Oral steroids for hearing loss associated with otitis media with effusion in children aged 2 8 years: the OSTRICH RCT

Article history

The research reported in this issue of the journal was funded by the HTA programme as project number 11/01/26. The contractual start date was in March 2013. The draft report began editorial review in October 2017 and was accepted for publication in April 2018. 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

Christopher C Butler is a National Institute for Health Research (NIHR) senior investigator. Kerenza Hood and Amanda Roberts are members of the NIHR Health Technology Assessment General Board. Kerenza Hood is a member of the NIHR Clinical Trials Unit Standing Committee.

Copyright statement

© Queen’s Printer and Controller of HMSO 2018. This work was produced by Francis et al. under the terms of a commissioning contract issued by the Secretary of State for Health and Social Care. 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.

2018 Queen’s Printer and Controller of HMSO

Importance of the problem

Otitis media with effusion (OME) is the commonest cause of hearing loss in children in the UK, and up to 80% of children are affected by OME by 4 years of age. 1 Overall, the prognosis for OME is good, with over 50% of OME episodes resolving spontaneously within 3 months and 95% resolving within 1 year. However, 30–40% of children have recurrent OME episodes, and 5% of preschool children (aged < 5 years) have persistent (> 3 months) bilateral hearing loss associated with OME. 2

Hearing loss from OME can have an important impact on children’s mood, communication, concentration, learning, socialisation and language development. This may affect other family members and family function. OME in early childhood can affect intelligence quotient (IQ), behaviour and reading into teenage years. 3

The UK’s National Institute for Health and Care Excellence (NICE) 2008 guideline for OME management recommends a ‘watchful waiting’ period of 3 months, with referral to an ear, nose and throat (ENT) department if hearing is significantly affected, if OME persists for > 3 months or if there is suspected language or developmental delay. 4 Similar recommendations come from the American Academy of Pediatrics and the American Academy of Otolaryngology–Head and Neck Surgery. 5,6 Treatment options for these children are limited to hearing aids or surgical insertion of ventilation tubes (grommets or tympanostomy tubes) through the tympanic membrane. Hearing aids are an effective treatment, but this intervention is not problem free; children often find them uncomfortable, may feel self-conscious and may become a target for bullying. 7

Although the diagnosis of OME in primary care has increased over the last decade, the number of grommet operations performed in England fell from 43,300 in 1994–95 to 25,442 in 2009–10, primarily as a result of the watchful waiting strategy. 8 However, OME remains the commonest reason for childhood surgery in the UK and comprises a considerable workload for hospital ENT departments. Furthermore, there is wide variation in the rate of grommet surgery between regions that is unlikely to be explained by variation in disease. In Wales, there is sixfold variation in the European age-standardised rates of grommet surgery between the highest and the lowest local authorities. 9

Both hearing aids and surgery require referral to secondary care with risks and major cost consequences. The Department of Health and Social Care-commissioned ‘McKinsey’ report10 stated that the NHS could save £21M per year by reducing grommet insertion, a procedure that was assessed as being ‘relatively ineffective’, by a further 90%. This position has been challenged. Deafness Research UK11 and ENT UK’s 2009 OME (Glue Ear) Adenoid and Grommet Position Paper12 conclude that reducing access to grommets will disadvantage thousands of children who are in genuine need of treatment.

Rationale for the current trial

Antibiotics, topical intranasal steroids, decongestants, antihistamines and mucolytics are all ineffective treatments for OME. 13–15 A rigorous evaluation of anti-inflammatory treatment for OME has been a priority for many years. 16 Cochrane systematic reviews have found insufficient evidence for the effectiveness of both oral steroids and autoinflation (AI) devices in resolving OME in children to recommend implementation, but sufficient evidence to recommend further research. 14

A recent trial of an AI device in children aged 4–11 years with OME has found a modest effect for some children. 17 However, 80% of children are affected by OME before the age of 4 years, at a time when language development is most rapid and hearing loss has its greatest effect on language development. 3 Alternative management options to hearing aids or surgery for children aged < 4 years (who are unable to use an AI device) are required.

Williamson et al. 18 evaluated topical intranasal steroids for children with OME in general practice and found that they are unlikely to be clinically effective for OME. This may be because topical steroids applied through the nose are unlikely to reach the middle ear. However, systemic steroids do reach the middle ear epithelium and modulate OME in animal models. 19

The evidence from in vitro and animal models suggests that steroids reduce middle ear effusions and middle ear pressure. 20–23 Various mechanisms have been proposed for a role for steroids in resolving middle ear effusions, including (1) reducing arachidonic acid and associated inflammatory mediators, (2) shrinking perieustachian tube lymphoid tissue, (3) enhancing secretion of eustachian tube surfactant with a resultant improvement in tubal function and (4) reducing middle ear fluid viscosity by its action on mucoproteins. 24

The latest update of the Cochrane review on oral or topical steroids for OME (last search conducted in August 2010) found no benefit from intranasal steroids. 14 However, the review did identify evidence of a statistically significant benefit from oral steroids plus antibiotics versus antibiotics alone for OME (five studies, 409 participants; 23% in the intervention group and 47% in the control group with persistent OME at follow-up) and a trend toward a significant benefit for oral steroids versus placebo in the short term (three studies, 108 participants). Oral antibiotics alone are not effective. The only study to assess the effect of oral steroids on hearing as an outcome was underpowered.

Studies included in the systematic review were short term, underpowered, often had poorly described inclusion criteria and/or did not assess hearing at the time of inclusion, used ears rather than children as the unit of analysis and used intermediate outcome measures, such as tympanometry results, rather than improved hearing. No cost-effectiveness studies of oral steroids for OME were found. Therefore, there is insufficient evidence to recommend oral steroids as a treatment for persistent OME because of inadequate evidence about the short-term effects on hearing and cost-effectiveness, and the absence of evidence about the longer-term effects.

Potential harms from oral steroids

No significant adverse effects from steroids were reported by the studies included in the Cochrane review. However, the numbers of participants were too small to rule out that possibility. Short courses of prednisolone are widely used in treating children with acute asthma and adverse events are extremely rare; when adverse events do occur, they are largely limited to behavioural disturbances and dyspepsia and resolve on withdrawal of the steroid drug. The safety of multiple short courses of oral steroid therapy has been evaluated. 25 Short courses of oral steroids, such as prednisolone, do not have lasting negative effects on bone metabolism, bone density, adrenal gland function, or weight or height, even if used on several occasions over the course of 1 year. 26

Summary

There is an important evidence gap regarding the clinical effectiveness and cost-effectiveness of short courses of oral steroid treatment for OME. Identifying an effective, safe, cost-effective, acceptable non-surgical intervention for OME in children (including those in the first 4 years of life) for use in primary care remains an important research priority.

The Oral STeroids for the Resolution of otitis media with effusion In CHildren (OSTRICH) trial aimed to determine the clinical effectiveness and cost-effectiveness of a 7-day course of oral prednisolone (steroid) in improving hearing over the short term in children with bilateral OME, as diagnosed at an ENT outpatient or paediatric audiology/audiovestibular medicine (AVM) clinic, who have had symptoms attributable to OME present for at least 3 months and currently have significant hearing loss (as demonstrated by audiometry). 27

Summary of the trial design

The OSTRICH trial was a double-blind, individually randomised, placebo-controlled trial involving children with persistent OME and significant hearing loss, which aimed to determine the clinical effectiveness and cost-effectiveness of a 7-day course of oral steroids in improving hearing in children with bilateral OME. The trial was based in secondary care, primarily in ENT outpatient clinics, but also included paediatric audiology and AVM clinics. Participating sites were asked to identify children (aged between 2 and 8 years) with persistent bilateral OME and significant hearing loss. Eligible, consented children were randomly assigned to one of the two treatment groups: oral steroids or matched oral placebo.

At the baseline visit, participants’ hearing was assessed and parents reported on quality of life (QoL), the impact of OME on the family and health status, using established assessment tools. Children aged > 5 years were also asked about their QoL. Parents were asked to complete a diary for the first 5 weeks following enrolment to record daily symptom severity, use of medication and health-care consultations. Participants were followed up at 5 weeks, 6 months and 12 months after the day of randomisation.

The main analysis compared hearing resolution at week 5 in the active treatment group (oral steroids) and the placebo group.

The schedule of events and participant flow for the trial is summarised in Figure 1.

FIGURE 1.

Trial schema and participant flow.

Clinical effectiveness objectives

Setting

Participants were recruited from ENT outpatient or paediatric audiology and AVM clinics across Wales and England.

Site recruitment

The trial was open to participant recruitment from 19 March 2014 until 31 March 2016. A principal investigator led each site.

Ear, nose and throat outpatient clinics and paediatric audiology/AVM clinics were considered to be sites for the purpose of the trial. The Clinical Research Network (CRN) in England and the Health and Care Research Wales Workforce in Wales supported site recruitment.

Clinics were invited to take part in the trial by e-mail or newsletter from the CRN. Interested practices were initially contacted by e-mail and asked to provide further information about their feasibility for conducting the trial. This was followed up by telephone from the trial team to discuss the trial in more detail.

Each site involved had a research nurse/co-ordinator and a local site pharmacy.

Participant selection

Children were eligible to join the trial if they attended a participating NHS site for their routine care, met the following inclusion criteria and did not meet any of the exclusion criteria.

Participant recruitment

Participating clinicians (in ENT or paediatric audiology/AVM) were asked to identify eligible patients with bilateral hearing loss and a diagnosis of OME during routine outpatient consultations, from current grommet surgery waiting lists or hearing aid review lists. In addition, potentially eligible children were identified in audiology, AVM, paediatric audiology and community audiology clinics and interested parents/legal guardians were directed to the participating OSTRICH trial clinician.

Randomisation, blinding and unblinding

Withdrawal and loss to follow-up

Parents were informed that they had the right to withdraw consent for their child’s participation in any aspect of the trial at any time. If a parent indicated that they wished to withdraw their child from the trial they were asked to give a reason for withdrawal.

Parents who wished to withdraw their child from the trial were asked to decide if they wished to withdraw their child from:

• further treatment, but allow the child to participate in all further data collection

• active follow-up, but allow existing data and their child’s medical records to be used

• all aspects of the trial, as well as requiring all data collected to date to be excluded from the analysis.

To minimise loss to follow-up, parents who had given permission to be contacted by short message service (SMS) text messaging were sent a reminder of their scheduled appointment, when possible, a few days before the appointment.

Trial interventions

Participants were randomised to the active treatment group [oral soluble prednisolone (oral steroid)] or control group (matched oral soluble placebo). Clinicians were blinded to allocation and so prescribed the trial intervention (either prednisolone or placebo); this was dispensed by the site pharmacy.

Trial procedures

Data management and monitoring

Patient and public involvement

The study had a patient and public involvement (PPI) representative who has a child who had long-standing OME. She joined the trial as a member of the trial management group. Our Trial Steering Committee (TSC) and Independent Data Monitoring Committee (IDMC) also had PPI representatives who had personal experience of children with OME. Our PPI representatives made important contributions to reviewing parent and child information sheets, providing feedback on the trial protocol and providing guidance on strategies for successful recruitment. One of our PPI representatives was also interviewed for a case study on the BBC (British Broadcasting Corporation) Wales News. 33 The trial has also benefited from our PPI representatives’ contribution during the analysis and dissemination of study results and they will continue to contribute to dissemination activities.

Outcome measures

Clinical effectiveness statistical considerations

Analysis

All of the clinical effectiveness analyses were by ITT without imputation, with outcome values compared between groups using mixed-effect two-level regression models to adjust for site and age of child (2–5 years and 6–8 years) as stratification variables.

Additional exploratory analysis

Additional exploratory analysis was conducted as part of a student research project assessing the association between baseline hearing threshold and HRQoL (including both overall HRQoL and scores for each domain). Pearson’s correlation coefficient (assuming that the distributions were sufficiently normally distributed) was used and 95% CIs and p-values were presented. Pearson’s correlation coefficient (r) was used to investigate the strength of any correlations.

Decisions about exploratory health economic analyses would be assessed following a review of the subgroup analyses, as described in Primary analyses.

Summary of changes to the trial

The main changes to the protocol that occurred during the conduct of the trial are summarised below.

A number of changes were made to the protocol to make it easier for sites to recruit children and schedule the follow-up appointments. For example, the study extended the site coverage into England, the eligibility criteria for audiometry-confirmed hearing loss was extended to 14 days preceding recruitment, follow-up visits were conducted in ENT or audiology outpatient clinics, and the time-frame windows for follow-up were extended to + 2 weeks for the 5-week follow-up and ± 2 weeks for the 6- and 12-month follow-ups. Paediatric audiology and AVM clinics were included as sites, and audiovestibular physicians were included as designated OSTRICH trial clinicians.

Additions were made to the exclusion criteria, such as ear infections, Kartagener syndrome or primary ciliary dyskinesia, existing known sensory hearing loss, undergoing cancer treatment, on a waiting list for grommet surgery and anticipated to have surgery within 5 weeks and unwilling to delay it, and live vaccines 4 weeks prior to recruitment if aged < 3 years.

A number of changes to the planned trial procedures were made because of time constraints resulting from the longer than anticipated recruitment period; for example, the removal of medical notes search and data linkage used to identify health-care consultations during the 12-month follow-up period in primary and secondary care. As a result of this, a specific assessment of resource use at baseline could not be collected.

Finally, a number of further amendments were made to the protocol, such as sending reminders for follow-up appointments, contacting parents regarding missed appointments and exploratory analysis to assess the association between baseline hearing threshold and QoL. In addition, the study undertook a qualitative substudy to explore parents’ understanding of the treatment options available to them, their views on shared decision-making in the context of managing glue ear (OME) and their views on the use of oral steroids for glue ear (OME). Further changes were made to the proposed longer-term modelling to be conducted as part of the health economic analysis as a result of the trial results, which are explained in detail in Appendix 1.

Site recruitment

The study initially planned to recruit participants from seven secondary care sites in Wales; however, it became apparent that it would not be possible to recruit all 380 children in the allotted time from Wales only, recruitment was therefore extended to sites in England. In total, 35 expressions of interest were received; 13 were from enquiries made by Comprehensive Local Research Network research leads and a further 15 resulted from a call from the Medicines for Children CRN and National Institute for Health Research (NIHR) ENT Specialty Group. Ten sites did not progress any further, as they felt unable to commit to the recruitment target or were no longer looking to take on new studies. The process of obtaining local R&D approval was started for two sites but was not progressed, as a result of the lengthy time taken for initial discussions and approvals. A breakdown of the number of sites that expressed an interest in participating, agreed to participate and were actively recruited is presented in Figure 2.

FIGURE 2.

Site recruitment flow. EOI, expression of interest.

Participant recruitment and trial flow profile

The first child was randomised on 20 March 2014 and the last on 5 April 2016. The flow of children through the trial is represented in the Consolidated Standards of Reporting Trials (CONSORT) trial profile diagram in Figure 3.

FIGURE 3.

The main trial CONSORT flow diagram. ITT, intention to treat.

A total of 1018 children were assessed for eligibility, with 389 children (38%) randomised into the trial over a period of 25 months. A total of 535 reasons for exclusion were given for 503 children. The main reasons for exclusion were insufficient hearing loss (49.3%) and having no diagnosis of OME on the day of recruitment or the preceding week (16.1%). The parents of 126 children (12.3%) declined participation, and a small number were unable to be consented at site. The majority (50%) gave no reason for declining. The most common reasons given were ‘having grommets fitted’ (12%) and ‘carer not wishing the child to have steroids’ (9%). After randomisation, a further six children were found to be ineligible and three withdrew from the trial (and declined the use of their data), leaving 380 children (193 in the oral steroid group and 187 in the placebo group). There was slight differential loss to follow-up for the 5-week clinic appointment (nine in the oral steroid group vs. three in the placebo group), but over 95% of the population were retained for the analysis of the primary outcome at 5 weeks.

Randomisation was remote and online, and was stratified by site and age of the child at recruitment (2–5 years and 6–8 years). For the majority of the recruiting sites and children aged 2–5 years, treatment allocation was well balanced (Table 3). There was, however, a slight imbalance of treatment allocation in the 6- to 8-years stratum and in a few sites. The explanation for this imbalance is that there were 53 incomplete blocks of allocations (29 in age stratum 2–5 years and 24 in age stratum 6–8 years). In one site, allocations were stopped mid-block as a result of the IMP expiring and requiring disposal or the study recruitment period ending mid-block.

Baseline characteristics

The baseline demographics of the randomised children were similar in the two groups (Table 4). Slightly more boys were randomised, and the majority were from a white ethnic background. Over 60% of children were randomised in the winter or spring. For over two-thirds of children, this was their first episode of OME, although around 10% in each group had had six or more episodes. Over 60% of children had had their current problem for > 12 months (see Table 4). Comparable numbers of children had previously had a tonsillectomy/adenoidectomy, and a marginally higher proportion in the placebo group had received surgery for the insertion of ventilation tubes or had been fitted with a hearing aid. Just under 30% of children were on a waiting list for ventilation tubes. Of those with a sibling, one-quarter either currently had or previously had OME. Around 65–70% of children did not have asthma, eczema, or hay fever and 15% were on long-term medications. Fewer than 10% in each treatment group had received antibiotics for an ear infection in the past month.

Ear-specific PTA was the favoured hearing assessment in the older age group (i.e. 6–8 years) and is the recommended standard method in children aged 3 years or older. The remainder of 6- to 8-year-olds had ear-specific VRA or play audiometry. Soundfield VRA or soundfield performance/play audiometry was used in only 19.3% of children aged 2–5 years. The method of audiometry was balanced across treatment groups (Table 5). Over 85% of ears were tested using ear-specific methods over all four frequencies (0.5, 1, 2 and 4 kHz), with slightly lower numbers (around 80%) using the soundfield VRA or soundfield performance/play audiometry. Hearing loss was slightly worse in the oral steroid group, with most children having mild to moderate hearing loss. Tympanometry and otoscopy were performed in almost all children, with the majority having type B tympanograms, and a small proportion having type C tympanograms. The tympanic membrane could be visualised in most ears and the appearance suggested the presence of middle ear infusion. QoL was high and comparable between treatment groups (Table 6).

Time of follow-up assessments

Follow-up assessments were scheduled at 5 weeks (35 days), 6 months (182 days) and 12 months (365 days) after randomisation. Parents were encouraged to return for assessment as close as possible to the scheduled date. Table 7 shows that the range of timings for the follow-up assessments were comparable between treatment groups. A total of 316 children attended all three assessments, 14 attended the 5-week follow-up assessment only, four attended only the 6- and 12-month follow-up assessments, 38 attended two of the three and four children missed their 5-week assessment, but returned for the 6- and 12-month assessments. The proportion of responders attending the clinic within the specified window (± 14 days) was high at 5 weeks (≈90% in each group) but then decreased over the next two time points. A greater proportion of children randomised to receive oral steroids attended their appointments within the window (71% in the placebo group vs. 79% in the oral steroid group at 6 months, and 65% vs. 75%, respectively, at 12 months).

Medication adherence

A 7-day course of oral steroids or matched placebo was given as a single daily dose of 20 mg for children aged 2–5 years or 30 mg for 6- to 8-year-olds. Over 90% of diaries were returned, in which over 98% of responders reported that they started medication and, hence, initiated treatment (Table 8). In total, 31 diaries were not returned, but all medication had been received according to pharmacy records and 79% had reported taking all medication for 7 days. One participant had not completed the adherence data, but had provided dates to enable a verification that medication had been taken for the 7 days. Of those initiating treatment, the majority initiated treatment the day after recruitment as instructed. More parents reported that the child did not find the oral steroid as palatable (did not like the taste or spat it out) as the placebo (21 vs. 10 children). Minimal numbers of children vomited after medication.

Main trial results

Subgroup analyses

No differences in treatment effects between subgroups were found (Figure 4), and the p-values for the interaction term (treatment group by subgroup) in the model ranged from 0.04 to 0.74 (see Appendix 2).

FIGURE 4.

Forest plot of subgroup analyses.

Secondary outcomes

Audiometry

The proportion of children who attended clinic for a hearing assessment was over 95% at 5 weeks, reducing to around 86% by 12 months. Acceptable hearing using audiometry was described by examining the proportion of children with acceptable hearing between treatment groups at 5 weeks and 6 and 12 months. A total of 306 children had audiology assessments at all three time points; 19 had no follow-up at 6 months and 26 had no follow-up at 12 months. Of the 306 children, 62 had acceptable hearing at each time point [placebo group 25 (13.4%) vs. oral steroid group 37 (19.2%)]. A repeated measures multilevel logistic regression model (adjusting for site and age of child) showed that there was a significant increase in acceptable hearing from 5 weeks to the 6 and 12 months’ time points, with a constant 7–8% difference between treatment groups (Table 11 and Figure 5). There was no overall difference in acceptable hearing between groups (oral steroids compared with placebo averaged across all follow-up time points) and no differential effect of treatment over time.

TABLE 11 - Proportion of children with evidence of (a) acceptable hearing at audiology and (b) tympanometric resolution of OME at 5 weeks, and 6 and 12 months, by treatment group

The treatment effect is oral steroids compared with placebo averaged across all follow-up time points.

Adjusted for site and child’s age group at recruitment (2–5 years and 6–8 years).

Outcome	Treatment group, n/N (%)		Effect				Treatment × time effect (p-value)
			Time		Treatmenta
	Placebo	Oral steroid	Adjustedb OR (95% CI)	p-value	Adjustedb OR (95% CI)	p-value
Acceptable hearing at audiometry
5 weeks	59/180 (32.8)	73/183 (39.9)	Reference		1.42 (0.91 to 2.21)	0.121	0.975
6 months	86/166 (51.8)	105/174 (60.3)	2.30 (1.47 to 3.60)	< 0.001
12 months	99/162 (61.1)	118/170 (69.4)	3.46 (2.19 to 5.46)	< 0.001
Tympanometric resolution of OME
5 weeks	13/178 (7.3)	7/182 (3.8)	Reference		0.51 (0.20 to 1.30)	0.159	0.007
6 months	17/147 (11.6)	26/152 (17.1)	1.69 (0.79 to 3.61)	0.179
12 months	9/144 (6.3)	31/159 (19.5)	0.85 (0.35 to 2.06)	0.724

FIGURE 5.

Proportion of children with audiometric-acceptable hearing over time, by treatment group.

Tympanometric resolution of otitis media with effusion over time

Of those children who attended a clinic, tympanometry was performed in over 98% of children at 5 weeks, a slight decrease at 6 months to 81% and increasing to 90% at 12 months. The main reason that tympanometry was not performed was that the child had ventilation tubes in situ or was about to have surgery to place them. Evidence of tympanometric resolution of OME is defined as moving from a type B or C tympanogram at baseline to a type A tympanogram in at least one ear at the 5-week follow-up. Of the children who had tympanometry performed, a small proportion had evidence of resolution in at least one ear at 5 weeks, 6 months and 12 months (see Table 11). Although there was no overall effect between treatment groups or over time, the rate of resolution in the oral steroid and placebo groups had a different trajectory (Figure 6).

FIGURE 6.

Proportion of children with tympanometric resolution (type A) of hearing over time, by treatment group.

Main clinical diary symptoms

For the 5 weeks post randomisation, the weekly scores were reported by parents on 10 symptoms on a scale of 0 (problem not present at all) to 6 (problem is as bad as it could be). Most symptoms were not present at all (see Figures 20–27, Appendix 3). The following eight problems were combined into a single symptom scale to avoid multiple outcomes: hearing, ear pain, speech, energy levels, sleep, attention span, balance, and being generally unwell. At 1 week post randomisation, Cronbach’s alpha for the eight symptom scores was 0.77 at week 1, indicating good reliability between the eight symptoms, suggesting that they could be combined into a single symptom score ranging from 0 (problems not present at all) to 48 (all problems are as bad as possible). The factor analysis also suggested that these symptoms could form a single scale. Cronbach’s alpha for the subsequent 4 weeks was > 0.80 for all symptom scores, suggesting a relatively high internal consistency over time. The distributions of the weekly overall symptom score was positively skewed, indicating no problems (Figure 7). The highest median scores were at the end of week 1 (7 in the placebo group and 6 in the oral steroid group), indicating that these symptoms were not a problem. When the scores were changed into binary outcomes (no symptoms vs. some symptoms), there was no difference between treatment groups over time (Table 15). Two categories of symptoms (nausea, vomiting or indigestion, and changes in behaviour and mood over time) were examined separately; a high proportion of children had resolution of symptoms over time, with no difference between treatment groups and over time (see Table 15).

FIGURE 7.

Box plot of weekly overall symptom score, by treatment group.

TABLE 15 - Summary statistics for weekly overall symptom score, nausea and behaviour change by treatment group

The treatment effect is oral steroids compared with placebo averaged across all follow-up time points.

Adjusting for site and child’s age group at recruitment (2–5 years and 6–8 years).

Hearing, ear pain, speech, energy levels, sleep, attention span, balance and being generally unwell.

Week	Treatment group, n/N (%)		Effect				Treatment × time effect (p-value)
	Placebo	Oral steroid	Time		Treatmenta
			Adjustedb OR (95% CI)	p-value	Adjustedb OR (95% CI)	p-value
No problem with any symptomc (score of 0)
1	19 (13.6)	13 (8.7)	Reference		0.59 (0.28 to 1.27)	0.178	0.801
2	20 (13.4)	18 (11.6)	1.00 (0.50 to 1.99)	0.990
3	22 (14.7)	19 (12.3)	1.09 (0.55 to 2.15)	0.807
4	17 (11.4)	21 (13.2)	0.83 (0.40 to 1.69)	0.600
5	21 (14.7)	22 (13.8)	1.10 (0.55 to 2.19)	0.789
No symptoms of nausea, vomiting or indigestion
1	132 (79.5)	124 (72.5)	Reference		0.67 (0.40 to 1.11)	0.116	0.806
2	136 (83.4)	138 (81.2)	1.30 (0.74 to 2.28)	0.355
3	145 (87.3)	142 (83.5)	1.78 (0.98 to 3.23)	0.057
4	138 (84.1)	144 (84.7)	1.37 (0.78 to 2.41)	0.278
5	136 (85.0)	143 (84.6)	1.46 (0.82 to 2.59)	0.201
No symptoms of changes in behaviour and mood
1	96 (58.2)	90 (52.9)	Reference		0.76 (0.49 to 1.19)	0.231	0.779
2	97 (59.9)	105 (61.4)	1.09 (0.69 to 1.70)	0.723
3	102 (62.6)	113 (66.9)	1.20 (0.76 to 1.89)	0.425
4	105 (65.2)	111 (66.1)	1.35 (0.85 to 2.13)	0.201
5	106 (67.1)	113 (67.7)	1.47 (0.92 to 2.34)	0.104

Health-related quality of life: Paediatric Quality of Life Inventory and Health Utilities Index, version 3

Table 18 displays the summary statistics for baseline and follow-up (5 weeks, 6 months and 12 months) for the total PedsQL scores and by each of the five domains. Higher scores indicate better QoL. For all domains, the QoL was high and increased over time with a negatively skewed distribution, with a high proportion of parents reporting higher QoL for their children. For all PedsQL outcomes, there were no differences between groups nor significant trends over time.

TABLE 18 - Median (25th–75th centiles) total PedsQL scorea and each domain, by treatment group

PedsQL = high scores indicate better QoL (maximum = 100).

The treatment effect is oral steroids compared with placebo averaged across all follow-up time points.

Adjusted for site, child’s age group at recruitment (2–5 years and 6–8 years) and baseline PedsQL score.

Squared transformation used on the raw scores. Parameter estimate = adjusted difference in squared means.

Outcome transformed to binary: perfect health (score = 100) vs. non-perfect health (< 100). Parameter estimate = adjusted OR.

No transformation used. Parameter estimate displayed as the adjusted difference in mean.

PedsQL domain	Treatment group				Effect				Treatment × time effect (p-value)
	Placebo		Oral steroid		Time		Treatmentb
	N	Median (25th–75th centile)	N	Median (25th–75th centile)	Adjustedc estimate (95% CI)	p-value	Adjustedc estimate (95% CI)	p-value
Total PedsQL score
Baseline	187	82.1 (69.0 to 90.5)	189	84.8 (73.8 to 92.7)			–85.11d (–420.65 to 250.44)	0.619	0.475
5 weeks	176	84.8 (73.8 to 92.7)	182	84.5 (72.4 to 91.7)	Reference
6 months	158	84.5 (75.0 to 90.7)	162	82.6 (72.6 to 94.6)	–32.11d (–378.89 to 314.67)	0.856
12 months	149	85.7 (77.7 to 92.9)	154	86.9 (75.0 to 95.2)	146.53d (–205.89 to 498.94)	0.415
Physical health
Baseline	187	90.6 (78.1 to 100.0)	189	90.6 (79.7 to 98.4)			0.84e (0.51 to 1.37)	0.480	0.554
5 weeks	176	90.6 (81.3 to 100.0)	182	90.6 (80.5 to 100.0)	Reference
6 months	158	93.8 (81.3 to 100.0)	162	93.8 (77.3 to 100.0)	0.90e (0.54 to 1.49)	0.675
12 months	149	93.8 (85.0 to 100.0)	154	93.8 (84.4 to 100.0)	1.39e (0.85 to 2.28)	0.195
Emotional functioning
Baseline	187	70.0 (60.0 to 85.0)	189	75.0 (55.0 to 85.0)			2.02f (–1.85 to 5.89)	0.306	0.778
5 weeks	175	75.0 (60.0 to 90.0)	182	75.0 (60.0 to 90.0)	Reference
6 months	158	70.0 (55.0 to 85.0)	162	75.0 (60.0 to 95.0)	1.17f (–4.46 to 6.79)	0.684
12 months	149	75.0 (60.0 to 90.0)	154	80.0 (65.0 to 100.0)	2.05f (–3.67 to 7.77)	0.482
Social functioning
Baseline	187	90.0 (75 to 100)	189	90.0 (72.5 to 100.0)			1.20e (0.75 to 1.92)	0.436	0.854
5 weeks	175	90.0 (80.0 to 100.0)	182	90.0 (73.8 to 100.0)	Reference
6 months	158	90.0 (80.0 to 100.0)	162	90.0 (70.0 to 100.0)	0.81e (0.50 to 1.32)	0.403
12 months	148	95.0 (80.0 to 100.0)	154	95.0 (78.8 to 100.0)	1.10e (0.67 to 1.79)	0.717
School functioning
Baseline	179	75.0 (58.3 to 90.0)	183	70.0 (58.3 to 85.0)			–240.53d (–718.72 to 237.65)	0.324	0.916
5 weeks	172	80.0 (65.0 to 91.7)	176	77.5 (60.0 to 90.0)	Reference
6 months	154	83.3 (66.3 to 95.0)	156	80.0 (66.67 to 90.0)	184.37d (–308.02 to 676.76)	0.463
12 months	143	83.3 (66.67 to 91.7)	147	80.0 (60.0 to 95.0)	217.95d (–286.59 to 722.49)	0.397
Psychosocial
Baseline	187	78.8 (63.5 to 87.5)	189	78.3 (63.4 to 87.1)
5 weeks	175	81.7 (69.2 to 90.0)	182	81.2 (67.3 to 90.0)	Reference		0.71e (0.26 to 2.00)	0.509	0.577
6 months	158	80.0 (70.0 to 90.0)	162	79.0 (67.5 to 93.3)	1.57e (0.64 to 3.90)	0.327
12 months	149	82.7 (71.4 to 91.7)	154	84.0 (69.8 to 93.6)	1.45e (0.57 to 3.66)	0.433

The HUI3 comprises a family of multiattribute preference-based utility measures, in which scores can range from –0.36 to 1.00, whereby higher scores indicate better HRQoL. At all follow-up time points, the HUI3 distribution was negatively skewed, with a high proportion of parents reporting a higher QoL for their children (Figure 8 and Table 19). The decision was taken to recode the score as a binary variable based on the maximum score of 1 (healthy) versus scores of < 1 (Table 20). There was an increase in the proportion over time but no evidence of a difference between treatment groups and no discernible difference between treatment groups over time.

FIGURE 8.

Box plot of HUI3 score by time and treatment group.

TABLE 19 - Information accompanying Figure 8

N	159	164	155	164	152	155	142	150
Median (25th–75th centile)	0.80 (0.63 to 0.93)	0.79 (0.66 to 0.92)	0.85 (0.66 to 0.95)	0.84 (0.64 to 0.97)	0.92 (0.75 to 1.00)	0.88 (0.73 to 1.00)	0.92 (0.77 to 1.00)	0.92 (0.73 to 1.00)

TABLE 20 - Summary statistics for the HUI3 score,a by treatment group

High scores indicate better HRQoL (maximum = 1.00).

The treatment effect is oral steroids compared with placebo averaged across all follow-up time points.

Adjusted for site, child’s age group at recruitment (2–5 years and 6–8 years) and baseline binary HUI3 score.

Time point	Treatment group, n (%)		Effect				Treatment × time effect (p-value)
	Placebo	Oral steroid	Time		Treatmentb
			Adjustedc OR (95% CI)	p-value	Adjustedc OR (95% CI)	p-value
N (%) perfect health score = 1
Baseline	22 (13.8)	22 (13.4)			1.23 (0.66 to 2.27)	0.511	0.790
5 weeks	33 (21.3)	37 (22.6)	Reference
6 months	49 (32.2)	52 (33.5)	2.36 (1.31 to 4.26)	0.004
12 months	44 (31.0)	51 (34.0)	1.99 (1.09 to 3.65)	0.025

Objectives

The objectives of the economic evaluation were to:

1. estimate the costs associated with a 7-day course of oral steroids at 12 months

2. assess the cost-effectiveness of a 7-day course of oral steroids to improve hearing over the short term (5 weeks) in children with bilateral OME

3. assess the cost-effectiveness of a 7-day course of oral steroids to improve hearing over 12 months

4. explore the longer-term cost-effectiveness of oral steroid treatment.

The objectives were designed to estimate the likely economic impact of treating OME in the target population. The maximum clinical effect for oral steroid use was expected to occur within 4–6 weeks. This could result in short-term relief of symptoms and subsequent reduction in costs as a result of reduced health care resource use. Over the longer term, these could translate into reduced morbidity as a result of OME and a positive impact on HRQoL, and could allow children to possibly avoid further treatment (e.g. surgical intervention). The protocol set out an original time horizon of 12 months, with the plan to undertake a model-based analysis to estimate potential longer-term cost-effectiveness beyond the trial period. In light of the clinical results presented in the preceding chapter, the revision to this objective is explained, with the modelling plans summarised in Appendix 1.

Methods

Missing data

Considerable effort was devoted to minimising missing data. Research nurses were trained in data collection and the questionnaires filled out by them were designed to minimise the amount of missing information.

The general problems associated with missing data are particularly relevant to health economic analysis, especially in a within-trial analysis in which costs and QALYs are based on cumulative measures collected over the trial duration. Missing items relating to health-care service usage may undervalue the total cost, whereas missing outcome data may bias effects, as those individuals without information may be systematically different from those for whom all information is observed.

The use of a complete-case analysis provides a useful first exploratory stage of an economic evaluation. However, according to recommended good practice,37,40 it is not sufficient for a robust base-case analysis in which missing data are identified. The health economic analysis plan took this into consideration and proposed MI as an additional approach to support the complete-case analysis. 50 Measures that include multiple components over time are particularly susceptible to missing data issues; for these outcomes, MI analysis will be the central approach.

The decision (and appropriate method) for missing data imputation was informed by conducting a descriptive analysis of resource use and outcome data by group at each assessment point. The pattern of missing data was examined to ascertain whether or not the missing data could be considered to be missing at random in order to employ suitable MI methods. The complete-case analysis and available analysis were conducted as additional sensitivity analyses to explore the impact of different imputation methods on the base-case findings.

The ‘MI’ command in Stata was used to impute missing data from the total costs and from the HUI3 data measures at baseline, 5 weeks, 6 months and 12 months. Age, gender and site, as explained in Chapter 2, were included as covariates in the imputation models for the CUAs. Costs and utilities were imputed using predictive mean matching, as this allows sampling within the observed values and is less dependent on the assumptions of normality. Fifteen data sets were created, based on the number of missing data. These imputed data sets were used to produce a summary statistic (i.e. mean cost and utility).

To ensure consistency across the statistical and economic analysis, a complete-case analysis was used to inform the base-case analysis for both the primary CEA and the secondary CEA of hearing resolution. The CUA, which includes the HUI3 and PedsQL, utilises the MI approach in the base-case analysis.

Incremental cost-effectiveness analysis

Two main incremental analyses were undertaken. The first comprised a CEA to calculate the incremental cost of achieving an acceptable level of hearing in one or both ears at 5 weeks, with an additional analysis undertaken to assess cost-effectiveness at 12 months. A secondary CEA investigated the incremental cost per improvement in the PedsQL score. The second incremental analysis was a CUA to assess the incremental cost per QALY, using the HUI3 to derive utilities gained as a result of oral steroid treatment at 12 months, with a secondary CUA conducted to assess the incremental QALY gained as a result of oral steroid treatment at 12 months based on utilities derived from mapping the OM8-30 to the HUI3.

The results of the comparative analyses of incremental costs and effects were expressed as incremental cost-effectiveness ratios (ICERs).

An ICER is calculated as:

⁽¹⁾ ICER = C 1 − C 0 E 1 − E 0 = ΔC ΔE .

C₁ and E₁ represent the costs and effects of the intervention group and C₀ and E₀ represent the costs and effects of the usual care group, with ΔC and ΔE being the incremental costs and effects of the intervention compared with usual care.

This results in four potential scenarios, which can be illustrated by the cost-effectiveness plane (Figure 9).

FIGURE 9.

Illustration of the cost-effectiveness plane.

If the intervention is less costly (negative incremental costs) and produces more effect (i.e. positive incremental QALY gain), then the intervention is considered to be dominant to placebo and thus cost-effective (the results would be located in the south-east quadrant). Conversely, if the intervention is more costly (positive incremental cost) and produces less effect (negative incremental QALY gain), the intervention can be considered to be dominated by the placebo and is thus not cost-effective (results would be located in the north-west quadrant). In these scenarios, no ICER is reported.

In the third possible scenario, the intervention is more expensive (positive incremental cost) but produces more effect (positive incremental QALY gain, result located in the north-east quadrant). In the fourth scenario, the intervention would be less costly but would produce less effect (negative incremental QALY gain, result located in the south-west quadrant). In these last scenarios, the ICER is computed to assess whether or not the net incremental health gain (or loss) is worth the incremental cost (or cost-saving). For the CUA, NICE guidelines regarding the societal willingness-to-pay (WTP) threshold were used. 51 Generally, an ICER below £20,000 per QALY is considered to be cost-effective.

For the ICER produced by the CEA, descriptions of the probability (%) of being within notional WTP thresholds of £1000, £2500 and £5000 per acceptable hearing resolution achieved were used.

Results

Incremental cost-effectiveness analysis

Table 26 presents the incremental CEA at 5 weeks. The base-case analysis showed an incremental cost of £546 to achieve an additional hearing resolution. This additional hearing resolution finding is the result of statistically significant differences in costs and statistically and clinically insignificant differences in hearing resolution in oral steroids compared with the placebo outcome and, as such, the ICER should be interpreted within this context. In the sensitivity analysis, oral steroids were dominated by placebo (i.e. oral steroids had a higher cost and lower net benefit) in two of the four scenarios (when costs or outcomes were at the lower-bound 95% CI); this suggests that the findings were not robust to changes in these parameters.

TABLE 26 - Base-case and sensitivity analyses for the primary CEA (incremental cost per acceptable hearing resolution at 5 weeks)

Based on the upper and lower CIs from the base-case costs and covariate-adjusted clinical resolution gain at 5 weeks.

Parameter	Incremental		ICER [cost (£) per additional resolution]
	Cost (£)	Effect
Base case	39	7.1%	£546 per additional hearing resolution
Upper 95% bounda of net cost and upper 95% bounda of % successful hearing resolution	71	16.4%	£432 per additional hearing resolution
Upper 95% bounda of net cost and lower 95% bounda of % successful hearing resolution	71	–0.0%	Oral steroid dominated by placebo
Lower 95% bounda of net cost and upper 95% bounda of % successful hearing resolution	6	16.4%	£37 per additional hearing resolution
Lower 95% bounda of net cost and upper 95% bounda of % successful hearing resolution	6	–0.0%	Oral steroid dominated by placebo

The results on the bootstrapped replications are presented in the cost-effectiveness plane (Figure 10). Although point estimates are displayed in all four quadrants of the cost-effectiveness plane, the greatest density of the point estimates reflects the scenarios in which oral steroids could be considered to be cost-effective (i.e. the north-east quadrant) on the cost-effectiveness plane or oral steroid is dominated by placebo (i.e. higher costs and less effect, as represented by the north-west quadrant).

FIGURE 10.

Cost-effectiveness plane of incremental costs per acceptable hearing resolution at 5 weeks.

The CEAC is presented in Figure 11. This showed that, at a WTP threshold of £1000, £2500 and £5000 per acceptable hearing resolution achieved, the probability of oral steroids being a cost-effective option at 5 weeks was 41%, 60% and 65%, respectively.

FIGURE 11.

Cost-effectiveness acceptability curve of incremental cost per acceptable hearing resolution achieved at 5 weeks.

Table 27 shows the results at 12 months. The base-case analysis indicates that the incremental cost of every additional hearing resolution was £3052 after 12 months. When costs and outcomes are varied, uncertainty again arises in the results, as a result of the statistically non-significant differences in costs and treatment benefit with wide CIs.

TABLE 27 - Base-case and sensitivity analyses for the primary CEA (incremental cost per acceptable hearing resolution at 12 months)

Based on the upper and lower CIs from the base-case costs and covariate-adjusted clinical resolution gain at 12 months.

Parameter	Incremental		ICER [cost (£) per additional hearing resolution]
	Cost (£)	Effect
Base case	177	5.8%	£3052 per additional hearing resolution
Upper 95% bounda of net cost and upper 95% bounda of % successful hearing resolution	487	17.2%	£2831 per additional hearing resolution
Upper 95% bounda of net cost and lower 95% bounda of % successful hearing resolution	487	–5.5%	Oral steroid dominated by placebo
Lower 95% bounda of net cost and upper 95% bounda of % successful hearing resolution	–132	17.2%	Oral steroid dominates placebo
Lower 95% bounda of net cost and lower 95% bounda of % successful hearing resolution	–132	–5.5%	£2400 saved per reduction in adequate hearing

Figures 12 and 13 present the cost-effectiveness plane and the CEAC to illustrate the incremental cost of achieving an acceptable hearing resolution at 12 months. Similar to the 5-week results, the results are spread across all four quadrants. There is a 12%, 46% and 70% probability of oral steroids being a cost-effective option at WTP thresholds of £1000, £2500 and £5000, respectively.

FIGURE 12.

Cost-effectiveness plane of incremental cost per acceptable hearing resolution at 12 months.

FIGURE 13.

Cost-effectiveness acceptability curve of incremental cost per acceptable hearing resolution achieved at 12 months.

When the analysis was extended to reflect a limited societal perspective, the results remained consistent with the analysis based on a NHS perspective (see Appendix 9).

Cost–utility analysis

Table 28 presents the findings from the primary CUA. The base-case analysis shows that oral steroid treatment was dominated by placebo, with oral steroids costing an additional £145, with a net loss of 0.015 QALYs per child treated. As shown earlier in Tables 22 and 24, neither the differences in costs nor QALYs were statistically significant, with only small, numerical differences in both measures. The uncertainty surrounding these findings is emphasised by the results of the sensitivity analyses undertaken. The complete-case analysis suggests that oral steroid treatment yields an ICER of £22,882. Varying the parameters for both costs and outcomes further highlights the variability in the results.

TABLE 28 - Results from the base-case and sensitivity analyses for the primary CUA (incremental cost per QALY gain at 12 months)

Based on the upper and lower CIs from the base-case costs and QALY gains at 12 months.

Parameter	Incremental		ICER
	Cost (£)	Effect
Base case (values using multiple imputed costs and QALYs)	145	–0.015	Oral steroid treatment dominated by placebo
Complete cases	389	0.017	£22,882
Upper 95% bound of net costa and upper 95% bound of QALY gaina	426	0.024	£17,750
Upper 95% bound of net costa and lower 95% bound of QALY gaina	426	–0.054	Oral steroid treatment dominated by placebo
Lower 95% bound of net costa and upper 95% bound QALY gaina	–136	0.024	Oral steroid treatment dominates placebo
Lower 95% bound of net costa	–136	–0.054	£2518 saved per QALY lost

Based on the bootstrapped replications, this uncertainty is clearly illustrated in the cost-effectiveness plane (see Figure 14). The distribution of results across all four quadrants is evident, with the greatest density reflecting the scenarios in which oral steroids are dominated by placebo (i.e. less effect and higher costs, as represented by the north-west quadrant).

FIGURE 14.

Cost-effectiveness plane of incremental cost per QALY gain at 12 months, based on values imputed from MI.

The CEAC is depicted in Figure 15. The probability of oral steroid treatment being cost-effective when compared with placebo at a £20,000-per-QALY threshold is 17%, increasing slightly to 22% at a £30,000-per-QALY threshold.

FIGURE 15.

Cost-effectiveness acceptability curve of incremental cost per QALY gain at 12 months.

When the analysis was extended to reflect a limited societal perspective, the results remained consistent with the primary analysis (see Appendix 10).

Subgroup analysis

As there was no evidence of a significant treatment effect in the subgroup analysis undertaken, this was not further considered within the economic analysis.

Secondary cost-effectiveness analysis: incremental cost per improvement in the Paediatric Quality of Life Inventory at 12 months

Table 29 shows the 12-month CEA. The incremental cost of achieving a point improvement in the PedsQL score at 12 months was £220 for the base case, and this remained consistent when complete cases were examined. Again, these results were sensitive to changes in varying the costs and outcomes.

Figures 16 and 17 display the cost-effectiveness plane and the CEAC curve. Again, these showed that point estimates were spread across all four quadrants. The probability of oral steroids being cost-effective based on notional WTP thresholds of £1000, £2000 and £5000 per point improvement in PedsQL score at 12 months was 50%, 51% and 51%, respectively.

FIGURE 16.

Cost-effectiveness plane of incremental cost per point improvement in PedsQL score at 12 months.

FIGURE 17.

Cost-effectiveness acceptability curve of incremental cost per point improvement in PedsQL score at 12 months.

Secondary cost–utility analysis: incremental cost per quality-adjusted life-year gain (based on utilities estimated from mapping the Otitis Media Questionnaire to the Health Utilities Index, version 3)

Table 30 shows the sensitivity analysis for the mapped OM8-30 to HUI3 scores for the MI approach and the figures from the complete-case analysis. The base-case approach found that there were insignificant differences between groups for both the incremental costs and effects. The base-case ICER of £26,750 should be viewed in the context of the sensitivity analysis, which finds estimates in each of the four cost-effectiveness quadrants.

Exploration of the longer-term cost-effectiveness of the intervention

The original protocol set out a third objective to explore the long-term cost-effectiveness of a 7-day course of soluble oral prednisolone as a treatment for bilateral OME in children. The health economics plan contains a summary of the analysis proposed. If feasible and indicated by the initial trial results, a decision-analytic model would be developed to assess the longer-term cost-effectiveness of the intervention. Prior to this, and on receipt of the initial analysis of the clinical effectiveness results, a feasibility check of undertaking the modelling would be discussed with the OSTRICH trial team based on the following parameters:

1. the intervention shows sufficient evidence of clinical effectiveness during the trial

2. the trial results (and supporting literature) provide a sufficiently robust source for the estimation of all data inputs attributed to the intervention compared with placebo

3. the model can be realistically expected to produce plausible estimates of the longer-term costs and outcomes associated with the intervention.

Preparatory work was undertaken, including a rapid review of the economic literature and spot searches to identify suitable data inputs supplementary to the trial results, to inform the model and a detailed modelling plan (see Appendix 1). Published model-based analyses of comparable interventions faced considerable challenges regarding outcome differences and available input data. 52 Similarly, the OSTRICH trial found no evidence of a statistically significant treatment effect or impact on health outcomes at 12 months. In addition, considering the paucity of evidence available within the literature on the longer-term impact of OME and its management, the time horizon would be severely constrained to 2 years. Even within this shorter duration, no reliable means for estimating the likely persistence of the small and statistically non-significant clinical and health outcome (QALY) effects beyond 12 months could be found.

The feasibility of modelling was fully discussed with the OSTRICH trial team and members of the TSC, in light of the presentation of the initial clinical effectiveness and cost-effectiveness findings. The consensus was that conducting the model-based analysis at this point was not feasible, but this could be revisited in the future, based on extending the meta-analysis reported in the Cochrane review14 to include the OSTRICH trial findings. On the basis of this further meta-analysis (i.e. whether or not there is stronger evidence of a significant treatment effect resulting from the use of oral steroids on OME), this would provide more robust and reliable data to inform a model-based analysis.

Conclusions

The economic evaluation found that oral steroids at 5 weeks cost more than a placebo, and there was no significant treatment effect. Interpreting the results of the CEA in light of the clinical findings is crucial, particularly as the findings failed to demonstrate a statistically or clinically significant treatment effect. The base-case CEA suggests that it would cost £546 to achieve an additional hearing resolution at 5 weeks, with uncertainty demonstrated in the sensitivity analysis; however, this finding should be fully considered within the context that there was no statistically significant difference in hearing resolution between the two groups.

The base-case incremental cost per QALY at 12 months estimated that oral steroids were dominated by placebo and oral steroids would not be considered to be a cost-effective option. However, there was again considerable uncertainty seen in the sensitivity analyses. The impact of the joint uncertainty in cost and outcomes is illustrated by the CEA curve estimating a 17% and 22% probability of oral steroids being considered to be cost-effective based on a societal WTP threshold of £20,000 and £30,000 per QALY gain, respectively.

Summary of the main results

A short course of oral steroids in children with symptoms of OME for at least 3 months, and proven significant bilateral hearing loss, was neither clinically effective nor cost-effective. A greater proportion of children in the group randomised to oral steroids, than those in the placebo group, had achieved satisfactory hearing at 5 weeks [39.9% vs. 32.8%, absolute difference of 7.1%, 95% CI –2.8% to 16.8%, number needed to treat (NNT) = 14]. However, this difference was not statistically significant. The secondary outcomes were consistent with the picture of a small or no benefit, and we found no subgroups that achieved a meaningful benefit from oral steroids. There was no significant increase in adverse events in the intervention group and we found no evidence to suggest important harms from taking oral steroids.

The results of the primary CEA suggest that the incremental cost of achieving an additional hearing resolution at 5 weeks as a result of oral steroid treatment was £690; this increased to £3052 at 12 months. However, the overall health economic analysis is more complex, with differences in costs and outcomes at 12 months being small and not statistically significant, and the CUA (incremental cost per QALY gain at 12 months) suggesting that oral steroids are dominated by placebo (i.e. they are less effective and more expensive).

The study did, however, identify a resolution of hearing in nearly one-third of those allocated to the placebo group by 5 weeks, more than half at 6 months and more than 60% at 12 months. These high rates of resolution may be useful in shared decision-making approaches to discussions concerning treatment options, including watchful waiting, with the carers of children who have OME.

Strengths and limitations

The OSTRICH trial is the first randomised controlled trial of oral steroids for OME, and one of only a few OME trials that used audiology-assessed resolution of hearing as the primary outcome. We found that the oral steroid and placebo groups had different trajectories in tympanometry resolution over time, but there was no difference in functional health status or hearing, underlining the importance of functional and patient-reported outcomes in addition to proxy measures. The hearing tests the study used are relatively objective, related to the resolution of the underlying pathology and to functional status and deemed to be a key outcome by the study’s PPI representatives. Clinicians, participants, parents, audiologists and members of the research team were all blind to the intervention allocation, and the study used matched placebos to help to maintain blinding. The study used remote, independent randomisation, and there was no indication of breaches in allocation concealment. Participants were recruited from a range of different hospital and audiology settings across the UK; all of these participants met strict inclusion criteria, including having symptoms for at least 3 months and audiology-confirmed significant hearing loss. Poor soundproofing of audiology testing rooms could affect the quality of audiology data. However, all sites received extensive training and monitoring, and over 60% of patients were recruited in university/teaching hospital sites, so it was believed that this was unlikely to be a problem in this study. The study measured many important secondary outcomes, including insertion of ventilation tubes and effect on QoL and functional health status, using well-validated instruments.

The trial was adequately powered to detect a 15% difference in recovery at 5 weeks, and is the largest trial of oral steroids for OME that has ever been conducted. However, more children than anticipated recovered by 5 weeks spontaneously (20% were used in the sample size calculation, but the actual rate in the placebo group was 32.8%), leading to the possibility of a type II error. The study was able to follow participants for up to 12 months and achieved high follow-up rates (over 90% at 5 weeks and 6 months, and close to 90% at 12 months). Reported adherence to trial medication was also high, suggesting that lack of effect was not caused by poor adherence to treatment.

The resource use questionnaire allowed participants to provide either a precise figure or a range, which resulted in some analytical challenges, and there were some differences in the way that resource use data were recorded at 5 weeks and at 6 and 12 months. Slower than anticipated recruitment resulted in the need to take a pragmatic decision to not collect routine health record data. As a result, the study was not able to validate self-reported health-care consultation data or collect baseline resource use data. This lack of baseline resource use data made it impossible for the study to assess any potential baseline imbalance in resource use and take this into account in the base-case and sensitivity analyses.

The study used a robust preference-based HRQoL measure (the HUI3) and the additional exploration of the impact of using alternative approaches (mapping of the OM8-30 ), which has already been investigated in detail within a comparable patient population. 47 Although there is evidence that the HUI3 can be used to collect self-reported HRQoL directly from children aged ≥ 5 years,46 the majority of children were too young to provide direct child-based HRQoL and, therefore, the study had to rely on parent proxy reporting, which has its own biases. 53

Generalisability

Participants were recruited from 23 hospitals and audiology clinics in Wales and England. Sites included large teaching hospitals, children’s hospitals and smaller general hospitals. Although only 38% of those participants assessed for eligibility were included in the trial, the main reason for not being included was not meeting the trial inclusion/exclusion criteria (in particular, not having bad enough, objectively measured hearing loss). Therefore, the study’s results are likely to be generalisable to the population meeting the inclusion criteria for this trial (i.e. symptoms of OME for at least 3 months and clinically important bilateral hearing loss). Most children included in the trial were ethnically white, and so the study’s results may not be generalisable to all ethnic groups.

The majority of children included in the trial (63% of 2- to 5-year-olds and 69% of 6- to 8-year-olds) had mild hearing loss (26–40 dBHL) and, therefore, our results may not be generalisable to children with more moderate hearing loss, although the study did not see any evidence of a beneficial effect in the subgroup with more moderate hearing loss. Furthermore, it is possible that children who did not meet the study inclusion criteria (those with Down syndrome or unilateral hearing loss, for example) may have responded differently.

The study recruited from secondary care sites (hospitals and audiology clinics) and, therefore, there needs to be caution when generalising these findings to a primary care setting. However, there is no reason to believe that the results are likely to be very different for children who meet the same criteria (including symptoms for at least 3 months and demonstrable hearing loss) but who are identified in primary care, especially as the majority of children in the trial had mild hearing loss.

Interpretation

The point estimate for the measure of effect suggests a small benefit from oral steroids in terms of resolution of hearing at 5 weeks. However, the CI crosses the null and, therefore, no effect or even a harmful effect cannot be excluded. Although no formal threshold exists for the WTP for an additional hearing resolution achieved, Williamson et al. 18 used a notional WTP threshold of £1000 per OME cured. 18 Using a similar notional value, oral steroids would have an 85% and 62% probability to be cost-effective at 5 weeks and 12 months, respectively.

The latest update of the Cochrane review on oral or topical steroids for OME (last search in August 2010) identified three trials (108 participants) of oral steroids versus placebo, which suggested a benefit from oral steroids, but did not demonstrate a statistically significant finding. 14 The review also identified five trials (409 participants) comparing oral steroids plus antibiotics with antibiotics alone, and found that the inclusion of oral steroids with antibiotics resulted in a statistically significant benefit in terms of resolution of OME. Studies included in the systematic review were short term, underpowered, often had poorly described inclusion criteria and/or did not assess hearing at the time of inclusion, used ears rather than children as the unit of analysis, and used intermediate outcome measures, such as tympanometry results, rather than improved hearing. No other cost-effectiveness studies of oral steroids for OME were identified. However, despite the poor quality of the prior evidence, the results of the OSTRICH trial are similar and suggestive of a small benefit from oral steroids. Whether or not such a small benefit is clinically important is another question. The study’s point estimate suggests a NNT of 14, which is not dissimilar to other medical interventions (e.g. a NNT of 15 for nicotine replacement therapy for smoking cessation54) and could justify the use of a low-cost and safe intervention to reduce the burden of an important problem, such as significant hearing loss. The study found little evidence to support the belief that a short course of oral corticosteroid therapy is associated with significant risks, either from the review of the literature or from the adverse effect and safety data from the current study. A subsequent trial found that oral steroids and oral steroids followed by intranasal steroids resolved OME more than watchful waiting at 6 weeks, but by 3 months this advantage disappeared. 55 The American Academy of Otolaryngology–Head and Neck Surgery Foundation and the American Academy of Pediatrics, informed by the flawed studies in the review, recommended against oral steroids for OME. 56 Despite this, adults diagnosed with OME are more likely to be prescribed oral steroids than those without. 57

A review of surgical treatments of OME found that ventilation tubes improved hearing and time to resolution of OME, but did not improve speech, language or other functional outcomes compared with watchful waiting or myringotomy, and that tubes increased the rate of otorrhoea and tympanosclerosis. 58

An overview of studies found that OME diagnosed by tympanometry of unknown duration had 28% spontaneous resolution by 3 months (95% CI 14% to 41%), rising to 42% by 6 months (95% CI 35% to 49%). 5 In the OSTRICH trial, higher rates of hearing resolution associated with OME were found.

Conclusions

Independent members of the Trial Steering Committee and Independent Data Monitoring Committee

We would like to thank our TSC and IDMC for their continued support.

Research networks

We also wish to thank the research nurses, clinical study officers and administrators who supported the study at each site. This includes staff at the Health and Care Research Wales Workforce (South East, South West and North Wales regions) and the local NIHR CRNs (West Midlands, Yorkshire and Humber).

Recruitment sites

We would like to thank all of the staff involved at our participating sites, including the principal investigators, ENT consultants and specialist registrars, audiologists, clinic staff, research nurses, research co-ordinators and administrators from the following sites:

• NHS health boards in Wales:

  • Cardiff and Vale University Health Board

  • Aneurin Bevan University Health Board

  • Cwm Taf Health Board

  • Abertawe Bro Morgannwg University Health Board

  • Hywel Dda University Health Board

  • Betsi Cadwaladr University Health Board.

• NHS trusts in England:

  • Royal Wolverhampton NHS Trust

  • Newcastle Upon Tyne NHS Foundation Trust

  • Bradford Teaching Hospitals NHS Foundation Trust

  • London North West Healthcare NHS Trust

  • Maidstone and Tunbridge Wells NHS Trust

  • Royal Free London NHS Foundation Trust

  • University College London Hospitals NHS Foundation Trust

  • University Hospitals of North Midlands NHS Trust

  • Isle of Wight NHS Trust

  • Sheffield Children’s NHS Foundation Trust

  • Central Manchester University Hospitals NHS Foundation Trust

  • Brighton and Sussex University Hospitals NHS Trust

  • Surrey and Sussex Healthcare NHS Trust

  • Blackpool Teaching Hospitals NHS Foundation Trust

  • Walsall Healthcare NHS Trust

  • Worcestershire Acute Hospitals NHS Trust.

Finally, we would like to thank the children and families who participated in the trial, without whom this trial would not have been possible.

Contributions of authors

Nick A Francis (GP and Clinical Reader, Primary Care and Public Health) was the lead applicant and a co-chief investigator for the OSTRICH trial. He had overall responsibility for the study design and implementation, and the writing of the background, methods, results and discussion sections of the report, with final approval of the report submission.

Cherry-Ann Waldron (Research Associate) was the trial manager, contributing to the trial design and implementation, and writing the methods section of the report. She also co-ordinated the compilation, formatting, proofreading and final approval of the report.

Rebecca Cannings-John (Research Fellow) was a co-investigator. She contributed to the trial design, conducted the statistical data analysis, cowrote the methods and results sections and gave final approval of the report.

Emma Thomas-Jones (Research Fellow) was a co-investigator and the lead for trial management. She contributed to the overall trial design and implementation, proofreading and final approval of the report.

Thomas Winfield (Health Economist) conducted the health economic analysis, co-writing the health economic evaluation section and gave final approval of the report.

Victoria Shepherd (Research Associate) was the research nurse for the trial, recruiting participants and providing research nurse guidance. She contributed to the trial management, design, implementation and final approval of the report.

Debbie Harris (Research Assistant) was the data manager, contributing to the design of the CRFs and to the data management of the trial. She also contributed to the methods and results sections and final approval of the report.

Kerenza Hood (Professor in Statistics and Director of Centre for Trials Research) was a co-investigator and contributed to the overall trial design and implementation, supervised the statistical analysis and gave final approval of the report.

Deborah Fitzsimmons (Professor of Health Outcomes Research and Academic Director, Swansea Centre for Health Economics) was a co-investigator and contributed to the overall trial design and implementation, supervised the health economic analyses, co-wrote the health economic evaluation section and gave final approval of the report.

Amanda Roberts (Paediatric Audiologist) was a co-investigator contributing to the overall trial design and implementation. She provided expert advice on paediatric audiology and final approval of the report.

Colin VE Powell (Paediatrician) was a co-investigator contributing to the overall trial design and implementation. He provided expert paediatric advice and final approval of the report.

Micaela Gal was a co-investigator and contributed to the overall trial design and implementation and final approval of the report.

Sarah Jones (Patient and Public Representative) contributed to the study design, study implementation in terms of PPI and final approval of the report.

Christopher C Butler (Professor of Primary Care) was the co-chief investigator of the trial. He contributed to the overall trial design and implementation, interpretation of the findings and final approval of the report.

Other members of the trial team

We would like to thank and acknowledge the input from Professor Ceri Philips (Health Economist), Dr Mathew Smith (Pharmacist) and Mr Alun Tomkinson (Consultant ENT Surgeon), who contributed to the trial design, and from Professor Mark Haggard and Ms Helen Spencer in respect of help with the OM8-30 mapping.

We would also like to thank Katy Addison (Trial Manager), Judith Evans (Research Administrator), Vincent Poile (Database Developer), Dr Jane Davies (Research Nurse), Hayley Prout (Research Nurse), Vasileios Gkiousias (intercalated BSc student) and Ellen Smith (MSc student) for their contribution to the trial implementation and management.

Publications

Waldron CA, Thomas-Jones E, Harris D, Shepherd V, Cannings-John R, Hood K, et al. Recruitment to oral steroids for the resolution of otitis media with effusion in children (OSTRICH) study: challenges of a randomised controlled trial in secondary care sites across Wales and England. Trials 2015;16(Suppl. 2):P119.

Waldron CA, Thomas-Jones E, Cannings-John R, Hood K, Powell C, Roberts A, et al. Study protocol: oral steroids for the resolution of otitis media with effusion (OME) in children (OSTRICH): a randomised double-blind placebo controlled clinical trial. Trials 2015;17:115.

Gkiousias V, Butler CC, Shepherd V, Kilgour J, Waldron CA, Thomas-Jones E, Francis N. Parental perceptions and understanding of information provision, management options and factors influencing the decision-making process in the treatment of children with glue ear. Int J Paediatr Otorhinolaryngol 2016;89:6–12.

Francis NA, Cannings-John R, Waldron CA, Thomas-Jones E, Winfield T, Shepherd V, et al. Oral steroids for resolution of otitis media with effusion in children (OSTRICH): a double-blinded, placebo-controlled randomised trial. Lancet 2018;392:557–568.

Data-sharing statement

All data requests should be submitted to the corresponding author for consideration. Access to anonymised data may be granted following review.

Patient data

This work uses data provided by patients and collected by the NHS as part of their care and support. Using patient data is vital to improve health and care for everyone. There is huge potential to make better use of information from people’s patient records, to understand more about disease, develop new treatments, monitor safety, and plan NHS services. Patient data should be kept safe and secure, to protect everyone’s privacy, and it’s important that there are safeguards to make sure that it is stored and used responsibly. Everyone should be able to find out about how patient data are used. #datasaveslives You can find out more about the background to this citation here: https://understandingpatientdata.org.uk/data-citation.

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 and Social Care. 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 and Social Care.

Aim of analysis

To estimate the cost-effectiveness of oral steroids and current standard practice versus standard practice in the treatment of persistent bilateral OME in children.

Population

Eligible children aged 2–8 years with symptoms of hearing loss for at least 3 months, attributable to OME, referred to NHS secondary care (ENT outpatient or paediatric AVM clinics). Entry criteria for the model would reflect the OSTRICH trial inclusion criteria.

Subgroups

No subgroups are considered in the health economic analysis.

Interventions

Short course of oral steroids as per the OSTRICH protocol.

Comparator

No treatment or other technologies.

Outcomes

Successful hearing resolution, treatment-related morbidity and HRQoL.

Time horizon

The time horizon was 24 months. Owing to the lack of literature available on the longer-term outcomes associated with OME and its treatment, this time horizon was agreed as the most plausible that could be considered within the model.

Analysis plan

A decision tree model will be constructed in Microsoft Excel^® (Microsoft Corporation, Redmond, WA, USA) using visual basic application. A decision tree was chosen as the most appropriate model structure to reflect the clinical pathway, based on discussion with the OSTRICH trial team and following a review of similar model-based analyses within a comparable population (i.e. Williamson et al. 52 and Mohiuddin et al. 59), with the latter model being used as the basis for adaption to the OSTRICH trial population. A schematic of the proposed model is set out in Figure 18.

FIGURE 18.

Schematic of the OSTRICH trial model.

Pathway assumptions

An eligible child will be assigned to receive either the steroid or the placebo for 1 week on model entry. After 4 weeks, the child will be reassessed to quantify the change in hearing loss and to determine whether or not hearing resolution was achieved. If hearing is not resolved, the next possible options are surgery, hearing aids or watchful waiting (as hearing loss in children with OME can spontaneously resolve within 2 years). Surgery carries the risk of complications, which can have a detrimental effect to patient QoL and cause costs to the NHS and society. Repeat surgery may be required if hearing resolution is not achieved or grommets fall out too early. Any treatment can be stopped at any time.

Although the NICE guidelines42 recommend a third surgery option, this cannot be modelled using the available trial data, as patients retrospectively reported data at three individual time points (5 weeks, 6 months and 12 months), and follow-up is not long enough to record third surgeries. It might be possible to obtain information regarding waiting list entries for third surgery at the third data capture point. In addition, having already had two surgical interventions, it can be assumed that this surgery would take place in the future, but there is no way of gauging the outcome of this procedure. In addition, the study cannot predict what will happen while the patients are waiting for this third surgery, whether or not they will experience some hearing resolution while being monitored or whether or not their hearing will deteriorate. Therefore, the study will not speculate about the outcomes at this stage, but it will include surgery within an upper-bound value of final costs for a sensitivity analysis.

Data inputs and sources

In the first instance, the transition probabilities, appropriate costs and outcomes (including HRQoL and OME resolution) will be obtained from the OSTRICH trial data. Costs and outcomes will be accumulated over the model horizon of 24 months. The cost-effectiveness of the steroid treatment will be calculated after 5 weeks, 6 months, 12 months and 24 months.

Discounting will be applied in the base-case analysis at 3.5%, with sensitivity analysis using a discount of 1.5% per annum for costs and outcomes, commensurate with the NICE reference case51 for cost-effectiveness at 24 months.

Analysis

In accordance with the within-trial OSTRICH economic analysis, two sets of base-case analyses will be undertaken:

1. incremental cost per successful hearing resolution achieved at 24 months

2. incremental cost per QALY gained at 24 months.

Sensitivity analysis

One-way sensitivity analyses will be undertaken to assess the effect of changing key parameters (e.g. utility values from the OSTRICH trial compared with literature-based inputs) on the base-case results. A probabilistic sensitivity analysis will be conducted to characterise the joint uncertainty in parameter estimates. CEACs will be generated to depict the probability of the intervention being cost-effective at different WTP thresholds.

Literature sources

A rapid review of the evidence was undertaken to gather suitable information to inform the data inputs required for the model pathways and the sensitivity analyses (e.g. by considering different parameter variations reported in the literature compared with the OSTRICH trial results).

Summary of rapid review

The objectives of the rapid review were to:

1. review the evidence of the cost-effectiveness of oral steroids in the management of bilateral OME

2. identify potential sources of information to inform a subsequent model-based analysis.

A population, intervention, comparison, outcome (PICO) was constructed based on the OSTRICH trial population and intervention. The study did not restrict for comparators in order to consider the full extent of the evidence (e.g. if oral steroids had been compared with another technology, such as intranasal steroids). Both primary and secondary care settings were considered in the study. For outcomes, we considered all full economic evaluations (cost-effectiveness, cost–utility, cost–benefit, cost-minimisation) and a cost–consequences analysis. Partial economic evaluations (e.g. in which only costs were presented) were excluded from objective 1, but considered for informing objective 2, if relevant.

A search strategy was developed (available from the authors). Searches were undertaken using the following electronic databases:

1. PubMed [via the National Centre for Biotechnology Information (NCBI)]

2. Web of Science (via Clarivate™ Analytics)

3. Scopus (via Elsevier)

4. The Cochrane Library (via Wiley Online Library), which includes the Cochrane Database of Systematic Reviews, the Database of Abstracts of Reviews of Effects, the Cochrane Central Register of Controlled Trials and the Cochrane Methodology Register

5. Cumulative Index to Nursing and Allied Health Literature [(CINAHL) via EBSCOhost].

In addition, reference lists were scanned and spot searches were undertaken to locate suitable evidence. No date or language limits were applied (Figure 19).

FIGURE 19.

Flow chart of rapid review search results.

For objective 1, no papers were identified.

For objective 2, we identified three full economic evaluations that could provide suitable inputs for the OSTRICH model (i.e. Williamson et al. ,52 Mohiuddin et al. 59 and Mohiuddin et al. 60), alongside the NICE clinical guidance (CG60)42 for the surgical management of OME in children aged < 12 years. Another study (i.e. Petrou et al. 61) reported the economic evaluation within the Williamson et al. 52 report. The study will use the Williamson et al. 52 report as the main source, because of the completeness of reporting.

One Cochrane review (i.e. Venekamp et al. 62) was identified as a key source for potential clinical inputs from the literature. No suitable papers were identified to provide longer-term data inputs on HRQoL/utilities.

All parameters will be extracted from the relevant papers. None of the studies selected for objective 2 met the full PICO criteria for the OSTRICH trial (i.e. different age population, setting or intervention). Therefore, a decision was made that estimation of additional inputs (such utilities beyond 12 months) would have to be obtained and validated from additional spot searches and clinical opinion. When applicable, sensitivity analysis would be used to take into account any parameter variation that could have an impact on the health economic results. A comprehensive table would be included in a final report of the model-based analysis, which would document all parameters used in the model and their source.

Primary outcome by age group

Primary outcome by atopy

Primary outcome by antibiotics received for ear problems in the last month

Primary outcome by number of previous episodes of otitis media with effusion

Primary outcome by duration of problems attributable to this episode of otitis media with effusion

Primary outcome by previous tonsillectomy or adenoidectomy

Primary outcome by household smoke present (> 5 hours a week)

Primary outcome by season of recruitment

Primary outcome by deprivation quintile