD after 3 months. Moderately wide confidence limits perhaps reflect the variability in the pragmatically provided treatment and therefore do not rule out a beneficial effect for NHS SLT or alternative protocolised SLT approaches, delivered to a heterogeneous PD population with dysarthria.

Although we found benefits for LSVT LOUD in the voice-related outcomes, there were no differences in the scores for all three treatment comparisons in any of the general quality-of-life secondary outcomes: EQ-5D-5L, ICECAP-O and PDQ-39.

Severity of speech disorder measured by VHI measure at baseline showed a significant interaction with treatment and, although hypothesis generating only, provides a signal which may be worth exploring further. There was no evidence that the intervention effect differed according to age, Parkinson’s disease severity and receiving treatment during the COVID-19 pandemic.

Process evaluation findings

Health economics findings

The base-case economic evaluation demonstrated that LSVT LOUD was not a cost-effective intervention when compared with either NHS SLT or no SLT (control). Over a 12-month follow-up period, primary analysis from the NHS perspective estimated the ICER for LSVT versus no SLT (control) to be £197,772 per QALY gained and £77,017 per QALY gained versus NHS SLT. NHS SLT was more expensive but lacked evidence that it was any more effective (in terms of QALYs) than no SLT (control) and was therefore by definition ‘dominated’. The LSVT LOUD arm had a high cost per participant and this cost was primarily driven by the predominantly therapist-led delivery of the intensive intervention, and while QALY gains were demonstrated with LSVT for both comparisons, the differences were small. Therefore, at the NICE willingness-to-pay threshold of £20,000–30,000 per QALY gained, both LSVT LOUD and NHS SLT were not cost-effective.

The high additional cost of delivering LSVT LOUD was not surprising given the intensity of the intervention which comprised of four sessions weekly for 4 weeks and a total of 16 sessions (if delivered as per protocol). The cost of therapist time to deliver the sessions contributed to over 80% of the total cost, consistent with results from a recent Australian LSVT study. 118 Analysis of alternative LSVT LOUD delivery scenarios revealed that the involvement of therapy assistants in LSVT LOUD sessions reduced the overall cost. Assuming the first eight sessions were supervised by a therapist with an assistant then conducting the remaining eight sessions, the ICER versus NHS SLT was reduced to £41,108 per QALY gained. Given that remote interventions have evolved in recent years, we also explored scenarios containing online sessions; however, these incurred extra one-off costs of companion software and laptop loan. This approach was also advocated by Saiyed et al. , who suggested that moving to online (supervised) delivery would be optimal to patients living in rural/remote areas where there is limited access to SLT services and significant travel costs incurred in traveling to urban centres. 118 The most cost-effective approach to delivering LSVT LOUD was a combination of supervised and unsupervised sessions where the LSVT LOUD intervention was delivered using the LSVT companion software, and resulted in an ICER of £27,808 per QALY. However, the scenario analysis assumed the same effectiveness as LSVT LOUD delivery in the trial and it is unknown whether unsupervised sessions are as effective as supervised sessions and have the same impact on the quality of life. Further analysis explored including only those participants who were fully compliant with the protocol, and this resulted in greater QALY gains and a lower ICER of £32,291 per QALY compared with NHS SLT.

When considering capability as the measure of effect, LSVT LOUD resulted in greater levels of capability compared with NHS SLT and no SLT (control), with an ICER of £46,210 per YFC versus NHS SLT. There is no agreed willingness-to-pay threshold per YFC, although preliminary work has suggested the threshold may be in the region of £33,500 per YFC,119 suggesting LSVT LOUD may not be cost-effective when using a broader measure of well-being.

The use of other healthcare services was similar for all participants in all treatment groups, and this was expected as the SLT interventions are unlikely to affect the physical health of patients with PD. The severity of PD is an important predictor of resource use and this study showed that patients randomised to different treatment groups had similar severity. 120 However, social services utilisation was higher in LSVT LOUD arm, and this was driven by the presence of a number of outliers, amplified by the relatively modest sample size of approximately 115 per trial arm. Social care utilisation is most likely to be related to physical symptoms and abilities unrelated to difficulty in communication; therefore, while the initial analysed proposed the NHS/PSS perspective, a decision was made to present the NHS perspective as the primary analysis. The SLT interventions remained not cost-effective when compared with no SLT (control) from the broader societal perspective. However, the cost per QALY gained for LSVT LOUD compared to NHS SLT was significantly reduced to £9889 due to less work absence, but this finding should be treated with extreme caution due to the small number of participants contributing to this and the presence of outliers.

Impact of voice problems

This trial does not provide any evidence about the relative benefits of NHS SLT and LSVT LOUD at any dosage apart from that delivered in the trial. For LSVT LOUD treatment, there is a protocol which must be followed and sets the dosage; however, NHS SLT reflects a multitude of possible approaches as chosen by the local therapist in response to the individual patient needs and thus did not have a defined protocol. Similarly, the dose was set by therapists though we did expect that the treatment would end within 3 months which may have curtailed long-term follow-up and review plans in this therapy arm. There is evidence that the dose of a therapy can alter the effectiveness of therapy interventions. The dose of NHS SLT which participants were receiving was not constrained, in order to replicate the clinical reality of using this intervention and no difference in the impact on voice problems was found compared to not receiving SLT. The extent to which any SLT approach is protocolised and the dosage regulated is an area which needs attention when considering whether to give SLT other than LSVT LOUD.

Suitability and acceptability

LSVT LOUD can be effectively delivered in the NHS context and appeared to be acceptable to most participants with few crossing-over from their allocated group, and a very small number citing the intensity of the intervention as a reason for cross-over or withdrawal. We are however also aware that LSVT LOUD has been recognised as ‘a selective and elective therapy’121 and the required intensity may have influenced some patients’ decisions about participation in the trial.

Thus, taking the limitations of the trial in account, on average, people with PD-related dysarthria will benefit more from LSVT LOUD than usual care and therapists should use the findings to challenge their perception about who is and is not suitable for LSVT LOUD. The choice about participation in LSVT LOUD should be given to people with Parkinson’s disease and UK SLT speech and language therapists should therefore offer LSVT to people with Parkinson’s who report problems with their speech or voice and avoid making that decision for the patient.

Health economics

The analysis of clinical effectiveness, concentrating on a speech-specific outcome, showed statistically significant results in favour of LSVT LOUD. In contrast, the EQ-5D-5L utility scores used as the health economic outcome measure and the total QALYs calculated from EQ-5D-5L showed positive but non-significant differences. Given that EQ-5D-5L dimensions focus on pain and discomfort, self-care and mobility and exclude a focus on speech and communication may thus be inadequately captured using this generic quality-of-life measure, and in turn the effectiveness of LSVT LOUD (and other speech and communication services) may be underestimated. 122 Previous studies have also questioned the validity of the EQ-5D-5L questionnaire in other patients with communication difficulties such as aphasia. 123,124 Therefore, there is a tension between using a frequently used questionnaire to calculate QALYs, a common outcome measure used by decision-makers for comparing interventions across diseases, and using an alternative outcome measure focusing solely on patient communication, but prohibiting across disease comparison of cost-effectiveness.

Trial design

Establishing a minimum clinically significant change for the primary outcome measure, the VHI, was one of the aims of this trial. We used a transitional outcome measure but did not conclusively establish an MCIC for VHI using anchor-based methods and ROC analysis curves as the sample size of the trial was not large enough. There was a signal that the MCIC for VHI may be a 3- to 4-point difference so the 8- and 9.6-point differences established for LSVT LOUD compared to NHS SLT and no SLT may be clinically relevant.

This is the first full economic evaluation evaluating the cost-effectiveness of SLT interventions within the UK NHS in patients with PD. The analysis has provided a comprehensive summary of different costs and outcomes from several perspectives and detailed resource use of health and social services follow-up costs and those concerned with delivery of the intervention. Different evaluative spaces have also been explored to assess the effect of SLT intervention on maximising not only health but also capability/well-being. The collection of patient-reported resource use (as primary and secondary care services) and SLT treatment logs completed by supervisors ensured all SLT sessions were captured and were not double counted. The analysis also explored a variety of LSVT LOUD delivery scenarios to provide information for commissioners of SLT services and the costs of each scenario were estimated to determine typical scenarios for LSVT LOUD delivery by local SLT service providers. These scenarios were validated by experienced therapists and clinicians working within different service providers within the NHS.

Pragmatic delivery of interventions

The applicability of these trial results is high as the pragmatic nature of the trial allowed for close replication of the interventions as they are, or can be, currently provided through the NHS. Large pragmatic trials in SLT for PD-related dysarthria are a relatively recent development, and PD COMM is the first large-scale trial to test SLT interventions which as closely as possible replicate those provided to people with PD-related dysarthria in a clinical setting.

However, LSVT LOUD is unusual within UK SLT interventions in a number of ways. Firstly, it is highly protocolised in content, dosage and delivery, which involves considerable reorganisation from standard practice. Secondly, therapists can only deliver it if they participate in mandatory training and 2-yearly updates (managed centrally by LSVT Global) and they receive ongoing support with access to trainers and an online community. Thirdly, patients are expected to meet certain criteria and demonstrate a high level of commitment to LSVT from the outset. Thus LSVT LOUD is a relatively ‘homogenous’ intervention which supported the intervention delivery and evaluation of its effectiveness.

Primary outcome choice

Most previous trials of PD-related dysarthria have used clinician-assessed vocal loudness as a primary outcome measure. This is the first large definitive RCT to use the participant-rated outcome measure of VHI which measures how the participant perceives the impact their voice problems are having on their daily activities and their quality of life.

Participant representation

People with communication impairment such as dysarthria related to PD are an under-researched population often excluded from research activities because of the challenges associated with their involvement. The trial population not only had a communication impairment, but additionally had a progressive disease which involves mobility and impacts on participation, raising questions about involvement and gatekeeping (participant health, cognitive concerns, support from others for attendance at therapy, etc., in addition to ability to participate). These concerns were taken into account when deciding the inclusion criteria including levels of PD severity and the details of trial design.

Steps taken included:

1. we conducted a successful pilot to demonstrate the feasibility of recruitment in this population;

2. an inclusive approach to recruitment where all levels of PD severity and dysarthria severity were eligible to enrol on the trial;

3. this was a UK-wide study, with the aim of recruiting regardless of geographical location;

4. the trial team worked with SLTs and gerontologists to highlight the need for high-level RCT evidence from trials such as PD COMM in order to inform clinical use of common interventions.

Speech-related problems associated with PD do not typically impact on language abilities and so accessible information sheets and consent forms were not necessary. Previous Cochrane reviews highlighted the paucity of data. Rapid searching identified few additional RCTs on treating PD-related speech problems and all of them were small and underpowered. People with dementia were excluded and are not routinely included in these clinical interventions and so further work is required to understand how to include people with substantial cognitive impairments.

Geographical areas with high proportions of different ethnic groups were included (Birmingham, Glasgow, London) and our previous work found that recruiting via the NHS was the most effective approach. 128 However, we did not record ethnicity in the screening logs and so cannot comment.

Future PD rehabilitation trials should refer to the recent Include Ethnicity Framework in the development of the trial, which aims to support the relevance and facilitate the inclusion of under-represented ethnic groups.

Reflections on your research team and wider involvement

There is a range of experience and expertise across the research team. The team was balanced in age and gender and included people from a range of ethnic and cultural backgrounds.

Trial management

The study design, collection, management, analysis and interpretation of the data and publishing the data were the responsibility of the chief investigator and the Collaborative Group. The trial was managed by the BCTU, with additional support for Scottish sites being provided by Glasgow Caledonian University. The process evaluation was managed by the University of Bangor and the intervention description analysis was managed by Glasgow Caledonian University. The funder did not have a role in the above activities. Oversight of the trial was performed by the Trial Management Group, an independent Data Monitoring Committee (DMC) and Trial Steering Committee (TSC). The DMC and TSC membership was agreed with the funder; these committees met at least annually to review the data and progress of the trial. Annual reports were submitted to the NHS REC and monthly updates against pre-determined milestones were sent to the funder.

Patient and public involvement

T Jefferys, L Sharp, T Farnsworth, C Jeffery.

Members of the PD COMM Collaborative Group

Members of the Trial Steering Committee

Lisa Shaw (Chair), Senior Research Associate, University of Newcastle upon Tyne

Mr Chris Jeffery, Patient and Public Involvement Representative

Michelle Collinson, Senior Medical Statistician, University of Leeds

Linsay Pennington, Senior Lecturer and Speech and Language Therapist, University of Newcastle upon Tyne

Catherine Sackley, Professor of Rehabilitation, King’s College London

Paul Worth, Consultant in Neurology, Cambridge University Hospitals NHS Foundation Trust

Members of the Data Monitoring Committee

Carl Counsell (Chair), Consultant Neurologist, University of Aberdeen

Katherine Deane, Senior Lecturer, University of East Anglia

Louise Hiller, Statistician, University of Warwick

Members of the Trial Management Group

Professor Catherine Sackley (Chief Investigator)

Professor Carl E Clarke (Deputy Chief Investigator)

Ms Pui Au

Mr Naveed Saeed

Ms Jane Futterer

Ms Gillian Beaton

Professor Marian Brady

Professor Sue Jowett

Ms Smitaa Patel

Ms Natalie Ives

Members of the Birmingham Clinical Trials Unit

F Dowling

S Tearne

L Sturdy

R Dhaliwal

D Hingley

P Au

Participating centres and PD COMM Collaborative Group members

*Indicates current principal investigator

#Indicates former principal investigator

Aneurin Bevan University Health Board: A Church*, A Davey, C Gallagher, A Conroy, S Bailey, B Done, D Davies

Basildon and Thurrock University Hospitals NHS Foundation Trust: S Sveinbjornsdottir*, M Kasti^#, K Allen, J Colnet, J Riches, L Kittridge, L Morris, C Waszkiewicz

Bath and North East Somerset Community Health and Care Services; V Lyell*, V Page, N Bassford, H Rayner, E Henderson

Betsi Cadwaladr University Health Board: S Abraham*, JV Hindle#, S Jones, P Martin-Forbes, C Watkins, A Roberts, E Newcombe, L Bibby, L Matthews, J Roberts, S Thomas, H Hawthorne, K Clewley, S Lord, J Roberts, G Bretag, S Noble

Blackpool Teaching Hospitals NHS Foundation Trust: D McGhee*, H Mcclure, A Ling Zhi Teo, R Wheeldon, C Wilkinson, M Oprea, G Van Duyvenvoorde, N Wilson, L Evans, R Belshaw, A Clarke, M Turner, C Thompson

Brighton and Sussex University Hospitals NHS Trust: R Saha*, J Aram, D Mullan, J Newman, K Micabel, H Robinson, K Chick, J Gaylard, J Cochrane

Cardiff and Vale University Health Board: E Christopher Thomas*, A Kissick, P Wright, B Mohamed, S Mahon, T Williams, S Appleton, N Elliott, L Evans, J Ridley

Central London Community Healthcare NHS Trust: A Funaki*, P Daly, J Hackworth, K Timms, L Evison, A Bajracharya

East Cheshire NHS Trust: M Silverdale*, K Andrews, A Davies, R Bedford, A Jones, D Mournfield, L Howieson, V Price, J Hunter, A Baggs, L Evans, R Norton, M Holland, K Pointon

Gloucestershire Hospitals NHS Foundation Trust: S Kulkarni*, S Bobeldijk, S Beames, J Cavanagh, S Sudlow

Great Western Hospitals NHS Foundation Trust: G Lennox*, R Whittaker, D Nelson, S Pegler, C Drayson, P Turner, J Stockwell Guy’s and St Thomas’ NHS Foundation Trust: T Andrews*, E White, E Turner, T Burnay, C Hickey, S Warner

Harrogate and District NHS Foundation Trust: R Buccoliero*, C Isles, C Stemp, J Guy, C Bennett, S Smith, P Randall, L Ware, J Cann

Homerton University Hospital NHS Foundation Trust: L Sautin*, S Grobler#, S Adjei#, A Hankin, D Ovayolu, M Dobbs, O Joyce, R Humphreys, A Jha, C Holbrook, K Rowsell, F Johnson, H Thornley, S Conway

Hywel Dda University Health Board: C James*, S Murrow, M Hughes, K Pope, C MacPhee, E Williams, R Hughes, A Evans, A Richmond, K Pilborough, K Campbell, S Elizabeth Davies, A Taylor, S Thomas, D Asandei

Lancashire Teaching Hospitals NHS Trust: T Majeed*, J Dawber, S Furey, A Oppetit, J Birt, M Hare, V Fleming, A Timoroksa, H Al-Nufoury, S Sharp, L Freimann

Leeds Teaching Hospitals NHS Trust: I Mavroudis*, J Alty#, E Richfield, J Harrison, V Smith, T Joyce, J Bamford, S Jamieson, J Cosgrove, S Butterworth, E Sacre, L Makawa, P Duggan-Carter, C Arnold, K Brown, P Mpofu, C Joyce, S Henderson, C Wiseman, L Hyne, B Stevens, A Wood, D Holland, V Smith

London North West Healthcare NHS Trust: J Juada*, S Molloy#, C Pavel, M Dhanarante, T Adedoyin, C Rowbottom, L Little, R Choudhury, L Prados, A Kelly, R Eggers, T Saifee, P Poku, L Niepage

Manchester University NHS Foundation Trust: D Ahearn*, A Fountain, A Curran NHS Ayrshire and Arran: A Watt*, M Wilson, A Anderson, M Graham, J Taylor, P Hewat

NHS Dumfries and Galloway: S Donaldson*, H Moores-Poole, C Angus, S Coull, R Davy, A Gilmour-Graham, T Mcilroy, A Kendall

NHS Forth Valley: S Pal*, E Messeder, D Thomson, V Johnston, P Raby, S Kinnear, P Faunce Smith

NHS Greater Glasgow and Clyde: S Wishart*, D Grosset, J Burns, A Louise Cunnington, E Newman, C Vennard, C Dalton, T Murphy, G Ralph, A Ritchie, C Nelson, A McEntee, P Fowley, H Hare, G Beaton

NHS Highland: M Wilson*, D McDonald, J Finlayson, A Donaldson, S Sutherland, S Bramley, C Dunn, M Wallis, S Hewitt

NHS Lanarkshire: H Morgan*, A Falconer#, L Peacock, A McAlpine, C McBrearty, A Lowe, N Findlay, A Adam, C Tearney, J Picken, K MacKenzie, L McCallum, L Smith, M Sidney, P Downie, L Donnelly, R McAllister, S Campbell, S Maclachlan, L Shearer, K Campbell

NHS Lothian: G Duncan*, S Marrinan, M Dewar, J Kerr, L Killin, A Peters, A Stewart, T Daniels, A Darbyshire, I McCoy, U Duff, F Young, S Orr, C Telford, D Fraser, S Borthwick, H Bailey, L Karbownicki, E McLeod

NHS Tayside: D Sutherland*, E Sammler, L Whyte, C Young, L Gillies, L Gall, J Dallas, L Cassidy, E Letham, V Salisbury, L Anderson, C Hutton, S Waggett, D Anderson, A Mcgee

Norfolk and Norwich University Hospitals NHS Trust: S Cooper*, G Mamutse, A Niruban, A Bath, A Wiltshire, M Harmer, C Wright, J Graham, K Richardson, J Tyler

Norfolk Community Health and Care NHS Trust: L Isaacs*, N Crow, S Pinnell, W Neale, L Maloney, R Weller, K Young, C Squire North Bristol NHS Trust: A Whone*, Y Hernandez, H Findlay, K King, L Gethin, S Ticehurst, A Swift, J Short, J Dean, C Westcott, K Thomas, S Cottrell, D Kruszynska

Northern Lincolnshire and Goole NHS Foundation Trust: S Kamath*, Q Ma, J Hall, R Wilson, H Goodhand, K Mellows, L Jayne Cottam, T Behan, J Gibson, E Lomas, J Kirk, L Smith, J Benson

Pennine Acute Hospitals NHS Trust: J Raw*, P Mulligan, A Ansari, R Irving, A Javed, S Hussain, L Johnson, R Joseph, J Brooke, J Melville, M McCormack, J Stockley, D Ganderton, A Cherriman

Powys Teaching Health Board: J Price*, C Douglas, C Cooter, J Bushell

Royal Devon and Exeter NHS Foundation Trust: R Sheridan*, C Browning, K Polverino, T Malone, S Jackson, A Foden, R James, S Hayes, L Roberts, E Davis

Sandwell and West Birmingham Hospitals NHS Trust: C Clarke*, D Nicholl, A Majeed, MT Oo, K Blachford, A Boughey, J Kaur, S Kaur, M Awan, S Rahman, J Round

St Helens and Knowsley Hospitals NHS Trust: D Gandecha*, S Williams, S Dealing, H Moss, L Talbot Taunton and Somerset NHS Foundation Trust: S Cooper*, R Sophia, J Allen, S Cox, C Moreira, D Woolven, D Sharratt, E Foster, H Hurren, J Watson, S Northover, D Green, A Treloggen, C Pawley, K Beesley, K Milne, L Howard, S Craw, A Lewis, A Whitcher, C Vickers, T Russell, A Sykes, H Meikle, N Loraine

The Walton Centre NHS Foundation Trust: M Steiger*, H Treloar, L Roebuck, M Taylor

Torbay and South Devon NHS Foundation Trust: R Nashed*, J Garfield-Smith, S Mills, H Griffin, C Marshall, G De Selincourt, V Queen, M Stone, M FarrowJones, E Sturdy, K Almedilla, F Fitzsimmons, M Alison, F Rogers, B Reed

University Hospitals Birmingham NHS Foundation Trust: M Pinkney*, S Jones#, S Muzerengi, M Johnson, S Stafford, E Parmar, J Albutt, S Kaur, M Awan, S Rahman, K Leahy

University Hospitals Bristol NHS Foundation Trust: T Allain*, M Sritharan#, A Daniell, K Kunsteinaite, S Slade, F Pimbblet, C Killourhey

Wye Valley NHS Trust: E Wales*, C Hughes, G Horsfield, L Mercer, Z Roberts, K Stock, M Evans, S Boyd, L King, J Birch, S Anderson, C Evans, N Stapleton, U Magennis, R Vernall