An open randomised study of autoinflation in 4- to 11-year-old school children with otitis media with effusion in primary care

Article history

The research reported in this issue of the journal was funded by the HTA programme as project number 09/01/27. The contractual start date was in September 2011. The draft report began editorial review in September 2014 and was accepted for publication in April 2015. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The HTA editors and publisher have tried to ensure the accuracy of the authors’ report and would like to thank the reviewers for their constructive comments on the draft document. However, they do not accept liability for damages or losses arising from material published in this report.

Declared competing interests of authors

James Raftery is a member of the National Institute for Health Research Editorial Board.

Copyright statement

© Queen’s Printer and Controller of HMSO 2015. This work was produced by Williamson et al. under the terms of a commissioning contract issued by the Secretary of State for Health. This issue may be freely reproduced for the purposes of private research and study and extracts (or indeed, the full report) may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NIHR Journals Library, National Institute for Health Research, Evaluation, Trials and Studies Coordinating Centre, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

Definition

The term otitis media with effusion (OME) is often used synonymously with ‘glue ear’. 1 The descriptive definition is based on the intraoperative findings of sticky mucinous secretions behind the eardrum that can hamper the free movement of the ossicles in the middle ear. The viscosity of the middle ear fluid has been found to vary and the fluid is sometimes described as serous or secretory. 1,2 Effusions are associated with notional conductive hearing losses of about 15–45 dB owing to damping effects on sound transmission to the inner ear. OME occurs usually in one ear, but frequently in both ears. 3 Probably the earliest and most relevant description of the condition from a primary care perspective was made by Dr John Fry, who wrote about the ‘catarrhal child’: a syndrome of coughs, colds, catarrh and subhealth, in which OME and frequent or persistent upper respiratory infections are the predominant features, and which presents most commonly in young children of early school age. 4

Natural history and scale of the problem

As many as 80% of children of all ages develop OME. 3 Most such episodes are anticipated to resolve naturally, with an average duration of 6–10 weeks, and with just 10% of episodes lasting ≥ 1 year. 5,6 The problem with such a common condition is that it is often regarded as normal. The prevalence rises to 46% (a secondary peak) in the early school years,6 when recurrent ear-related symptoms and broader developmental concerns most often bring the condition to light,7,8 and not infrequently results in surgical referral for grommets. 1,9,10 Time to resolution of effusions remains clinically unpredictable; many last ≥ 3 months and 30% are recurrent. 3,11 In the UK, about 200,000 children are seen every year with this problem in primary and community care. 1,12 The full scale of the total health burden is only partly reflected in high international rates for grommet surgery,13–15 but with falling rates observed in the UK. 16 In the USA, in 2004, as many as 2.2 million people were diagnosed, at an estimated cost of US$4B. 17

The impact of this very common condition on child physical health, hearing, speech, behaviour, development and mapped quality of life (QoL) has been found to be just as great in a primary care sample as in hospital samples. 9,18–21

Clinical management

Conclusion

Temporising medical management is frequently given in general practice, and often includes prescribing antibiotics, decongestants and antihistamines, all of which have been shown to be clinically ineffective, and, worse still, have significant harms. 1,12,34,38,64 Furthermore, these interventions are associated with substantial NHS costs. Antibiotic prescription in primary care is rising progressively again, and has now exceeded the peak in the late 1990s, further driving the development of antibiotic resistance, which may lead to serious infections becoming untreatable. 47,50 The high prevalence of OME, and the fact that it is estimated that one-third of all cases of otitis media are primarily OME related,2,12 means that estimated rates of 80% antibiotic prescribing for all types of otitis media episodes in primary care are potentially reducible by a further 20–30%. 12 Finding an appropriate, feasible and low-cost management option for primary care for the majority of affected children must therefore be seen as an urgent priority, with the status quo of ‘watch and wait’ sometimes interpreted by parents as ‘doing nothing’ or as a form of demand suppression. 67

Autoinflation has been identified by a systematic search through the evidence as the best potential option for primary care. If found to be the case from research, a low-cost, safe and clinically effective treatment might resolve effusions and related symptoms, concerns and global impact on the child’s life and development. There is thus potential to improve child and parent QoL, increase satisfaction and adherence to a recommended watch-and-wait strategy, and also to reduce the harms of overprescribing antibiotics and other presently misapplied treatments.

A simple autoinflation method (Otovent) used for 1–3 months is proposed here in a pragmatic, open, randomised two-arm controlled trial in UK primary care, in which both arms receive usual (routine or standard) management, in order, primarily, to evaluate both clinical effectiveness and cost-effectiveness of the intervention. Health economic outcomes are proposed to be collected up to 12 months post randomisation.

Primary aim and objectives

The primary aim of the study is to evaluate the clinical effectiveness of autoinflation in resolving OME at 1 and 3 months by assessing the proportions of children showing rigorously defined improvement as accepted by Cochrane, that is, tympanometric resolution in at least one ear per child of a type B tympanogram (fluid) to normal, type A/C1, tympanograms. 35

Secondary aims and objectives

1. Tympanometric: assessment of the proportions of ears showing resolution from B to A/C1 types at 1 and 3 months. 35

2. Clinical outcomes: evaluation of the clinical effectiveness of the intervention on total symptoms (e.g. hearing difficulty, earache and difficulty concentrating) using a total diary score. We also used a total ear problem (mapped) QoL measure using a PROM [a 14-point questionnaire on the impact of OME (OMQ-14), which is a subset of the Medical Research Council-developed 30-point questionnaire on the impact of OME (OM8-30)].

3. Health economic (HE) assessment of the cost-effectiveness of autoinflation in terms of the cost per additional child achieving resolution of OME at 1 and 3 months, and also in terms of cost per quality-adjusted life-year (QALY). Evaluation by notes audit30 of the 12-month HE outcomes.

4. Qualitative: to describe the experience of using autoinflation, including nurse-observed competence and reported compliance, and develop an easy-to-use training package for everyday practice.

Introduction

A pilot study was proposed in advance of a main trial of autoinflation to test the feasibility of recruitment rates, randomisation procedures, training of practice staff, acceptability to patients and compliance with the autoinflation device. The pilot also improved costing estimates for the main trial.

Methods

Assessments

The primary outcomes assessed were tympanometric, resolution of type B (effusions) at 1 and 3 months. Tympanometric outcomes were assessed blind to intervention group by the chief investigator.

Compliance with the autoinflation device over the study period was recorded using a daily sticker reward chart completed by the child and was also parent reported (as use of autoinflation on a 4-point Likert scale and number of times per day). The ear-related QoL questionnaire (OM8-30) and Health Utilities Index (HUI) were also evaluated in the pilot study (Figure 1). 30

FIGURE 1.

Flow of participants through the pilot study.

Results

Patient recruitment

Practices commenced their computer searches in November 2009, with 357 children invited for screening from four practices between January 2010 and May 2010. Fifty-eight children were consequently screened for eligibility and 21 were randomised into the pilot.

The pilot Consolidated Standards of Reporting Trials (CONSORT) diagram, illustrated in Figure 2, details the numbers of children that progressed or otherwise through the study. Two children had to be excluded after randomisation, both in the autoinflation group, one because of an existing perforation of the eardrum and the second because of a tympanometric misclassification error, leaving 19 patients who were correctly randomised into the pilot.

FIGURE 2.

The pilot CONSORT diagram.

The baseline characteristics of children screened for the study are presented in Table 1 and all correctly randomised children are presented in Table 2.

TABLE 1 - Characteristics of screened children (n = 58)

Variable	n (%)
Age at screening (years)
4–5	12 (21)
5–6	32 (55)
6–7	4 (7)
7–11	10 (17)
Sex
Male	30 (52)
Female	28 (48)

TABLE 2 - Baseline characteristics of randomised children

Variable	Autoinflation (n = 9), n (%)	Standard care (n = 10), n (%)
Sex
Female	5 (56)	5 (50)
Male	4 (44)	5 (50)
Age (years)
4–5	3 (33)	2 (20)
5–6	4 (45)	8 (80)
6–7	–	–
7–11	2 (22)	–
Ethnicity
White	8 (89)	9 (90)
Oriental	–	–
Afro-Caribbean	–	–
Bangladeshi/Indian	–	1 (10)
Mixed	–	–
Other	1 (11)	–

Discussion

The pilot study under-recruited compared with the target set of 15 children per group. The main reasons for under-recruitment were twofold. First, delays in obtaining necessary ethics, governance and site permissions resulted in a late recruitment start date of January 2010, leaving only half the anticipated recruitment time frame (OME is a seasonal condition that substantially tails off around April). Second, the lowest recruiting of the four practices involved an enthusiastic GP who, because of time pressures, started recruitment extremely late.

A previous primary care trial estimated the need for 2940 children to be invited, in order to screen 1176 (40%), of whom 294 (25%) would be randomised. 30 Using the pilot CONSORT figures for a revision of the estimates, it was predicted that, of 5570 invitations for screening, 891 (16%) would respond and 294 (33%) would be randomised. This enabled a more precise estimate of practice recruitment and costs for the main study. More 4- to 6-year-olds than anticipated were identified, and approximately half of all those with symptoms who were screened were eligible and subsequently randomised. No parent/guardian of an eligible child refused consent for screening or to participate in the pilot, revealing very high acceptance rates for the study, and overall retention was good.

Web-based randomisation was the preferred option for site staff, and considered a more robust and inclusive system for randomisation, and was therefore implemented in the main study.

Children randomised to autoinflation were noted to be fully compliant with the method of autoinflation. No withdrawals occurred once children had started using the device. Feedback obtained from both parents and children about using the device was incorporated into subsequent training days for the main trial. Feedback included what to expect from the technique; the importance of prior stretching of new balloons; the need to involve the parents in demonstrating the method; and the need for persistence, especially over the first few days.

One child did experience nosebleeds while using autoinflation. The parent reported that the child had suffered from previous recurrent nosebleeds, but chose to continue with the study anyway. Dr Stangerup (Otovent inventor) told us that there had been no previous reports of nosebleeds as a complication of nasal balloon inflation (Professor Sven Eric Stangerup, University of Copenhagen, 2011, personal communication). However, it was decided on review that it would be best to avoid any such prior histories based on a degree of hypothetical risk. A new exclusion criterion was therefore added (Box 2).

The trial materials and operating manual were deemed satisfactory and were well accepted. The pilot study highlighted that not all nurses were confident with tympanometry, and three nurses would have liked additional training in use of the machine and interpretation (Box 3).

Conclusion

This small-scale study, once under way, was successful in recruiting patients, compliance was excellent and no major protocol changes were needed for the main study and costing was improved. A great number of useful and practical points were learned from the study, in both a systematic and informal way, from parents and nurses. The overall performance of the pilot was encouraging and appeared sufficient in terms of recruitment and in answering what was felt to be the major unknown issue about whether or not children were able perform the technique and achieve sufficient compliance over 1 month in a primary care setting. 35

Study design

We conducted an open, pragmatic RCT in primary care. We examined the difference in effectiveness between regular autoinflation with standard care and standard care alone.

Setting

The study was set in primary care in three regions of England: the South West, Thames Valley and Cheshire regions. The main study recruited children from 43 family practices from 17 UK PCTs, between January 2012 and February 2013.

Ethics approval and research governance

Ethics approval was awarded by the NRES on 10 August 2009, reference number 09/H0504/75 (see Appendix 1), and local research governance approval was obtained from all participating PCTs.

Recruitment and training of research nurses

Practices were recruited to the study by the PCRNs. At each practice a GP acting as principal investigator and a RN were assigned to the study. Practices were reimbursed for nurse time by the Department of Health service support costs.

Recruitment of practices

A total of 50 general practices were recruited to the study, of which 43 practices actively screened and randomised children during the study period (Tables 5 and 6). During winter 2011, 39 practices received training; seven did not continue to patient screening because of practice withdrawal (n = 3), slow start (n = 2), staff illness (n = 1) or research governance delay (n = 1). In order to maximise recruitment, an additional 11 practices were trained in autumn 2012 to replace nine low-/non-recruiting practices from the first season. Reasons for low recruitment included change in nursing staff (n = 3), practice withdrawal owing to other commitments (n = 2), recruitment problems (n = 1) and small practice list size that limited further recruitment (n = 1).

Training of research personnel

Five regional training days took place between November 2011 and January 2012 in Southampton (n = 2), Chippenham (n = 1), Oxford (n = 1) and Nantwich (n = 1). This provided convenient training locations for RNs and helped establish good and effective communication between RNs and the study team. One practice received on-site training and three practices requested additional pre-study visits from the study manager. One further training day took place in September 2012 (Oxford) for the 11 additional practices.

The training days involved the chief investigator, the trial co-ordinator, an audiologist, PCRN observers and a company representative. Study aims and methods were comprehensively covered. These included procedures for identifying and screening patients, taking consent, how to perform web-based randomisation, patient assessment and data management. A training manual was provided giving full details of the study methods and outcome assessments. Training in otoscopic examination and tympanometry was provided by the chief investigator and an audiologist from Starkey Laboratories Ltd, who gave detailed information and a practical demonstration of the technology and interpretation of the tympanometric results. This included classification of the four main types (A, C1, C2 and B) of tympanograms, recognition of obstructing wax, perforations, grommets and appropriate canal volumes for age, etc. Nurses had the opportunity to practise and refine their techniques during the training day. Brief training in the correct use of the autoinflation device (Otovent) was given by a representative from the suppliers (Kestrel Medical Ltd). All nurses or recruiting GPs had current training in GCP by the start of the trial.

As a means of additional support and training, nurses were invited to fax their initial screening tympanograms to the co-ordinating centre, where they were independently reviewed for categorisation based on all available parameters and observations. Any discrepancies in tympanometric classification were then fed back to nurses to help improve their precision of diagnosis in the field. Extra on-site training was offered by two experienced members of the study team where requested. Regional meetings were scheduled at the start of the second recruitment season to provide a study update and a further review of tympanometry and interpretation.

Recruitment of children

Participating practices were asked to invite 140 children as a target for screening. It was conservatively estimated from the pilot study that the proportion of invited children who would be recently symptomatic and/or whose parents would have concerns and who would, therefore, attend for screening would be 16%, and that one-third of these would be eligible for randomisation (140 children invited; 22 screened; and seven randomised in each participating practice, i.e. 1 in 20 children).

Children were identified by practice-based computer search or opportunistic case finding by practitioners, nurses and health visitors as follows.

Computer searches

• High-risk children, that is those aged 4–6 years, were identified from practice age/sex registers and those with one or more OME-related symptoms or concerns in the previous 3 months were invited for screening.

• An audit of the attendance records of 7- to 11-year-old children identified those with ear-related problems in the previous year.

Opportunistic case finding

• General practitioners, nurses and health visitors identified children leading to an in-practice referral to the RN.

The parents of children identified for the study received an invitation letter and information sheet, and children received an age-appropriate information sheet (see Appendices 2 and 3). Parents gave written informed consent for screening and children were invited to give written assent if deemed appropriate by the RN (see Appendix 4). Table 7 shows features of study entry by practice and PCT.

At the screening visit, the RN checked both ears for any obstructing wax, perforations or grommets using otoscopy. If the ear canal was occluded with wax, olive oil ear drops were recommended to soften and disperse the wax, and rescreening was rescheduled. Tympanometric screening was performed to assess full eligibility for the study, that is, every symptomatic child had to have one or two type B tympanograms confirming the presence of uni- or bilateral OME (Table 8 shows details of tympanometric classification used). This examination was requested to be repeated wherever the interpretation was unsatisfactory or inconclusive. The inclusion and exclusion criteria are shown in Box 1.

Randomisation and masking

Eligible children were individually randomised to autoinflation plus routine care or routine care alone within 1 week of screening. An independent external agency provided a centralised web-based randomisation system (www.sealedenvelope.com) for nurses to access while recruiting children to the study. The Oxford Primary Care CTU independently managed, co-ordinated, analysed and checked the data validity. The randomisation used an algorithm with minimisation based on three potential effect modifiers/confounders: age (< 6.5 years vs. > 6.5 years), sex and baseline severity of OME (one vs. two baseline type B tympanograms). 18 Owing to the nature of the intervention, use of placebo was not possible and therefore nurses, children and families were not masked to treatment allocation.

Trial intervention

The simple autoinflation treatment used in this trial involved inflating a purpose-manufactured balloon (Otovent), by blowing through each nostril into a connecting nozzle three times per day for 1–3 months (Figure 3). 55,56,60,61,73 Children in the treatment arm were instructed by watching the nurse and/or parent demonstrate the procedure, starting with stretching the balloon (by hand or mouth blowing). A website, which included a short instruction video, was also available as a back-up for parents and children (www.gluear.co.uk). A sticker book was provided to encourage the child’s ongoing participation. Children still recording a type B tympanogram in either ear at 1 month were advised to continue with nasal balloon autoinflation for a further 2 months. All study children (both arms) received their usual/routine clinical care as normal. At the end of the 3-month clinical study period, children in the standard care arm with tympanometric evidence of glue ear were offered a 1-month supply of nasal balloons.

FIGURE 3.

Nasal balloon autoinflation (Otovent). Reproduced with permission from Kestrel Medical Ltd.

Assessments

Baseline assessment was conducted within 1 week of screening. Parents of children in the intervention group were routinely contacted by telephone after 3 days to provide additional support for autoinflation, if required. All children were followed up at 1 and 3 months (Figure 4).

FIGURE 4.

Flow of participants through the main study.

Advanced appointments were made for the next follow-up assessment at each visit and an appointment card given. A postcard reminder was sent 1 week before the next appointment to encourage attendance. In the case of non-attendance, the RN was asked to attempt contact twice by telephone and twice with a missed appointment card, after which the patient was considered lost to follow-up.

Withdrawals

In accordance with GCP, parents/guardians were free to withdraw their children from the study at any time without affecting their medical care. Children were withdrawn from the study in the event of incorrect diagnosis at the time of randomisation (no type B tympanogram confirmation).

Outcome measures

Children were assessed for the most important clinical outcomes at 1 and 3 months post randomisation over a recommended 3-month waiting or monitoring period, over which time natural resolution effects would be expected to occur in some children. 3,5,6 It was anticipated that, if the method was effective, it would most likely be at 1 month, when taking into account poor compliance reported in one secondary care study,58 and implied or suggested concerns for general use and durability of the method in primary care. 1,35,37,74 Health economic outcomes were collected for 1, 3 and 12 months. With the protocol indicating that routine care should not be affected in any way during the clinical phase (3 months) of the study, referrals in particular would be difficult to assess without longer-term follow-up (12 months). The flow diagram for the main study was essentially unchanged from the pilot except that (1) the OMQ-14 replaced the OM8-30 at baseline and 3 months; (2) the TADAST hearing performance test (see Two alternative auditory disability and speech reception tests hearing test) was added at baseline and 1 month; and (3) a notes audit for HE purposes was completed up to 12 months post randomisation.

Main outcomes

Changes to the protocol

The majority of these were made as a result of the pilot study and are described in Chapter 2. During the main study a total of four substantial amendments were approved by the ethics committee and comprised minor changes to the study documentation, addition of a qualitative evaluation, a 12-month notes review (instead of 6 months post baseline) and a refinement to the study closure strategy, which allowed for a slight overshoot of recruitment. Full details are presented in Appendix 1.

Statistical methods

A detailed statistical analysis plan was developed at the start of the study.

Sample size calculation

A 45% control resolution (improvement) rate at 1 month was anticipated in the calculations, as found in the previous GNOME trial. 18 The best estimates for the expected difference at 1 month were based on a meta-analysis of four small secondary care trials that used Otovent, included in the update for Cochrane. 35 Thus, for resolution at 1 month, the most conservative evidence-based estimate of effect size was an OR of 2.4. Given this effect size, 250 children were required (125 in each group) for a standard α = 5% and power = 90%. With 15% lost to follow-up, 295 were needed in total (for power = 80%, 226 were needed in total). The sample was also powered to detect a ≈0.3 SD effect on continuous variables such as the OMQ-14 total score at 3 months, which was deemed clinically significant.

Changes to the statistical analysis plan

1. Analyses were not completed using Stata version 11.2 (which is now an old version) but rather SAS version 9.3 (SAS Institute Inc., Cary, NC, USA) and Stata version 13 (StataCorp LP, College Station, TX, USA).

2. Twenty multiply imputed data sets were created for the sensitivity analysis of the primary end point rather than the five data sets specified, as more data sets result in more robust estimates.

Recruitment and trial flow profile

Screening commenced in December 2011. The first patient was randomised in January 2012 and the last child was randomised in February 2013. A total of 1235 children were screened, with 320 children (26%) randomised into the study over a period of 13 months. Table 10 displays the characteristics of screened children. The main reasons for ineligibility were no type B tympanogram, not currently at school or a recent nosebleed. Ineligible children reported fewer symptoms associated with OME in the preceding 3 months, and had fewer consultations for otitis media and OME in the previous 12 months.

The study population comprised 167 (52%) boys and 153 (48%) girls, with an age range of 4–10 years (mean 5.40 years; median 5.71 years); 181 (57%) children had unilateral OME and 135 (42%) had bilateral OME [on review, four children (1%) recruited were deemed not to have OME on tympanometry]. Two hundred and fourteen (67%) were recruited between October and March, while 106 (33%) were recruited between April and September.

The baseline characteristics of randomised children were well balanced between the two groups (Tables 11 and 12) with six being the median number of symptoms and concerns [interquartile range (IQR 4–8)] in the routine care arm and seven (IQR 5–9) in the intervention arm. The trial demographic data were comparable to national figures, but 33% of participating parents (vs. 27% nationally) were educated to degree level or higher. 88

Exclusions, withdrawals, loss to follow-up and missing outcomes

Details of all these with reasons are included in the CONSORT trial profile provided (Figure 5). Retention in the study was good, with 27 of 320 (8.4%) lost to follow-up at 1 month and 39 of 320 (12.2%) by 3 months. Uninterpretable tympanograms owing to poor technique (leakage or low canal volume) and clinical problems (wax or perforation) were similar in both groups, leaving 131 children in the autoinflation arm and 132 in the routine care arm in the ITT analysis for the primary outcome.

FIGURE 5.

Main study CONSORT diagram.

Main trial results

Secondary outcomes

Proportion of children showing clearance/resolution of at least one type B tympanogram (effusion) back to normal A/C1 pressures at 3 months

The resolution of at least one B-type ear at 3 months was analysed in the same way as the 1-month primary end point. There were 108 children with resolution of at least one B-type ear at 3 months. At 3 months, those in the autoinflation group were 37% more likely to have resolution of at least one B-type ear (RR 1.37, 95% CI 1.03 to 1.83; p = 0.0283) (Table 15).

TABLE 15 - Summary of the main tympanometric outcomes at 1 and 3 months

Adjusted for baseline severity (one or two B-type ears), age and sex, and PCT.

Adjusted for baseline severity (one or two B-type ears), age and sex (not adjusted for centre effects because of non-convergence).

Tympanometric resolution in children of at least one affected type B ear	n	Standard care, n/N (%) resolved	Autoinflation, n/N (%) resolved	NNT	Adjusted RR (95% CI)	p-value
1-month analysisa	263	47/132 (35.6)	62/131 (47.3)	9	1.36 (0.99 to 1.88)	0.06
3-month analysisb	245	46/120 (38.3)	62/125 (49.6)	9	1.37 (1.03 to 1.83)	0.03

Proportions of children’s ears showing resolution of a type B tympanogram back to A/C1 normal pressures at 1 and 3 months

An analysis of each ear separately was conducted, adjusting for the correlation between ears from the same child using generalised estimating equations (Table 16). Results were very similar to the main per-child analyses.

TABLE 16 - Per-ear resolution at 1 and 3 months

Adjusted for baseline severity (one or two B-type ears), age and sex (not adjusted for centre effects because of non-convergence).

Generalised estimating equation model adjusts for correlation between ears for each child.

Tympanometric resolution in B-type ears	n	Standard care, n/N (%) resolved	Autoinflation, n/N (%) resolved	Adjusted RR (95% CI)	p-value
1-month analysisa,b	375 B-type ears/263 children	52/187 (28)	73/188 (39)	1.38 (1.01 to 1.87)	0.0420
3-month analysisa,b	348 B-type ears/245 children	52/166 (31)	74/182 (41)	1.41 (1.5 to 1.88)	0.0226

Tympanometric deterioration

There were 92 A- or C1-type ears at baseline for which valid tympanograms were also available at 1 month (Table 17). At 1-month follow-up, three A-type ears had deteriorated to B-type ears (one in the standard care group and two in the autoinflation group). Six of the C1-type ears had deteriorated to become B-type ears (four in the standard care group and two in the autoinflation group). Owing to the small number of deteriorations there was no further analysis on this end point.

TABLE 17 - Tympanometric deterioration of A- and C1-type ears at 1 month

Group	Ears with no deterioration	Ears which deteriorated	Total
Standard care	38	5	43
Autoinflation	45	4	49
Total	83	9	92

OMQ-14 impact measure

Table 18 displays summary statistics for baseline and follow-up (3-month) standardised OMQ-14 scores. Higher (more positive) scores represent worse outcomes. The average change from baseline in the standard care group was a decrease of 0.2 points, compared with a decrease of 0.7 points in the autoinflation group (p < 0.0001).

TABLE 18 - Summary statistics for standardised OMQ-14 scores

Note

A lower/more negative score indicates a better outcome.

Group	Time point	n	Number missing	Mean	SD	Minimum	Maximum
Standard care	Baseline	153	7	–0.04	0.95	–1.93	2.90
	3 months	121	39	–0.37	1.06	–2.15	3.06
	Change from baseline	121	39	–0.21	0.90	–3.11	1.90
Autoinflation	Baseline	153	7	–0.07	1.00	–1.81	2.88
	3 months	127	33	–0.70	1.01	–2.12	3.11
	Change from baseline	126	34	–0.69	0.84	–2.58	1.44

A mean change in baseline score of –0.69 points (SD 0.84 points) at 3 months in the treated arm represents a large improvement.

At 3 months, the adjusted mean change from baseline in the standardised OMQ-14 total scores was greater in the autoinflation than in the routine care arm (Table 19). The difference between groups was –0.42 (95% CI –0.63 to –0.22) points. This score difference represents an adjusted effect size of 0.48 (of a SD; p < 0.0001) favouring the intervention (Figure 6).

TABLE 19 - Change in standardised OMQ-14 scores

Adjusted for sex, age, centre (PCT), baseline values and baseline severity.

A lower/more negative score indicates a better outcome.

Group	n	Adjusted meana change from baseline	95% CI	Mean difference	95% CI	p-value
Standard care	121	–0.22	–0.40 to –0.04
Autoinflation	126	–0.64	–0.81 to –0.47	–0.42	–0.63 to –0.22	< 0.001

FIGURE 6.

Standardised OMQ-14 scores at the 3-month follow-up. A lower/more negative score indicates a better outcome.

Main clinical diary symptoms

Days during which parents reported their child had hearing loss, earache, sleep disturbance or problems concentrating, and days requiring pain relief and days off school were summarised in accordance with the statistics plan (to avoid multiple outcomes) as days with any problem (Table 20). Overall, children in the autoinflation arm had fewer days/weeks with any symptom/problem at 1 month (OR 0.66, 95% CI 0.41 to 1.05; p = 0.08) and at 3 months (OR 0.58, 95% CI 0.37 to 0.90; p = 0.02).

Compliance

Of the 130 parents recorded, 116 (89%) reported the use of autoinflation as ‘most’ or ‘all of the time’ during the first month of treatment, consistent with the daily compliance charts. This level of compliance appears to be maintained in those continuing treatment up to 3 months (68/85, 80%) (Table 22). Figures 7 and 8 show median comparator estimates for the two methods used to assess compliance (diary sticker charts and parental report).

FIGURE 7.

Compliance for months 0–1 (diary assessment vs. parental report of compliance).

FIGURE 8.

Compliance for months 1–3 (diary assessment vs. parental report of compliance).

Adverse events

As there were no previously reported adverse events for autoinflation, apart from anecdotal otalgia, the events recorded here are otherwise biased towards overinclusivity, and some overlap with the additional symptoms reported in Table 21. There was very little difference between treatment arms in terms of numbers of children with a nosebleed (15% vs. 14%), but there were more reported RTIs in the treatment group (18% vs. 13% of children, 37 vs. 18 episodes). Most of the RTIs were classified, however, as mild afebrile rhinorrhoea. Eight children in the autoinflation arm (compared with two in routine care) reported otalgia (Table 23). One 8-year-old child in the autoinflation arm was hospitalised with mild/early mastoiditis on day 10 of treatment and made a full recovery after treatment with intravenous antibiotics.

No bacteriology was available to confirm the cause. Clinical details of relevance were obtained in full and revealed that the child had been investigated by a paediatrician for recurrent RTIs in the previous 12 months. The case was reviewed independently by the DMEC (see Appendix 7). It was noted that no previous cases of mastoiditis had been reported with the use of Otovent, a device available over the counter in the UK and abroad, and no cases had been reported in any of the previous trial literature (≈500 receiving autoinflation). The DMEC did a full risk assessment and advised that it was safe to continue with the study.

Additional secondary analyses

Tables 24–29 deal in more detail with specified subgroups and have been included because of clinical importance and interest to the general literature. The analyses are not sufficiently powered and show a number of non-significant p-values for each headed subgroup. Details of the overlapping CIs between the two study arms for resolution rates (primary outcome) are also shown.

We have also, for reasons of clinical interest, added to this section the pre-baseline symptom (history) predictors of the finding of at least one type B tympanogram in the screened child. A univariate analysis was performed on all the parent-listed ear symptom/concerns items in the previous 3 months and those items that remained significant (Table 30) were then added in a stepwise multivariate analysis to derive a receiver operator characteristic (ROC) curve (Figure 9).

FIGURE 9.

Area under the ROC curve based on independent historical variables.

Meta-analysis for research in context

A meta-analysis with trials identified in the recent Cochrane review35 using similar outcomes at 1 month (ear-based analysis B to A/C1) favoured autoinflation (RR 1.61, 95% CI 1.26 to 2.06) (Figure 10). When the pilot study was combined with the main study as per the statistical plan, the primary care setting studies, when pooled, found a RR of 1.37 (95% CI 1.00 to 1.87) (Figure 11).

FIGURE 10.

Meta-analysis of 1-month outcomes (ear-based analysis). AIRS, AutoInflation Randomised Study; D + L, DerSimonian and Laird method; I-V, inverse variance method; SE, standard error.

FIGURE 11.

Meta-analysis primary care studies (per-child analysis) at 1 month. AIRS, AutoInflation Randomised Study; D + L, DerSimonian and Laird method; I-V, inverse variance method; SE, standard error.

Introduction

The AutoInflation Randomised Study (AIRS) was designed to estimate the clinical effectiveness and cost-effectiveness of autoinflation alongside standard care compared with standard care alone in children with OME in primary care. An economic analysis was part of the design of the study. Data on resource use and health-related QoL using the HUI376,77,89 were collected during the trial. Outcomes were expressed both as cost per QALY gained and as cost per additional proportion of tympanic resolution of the intervention compared with standard care. This section reports that analysis.

Methods

The economic evaluation was taken from both the NHS and a Personal Social Services perspective. The baseline analysis was at 3 months. Resource usage beyond 3 months and the effects of including travel costs and those due to parents’ time off work were explored in scenario analysis.

Cost estimation

Quality of life

Self-completed questionnaires using the HUI3 were recorded at baseline, 1 month and 3 months during the trial. The data in eight categories were used to estimate utility scores for each individual child. The mapping algorithm and score functions were purchased from Health Utilities Inc. The algorithms map data from the 17-item interviewer-administered questionnaire to each of eight attributes (vision, hearing, speech, ambulation, dexterity, emotion, cognition and pain) of the HUI3 classifications. Utility scores for each patient were calculated based on these algorithms.

Results

The NHS costs at 3 months showed no statistically significant differences between arms (Tables 33 and 34). The mean cost per patient at 3 months was £17.61 (95% CI £10.42 to £24.81) for the routine care arm, compared with £31.94 (95% CI £23.69 to £40.19) for the Otovent plus routine care arm. The higher mean cost in the intervention arm was almost entirely caused by the cost of the intervention at £14.22. Only small differences were found between arms in use of the other resource headings.

Sensitivity analysis

The cost-effectiveness acceptibility curve (probabilistic sensitivity analysis) for the cost per QALY put the probability of the intervention being cost-effective at 50.1% and 50.2% at a willingness-to-pay threshold of £20,000 and £30,000 per QALY respectively (see Figures 12 and 13).

FIGURE 12.

Scatterplot with 95% confidence ellipse. 95% confidence ellipse of difference in cost vs. QALY for Otovent vs. control.

FIGURE 13.

Cost-effectiveness acceptability curve.

Discussion

The intervention improved outcomes at a low cost, leading to just over 50% probability of it being cost-effective at conventional cost per QALY thresholds. This analysis was based on ITT, a cost difference that was statistically significant when bootstrapped and a QALY difference that was not statistically significant.

The cost-effectiveness analysis was based on a smaller subset of patients for whom data were available. This showed a statistically significant difference in cases resolved at 3 months and put the cost per case resolved at £132.

The difference in mean cost was almost entirely caused by the cost of the intervention. The improved outcomes were not associated with any change in NHS service use, or with any privately borne costs. This may be because the improvement was insufficiently large to change patterns or service use, or it may be because of individual practitioners being reluctant to change their behaviours until evidence for an intervention is proven. Given that all unresolved children in the standard care arm were offered Otovent at 3 months as per the protocol (an ethical consideration), the effects are confounded beyond the end of the trial at 3 months.

Limitations of the CEA include:

• The duration was short, with outcomes followed up only to 3 months and with selected resource use up to 12 months. Follow-up on service use to 12 months showed no difference in hospitalisations for related conditions, the main one being grommet surgery (± adenoidectomy).

• HUI used US population weights. This was because of the greater experience of this instrument with children and its prior use in a similar study by the same team. 30 The QALY differences were reassuringly in the same direction as the primary outcome. Although US population weights were used, these are likely to be similar to those for the UK, as indicated by the similarity of US and UK weights for European Quality of Life-5 Dimensions. 94 The potentially beneficial effects of improvement in case resolution may over antibiotic prescribing have not been included.

• The potentially beneficial effects of improvement in case resolution over antibiotic prescribing have not been included.

• Although imputation was used in the clinical analysis of outcomes at 1 month, given that this made little difference to the results, a decision was made not to combine the methods of multiple imputation and bootstrapping.

Conclusions

The cost-effectiveness analysis based on the statistically significant difference in cases resolved puts the cost per case resolved at £132. Although the cost difference was not statistically significant, it was based almost entirely on the cost of the intervention.

The cost per QALY analysis showed the Otovent device to be just likely to be a cost-effective intervention. The uncertainty reflects the small and non-statistically significant difference in QALYs, a generic rather than condition-specific measure of outcome than the OMQ-14.

Background

Autoinflation is a promising non-surgical treatment for OME, which has potential to improve natural resolution rates and QoL for children with OME-related concerns and symptoms, some of whom may be considered for ENT referral. The reliability of children inflating the nasal balloon and longer-term compliance with treatment has remained a concern regarding whether or not it could be a suitable treatment in primary care. 35 Although overall compliance has been assessed in the main trial, no previous qualitative work has been carried out with families or health-care professionals to explore facilitators and barriers to UK primary care management of OME, and the practicalities of use of autoinflation during this period.

This chapter reports a nested qualitative study, which is designed to inform the wider implementation of autoinflation in the primary care setting, including the monitoring process.

Objective

The qualitative study aims to explore the views and experiences of parents and practice nurses of both autoinflation and monitoring in primary care.

Methods

Findings

Discussion

This nested qualitative study of primary care monitoring and autoinflation in children with OME highlights the potential for an improved and more proactive role for general practice in the earlier diagnosis and treatment of this common childhood condition.

Strengths and limitations of the qualitative study

This research is the first to provide pragmatic, experiential data about use of autoinflation from a primary care setting, and includes both nurse and parent participant perspectives. It covers screening, AM and the use of the nasal balloon from the views of both the parents and primary care nurses. Using more than one data source to obtain different perspectives allows triangulation of the findings to check and establish study validity. 110

The study has also given insight into day-to-day, real-life experiences of children using the nasal balloon, which has not hitherto been formally captured in previous studies of autoinflation. This study information should help identify the common enablers and barriers to the wider implementation of autoinflation in a community setting.

It was not possible to recruit parents of children who withdrew or dropped out of the AIRS. Their experiences of screening, monitoring and treatment may well have differed to the study group, such that possible problems associated with AM and compliance with the autoinflation treatment may be missing. It might have also been useful to gather views and opinions from the participating GPs, especially regarding primary care-led services. Also missing were the direct voices of the children themselves. Including children in research can enhance the scope and findings of a study;111 however, in this instance the children were individually considered too young to be able to separately contribute to this study (predominantly 4–6 years).

Implications for clinical practice

The findings suggest that primary care professionals are eminently capable of engaging families early on in the process of AM with autoinflation and can provide good-quality information while drawing in parents and children in co-operative management decisions. Good demonstration/training with the autoinflation method, together with positive reinforcement by the health professional, will enhance child co-operation and improve overall adherence to the treatment schedule. Parents reported autoinflation to be acceptable to their children and compliance was improved by making the treatment part of the daily routine. However, the sample of parents had a somewhat higher than average educational level and were from areas of low social deprivation. Therefore, it would be of much interest to explore the potential barriers to autoinflation in lower socioeconomic groups where OME may have disadvantageous impacts.

Parents viewed practice nurses as accessible, local and able to provide continuity of care for OME. However, it remains uncertain as to exactly how a nurse-led service would work in the wider context of general practice, and this requires further research.

Principal findings

We report the results of the first pragmatic trial of the clinical effectiveness and cost-effectiveness of autoinflation that is generalisable to primary care, that is, to the majority of children attending practices with typical symptom clusters and impaired QoL linked to OME. It is the largest of the relatively few RCTs to date reported from either primary care or the community for any type of medical/non-surgical intervention,1,45 and the largest trial on autoinflation to date performed in any health setting. 35 There are currently no proven non-surgical interventions for glue ear, which often leads to inappropriate treatment with antibiotics and other ineffective remedies. 1 A NNT of 9 for autoinflation shows it to be a reasonably effective method for clearing middle ear effusions when using stringent tympanometric criteria of resolution. The method also significantly and importantly reduces the level of ear concerns and symptoms that include reported hearing loss, earache, difficulty concentrating and consequent impaired QoL for both child and family over a 3-month period.

We found autoinflation to be a simple, low-cost procedure that appears to be moderately easy to teach to appropriate-age selected children (attending school) in a primary care setting, with good reported compliance. Blowing up a balloon through the nose is an acceptable relatively non-invasive option with potential to add benefit to guideline recommended ‘watch and wait’ or surveillance processes, which are commonly performed in primary care and the community. It is thus an intervention with considerable potential to be used at scale in the NHS.

Research in context of other studies

The most recent Cochrane review of autoinflation,35 which highlighted the need for a large primary care trial, included three hospital studies of the same low-cost device trialled here. Adding our data to the meta-analysis more than doubles the available sample size of studies using similar outcomes with an estimated aggregate effect size (RR of improvement) of 1.61 (95% CI 1.26 to 2.06; I2 heterogeneity 0.0%).

Tympanometry findings can be misleading when comparing studies, particularly because different studies can class type C2 as resolved or improved. As we regard C2 as poorly predictive of effusion status, we have not considered such cases as sufficiently resolved and hence our resolution rates will be lower than in studies that present C2 as improved/resolved.

Effect on quality of life

Although clearance of effusions is a necessary and important physiological outcome, it is known that there is only poor correlation between tympanometry and audiometry (hearing level),2 and between audiometry and QoL. 19,21 For the child and parent, the most important issues are the expressed concerns about the consequential impacts caused by the OME. 75 Significant impacts of OME have been found in general practice settings to rival the impacts seen in UK secondary care settings. 21,30 These impacts include recurrent physical illness, hearing, speech and developmental impairments, and total effects on child and parent QoL. Taken from this perspective, the improved difference in the global OMQ-14 score of –0.42 points between arms (representing a moderate effect size) is both important and encouraging.

Feasibility and compliance

Children found autoinflation fun to do. In addition, the training method of demonstration by the nurse, then parent and then the child, was associated with good acceptability and engagement. With practice, nearly all the children mastered the technique: 89% of parents reported good compliance in their child’s use of the balloon at 1 month, and 80% at 3 months. Making the treatment part of daily routine could enhance compliance, especially over a longer period.

Adverse events and safety

Parent-reported adverse events were similar between groups. There were, however, more mild URTIs and mild to moderate otalgia, which usually settled quickly, in the treated arm. This contrasts with fewer URTIs found in two hospital studies. 55,56 A single case of mastoiditis in the treatment arm was reported to the DMEC, who conducted a full, independent review. They concluded that the case of mastoiditis could not be attributed to autoinflation and that it was safe to continue the trial (see Appendix 7).

Strengths

A key strength is that the study population is representative of typical cases of OME seen in primary care, that is, in most cases parents had recently expressed relevant concerns that suggested OME, and OME had been clinically confirmed at the point of initiating treatment. The findings are therefore likely to be generalisable to a majority of affected children. We think that the observed aggregate effects, in terms of both consistent direction and the magnitude of effect sizes, across a range of repeated tympanometric and clinical outcomes, strengthen the plausibility and reliability of our findings.

The trial methodology was rigorous in other ways; for example, a power calculation was performed and a large sample achieved within the allocated time frame, web randomisation was used, and the executor and generator of randomisation were kept entirely separate. The execution and generation of randomisation was done by a company using computer-generated randomised sequences and stratified according to an analysis plan. The trial was analysed on an ITT basis (and PP), with objective evidence of OME (both at trial entry and as an outcome), good treatment compliance ≥ 80%, and a very modest loss to follow-up of ≈10%. Patient and public involvement contributed to various aspects of the trial. Feedback about the practicalities and training of the treatment method from parents whose children participated in the pilot was incorporated into the main study. A lay member of the Trial Steering Committee, also a parent of children with glue ear, contributed to study recruitment strategies and had input into the qualitative evaluation.

All practice nurses had training in trial protocol and methods used, for example otoscopy and tympanometry. Two authors and one external audiologist, who were all blind to treatment allocation, independently reviewed the outcome assessments at 1 and 3 months. Nurses showed a substantial level of agreement with expert interpretation of tympanometry as a relative standard (κ > 0.7), which improved as the study progressed. Completion rates of trial forms (case report forms) were very high and multiple imputation methods were used for all missing data both at 1 and 3 months. A CONSORT diagram is provided (see Figure 5), with separate baseline tables for both the screened and entered populations.

Limitations

The main limitation of the study was that using a nasal balloon is a method that cannot be blinded and Hawthorne effects are possible. 112 However, the lack of blinding is unlikely to affect the primary tympanometric outcomes and, even if symptom and mapped QoL scores are affected by performance bias, the effects observed are still likely to be commensurate with those found in routine clinical practice. The HE analysis suggests that trial behaviour was very similar in both arms, and the PP analysis was not different from the ITT analysis in terms of effects. The study population included children who were deemed likely to be able to reliably perform autoinflation (≥ 4.5 years), whereas the usual presentation pattern to primary care and ENT clinics in the UK is from approximately 3.5 to 8 years. This does not mean that younger children cannot perform the technique – some children as young as 3 years have been able to use the device in secondary care,28,29 and even younger than 3 years when a novel counter-pressure method is used. 65

Clinical implications

At the time of writing there are no known effective non-surgical treatments that have been proven to be satisfactory to apply in primary care and the community for children with the tenacious cluster of symptoms and impacts that characterise OME. 2,7,12,23 The study findings of a treatment, a device, that actually works for OME in primary care has potential when judiciously applied to fill the ‘management gap’ in current practice that exists between either doing nothing effective or referring the worst cases for surgery (often after incurring long delays). Temporising strategies, an attempt to let nature take its course, are often seen as unreasonable delay(s) by parents. Some strategies, such as prescribing antibiotics, are a misplaced attempt to fill the therapeutic vacuum, because they are inappropriate, ineffective and harmful, and contribute a major threat to public health in the form of antibiotic resistance.

Although finding the method effective over 3 months, because fluid in the ear does not fully resolve in about half of all children who use autoinflation in the short term, and with the known tendency to recur, continued vigilance with consideration of surgical referral must remain a central consideration in evidence-based management of children presenting with OME.

In this study we have used relatively rigorous measures of diagnosis, impact and outcome, but there is no reason to suppose that GPs’ and practices’ own routine criteria for identifying cases of OME are generally inadequate or insufficient, given the time and demand pressures on the NHS. Like many areas of current health care, however, there is always scope for better definition of the problem needing fixing. The characteristic symptoms and concerns of OME are presented in Box 1 and the OMQ-14 that was used in the study (see Appendix 5). They are distinct from acutely presenting otitis media with fever and pain, and more problematic to child and family than those of simple self-limiting viral illnesses. The more frequently the child attends with relevant ear symptoms and concerns, the clearer that the case becomes for treatment. A secondary analysis found three symptoms/concerns: any ear concern in the previous 3 months; mis-hearing what is said; and needs the television turning up. These three symptoms together produce 70% of the area under the ROC curve, which is a reasonably good indication for management purposes where tympanometry is not available. However, it must be accepted that tympanometry is only a relative and not a gold standard. Treatment criteria clearly depend on the case being considered, and secondary care criteria for intervention, somewhat embedded in current clinical culture (perhaps because grommets are the only known effective treatment), cannot be used as a basis for any treatment applicable to an earlier case stage in a primary care context, where the therapeutic aim is timely and proportionate remediation.

Although autoinflation is generally acceptable with brief instruction, it may not be suitable for all, particularly children aged under 4 years, and does require regular ongoing commitment to treatment. The method is deemed to have scope to be used in the majority of routine symptomatic cases, and thus is capable of improving satisfaction with management and outcomes in primary care. It should be more widely used.

Implications for practice

This clinical trial is the first of its type in a primary care setting. It is one of the largest trials reported for interventions for OME from any setting. Considering the current evidence base for non-surgical interventions systematically, one is led to the conclusion that there is no prior justification for any cost-effective treatment for OME, at the point in the NHS where the majority of children are initially identified and treated. Furthermore, medical treatments that are presently applied as part of temporising management are not only ineffective but also harmful. The latter is particularly the case for antibiotics.

Our findings reveal that autoinflation using the balloon method is feasible in primary care. A NNT of 9 at both 1 and 3 months was found for improved clearance of middle ear effusions, beyond what can be expected from natural resolution effects alone (standard care). The symptom diaries (hearing loss, earache, etc.) showed significant and encouraging improvements by 3 months, a recommended waiting time, as did the mapped ear-related QoL measure (PROM). The effect size for effusion clearance is comparable to that achieved when smaller secondary care studies are combined in Cochrane and makes the case for wide use of autoinflation more robust.

Our sample characteristics are considered generalisable to primary care populations, of which they are reasonably representative in terms of both the baseline severity of the effusions and the prior number of typical symptoms and concerns expressed. There are relatively few exclusion criteria and the sample, although heterogeneous, has demonstrated important clinical effects. Baseline severity markers and method of recruitment were shown in the statistical models to have no effect on the primary treatment outcomes, which is an important clinical finding. The main limitation in terms of generalisability is the age of the children recruited in to the study because of age-related limitations with the method. Surgical studies have demonstrated effectiveness when recruiting children as young as 3 years,55,56 but compliance is likely to be poorer in such age groups.

In terms of capacity to change clinical practice, we have demonstrated that this method can augment the current natural resolution process in a beneficial, inexpensive, safe and timely fashion for the majority of children with symptomatic OME. It should therefore be an attractive initial option when one considers the unsatisfactory nature of present ‘temporising’ or available management options, which include usually either offering advice only or giving a ‘known’ ineffective and harmful treatment such as antibiotics or a decongestant, or referring the child prematurely for further evaluation for surgery.

Clinically, assuming the status quo of children identified in primary care across the UK with a working diagnosis of OME, the majority will be eligible for empirical management – the modus vivendi of primary care practitioners. Thus, although there are inevitable limits to what one can conclude from a large open pragmatic trial, the sum of the new evidence appears sufficiently strong to justify far wider use of this intervention than is currently the case.

Recommendations for future research

• The findings from this study should be reviewed in wider terms, for example relevance to current practice by multiprofessional groups to aid positioning and prudent application. There is potential to be considered in updates to the Cochrane meta-analysis of autoinflation and review of future NICE guidelines for OME.

• Further pragmatic research should be undertaken to evaluate the relative benefits of:

  • usual care in practices plus or minus autoinflation using HE and standardised outcomes

  • improved diagnostic care for recurrent otitis media/OME using trial-developed symptom predictors of effusions/impact measures or by selective use of tympanometry in primary care practices

  • shared care with other agencies.

  Such research work should aim to offer more effective integrated approaches for children recognised with OME symptoms and concerns in primary care.

• To improve the management of OME in primary care by the development of tools that encourages and promotes self-efficacy. Development of a web-based support intervention could provide evidence-based patient information, practical support for use of nasal balloon autoinflation to enhance uptake and compliance.

• To evaluate selective screening and monitoring of at-risk children in primary care, using age, attendance records, near patient hearing tests and/or short-form QoL questionnaires. This has potential to address Tudor Hart’s inverse care law,33 but needs to be shown to be cost-effective.

• Treatment failures are an important group for further research in primary and secondary care settings. Randomised trials would be helpful to determine the probable effectiveness of autoinflation as a re-treatment, treatment before surgery in hospital and also effectiveness in prevention of second operations for grommets.

• The youngest children are an important group for further research. To evaluate different autoinflation methods that evaluate comparative effectiveness and feasibility in relation to the age-related competence of the child. New methods using counter pressure may be promising for the very youngest children.

• There is potential scope for basic science technical development of the study device in relation to drug delivery to the Eustachian tubes.

• To evaluate the clinical effectiveness of oral steroids, which may be considered appropriate for young age groups or for treatment failures (of autoinflation), or for children seen in hospital prior to surgery. [Current research in progress includes a randomised trial funded by the NIHR HTA programme that addresses relevant issues (HTA reference number 11/01/26).]

• To update the epidemiology of otitis media with use of databases, for example Clinical Practice Research Database, with better differentiation of OME from recurrent AOM.

Contributions of authors

Ian Williamson (Associate Professor, University of Southampton and chief investigator) conceived and designed the study, led protocol development and the funding application. Also contributed to the analysis and interpretation, and led the drafting of all chapters of the report.

Jane Vennik (Trial Manager, University of Southampton) provided day-to-day management, co-ordinated recruitment and contributed to the data collection, analysis and interpretation. Also led the design, data collection, analysis and reporting of the qualitative study. Contributed to drafting of the all chapters of the report.

Anthony Harnden (Professor of General Practice, University of Oxford) contributed to protocol development and funding application, and contributed to the drafting of all chapters.

Merryn Voysey (Senior Statistician, Oxford PCVC-CTU) led the statistical analysis of the study and contributed to drafting Chapters 3 and 4.

Rafael Perera (Associate Professor, University of Oxford) contributed to the protocol statistics section and advised on drafts.

Maria Breen, Brendan Bradley and Sadie Kelly (Data Officer, Senior Data Specialist and Head of Trials, Oxford PCVC-CTU, respectively) provided CTU oversight for both the pilot and the main study, and contributed to the main trial design. Also supervised and managed the randomisation process, data collection, cleaning and validation, and commented on drafts of all chapters of the report.

Guiqing Yao (Associate Professor of Health Economists, University of Southampton) co-led the HE analysis and drafting of Chapter 5.

James Raftery (Professor of Health Economists, University of Southampton) co-led the HE analysis and drafting of Chapter 5.

David Mant (Emeritus Professor of General Practice, University of Oxford) contributed to protocol development and funding application, contributed to the drafting of all chapters.

Paul Little (Professor of Primary Care Research, University of Southampton) contributed to protocol development and funding application, contributed to the drafting of all chapters.

Contributors to protocol and pilot study

Dr Sarah Benge (Trials Unit Manager, Oxford PCVC-CTU) contributed to protocol development for the funding application.

Ms Michelle Caneppele (Trial Co-ordinator, PCVC-CTU) provided day-to-day management of the pilot study, including interpretation and reporting.

David Turner (Health Economist, University of Southampton) developed the HE plan.

Jing Jin (Statistician, Oxford PCVC-CTU) contributed to trial design and developed the statistical analysis plan.

Members of the Trial Steering Committee

Mr James Ramsden (chairperson), Professor Chris Frost, Dr Anthony Harnden, Dr Tim Whelan, Dr Ian Williamson (chief investigator) and Dr Rachel MacDonald (lay member).

Members of the Data Monitoring and Ethics Committee

Professor Kerry Hood (chairperson) and Mr Emeka Okpala.

Data sharing statement

All available data can be obtained from the corresponding author.

Disclaimers

This report presents independent research funded by the National Institute for Health Research (NIHR). The views and opinions expressed by authors in this publication are those of the authors and do not necessarily reflect those of the NHS, the NIHR, NETSCC, the HTA programme or the Department of Health. If there are verbatim quotations included in this publication the views and opinions expressed by the interviewees are those of the interviewees and do not necessarily reflect those of the authors, those of the NHS, the NIHR, NETSCC, the HTA programme or the Department of Health.