Anaesthetic analgesic ear drops to reduce antibiotic consumption in children with acute otitis media: the CEDAR RCT

Article history

The research reported in this issue of the journal was funded by the HTA programme as project number 13/88/13. The contractual start date was in January 2015. The draft report began editorial review in January 2018 and was accepted for publication in June 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

Alastair D Hay and William Hollingworth are members of the Health Technology Assessment Clinical Trials Board. Chris Metcalfe is a member of a clinical trials unit in receipt of National Institute for Health Research (NIHR) support funding. Desmond Nunez is an author of a related Cochrane review protocol. Paul Little is the Director of the NIHR Programme Grants for Applied Research programme and a member of the NIHR Journals Library Board.

Copyright statement

© Queen’s Printer and Controller of HMSO 2019. This work was produced by Hay 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.

2019 Queen’s Printer and Controller of HMSO

Structure of this report

In this chapter, the rationale for reducing antibiotic prescriptions by treating ear pain in acute otitis media (AOM) with anaesthetic–analgesic is presented. Chapter 2 describes the study population and gives details of the interventions being compared and the study design. Chapter 3 presents the results of the study, followed by discussion and conclusions in Chapter 4.

Background

There is broad agreement that a high proportion of prescriptions for antibiotics are unjustified1,2 and likely to be harmful given the clear relationship between primary care prescribed antibiotics and bacterial resistance. 3 Furthermore, resistant bacteria are transmitted to social contacts, which is a problem with young children, who are unaware of hygiene conventions and who have high contact rates with other children, parents and grandparents. 4,5

In September 2013, the Department of Health and Social Care published the UK Five Year Antimicrobial Resistance Strategy 2013 to 2018,6 which calls for change in the understanding and response to antimicrobial resistance by the public, the NHS and the government in the UK. Its overarching goal is to slow the spread of antimicrobial resistance through three strategies: (1) improving the knowledge and understanding of antimicrobial resistance, (2) conserving and stewarding the effects of existing antibiotics and (3) stimulating the development of new antibiotics. Conserving and stewarding the effects of antibiotics can be achieved in five ways: (1) reducing the overall quantity of antibiotics prescribed and consumed; (2) when antibiotics are needed, promoting the use of narrow-spectrum agents; (3) when antibiotics are in demand but are ineffective, providing alternatives; (4) reducing the transmission of antibiotic-resistant bacteria; and (5) vaccinating against antibiotic-resistant bacteria.

Acute otitis media is a common, painful condition of childhood that has an impact on the family because of disrupted sleep and time off work and school. Primary care consultation and prescription of an antibiotic have been the mainstay of management; one UK study found that between 80% and 84% of children presenting to primary care with AOM were prescribed an antibiotic during the years 1995–2011. 1,2 This is despite the available evidence of benefit being restricted to children < 2 years old with bilateral AOM and those with otorrhoea,7 leading the National Institute for Health and Care Excellence (NICE) to conclude that these are the only children warranting a same-day full course of antibiotic treatment. 8 For other children, the use of a ‘wait-and-see’ strategy, with or without a delayed prescription for an antibiotic, has been shown to be safe in terms of treatment failure and in the frequency of complications of AOM including mastoiditis. 9,10

Evidence of the effectiveness for alternatives to antibiotics is urgently needed to reduce the largely inappropriate reliance on antibiotics and to relieve the most common and distressing symptoms of AOM. The results of the Children’s Ear Pain Study (CEDAR) are presented in this report, which investigated if anaesthetic ear drops are effective in reducing the consumption of antibiotics by children with AOM.

Literature review

Three previous trials have assessed the effectiveness of topical analgesia against placebo in relieving pain due to AOM. 11–13 All were included in a Cochrane review (updated in 2011) that concluded that ‘the evidence from [these] RCTs is insufficient to know whether ear drops are effective’. 14 In any case, those studies did not assess the impact of topical analgesia on antibiotic prescribing when used as an approach to the management of an episode of AOM, and so it is unknown whether or not parents will see topical analgesia as an effective and preferable alternative to antibiotics and, therefore, reduce unnecessary antibiotic use.

Benzocaine–phenazone otic solution (Auralgan^®, currently manufactured by Pfizer Consumer Healthcare) drops have been used in Australia for > 40 years with little or no evidence of harm. Prior to commencing this study, as part of the trial risk assessment and because of the theoretical potential for induced methaemoglobinaemia, we undertook a search of the Australian Therapeutic Goods Administration Database of Adverse Event Notifications (DAEN) (www.tga.gov.au/database-adverse-event-notifications-daen). On 2 December 2015, the DAEN was searched for adverse event reports associated with Auralgan between 1 January 1971 and 19 August 2015. A total of 11 cases were reported, all of which appear to be minor adverse effects (some patients had more than one adverse effect): ear pain (two cases), discomfort (two cases), pruritus (one case), hyperacusis (one case), hearing impaired (one case), eye pain (one case), tinnitus (one case), burning sensation (two cases), ear canal erythema (one case), accidental exposure and overdose by child (one case), deafness (two cases), depressed mood (one case). This suggests a low risk of adverse reactions to Auralgan ear drops. There is a further potential for adverse effects arising from reduced antibiotic prescribing to children with AOM. However, previous observational data suggest that the frequency of such adverse effects is likely to be extremely low. 1 Moreover, in a survey of Australian pharmacies undertaken as part of preparation for the trial, it was established that these drops are widely available over the counter and sometimes used to avoid the need for a GP consultation (Harriet Downing, University of Bristol, 2015; unpublished survey data).

Study aim

The aim of the CEDAR trial is to investigate the clinical effectiveness and cost-effectiveness of benzocaine plus phenazone (active) ear drops compared with usual care (no drops) for reducing antibiotic consumption in children aged between 12 months and 10 years presenting to primary care with AOM.

Specific research questions

Following a presentation to primary care with AOM and compared with usual care:

1. Do active drops lead to a lower proportion of children consuming antibiotics in the first 8 days?

2. Do active drops lead to reduced oral analgesic consumption in the first 8 days?

3. Do active drops provide superior pain relief in the first 24–36 hours?

4. Do active drops provide superior pain relief in the first 24 hours?

5. Do active drops alter the number of days before starting antibiotics in the first 7 days?

6. Do active drops reduce the overall symptom burden, including episodes of crying or distress, disturbed sleep, interference with normal activity, appetite and fever, in the first 7 days?

7. Do active drops alter overall illness duration?

8. What are the net incremental costs to the NHS and society in the short (7 days) and medium (3 months) term?

9. Are the net incremental costs of active ear drops justified by improved pain relief, symptom burden, antibiotic use or quality of life?

Further objectives were to use qualitative methods to investigate parents’ and clinicians’ views and experiences of AOM in children in the CEDAR trial and to investigate the representativeness of the CEDAR trial sample by describing the presentation, management and outcome of children with AOM in primary care.

Study design

The CEDAR trial was a multicentre randomised controlled trial (RCT) and the participating children were randomly allocated between three study groups: active treatment (benzocaine-based ear drops), placebo treatment (drops without the ‘active’ ingredient) and no drops (current standard care). Parents who did not want their children to take part in the trial but were willing to complete the study diary were offered enrolment in an observational cohort study (Figure 1).

FIGURE 1.

The CEDAR trial three-group randomised study with observational cohort study. OCS, observational cohort study.

Recruitment sites and site training

General practitioners at primary care research network practices in Bristol, Cardiff and Southampton were invited to take part in the study.

Training was delivered to each clinician who was involved in the trial. The training provided an outline of all trial recruitment and baseline data collection procedures, including how to train parents in administering the ear drops (for parents of children allocated to one of the treatment groups) and completing the Symptom and Recovery Questionnaire. A clinician training log was maintained at sites and within the Trial Master File reflecting the staff who had been trained. Clinicians at all participating primary care sites were also offered ongoing support, recruitment advice and refresher training on request by the study team.

Participants and recruitment

Interventions

The investigational medicinal product (IMP) for this trial was a benzocaine and phenazone otic solution. This is an oil-based, combined local anaesthetic (benzocaine) and analgesic (phenazone) ear drop. One millilitre contains 14 mg (1.4%) of benzocaine and 54 mg (5.4%) of phenazone suspended in a glycerine-based liquid along with a preservative (hydroxyquinolone sulphate). Despite an absence of published evidence of effectiveness, it is available as a pharmacy medicine in Australia and New Zealand, and has been marketed since 1947 under Auralgan and other brand names. For this trial we intended to test Auralgan sold in 15-ml bottles.

The placebo was glycerine, with packaging as close in appearance as possible to that used for the active drops.

Outcomes

Secondary outcomes

Ear pain score on day 2

The 8-day Symptom and Recovery Questionnaire also asked the parent to rate their child’s ear pain on a scale of 0–10 (Figure 2). Although active drops should have an almost immediate effect, the evening of day 2 was chosen as the time at which the drops had enough time to ease ear pain if they were going to. Active and placebo drops were compared to see whether or not the active ingredient provided relief over and above the soothing effects of the oily liquid in which they are contained. Placebo drops were also compared with usual care to test whether or not the oily liquid itself was more effective than nothing at all. The key secondary outcome tested the difference in ear pain scores between active and placebo drops, whereas an additional secondary outcome tested difference in ear pain between placebo and usual-care groups.

FIGURE 2.

Parent-completed assessment of pain.

Ear pain score on day 1

Another objective was to see whether or not the drops had an immediate effect on pain; therefore, parents were asked to score their child’s ear pain 1 hour after giving the first dose of ear drops. Active and placebo drops were compared to see whether or not the active ingredient provided relief over and above the soothing effects of the oily liquid in which they are contained. Pain scores were on a scale of 0–10. This was then compared between active and placebo groups.

Analgesic consumption

On each day of the symptom diary, parents were asked whether or not they had given their child any analgesics (e.g. ibuprofen or paracetamol). The number of doses was also recorded. A binary variable was created to identify children who had received analgesics on any of the first 8 days versus children who had not received any oral analgesics during any of the first 8 days; this measure was not derived for children with any missing oral analgesic data during the first 8 days. This measure was then compared between active and placebo groups.

Number of days before taking antibiotics

The number of days before taking antibiotics was obtained from the child’s symptom questionnaire. This was then compared between active and placebo groups, and between active and usual-care groups.

Mean overall symptom severity on days 1–4

Parents were asked to complete questions concerning other symptoms, such as episodes of crying, disturbed sleep and fever in their symptom diaries. Symptoms were rated on a scale of 0 (no problem) to 6 (extremely bad). All symptom scores were combined for each patient per day before calculating the area under the curve for days 1 to 8. This was then compared between active and placebo groups, and between active and usual-care groups.

Overall illness duration

The number of days until the parent rated score was 0 for 2 consecutive days was calculated for all children. This was then compared between active and placebo groups, and between active and usual-care groups.

Adverse events

Details of serious adverse events were collected, and the trial team notified immediately, using adverse event forms. Parents were also asked on the last day of their questionnaire whether their child had experienced any new or worsening symptoms during the trial.

Patient satisfaction

Parents were asked about their satisfaction with, and opinion of, treatment allocation and future intention to use drops (with/without prior GP consultation if drops were to become available over the counter) 8 days post randomisation.

NHS and societal costs

The net incremental costs to the NHS and society of using active ear drops compared with no drops (usual care) in the short (8 days) and medium term (3 months). NHS resource use (e.g. antibiotic or analgesic use, GP visits) and societal costs (school/nursery absences, parent lost productivity and other family expenses) were reported by parents in the 8-day questionnaire. Medium-term NHS resource use was collated from a review of the child’s GP records.

Child’s quality of life

Preference-based quality of child life was measured (baseline, 24–36 hours, 8 days and 3 months post randomisation) using the CHU-9D. 16 The CHU-9D is not designed for, or validated for, preschool children; therefore, this measure was used only in children aged ≥ 5 years. The OMQ-1417 was given at baseline and 3 months post randomisation for children aged ≥ 2 years (clinical report form and postal/online questionnaire at 3 months).

Sample size

Previous studies have suggested that 80–90% of children presenting to UK GPs with AOM receive a prescription for an antibiotic. 18,19 It has also been argued that a 20% change in consumption could have important effects on antimicrobial resistance. 3 The number of children needed in each group (active ear drops and usual care) to demonstrate a true 20% fall in antibiotic consumption, from 80–90% in the control group to 60–70% in the active ear drop group, with 90% power and two-sided significance level of 0.05, ranged from 92 to 119. Using the more conservative estimate of 119 children, and making provision for 20% attrition, this gave a target sample size of 149 children per group.

Our key secondary outcome was pain on day 2, measured using a validated 0–10 numerical rating scale. Based on the findings of a previous study using this measure,20 and the opinions of our patient and public involvement group, it was felt that a reduction in pain of 1 point was a clinically important difference. With 90% power and α = 0.05, we can detect a true mean difference of 1 point [standard deviation (SD) 2.5 points] in pain score between the active and placebo ear drop groups, detected with 90% of power at the two-sided significance level of 0.05, with 133 children per group. Accommodating 20% attrition, and assuming equal numbers in the three groups, we needed 167 children per group (more than for the primary outcome) and 501 children in total.

Random allocation

Each pack contained two bottles of active medicine, two bottles of placebo medicine or no bottles but a non-medicinal item of similar weight. Randomisation was stratified by centre in blocks of 30 packs, each block having the packs arranged in a random and consecutively numbered sequence. The Bristol Randomised Trials Collaboration provided the pharmaceutical supplier with a set of patient IDs, which were allocated to a pack according to the sequence for the current block for a centre. The supplier provided the pharmacy with medicine packs that had each been prelabelled with the patient ID and medicine pack ID numbers.

Patients were enrolled by their GP or research nurse who, at the stage of enrolment, was unaware of the contents of the next treatment pack in the sequence (allocation concealment). When the informed consent process was completed and signed, the trial pack was opened and allocation to drops (active or placebo) or not was confirmed.

Blinding

Although parents were aware if their child had been allocated to the usual-care group, if they received drops they were not aware if they had received active or placebo drops. An unblinded comparison of parents provided with ear drops and those allocated to usual care was the appropriate approach to establish if the provision of ear drops would reduce antibiotic consumption in children with ear pain due to AOM. All parents, researchers and clinicians were blinded to the treatment allocation between active and inactive drops as far as possible. The statistician had unblinded access to the data in February 2017, to report to the Data Monitoring Committee, when approximately 50 patients had been recruited.

Statistical methods

The main statistical analyses were prespecified in a statistical analysis plan. The final version of the statistical analysis plan was signed off on 19 September 2017. A confidential interim analysis was completed in February 2017 for the Data Monitoring Committee. Final data analysis started in October 2017 and finished in January 2018. Stata^® 14.2 (StataCorp LP, College Station, TX, USA) was used for all statistical analyses.

Health economic methods

The prespecified economic analysis consisted of a cost-effectiveness analysis and cost–consequence study. As antimicrobial resistance is such an important concern,21 our primary economic analysis was a cost-effectiveness analysis with antibiotic use as a proxy for this outcome. The economic case for or against the active drops is most likely to hinge on externalities, that is antibiotic use and resistance, in the population, rather than purely on within trial estimates of costs and quality of life. The cost–consequence study (including quality of life) was a secondary economic analysis.

The primary comparison was the NHS and societal cost per antibiotic consumption avoided during the acute episode between those children receiving active ear drops and those receiving usual care. The active treatment (Auralgan) is currently not in the British National Formulary (BNF)22 or available in UK pharmacies and, therefore, there is no national unit cost available. Auralgan is available to purchase over the counter in Australian pharmacies. The median retail cost for a 15-ml bottle in December 2017 at 10 online Australian pharmacies was AU$10.99, equivalent to approximately £5.20 when converted using 2016 Organisation for Economic Co-operation and Development purchasing power parity statistics23 or £10.40 retail cost for the two 15-ml bottles given to each participant in the active drops group of the trial. We assumed that, if used in routine UK clinical practice, Auralgan would be prescribed by a GP and, therefore, the cost would be borne by the NHS rather than by the family. For many medicines, the NHS drug tariff paid to pharmacies is much lower than the retail cost. For example, for 120 mg/5 ml of oral suspension paracetamol the drug tariff cost is approximately one-third of the retail cost. Therefore, in our primary analysis we assumed that the cost to the NHS of two 15-ml bottles of Auralgan would be £3.54 and we conducted a sensitivity analysis assuming that the NHS would pay the retail cost of £10.40.

The incremental NHS treatment costs during the first 8 days after randomisation were ascertained from the symptom and recovery questionnaire and valued using national unit costs22,24 using 2016/17 values, when available. Parents recorded whether or not their child received an antibiotic on each day that symptoms persisted. We assumed that the antibiotic prescribed was 250 mg of amoxicillin, three times a day (age 1–4 years) or 500 mg three times a day (age ≥ 5 years) for 7 days. Using the NHS drug tariff price,25 this equates to a cost of £1.17 (age 1–4 years) or £2.34 (≥ age 5 years) per antibiotic prescription.

Parents also recorded daily use of ibuprofen or paracetamol during the first 8 days. When use was reported, we assumed this was prescribed (i.e. it was a cost borne by the NHS) and that age-appropriate doses were given. For example, based on national drug tariff prices,25 the cost of a 2-year-old child taking 100 mg/5 ml of ibuprofen three times a day for a total of 20 doses is £1.28.

For Auralgan and antibiotics, we assumed that the whole cost would be incurred no matter how frequently parents administered the medicine (i.e. the remaining medication would be disposed of after the acute episode rather than reused during later episodes). For paracetamol and ibuprofen, we assumed that the cost of the medication would be proportional to the number of times it was administered (i.e. the remaining medication would be stored and reused during later episodes).

On the eighth day after randomisation, parents were asked to report any further consultations with their GP, the NHS 111 telephone service or hospital ambulatory care relating to their child’s ear symptoms. Parent productivity costs were measured on the 8-day questionnaire and valued using average earnings. Bootstrapping was used to calculate confidence intervals (CIs) around the point estimate of the incremental cost-effectiveness ratio and cost-effectiveness acceptability curves.

In the secondary economic analysis, we prespecified a cost–consequence study tabulating 3-month treatment costs, including data from the 3-month GP note review, alongside other important outcomes including 24- to 36-hour ear pain relief, overall symptom burden, acute illness duration and CHU-9D utility scores. At the request of the funder, the GP note review was not conducted; therefore, we did not estimate 3-month treatment costs and instead present only CHU-9D utility scores at 3 months.

Qualitative methods

To examine the views and experiences of AOM and its treatment, qualitative interviews were conducted with parents and clinicians involved in the care of trial participants. The interview topic guide included the perceived effectiveness and acceptability of treatment and exploration of any barriers to treatment uptake outside the trial.

Study progress

Study recruitment had been planned to take place between June 2015 and June 2017. Owing to a delay in the supply of a suitable placebo, recruitment began in October 2016 across 27 GP practices with a two-group internal pilot phase. Children were randomly allocated to the active treatment group or the usual-care group. The protocol change necessary for the two-group trial (protocol version 1.4) received approval from the ethics committee in August 2016 (NHS HRA reference 15/SC/0376). When the placebo became available in March 2017, the three-group study began recruiting across 35 GP practices. Although the two- and three-group designs ran concurrently for 3 months, the two studies did not run at the same time within any given practice. Owing to the late start of the three-group trial, the observational cohort for parents who did not want their children to take part in that trial but were willing to complete the study diary (see Figure 1) was not initiated. Recruitment to the trial closed as planned in June 2017, with a recruitment period of 9 months having been achieved rather than the planned 24 months.

In April 2017, a Data Monitoring Committee meeting was held that highlighted that antibiotic prescribing rates were much lower than assumed in all study groups combined and, consequently, that the sample size target would not provide 90% statistical power. This may in part have been due to children requiring immediate antibiotics, and so ineligible for the trial, accounting for a bigger than assumed proportion of those prescribed antibiotics in the studies used to inform the sample size calculation. Following this finding, the Trial Management Group, along with the Trial Steering Committee, jointly agreed that the data collected should serve as internal pilot data (i.e. the data would be included in the primary analysis) and an application made to extend the study to allow further recruitment with a revised exclusion criterion relating to antibiotic prescribing.

In Autumn 2017, this application to extend the study was declined by the funders, and the study was required to close by the 31 December 2017, several months earlier than originally planned. Owing to this, no 3-month reviews of primary care records were conducted and so planned analyses relying on those reviews were not possible. The following presentation of results has adhered as closely as possible to the prespecified statistical analysis plan, including an amendment written as part of the study close-down plan, which adapts the original plan to data available from the internal pilot and full three-group stages.

Recruitment and completion of follow-up assessments

Seventy-four children were recruited during the two-group internal pilot, which ran from October 2016 to the end of April 2017, in 27 GP practices situated in all three centres. Thirty-two children were recruited to the three-group study, which ran from March 2017 to June 2017, in 35 GP practices in Bristol and Southampton. All 106 patients were recruited by 17 GPs, 18 nurses and 3 other health-care professionals, with each recruiting between 1 and 19 children. Our interim recruitment target at the end of June was 150 participants. After initially recruiting to target, recruitment fell short of the June target in part because of the time taken to switch practices from the two-group to the three-group trial and in part because of the limited supply of treatment packs (see Appendix 1 for the chart used to monitor recruitment).

Figure 3 shows the stages of the trial and the different levels of dropout that led to the final analysis sample. Overall, five patients randomised into the trial were found to be ineligible post randomisation. For three of these patients (two active two-group patients and one usual-care two-group patient), this was discovered shortly after randomisation and the patients were advised to return the drops and not complete the diary. For the other two of these patients, this was noticed only after follow-up had been completed. Given that our analysis was on an ITT basis, these patients were included in all analyses but excluded in a sensitivity analysis. 32 Further detail of the availability of data on the two key outcomes of antibiotic use over eight days and pain reported is given in Table 1.

FIGURE 3.

The CEDAR trial CONSORT (Consolidated Standards of Reporting Trials) flow chart. a, Been using dexamethasone/neomycin sulphate/acetic acid spray (Otomize^®; Teva UK Ltd.), temporary resident so clinician decided not to recruit, previously taken part in the CEDAR trial, grommets, history of bloody discharge, dental issue causing ear pain, patient seen at branch surgery and unable to come to main surgery for appointment; b, child not happy about taking part, parent did not want to risk getting placebo, parent did not want child to participate; c, child too unwell (clinician’s view), risk of complications due to pre-existing comorbidities, right ear otitis externa, child too old, no reason stated; d, child too unwell (clinician’s view), child not happy about taking part, parent did not have time for recruitment, no reason listed; e, recruitment to the two-group pilot ran from October 2016 to the end of April 2017 (in 27 sites) while recruitment to the three-group pilot ran from March 2017 to June 2017 (in 35 sites); 12 sites recruiting to the three-group pilot were previously involved in the two-group pilot. Although the two-group and three-group trials ran concurrently between March 2017 and June 2017, the two trials did not run at the same time within a given site; f, one patient who completed follow-up was later found to be ineligible (bilateral AOM and < 2 years of age); andg, one patient who completed follow-up was later found to be ineligible (otorrhoea in right ear).

Baseline data

A total of 106 patients were randomised to the CEDAR two- and three-group studies and baseline characteristics are shown in Tables 2–5, respectively. The team prespecified in the analysis plan that any baseline characteristics that differed by at least 10%/0.5 SDs would be adjusted for in a sensitivity analysis. However, as the study recruited only one-fifth of the target sample size, the criterion was increased to 20%/0.75 SDs to counteract the increased chance of finding an imbalance.

In the two-group study, the only measure to have a difference exceeding the prespecified criterion of 20% or 0.75 SDs was receipt of a delayed prescription for an antibiotic (see Table 3). Those prescriptions should have been issued, and recorded, prior to randomisation and so it should not be the case that allocation to study group influenced antibiotic prescription, although we are unable to confirm this for sure.

Prior to analysis, the criterion for imbalance for the three-group study was increased to 30%/1 SD to counteract the increased chance of finding an imbalance given the small sample size. Variables that exceeded this criterion were sex, accompanying adult’s age and employment status, living in an area of deprivation, breastfeeding status at 3 months (see Table 4), episodes of distress/crying and disturbed sleep (see Table 5).

Outcomes and estimation

Secondary outcome: ear pain on day 1

There was no evidence to suggest that pain was reduced in the active drops group compared with the placebo drops group on day 1 (Table 10).

TABLE 10 - Secondary analysis: ear pain on day 1 (approximately 1 hour after administering the drops), by group

Adjusted for the parent-reported pain score at consultation.

Group	Treatment arm, mean (SD); n		Mean difference (95% CI)	p-value	Mean difference (95% CI)a	p-valuea
	Active drops	Usual care
Ear pain	2.70 (1.16); 10	3.42 (1.62); 7	–0.73 (–2.16 to 0.70)	0.295	–0.74 (–2.32 to 0.85)	0.338

Combining the two- and three-group study data, Figure 4 shows the average daily pain scores for each of the three groups. Over the 8 days, lower levels of pain were reported by the active drop and placebo drop groups than by the no drops group.

FIGURE 4.

Mean pain scores by day and group, combining the two- and three-group children and including only those with full 8 days of data on ear pain.

Secondary outcome: oral analgesic consumption

Oral analgesic consumption was high, with 88% of children taking paracetamol or ibuprofen during the first 8 days after consultation. There was no evidence to suggest a difference in oral analgesic consumption between those taking active and placebo ear drops (Table 11).

TABLE 11 - Secondary analysis: analgesic consumption, by group

OR, odds ratio.

Adjusted for the parent-reported pain score at consultation.

Group	Treatment arm, n/N (%)		OR (95% CI)	p-value	OR (95% CI)a	p-valuea
	Active drops	Placebo drops
Analgesic consumption	8/9 (89)	6/7 (86)	1.33 (0.07 to 25.91)	0.849	1.21 (0.04 to 34.00)	0.911
Ibuprofen	4/9 (44)	3/7 (43)
Paracetamol	6/9 (67)	6/7 (86)

During the recruitment process, parents were asked if their child had received any painkilling medicine prior to the recruitment consultation and 56 out of 79 (71%) participants said that they had. All 56 of these children subsequently received analgesics during the study. Out of the 23 who did not give their child painkilling medicine before taking part in the study, five continued to not use analgesics during the trial (active drops, n = 2; placebo, n = 1; and usual care, n = 2).

Considering the number of doses of oral analgesic that the children were receiving over the 8 days, there were no clear differences in the two-group study (Figure 5). In the three-group study, children allocated to the active drops appeared to take fewer doses than their placebo and usual-care counterparts (see Figure 5). However, these observed differences were consistent with chance (Table 12).

FIGURE 5.

Analgesic consumption up to day 8 for the two- and three-group studies. The box and whisker plots show the distributions of analgesic consumption. The box represents the IQR, the line within the box is the median, the upper and lower ‘whiskers’ are defined as the 75th percentile + 1.5 × IQR and the 25th percentile – 1.5 × IQR, respectively. The dots outside these ‘whiskers’ are the outliers observed in the data. IQR, interquartile range.

TABLE 12 - Secondary analysis: number of doses of oral analgesic, by group

IQR, interquartile range.

Calculated using the square root of the number of doses.

Group	Treatment arm, median (IQR); n		Difference in meansa (95% CI)	p-valuea
	Active	Placebo
Number of doses	2.0 (1.0–5.0); 9	6.5 (2.0–9.0); 6	–0.54 (–1.86 to 0.787)	0.397

Parents were also asked if they had given their children any other painkilling remedies. Only one additional treatment was reported for one child: xylometazoline (Otrivine, GlaxoSmithKline) nasal drops over 2 days.

Secondary outcome: overall symptom burden

For a first exploration of symptom burden, each child’s scores (high scores indicate greater severity, maximum score of 36 on each day) on each of the symptom measures were added together to give a total for each of the 8 days following recruitment. An ‘area under the curve’ approach (broadly equivalent to taking the average score for each child over the 8 days) was taken, with the resulting scores being positively skewed (Figure 6) and so were transformed (by taking the square root) to allow analysis with linear regression.

FIGURE 6.

Mean symptom burden in the two-group [(a) and (b)] and three-group studies [(c), (d) and (e)].

For both the two- and the three-group studies, the active drops appeared to lead to a faster resolution of symptoms, and so a lower symptom burden over the 8 days (see Figure 6), with modest evidence in support of these differences (Table 13).

TABLE 13 - Secondary analysis: overall symptom burden over 8 days, quantified as the area under the curve, by group

IQR, interquartile range.

Calculated using the square root of the area of the curve.

Treatment arm (study)	Symptom burden, median (IQR); n	Difference in meansa (95% CI)	p-valuea
Active drops (three-group)	15.8 (8.5–21.5); 10	Comparison
Placebo drops (three-group)	24.5 (10.5–50.5); 7	1.81 (–0.28 to 3.90)	0.085
Usual care (three-group)	28.5 (14.0–42.0); 9	1.35 (–0.13 to 2.84)	0.072
Active drops (two-group)	11.5 (5.8–33.5); 32	Comparison
Usual care (two-group)	30.3 (6.3–45.0); 28	1.14 (–0.20 to 2.49)	0.094

Secondary outcome: overall illness duration

Children in the active group appeared to be recovering at a faster rate than those in the usual-care group (Figure 7), but this difference was consistent with chance (Table 14). There was no evidence against the proportional hazards assumption.

FIGURE 7.

Kaplan–Meier curve for the time to recovery from pain in the two-group trial. Time to recovery (score = 0).

TABLE 14 - Overall duration of pain

IQR, interquartile range.

Missing 75th percentile owing to large proportion of censoring at 8 days.

Treatment arm (study)	Pain duration, median (IQR); n	Hazard ratio (95% CI)	p-value
Active drops (three-group)	3 (3–5); 10	Comparison
Placebo drops (three-group)	2 (2–4); 7	1.70 (0.61 to 4.75)	0.31
Usual care (three-group)	3 (2–6); 9	0.94 (0.38 to 2.61)	0.90
Active drops (two-group)	3 (2–5); 34	Comparison
Usual care (two-group)	4 (3–Xa); 31	0.62 (0.34 to 1.11)	0.11

Secondary outcome: parent satisfaction

When parents were asked if they were satisfied with the trial ear drops, 93% (27/29) of parents receiving active drops in the two-group trial said that they were satisfied and 7% (2/29) of parents reported that they were neither satisfied nor dissatisfied. Parents of children allocated to active drops in the three-group trial reported 90% (9/10) satisfaction, with 10% (1/10) reporting that they were neither satisfied nor dissatisfied. Only 57% (4/7) of parents with children allocated to placebo drops reported satisfaction, with 29% (2/7) reporting that they were neither satisfied nor dissatisfied and 14% (1/7) reporting that they were not satisfied.

In the three-group trial, parents of children recruited to the active or placebo drops were asked on day 8 if they thought that their child was taking the painkilling ear drops. In the active group, 60% (6/10) thought that their child was taking the painkilling ear drops, and in the placebo group 29% (2/7) thought that they were.

When asked if they would use the drops if they were to be available over the counter, all parents in the two-group trial said ‘Yes’, with 69% in the active group saying that they would take them without GP advice compared with 50% in the usual-care group. Of the 26 children recruited in the three-group trial, two parents said that they would not use the drops if they became available; both were in the blinded placebo group. When asked what they would use in the future, 79% (23/29) of those in the active group of the two-group trial said just the trial drops, compared with 40% (4/10) of those active group patients in the three-group trial; these results are given in Figure 8. When comparing the answers to this question between those who gave their children antibiotics during the study and those who did not, it was found that those who gave antibiotics were more likely to say that they would use antibiotics again (20%, 3/15) than those who had not (3%, 2/69).

FIGURE 8.

By trial group: what would you want to use for your child for a similar illness in the future?

Sensitivity analyses

Exploratory analysis: reported pain by antibiotic use

For those with pain scores for all 8 days, children who took antibiotics (n = 15) appeared to have higher pain scores than those who did not take antibiotics (n = 69) during the study (Figure 9). Although the direction of cause and effect could not be established from these data, it is possible that the use of antibiotics resulted from the higher pain.

FIGURE 9.

Mean pain scores by day and antibiotic consumption, combining the two- and three-group children and including only those with full 8-day data on ear pain.

Subgroup analyses

The evidence for subgroup effects is difficult to interpret in the context of low statistical power, which was the case for this study because of the modest sample size realised and the dimensions investigated having one subgroup much larger than the other. Consequently, the summary statistics are presented in Table 18.

Health economic results

Qualitative data results

Summary of main findings

Despite the CEDAR trial falling short of its sample size target, and demonstrating a lower than expected antibiotic consumption rate, the results provide evidence of a reduction in antibiotic consumption with the use of anaesthetic–analgesic ear drops compared with children who were not provided with ear drops.

The primary analysis (active drops vs. usual care) found modest evidence of a difference between groups when combining the results of the two- and three-group studies, with 11 out of 38 (29%) children in the usual-care group consuming antibiotics, compared with 1 out of 39 (3%) children in the anaesthetic–analgesic drops group. This represents an 87% reduction in the odds of consuming antibiotics (odds ratio 0.13, 95% CI 0.01 to 1.13). If this reflects a true effect of the ear drops, the most likely explanation for reduced antibiotic consumption is reduced pain. However, secondary analyses showed both placebo (glycerol) and active drops to be superior to no drops for pain relief at day 2. The former could be explained by the placebo effect or a soothing anti-inflammatory effect of glycerol. Secondary outcome measures provided modest evidence of reductions in oral analgesic consumption, symptom burden and illness duration in the active drops group compared with the usual care group.

Our economic analyses demonstrated that the use of active drops is unlikely to reduce medication costs in the short term, primarily because antibiotics are cheap. The more substantial economic benefits are likely to come in the longer term, via reduced antimicrobial resistance. Nevertheless, the ‘cost per antibiotic avoided’ of active drops is low and, if parental productivity costs are included in the analysis, it may even save money. Weak evidence was found that active drops improved health-related quality-of-life scores, measured by the CHU-9D at 8 days.

Interviews conducted with three participating parents revealed their views and experiences of AOM, its treatment and the CEDAR trial. The study aim was viewed positively by parents, who felt the treatment of pain to reduce the use of antibiotics was a good idea and who were happy for their child to participate. Parents’ main reason for taking their child to the doctor was that their child was upset and in pain and they wanted pain relief for their child. Parents did not express any preconceptions of whether or not the child should receive antibiotics and were happy to discuss treatment options with the doctor. Parents spoke positively about the trial drops, and all believed that they reduced pain in the child, although this was also the case for parents whose children received placebo drops. Parents stated that they would be happy to purchase the drops over the counter from a pharmacy, provided that pharmacist advice was available.

Strengths and weaknesses

The CEDAR trial was a pragmatic RCT conducted in UK primary care practices; allocation was concealed during recruitment and around 80% of parents provided full data on their child’s antibiotic consumption during the key 8-day follow-up period. However, the study was subject to a number of limitations.

The clear weaknesses of the CEDAR trial were the small sample size achieved (particularly with the planned three-group study), not being able to collect data at the planned 3-month follow-up point and not being able to commence recruitment in emergency departments as planned. Although comparisons of the raw data suggest a substantial treatment effect on antibiotic consumption, the results need to be interpreted with caution as estimates are imprecise and we are not able to confirm the probable mechanism through reduced pain. This weakness is likely to be purely one of low statistical power; although the study did close early, this is unlikely to have caused bias as early closure did not curtail the planned recruitment period and was a decision of the funder, who was unaware of the findings of the confidential interim analysis (although it had been made aware of the lower than expected antibiotic prescription rate across the three study groups combined). The lack of data at the 3-month follow-up means that the safety of the intervention in the period immediately after the infection, and the incidence of recurrent infection, remain unknown.

We observed a greater proportion of children in the usual-care group than in the drops groups who received a delayed prescription for an antibiotic. We had intended for children to receive these prescriptions prior to being allocated to a study group, although we cannot guarantee that this always occurred and, in hindsight, whether or not the child had received a delayed prescription should have been included in the information provided when logging the child into the study, in preparation for allocation to a study group. Those children provided with a delayed prescription were more likely to consume antibiotics during the 8-day follow-up; our prespecified primary analysis attempted to control for any confounding through this route. Of the small number of children in one of the drops groups who received a delayed prescription, none subsequently consumed antibiotics; this suggests that if receipt of a delayed prescription been equally low in the usual-care group, then this may have reduced antibiotic consumption in that group and hence reduced the difference in antibiotic consumption between study groups. However, taking a different view, although the intention had been to standardise antibiotic prescribing across the study groups, what actually happened in the trial would be closer to routine clinical practice, in which GPs might prescribe fewer antibiotics if they can provide anaesthetic–analgesic ear drops instead, with the lower prescription rate resulting in lower antibiotic consumption.

In fact, antibiotic consumption was already lower than anticipated in the usual-care group, potentially owing to the overcautious exclusion of children considered to need immediate antibiotics, to reduced use of delayed prescriptions by participating GPs who are perhaps already keen to reduce antibiotic consumption, or to informed consent procedures including the study rationale, which would have given parents the chance to opt out if they were keen to use antibiotics (the planned but unrealised observational cohort would have addressed this) or may have convinced some parents of the unsuitability of antibiotics for the treatment of AOM ear pain.

The larger number of participating children were recruited to the two-group study, which did not include the placebo drops group and so could not provide reliable evidence that the anaesthetic–analgesic drops were reducing antibiotic consumption through a painkilling mechanism. However, despite the modest sample size realised, the study does provide weak evidence that parents of children presenting to primary care with ear pain due to AOM find ear drops to be an acceptable treatment and are consequently less likely to use antibiotics. It seems unlikely that a different pattern of results would have been observed had children in the drops groups been more often provided with a delayed prescription for antibiotics, and indeed the study shows that parents of children in the drops group found it acceptable not to receive such a delayed prescription.

From the perspective of the economic analysis, as Auralgan is not in the BNF,22 it is currently unclear how much the NHS would pay for it if GPs were to prescribe it. Plausible assumptions have been made based on drug prices in other high-income countries and potential NHS discounts. However, if the actual procurement price were higher, the intervention would be less cost-effective than we have estimated. Owing to falling short of the recruitment target, the evidence of a reduction in health-care use up to 8 days was very weak, and with the review of medical records not going ahead we were unable to evaluate whether or not the possible reduction in health-care use might have been amplified over a 3-month period.

The strengths of the qualitative study are that it included parents who took part in the trial and it was therefore able to provide views and experiences from those directly involved. Reported themes were present across all interviews, strengthening their standing. All parents had previous experience of children with AOM and treatment and were therefore well placed to comment on whether or not the treatment was acceptable and whether or not they believed it worked. The main limitation is the small number of interviews conducted. Although more were planned, only three interviews with trial parents could be conducted within the timeframe of the three-group trial. This does provide useful insight into how parents experienced the trial and using the trial drops. However, caution needs to be exercised when taking findings forward. Views were captured only from those who agreed and continued to participate in the trial. Further interviews need to be conducted with a larger number of parents to determine whether or not findings continue to be prevalent and important. Further issues including divergent views and experiences may also be uncovered. As the sample was small, it was not possible to utilise a maximum variation approach and parents with different characteristics, for example socioeconomic backgrounds, child age, trial centre and trial group, need also to be interviewed. Owing to the delay with the trial it was also not possible to capture views and experiences of health-care professionals. Therefore, important insights into how the people directly involved in the care of the children and who were key to successful trial implementation were not gleaned. Future research needs to capture the views and experiences of health-care professionals from a range of trial sites.

Comparison with other literature

No new randomised studies of anaesthetic–analgesic ear drops have been published since the last Cochrane review in 2011. The three previous studies have assessed the effectiveness of topical analgesia against placebo in relieving pain due to AOM;11–13 all were included in the Cochrane review (updated 2011), which concluded that ‘the evidence from [these] RCTs is insufficient to know whether ear drops are effective’. 14 These studies address the efficacy of drops for pain and not the impact on case management with antibiotics; none of the trials included an economic evaluation. That said, pain relief is important to parents, and this emerged in the qualitative interviews completed for this study. In a post hoc and unblinded comparison of ear pain between this study’s active and usual-care groups, there was moderate evidence to suggest that active ear drops reduced ear pain, and this was strengthened after adjusting for baseline ear pain (at consultation). After adjustment, the combined mean difference was almost 2 points on the 0- to 6-point scale. In the analysis of pain relief following first instillation of drops in the small sample of 17 patients, there was no evidence to suggest that pain was reduced in active compared with placebo drops groups. However, both unadjusted and adjusted results were in favour of the active drops; those in the active groups had pain scores that were, on average, 0.73 points lower than in the placebo group. Thus, despite the small sample size, the present study may make a further contribution to the Cochrane review.

There is a well-recognised variation in prescribing rates for otitis media: the median rate being 60%, while in the highest decile of prescribing practices the figure is 90%. 1 A previous study in AOM investigated consumption,20 finding that 24% of children consumed antibiotics following a delayed prescription, compared with 33% in this usual-care group.

Our analyses suggest that anaesthetic–analgesic drops might increase medication costs during the acute episode. However, the wider societal costs of antimicrobial resistance were not quantified here. A literature review concluded that current estimates of the costs of resistance are likely to be underestimates and that an accurate forecast of costs may not be possible. 21 A recent trial evaluating the use of antibiotics in lower respiratory tract infections demonstrated that strategies to avoid antibiotic use become dramatically more cost-effective once the costs of resistance are included in the estimate. 35

Implications for clinical practice and research

Tackling antibiotic resistance is an international priority and reducing unnecessary antibiotic use for self-limiting illness is one potential mechanism to address this. This study suggests that substantial reduction in antibiotic use might be achieved in AOM in children by combining a no or delayed prescribing strategy with anaesthetic–analgesic ear drops.

A benefit of the analgesic effect of the active drops remains unproven when compared with placebo, although both active and glycerol drops may have important pain-relieving properties. Further research is needed to establish if effects are mediated by the active ingredients or if they could be achieved using a non-pharmacologically active oil.

Conclusion

This study has provided evidence that anaesthetic–analgesic ear drops result in a substantial reduction in antibiotic consumption in children presenting to primary care with AOM. The small number of children allocated to the placebo drops group was insufficient to establish whether or not the anaesthetic–analgesic ear drops reduced pain. Replication of these findings in a larger study would give greater confidence for informing clinical guidelines, would establish the safety of the intervention beyond the first 8 days and, if the mechanism of action can be established, may allow refinements to the intervention.

Funding history

The CEDAR trial was funded by the NIHR Health Technology Assessment programme. The project start date was 1 January 2015 and the planned study end date was 31 March 2018.

The project closed early, on 31 December 2017, as required by the funders. Issues were encountered with a lengthy delay to the IMP supply. Although the IMP supplier was identified through a competitive procurement process, the supplier failed to deliver the active drops and placebo in line with expected and revised timeframes. There were particular problems with the production of the matched placebo and the ability of the supplier to source materials to replicate the active drops. This contributed to the early closure of this important trial.

Contributions of authors

Alastair D Hay (University of Bristol) was the chief investigator chaired study management group, and contributed to the primary care expertise, study design, protocol development, trial monitoring, writing and approval of final report.

Harriet Downing (University of Bristol) was the trial manager and study management group member, and contributed to the protocol development and trial monitoring.

Nick A Francis (Cardiff University) was the principal investigator for Cardiff Centre and study management group member, and contributed to the primary care expertise, protocol development, trial monitoring, writing and approval of final report.

Grace J Young (University of Bristol) was the study statistician, and contributed to the statistical analysis, reporting to the data monitoring committee, writing and approval of final report.

Clare Clement (University of Bristol) was the qualitative researcher, and contributed to the qualitative interviews and analysis, writing and approval of final report.

Sue D Harris (University of Bristol) was the study nurse and study management group member, and contributed to the primary care expertise and approval of final report.

Aideen Ahern (University of Bristol) was the health economics researcher, and contributed to the health economics analysis and writing and approval of final report.

Behnaz Schofield (Cardiff University) was a study management group member, and contributed to the primary care expertise, protocol development, trial monitoring and approval of final report.

Tammy E Thomas (Southampton University) was a study management group member, and contributed to the primary care expertise, protocol development, trial monitoring and approval of final report.

Jeremy Horwood (University of Bristol) was the senior qualitative researcher, and contributed to the protocol development, study management group member, qualitative analysis, writing and approval of final report.

Peter S Blair (University of Bristol) was the senior statistician, and contributed to the protocol development, study management group, statistical analysis, writing and approval of final report.

William Hollingworth (University of Bristol) was the senior health economist, and contributed to the protocol development, study management group, health economic analysis, writing and approval of final report.

Victoria Wilson (University of Bristol) was the trial manager and study management group member, and contributed to the trial monitoring, writing and approval of final report.

Chris Metcalfe (University of Bristol) was the senior trials methodologist, and contributed to the study management group, writing and approval of final report.

Peter Stoddart (Bristol Royal Hospital for Children) was a study management group member, and contributed to the secondary care expertise, protocol development and approval of final report.

Desmond Nunez (University of British Columbia) was a study management group member, and contributed to the primary care expertise, protocol development, writing and approval of final report.

Mark D Lyttle (Bristol Royal Hospital for Children) was a study management group member, and contributed to the secondary care expertise, protocol development, approval of final report.

Paul Little (Southampton University) was a study management group member, and contributed to the primary care expertise, protocol development and approval of final report.

Michael V Moore (Southampton University) was the principal investigator for the Southampton Centre and was a study management group member, and contributed to the primary care expertise, protocol development, trial monitoring, writing and approval of final report.

Publication

Venekamp RP, Prasad V, Hay AD. Are topical antibiotics an alternative to oral antibiotics for children with acute otitis media and ear discharge? BMJ 2016;352:i308.

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.