The SINFONIA Study: A Randomised Controlled Trial to Determine Feasibility and Effectiveness of Online Group Singing for People With COPD or ILD and Breathlessness

ABSTRACT

Background and Objective

People with chronic obstructive pulmonary disease (COPD) and interstitial lung disease (ILD) experience chronic breathlessness and reduced quality of life (QOL). Specialised singing programs for people with lung disease may improve QOL and symptoms.

Methods

A randomised controlled trial conducted in Australia to determine the feasibility and effectiveness of 12 weeks of online, guided group singing on QOL (primary outcome, as measured by SF36), as compared to usual care in people with COPD or ILD. Secondary outcomes: anxiety, depression, dyspnoea and loneliness. Intervention effectiveness was assessed using mixed effects linear regression.

Results

One hundred and one participants were randomised (50 singing, 51 control; 64 COPD, 37 ILD; mean age 69.6 years; 74% women; 42% mMRC = 2, 58% mMRC ≥ 3). The results of the primary endpoint were a mean treatment effect (MTE) of singing of 2.5 points (95% CI −1.9 to 6.8; p = 0.266) on SF‐36 mental health score (MCS) and 2.6 points (95% CI −0.9 to 6.1; p = 0.138) on the physical component score (PCS), neither of which were statistically or clinically significant. In the per‐protocol analysis, significant improvement in MCS was detected in participants who attended 8 or more sessions (MTE:4.8; 95% CI 0.1 to 9.5; p = 0.047). No significant effects on secondary outcomes were detected. No adverse events were reported.

Conclusions

This RCT demonstrated that online group singing is a feasible intervention for people with COPD or ILD and breathlessness. Although the intention‐to‐treat analysis did not demonstrate effectiveness, per‐protocol analyses suggest that singing may improve mental health components of QOL in people with COPD or ILD.

Short abstract

SINFONIA is an Australian‐first clinical trial of 12 weeks of online group singing for people with COPD or ILD. No effect of online guided group singing on quality of life was detected in the primary intention‐to‐treat analysis. People who attended ≥8 sessions had clinically and statistically significant improvements in quality of life in the per‐protocol analysis. Online group singing represents a simple and feasible approach that may support quality of life.

1. Introduction

People living with chronic obstructive pulmonary disease (COPD) or interstitial lung disease (ILD) often experience prolonged and distressing respiratory symptoms, including breathlessness and cough, which frequently drive other non‐respiratory symptoms, including fatigue, loss of functional capacity, anxiety or depression, and overall reduced quality of life [1, 2]. In Australia, COPD and ILD constitute 43% and 8% of the overall burden of chronic respiratory disease respectively [3].

Managing breathlessness in chronic respiratory conditions is notoriously difficult [1, 4], with breathlessness poorly correlating with the severity of lung function impairment [5] and often persisting despite optimal disease‐directed treatment [6]. Alongside disease‐directed pharmacotherapies, pulmonary rehabilitation (PR) is recommended as a gold standard exercise, education and counselling program for people with chronic respiratory conditions [7, 8]. Although PR is an established and well‐evidenced model of care, the capacity and accessibility of these services can be a barrier to engagement [9]. Furthermore, benefits of PR are not always sustained following the conclusion of the program [10]. Currently, there is a lack of safe and effective pharmacotherapies for managing breathlessness [11, 12]. There is thus an urgent need for sustainable, novel, community or home‐based interventions that can provide an alternative or adjunct to existing models of care to better support the diverse needs of people with COPD, ILD and other chronic respiratory conditions.

Group singing is a low‐cost and low‐risk activity that can support social connection, boost mood, and improve overall quality of life [13]. Group singing has been trialled in a range of settings and across different health conditions, including neurological [14], mental health [15], and increasingly, chronic respiratory conditions [13, 16, 17]. A recent systematic review of five randomised controlled trials of singing for people with COPD demonstrated significant improvements in physical health components of quality of life and maximum expiratory pressure [18]. To date, however, there have been no trials of group singing for people with chronic respiratory conditions in Australia, and no trials with large cohorts of people with ILD. Furthermore, there is a need to determine the feasibility and acceptability of fully online delivery models, which can enhance access to care for people who are not readily able to access traditional in‐person models of care. This is particularly relevant in a geographically vast country like Australia, where service delivery is scarce in regional and remote areas [19].

The SINFONIA study (a clinical trial examining the benefits of SingINg For breathing in COPD aNd ILD pAtients) aimed to determine the feasibility, acceptability, and effectiveness of guided, online, group singing as an intervention to improve quality of life in people with COPD or ILD with significant breathlessness living in Australia.

2. Methods

2.1. Study Design and Setting

The full methodology of this randomised controlled trial has been published separately [20]. In brief, a parallel, double‐arm, randomised, blinded‐analysis, mixed‐methods phase II/III trial of online guided group singing was conducted between April 2022 and October 2024. Feasibility outcomes are partially reported in the current manuscript (recruitment and completion rates); however, additional phase acceptability outcomes are reported in full elsewhere [21, 22].

2.2. Participants

Adults with a confirmed diagnosis of COPD or ILD with a modified Medical Research Council (mMRC) dyspnoea score ≥ 2, who were stable on treatment at the time of recruitment, were eligible to participate. Eligibility was confirmed by medical review from the site principal investigator for participants recruited from participating hospitals, or the coordinating principal investigator for self‐referred participants. Recruitment from participating hospitals was via respiratory outpatient clinics, and self‐referrals were via expressions of interest distributed by patient facing organisations such as the Lung Foundation Australia.

Stable treatment was defined as no new or changed doses of cardiorespiratory medications including oral steroids or antibiotics, and no hospital admissions for exacerbation of the underlying condition within the past 4 weeks. Concurrent enrolment in a PR program was not permitted; however, prior completed participation was allowed. One nominated adult carer per participant was eligible to participate; however, this was not required.

2.3. Randomisation and Masking

At first visit, participants were randomly allocated using a permuted block design stratified by disease type (COPD versus ILD) generated in Stata 17.0 [23] and housed in REDcap [24] in a 1:1 ratio to intervention (singing) or usual care. Allocation was concealed until the time of randomisation.

2.4. Procedures

Participants allocated to the intervention arm were invited to join one 90‐min online group singing class per week for 12 weeks. Online group singing sessions were delivered via Zoom and facilitated by a registered music therapist trained in delivering singing interventions for people with lung conditions. Up to 15 participants were enrolled in the group on a rolling basis. Participants allocated to the singing intervention who did not have access to a suitable device were provided with loaned iPads with data plans for the duration of the study. The structure of each session is outlined in Table 1. In addition, participants were provided resources (including a Spotify playlist developed by the study music therapist and a ‘Singing for Breathing’ CD developed by the UK Singing for Lung Health investigator, Phoene Cave) and were encouraged to sing regularly at home. Further details on the content and development of the singing program are available elsewhere [25].

TABLE 1.

SINFONIA group singing structure.

Session element	Time (min)	Description
Welcome & Social Check‐in	15	Ice breaker and relaxed chatting
Relaxation	5	A moment of pause to connect to the body
Physical Warmup	5	Seated physical stretches and movement activities with and without musical instruments
Postural Work	5	Education around supportive posture for singing and breathing
Breathing Exercises	10	Education around diaphragmatic breathing and accompanying exercises
Vocal Warmups	10	A range of warmups to work on pitch, range, diction, and expressive techniques
Song singing	20	Four participant‐selected songs from a limited list, chosen the week prior
Vocal Cool Down	5	Gentle vocal stretches and education around vocal care
Guided Relaxation	5	Progressive muscle relaxation accompanied by recorded music
Social time	10	Questions and reflections

Participants allocated to usual care continued to receive routine respiratory care delivered by their usual treating team. Usual care for people with COPD or ILD is varied and may include, but is not limited to, routine reviews with respiratory specialists or primary care clinicians, vaccinations, inhaler therapy, oxygen therapy, smoking cessation advice or pharmacotherapy, etc. The singing intervention was provided as an adjunct to usual care, and as such, intervention arm participants also continued to receive routine respiratory care.

2.5. Outcomes

Phase II aimed to determine feasibility and acceptability of the singing intervention measured via the number of participants that were recruited, enrolled, and who completed the study within a 6‐month period. Completion was defined as attendance at ≥ 8 of 12 available sessions for the intervention arm, and completion of the 12‐week assessment for the usual care arm. Whilst no predefined threshold for recruitment or completion was defined in the phase II outcomes, the recruitment and completion rate were used to determine if the full sample size for phase III was feasible. Patient and carer perceptions of group singing, including acceptability, were also assessed using semi‐structured interviews, along with an exploration of social dynamics and interactions within the groups using ethnographic observations. These qualitative results which further address the phase II outcomes of acceptability are reported elsewhere [21, 22].

Phase III aimed to determine the effect of online group singing on quality of life (QOL) with the primary outcome being between‐arm difference in QOL as measured by the 36‐Item Short Form Survey (SF‐36 [26]) at 12 weeks (reported as mental component score (MCS) and physical component score (PCS) [27]). Secondary outcomes in the phase III study included between‐arm differences at 12 weeks in anxiety and depression (Hospital Anxiety and Depression Scale (HADS [28])), breathlessness severity (Dyspnoea‐12 [29]), breathlessness mastery (Chronic Respiratory Disease Questionnaire—Self Administered Standardised mastery subdomain (CRQ‐SAS [30, 31])), loneliness (University of California Los Angeles Loneliness Scale version 3 Short Form (UCLA‐3 [32])), healthcare utilisation (self‐report), and carer quality of life (SF‐36 and CareQOL [33], optional). Exercise tolerance as measured by the 6 Minute Walk Test (6MWT [34]) was planned as a secondary outcome; however, collection of this outcome was not feasible in the phase II study (due to the COVID‐19 pandemic) and thus was not measured either in the phase III study.

2.6. Sample Size

We had initially based the sample size calculation on the total SF‐36 score, aiming to detect a minimal clinically important difference (MCID) of 10 points with an estimated standard deviation of 16 [35, 36] on the primary outcome of the total SF‐36 score at 12 weeks. This approach has been adopted in other trials [37, 38, 39]. Assuming 80% power and a 5% significance level, at least 42 participants per arm were required. A target of 140 participants was set to account for up to 40% attrition rate.

The reporting of the primary outcome was revised during analysis to utilise the PCS and MCS summary scores recommended by the developers rather than total SF‐36 score [39]. The MCID and sample size were revised to reflect this change with an estimated MCID of 3 to 6 points for the MCS and PCS summary scores [40]. Assuming 80% power and 5% significance level with an SD of 8.4, 56 participants per arm were required to detect a 4.5‐point difference. Of note, this was a deviation from the original protocol sample size calculation.

2.7. Data Collection

Participants and their carers in both arms were contacted by phone at 4‐week intervals by a member of the research team to collect study assessments. Study assessments were completed directly by participants using a secure web‐based REDcap survey or via phone with a member of the research team. During the 4‐weekly phone calls, participants were prompted to report their healthcare utilisation over the previous 4 weeks. Participants in the singing arm were asked to report whether they had engaged in additional singing activities outside of the group classes during the 4‐weekly phone calls.

2.8. Statistical Analyses

Baseline demographics are reported descriptively as mean and standard deviation or median and interquartile range for normally distributed and non‐normally distributed continuous variables, respectively. Categorical variables are reported as frequency and percentage. Raw outcome data are also reported descriptively as mean and standard deviation or median interquartile range as appropriate for continuous variables.

The phase II feasibility and acceptability outcomes were analysed and described descriptively and informed the delivery of phase III.

The phase III primary outcome of SF‐36 PCS and MCS scores and all secondary outcomes were analysed on an intention‐to‐treat basis using mixed effects linear regression with restricted maximum likelihood estimator. Participants were entered as random effects and diagnosis (COPD vs. ILD), week (categorical variable: four timepoints), group allocation, and interaction between week and allocation were entered as fixed effects. An additional per‐protocol analysis of the primary outcome was conducted for participants who did or did not attend the pre‐specified minimum dose of 8 sessions. Results of the mixed effects linear regressions are presented as mean treatment effect (MTE) of singing intervention at 12 weeks with corresponding 95% confidence intervals (CI) and p‐value. Statistical significance was declared at p < 0.05. All secondary outcomes were analysed using the same mixed effects linear regression approach.

Exploratory subgroup analyses of the primary outcome were conducted, with results stratified by: sex (male vs. female), age group (< 60 years vs. 60–70 years vs. > 70 years), severity of disease (mild–moderate with DLCO > 50% vs. severe with DLCO ≤ 50%), primary diagnosis (COPD vs. ILD), prior psychological disease (self‐reported anxiety and/or depression present vs. absent), breathlessness severity (mMRC 2 vs. 3 vs. 4), and previous pulmonary rehabilitation (yes vs. no). Results of subgroup analyses are reported as MTE of singing intervention on total SF‐36 score at 12‐weeks, 95% CI, and p‐value in a Forest plot.

3. Results

3.1. Recruitment and Feasibility

Within the first 6 months of the study (21st April to 21st October 2022), 42 participants were enrolled (22 singing, 20 control). Of these, 90% of the control arm (n = 18) and 82% of the singing arm (n = 18) completed the final assessment. Of 22 participants enrolled into the singing arm within the first 6 months, 68% (n = 15) attended the predefined minimum dose of eight sessions. The phase II outcomes were deemed to be feasible for continuation into phase III. The initial target sample size of 140 was reduced due to the lower than anticipated attrition in the phase II period (15% compared to 40%) with an estimate of 98 participants needed. The phase II primary outcome of acceptability was assessed through qualitative semi‐structured interviews with participants which have been reported in detail elsewhere and broadly identified that online group singing was acceptable to people with COPD or ILD and breathlessness [22, 23].

A total of 101 participants were enrolled and randomised between 21st April 2022 and 29th July 2024 in the combined phase II/III trial (Figure 1), of which 82 (81.2%) completed the final primary outcome. Primary outcome completion rates were lower in the singing arm (n = 36; 72.0%) compared to control (n = 46; 90.2%). Seven participants allocated to the intervention arm did not own a suitable device and were provided with loaned devices and a data plan for the period of the trial to enable them to join sessions. One participant in each group opted to discontinue the study after randomisation but prior to completing any baseline outcome measures; two participants in the singing arm and one in the control arm died during the follow‐up period due to causes unrelated to the intervention. In total, five participants withdrew from the intervention arm (three who had other commitments at the time the session was scheduled, one due to recent bereavement, one due to exacerbation of unrelated health issues) and two withdrew from the control arm (one due to unrelated treatment side effects which they felt would impact their participation, one due to a change of mind). An additional 10 participants (8 intervention and 2 control) were lost to follow‐up despite not formally withdrawing consent.

FIGURE 1.

CONSORT participant flow diagram.

3.2. Participant Characteristics

Of 101 total participants enrolled and randomised, 50 participants were randomised to singing and 51 were randomised to control (Figure 1; Table 2). Participants were predominantly female (n = 75, 74.3%) with a mean age of 69.6 years (SD 9.2). Two thirds had a primary diagnosis of COPD (n = 64, 63.4%) and one third had a primary diagnosis of ILD (n = 37, 36.6%). Baseline breathlessness severity was broadly distributed (mMRC 2: n = 42, 41.6%; mMRC 3: n = 34, 33.7%; mMRC 4: n = 25, 24.8%).

TABLE 2.

Participant characteristics.

	Singing (n = 50)		Control (n = 51)		Total (n = 101)
	n	%	n	%	n	%
Age (years) a	68.5	9.6	70.6	8.7	69.6	9.2
Sex
Male	10	20.0	16	31.4	26	25.7
Female	40	80.0	35	68.6	75	74.3
Primary diagnosis
COPD	32	64.0	32	62.8	64	63.4
ILD	18	36.0	19	37.8	37	36.6
mMRC
2	23	46.0	19	37.3	42	41.6
3	19	38.0	15	29.4	34	33.7
4	8	16.0	17	33.3	25	24.8
Baseline pulmonary function
FEV1 percent predicted a	57.78	27.9	57.2	24.2	57.51	26.0
FVC percent predicted a	73.70	21.4	73.6	23.1	73.63	22.1
DLCO percent predicted a	52.69	10.1	52.4	19.5	52.56	19.7
Completed pulmonary rehabilitation	30	60.0	33	64.7	63	62.4
Psychological disease
Anxiety	13	26.0	12	23.5	25	24.8
Depression	14	28.0	8	15.7	22	21.8
Availability of carer (n = 100)
Lives with carer	15	30.6	13	25.5	28	28.0
Lives alone but carer available	9	18.4	10	19.6	19	19.0
Lives with non‐carer	12	24.5	12	23.5	24	24.0
Lives alone	13	26.5	16	31.4	29	29.0
Carer enrolled	3	6.0	5	9.8	8	7.9
Highest level of education
Primary	5	11.6	3	6.1	8	8.7
Secondary	12	27.9	15	30.6	27	29.4
Tertiary	26	60.5	31	63.3	57	62.0

^a Data reported as mean and SD.

3.3. Primary Outcome: Quality of Life

3.3.1. Intention‐To‐Treat Analysis

In the intention‐to‐treat analysis, no clinically or statistically significant effects of singing on SF‐36 MCS or PCS were detected. Singing had a mean intervention effect of 2.5 points (95% CI −1.9 to 6.8; p = 0.266) on the MCS score and 2.6 points (95% CI −0.9 to 6.1; p = 0.138) on the PCS score (Table 4; Table S1). Although no intention‐to‐treat effects were observed on the MCS and PCS summary scores, statistically significant improvements were detected on the role limitations due to physical health and role limitations due to emotional problems subdomains of the SF‐36 (Table 3). These improvements were equivalent to small to moderate clinical changes [41].

TABLE 4.

Singing attendance.

	Singing group (n = 50)
	n	%
Per‐protocol intervention completion
< 8 sessions	23	46.0
≥ 8 sessions	27	54.0
Reasons for non‐attendance (n = 256 non‐attendances)
Other illness	57	22.3
Other commitment	60	23.4
Technical issues	7	2.7
Exacerbation of primary diagnosis	1	0.4
Forgot	2	0.8
No reason provided	129	50.4
Additional singing activities (self‐reported)
Weeks 1–4 (n = 39)
Yes	29	74.4
No	10	25.6
Time per week (minutes)	Median 70	IQR I30–120
Weeks 5–8 (n = 38)
Yes	29	76.3
No	9	23.7
Time per week (minutes)	Median 70	IQR 30–105
Week 9–12 (n = 37)
Yes	22	59.5
No	15	39.5
Time per week (minutes)	Median 65	IQR 30–120

TABLE 3.

Intervention effectiveness on outcome measures at 12 weeks (mixed effect linear regression analyses).

Outcome measure	n		Mean intervention effect (between group difference at 12 weeks)	95% confidence interval	p
	Singing	Control
SF‐36 MCS
Intention‐to‐treat analysis	36	46	2.5	−1.9 to 6.8	0.266
Per‐protocol analysis (attended ≥ 8 sessions)	25	57	4.8 a	0.1 to 9.5	0.047*
SF‐36 PCS
Intention‐to‐treat analysis	36	46	2.6	−0.9 to 6.1	0.138
Per‐protocol analysis (attended ≥ 8 sessions)	25	57	3.3	−0.5 to 7.1	0.088
SF‐36 subdomains (intention‐to‐treat analysis)
Physical function	36	46	6.4	−2.4 to 15.3	0.154
Role limitations due to physical health	36	46	14.6	0.6 to 28.6	0.040*
Role limitations due to emotional problems	36	46	17.7	0.3 to 35.1	0.046*
Energy/fatigue	36	46	1.8	−6.6 to 10.1	0.679
Emotional well‐being	36	46	4.6	−3.0 to 12.2	0.234
Social functioning	36	46	4.1	−6.6 to 14.9	0.452
Pain	36	46	9.7	−1.6 to 21.0	0.093
General health	36	46	0.5	−6.6 to 7.6	0.889
HADS
Anxiety	35	46	−0.7	−2.1 to 0.8	0.371
Depression	35	46	−1.4	−2.9 to 0.1	0.059
Dyspnoea‐12	35	46	−1.8	−5.3 to 1.8	0.330
CRQ‐SAS
Dyspnoea	35	46	0.4	−0.2 to 1.0	0.203
Fatigue	35	46	0.2	−0.3 to 0.7	0.516
Emotional function	35	46	0.2	−0.3 to 0.7	0.388
Mastery	35	46	0.5	−0.1 to 1.0	0.083
UCLA‐3	35	46	−0.3	−1.1 to 0.5	0.439

Note: Adjusted model for all analyses includes participants as random co‐efficient, with diagnosis (COPD vs. ILD), week, group allocation, and interaction between week and allocation as fixed effects. Direction of effect indicating improvement: SF‐36 higher score; HADS lower score; Dyspnoea‐12 lower score; CRQ‐SAS higher score; UCLA‐3 lower score. *p ≤ 0.05.

^a Mean treatment effect exceeds minimal clinically important difference of 3 to 6 points.

3.3.2. Per‐Protocol Analysis

In the per‐protocol analysis, singing had a mean intervention effect of 4.8 points (95% CI 0.1 to 9.5; p = 0.047) on the MCS score and 3.3 points (95% CI −0.5 to 7.1; p = 0.088) on the PCS score (Table 3). This exceeded the MCID of 3 to 6 points for the MCS score.

3.3.3. Exploratory Subgroup Analyses

In the exploratory subgroup analyses of between treatment arm difference in SF‐36 MCS scores at 12‐weeks (Figure 2A; Table S2), mean treatment effects of singing exceeding the lower end of the estimated MCID were detected in people aged over 70 years (MTE 6.1; 95% CI 0.5 to 11.7; p = 0.033), those with severe disease (MTE 6.6; 95% CI 0.7 to 12.5; p = 0.030), and those with a self‐reported diagnosis of anxiety and/or depression (MTE 8.1; 95% CI 1.1 to 15.1; p = 0.023).

FIGURE 2.

(A) Exploratory subgroup analyses on SF‐36 MCS; (B) Exploratory subgroup analyses on SF‐36 PCS; Dashed line indicates the lower end of minimal clinically important difference of 3 to 6 points.

In the exploratory subgroup analyses of between treatment arm difference in SF‐36 PCS scores at 12‐weeks (Figure 2B; Table S3), mean treatment effects of singing exceeding the lower end of the estimated MCID were detected in people who had not previously attended pulmonary rehabilitation programs (MTE 5.8; 95% CI 0.7 to 10.9; p = 0.027).

Effectiveness amongst people with COPD as compared to those with ILD was similar, as was the effectiveness across baseline breathlessness severity levels for both MCS and PCS.

3.4. Secondary Outcomes

No clinically or statistically significant differences were detected in the secondary outcomes of depression and anxiety (HADS), breathlessness severity (Dyspnoea‐12), breathlessness mastery (CRQ‐SAS), and loneliness (UCLA‐3) (Table 4). Carer outcomes are reported descriptively in Table S1 but were not analysed further due to low recruitment of carers. Self‐reported healthcare utilisation was very low for all participants with no differences between groups detected (Table S4). No adverse events were reported.

3.5. Intervention Adherence

Of the 50 participants randomised to singing, 54% attended the minimum dose of 8 or more of the 12 available sessions (median attendance: 8 sessions, IQR 3–10; Table 3). Self‐reported additional singing ranged from 74.4% at week 4, to 59.5% at week 12. The median self‐reported time per week spent on additional singing activities outside of group sessions ranged from a median of 70 min (IQR 30–120) at week 4 to a median of 65 min (IQR 30–120) at week 12.

Reasons for non‐attendance were provided for 127 of 256 instances, with the most commonly cited reasons being other illness (n = 57, 22.3%) or other commitments (n = 60, 23.4%). Of note, although illness was a commonly cited reason for non‐attendance, no occurrences of illness were reported during formal assessment of adverse events. Non‐attendance reasons were collected by the music therapist whilst adverse events were assessed by an independent researcher. Technical issues were cited for seven instances of non‐attendance (2.7%), six of which occurred during the first 3 weeks for the given individual.

4. Discussion

The study reports the results of an Australian‐first clinical trial of group singing for people with COPD or ILD. This trial also provides new insights into fully online delivery and outcomes specific to people with ILD. The SINFONIA intervention was found to be feasible. Effectiveness of group singing on the mental or physical components of quality of life measured at 12 weeks was not statistically or clinically significant in the intention‐to‐treat analysis. The per‐protocol analysis demonstrated a clinically and statistically significant improvement in mental components of quality of life for those who attended eight or more singing sessions during the 12‐week intervention period.

The exploratory subgroup analyses in the current study suggest differential outcomes across various participant characteristics, with possible mental health improvements observed amongst people aged over 70 years, those with severe disease, and those with a psychological comorbidity, and possible physical health improvements for those who had not previously participated in pulmonary rehabilitation. These findings are broadly reflective of the varied results in the literature regarding either physical or psychological effects of group singing [18]. It should also be noted that the underlying mechanism of group singing as a therapeutic option remains poorly elucidated. Although these findings are preliminary and should be interpreted with caution, they highlight the potential value of group singing as a community‐based symptom management approach for some people with COPD or ILD. Importantly, these are groups who experience common barriers to engaging with pulmonary rehabilitation, including poor psychological health and disease exacerbations [42, 43]. Whilst group singing is not proposed as a replacement for existing evidence‐based models of care, it may provide a valuable adjunct or alternative for those who are unable or unwilling to engage with pulmonary rehabilitation. Indeed, a non‐inferiority trial of 270 people with COPD, which substituted group singing in place of graded exercise therapy within a comprehensive pulmonary rehabilitation program found that group singing resulted in equivalent improvements in functional capacity and quality of life [16]. Importantly, the exploratory subgroup analyses in the current study also demonstrated comparable treatment effects across disease type (COPD or ILD) and breathlessness severity (mMRC 2 vs. 3 vs. 4). These findings support group singing as a disease‐agnostic model of care, which reflects its growing use in other non‐respiratory conditions [14, 15].

SINFONIA utilised a fully online delivery model from the outset. A previous group singing study, which had to switch to online‐delivery due to restrictions associated with the COVID‐19 pandemic, reported considerable drops in attendance rates from 90% to 53% [44]. A later randomised controlled trial of fully online group singing for people with COPD reported improvements in the physical health, but not mental health, components of quality of life, albeit with reduced effectiveness on outcomes compared with in‐person intervention delivery [17]. The current study similarly reported modest improvements in overall health‐related quality of life in the intention‐to‐treat analysis, particularly in domains related to role limitations due to physical health and emotional problems. When this was adjusted to reflect actual attendance in the per‐protocol analysis, a clinically significant improvement was detected. Whilst online delivery can be a means of increasing accessibility for some people, such as those in regional areas or with lack of viable transport options, there is a risk of excluding other groups who have reduced familiarity with, or access to, technology [45]. The current study attempted to mitigate these issues through the provision of loaned tablets and data plans for those without a suitable device, as well as on‐call support from a member of the research team to assist with technical issues joining the Zoom calls. The low frequency of ‘technical issues’ as a stated reason for missed sessions reflects the success of these supports, however, it should be acknowledged that people who are reluctant to engage with online programs are unlikely to have consented to joining the trial so may be under‐represented here.

Whilst SINFONIA was limited to a 12‐week intervention for pragmatic reasons, future implementation should examine opportunities for longer participation in singing groups. A reflection on the development and delivery of the SINFONIA program, as well as other implementation considerations, has been published elsewhere [25]. Attendance varied broadly across participants with some disengaging from the sessions early on and utilising little if any additional singing resources outside of group sessions. Amongst those who achieved the minimum dose, many expressed a desire to continue attending where possible, and it is amongst this group that the greatest improvements in quality of life were detected. These findings suggest that group singing unsurprisingly varies in acceptability and subsequently in effectiveness across different patient groups. Non‐therapeutic singing activities, such as choirs, also have a skewed demographic distribution with women participating more than men [46]. However, this does not detract from the value of group singing as an option for symptom management for those who may benefit. Future work is needed to better understand how group singing for people with COPD or ILD can be scaled and implemented more broadly.

SINFONIA is an Australian‐first trial of fully online group singing for people with COPD or ILD. This trial adds to the existing literature and enhances understanding of disease‐specific effectiveness with a large proportion of participants with ILD and an exploratory subgroup analysis showing comparable outcomes. These findings support the concept of group singing as a disease‐agnostic approach for conditions with comparable symptom burden but distinct underlying aetiology. Further, the exploratory subgroup analyses help to highlight groups which may benefit most from group singing. Although efforts were made to support participants and total attrition was lower than anticipated, there was unequal attrition in the singing arm which is not uncommon in behavioural interventions [47]. Additionally, 10 participants who did not formally withdraw consent and some of which continued to attend group sessions, were unable to be reached for completion of all outcome measures within the allocated timeframe. This is a limitation of all research trials but particularly fully remote trials. The exploratory per‐protocol analysis may reflect increased attrition amongst singing arm participants with lower quality of life scores. Additionally, revisions to the statistical analysis of the primary outcome to utilise the MCS and PCS summary scores as recommended by the tool developer [39] impacted the relevant MCIDs and sample size calculations. Similarly, despite a robust randomisation protocol, baseline differences in the primary outcome and other baseline variables occurred. Therefore, the logistic regression model was adjusted to account for baselines differences. The results of this study should be interpreted in light of these limitations. The current study population had a high attainment of tertiary education, however this was broadly consistent with the Australian general population where 64% hold tertiary qualifications [48]. Whilst carers were eligible to participate in the current study, few chose to participate and therefore it is difficult to ascertain the impact of a group singing intervention on carers of people with COPD or ILD. Similarly, future research may examine the impacts of living alone compared to living with others on outcomes.

In conclusion, online group singing was found to be feasible for people with COPD or ILD. In the primary analysis, no improvements in quality of life were detected due to singing. However, attending eight or more sessions of group singing over a 12‐week period may improve mental health components of quality of life. Further, exploratory subgroup analyses suggest mental health related quality of life improvements amongst people aged over 70 years, those with severe disease, and those with psychological comorbidity (anxiety or depression), as well as physical health related quality of life improvements amongst those who had not previously participated in pulmonary rehabilitation. Further studies are needed to confirm these findings. Group singing represents a simple approach that may help to actively manage challenging symptoms and support people with COPD or ILD.

Author Contributions

Nicole Goh: conceptualization, methodology, writing – original draft, writing – review and editing. Amy Pascoe: methodology, investigation, data curation, formal analysis, writing – original draft, writing – review and editing. Lena Ly: investigation, data curation, writing – original draft, writing – review and editing. Stefanie Zappino: investigation, data curation, writing – review and editing, writing – original draft. Natasha Smallwood: conceptualization, methodology, investigation, writing – original draft, writing – review and editing, funding acquisition. Catherine Buchan: investigation, data curation, writing – original draft, writing – review and editing. Jennifer Philip: conceptualization, methodology, writing – original draft, writing – review and editing, funding acquisition. Anna Spathis: conceptualization, methodology, writing – original draft, writing – review and editing. Sara Vogrin: formal analysis, writing – review and editing, writing – original draft.

Funding

This study was funded by the Windermere Foundation.

Disclosure

No generative AI tools were used in the preparation of this manuscript.

Ethics Statement

This trial was sponsored by the St Vincent's Hospital Melbourne. All trial activities were approved by, and conducted in accordance with, St Vincent's Hospital Melbourne Human Research Ethics Committee (LRR 237/21). Governance approvals were additionally obtained from all trial sites (Alfred Health 720/21; Austin Health 73427; Royal Melbourne Hospital 2022/020). Written consent was obtained from all trial participants. Trial details were prospectively registered on the Australian and New Zealand Clinical Trials Register (Phase II—ACTRN 12621001274864, registered 20th September 2021; Phase III—ACTRN 12621 001280897, registered 22nd September 2021).

Conflicts of Interest

Natasha Smallwood and Nicole Goh are Editorial Board members of Respirology and co‐authors of this article. They were excluded from all editorial decision‐making related to the acceptance of this article for publication.

Supporting information

Table S1: Patient reported outcomes.

Table S2: Subgroup analyses of SF‐36 MCS score.

Table S3: Subgroup analyses of SF‐36 PCS score.

Table S4: Self‐reported healthcare utilisation.

Smallwood N., Pascoe A., Goh N., et al., “The SINFONIA Study: A Randomised Controlled Trial to Determine Feasibility and Effectiveness of Online Group Singing for People With COPD or ILD and Breathlessness,” Respirology 31, no. 10 (2026): 1003–1013, 10.1002/resp.70281.

Data Availability Statement

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.