Methods

Text in this chapter is reproduced with permission from Chalmers et al.46 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attrition (CC BY 4.0) licence, which permits others to distribute, remix, adapt and build on this work, for commercial use, provided the original work is properly cited. See https://creativecommons.org/licenses/by/4.0/. The text below includes minor additions and formatting changes to the original text. The full barrier enhancement for eczema prevention (BEEP) trial protocol (final version 7.0, 26 February 2021) is available on the NIHR Funding and Awards website47 and a summary protocol has been published.46 The main changes to the protocol after the initial approval in June 2014 was the addition of secondary outcomes for confirmed diagnosis of food allergy and allergic sensitisation and the associated assessments after separate funding was obtained to complete these (added May 2016).

Trial objectives

Primary objective

To determine whether advising parents to apply emollient daily to the entire body surface area for the first year of life can prevent AE in high-risk children.

Secondary objectives

• To determine whether emollients can delay the onset and/or reduce the severity in those who develop AE.
• To determine whether emollients can prevent other allergic diseases developing.
• To determine the safety and cost-effectiveness of the prevention strategy.
• To determine whether any preventative effect is sustained into later childhood.

Trial design and setting

This was a randomised, controlled, two-arm (skin-care advice plus emollient vs. skin-care advice alone), parallel group, multicentre, assessor blind trial with 5-year follow-up and primary outcome assessed at 2 years (Figure 4). Recruitment took place in 12 secondary care sites and four primary care sites in England (see Appendix 1, Figure 12).

FIGURE 4 Schematic diagram showing the trial design and duration for participating families. a, Screening took place during pregnancy or within 21 days of delivery. Reproduced from Chalmers et al.46 under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/). (Figure not reproduced here; see the original article.)

A methodological two-by-two factorial substudy was also nested within the trial to investigate the effectiveness of two interventions on the rates of follow-up data collection. The interventions were: (1) short message service (SMS) notification prior to sending questionnaires at 3, 6, 12 and 18 months versus no SMS notification and (2) sending the £10 voucher for the primary follow-up visit at 24 months with the invitation letter versus giving the voucher at the visit. Full details can be found in the studies within a trial (SWAT) registry (SWAT Repository Store ID 25).48

Participants and eligibility

The eligibility criteria for the participants are shown below.

Inclusion criteria

• Child had a first-degree relative with parental-reported doctor diagnosis of eczema, allergic rhinitis or asthma.
• Child up to 21 days old.
• Mothers must be at least 16 years old.
• Consenting adult had the ability to understand English.

Exclusion criteria

• Preterm birth (defined as birth prior to 37 weeks’ gestation).
• Sibling (including twin) previously randomised to this trial. If multiple birth, the first-born child was randomised into the trial.
• Child had a severe widespread skin condition that would make the detection and/or assessment of eczema difficult.
• Child had a serious health issue which, at parent or investigator discretion, would make it difficult for the family to take part in the trial.
• Any condition that would make the use of emollient inadvisable or not possible.

Recruitment to the trial came from a variety of sources, including primary and secondary care and advertising. Poster and flyers were displayed at secondary care and primary care sites. If interested in taking part, expectant mothers/parents of the newborn were asked to contact the research team directly. Expectant mothers were identified through antenatal and secondary care clinics in which case relevant healthcare professionals approached parents directly about the study or sent invitation letters. In addition to the general practitioner (GP) surgeries used as sites, GP surgeries in areas with a secondary care site were used as Participant Identification Centres (PICs). In the PICs, invitation letters and information sheets were sent to expectant parents. The trial was also promoted through local radio, television, newspapers, the Mumsnet website and the public display of posters and flyers at venues that families and expectant mothers frequent including nurseries, libraries, supermarkets and Sure Start centres.

Screening and consent

Pregnancy status and family history of atopic disease were checked for parents initially expressing an interest in taking part and, if eligible, the parents were sent the participant information leaflet and a screening visit was arranged. The screening visit took place either in the family home or at the recruiting site, depending on parent preference and was conducted either in the third trimester if the baby had not yet been born or as soon as possible after birth and within 21 days of being born. During the screening visit, the research nurse obtained informed consent and checked eligibility. Consent was also sought for the optional genetic component and for use of samples in potential future research. Parents who consented agreed to their child providing a saliva sample at the 24-month visit for FLG genotyping in order to conduct subgroup analysis to explore the effect of emollients on eczema prevention according to genetic risk of atopic disease (FLG null mutations).

Outcomes

Full details of definitions and derivation of outcomes are given in the statistical analysis plan (SAP).47

Primary outcome

The primary outcome was a diagnosis of AE at 24 months defined as meeting the United Kingdom Working Party (UKWP) Diagnostic Criteria for Atopic Eczema,4 which assesses signs and symptoms present over the past year, assessed by a trained research nurse blinded to treatment allocation.

Applying the criteria at 24 months was chosen to ensure that any observed effect on reducing AE prevalence could be considered a true preventative effect rather than masking the emergence of mild eczema in the first year due to the use of the emollient.

Secondary outcomes

1. Presence of eczema between birth and 24 months:
   • Any parental report of a clinical diagnosis of eczema.
   • Completion by parents of UKWP Diagnostic Criteria for Atopic Dermatitis at 12 and 24 months.
2. Presence of visible eczema at 24 months (skin examination by researcher).
3. Time to onset of eczema:
   • First parental report of a clinical diagnosis of eczema.
   • First topical corticosteroid and/or immunosuppressant prescription for eczema.
4. Severity of eczema:
   • Eczema Area and Severity Index (EASI)49 at 24 months.
   • POEM24 at 12 and 24 months.
5. Presence of other allergic diseases:
   • Parental-reported wheezing and allergic rhinitis between 12 and 24 months.
   • Parental report of a clinical diagnosis of food allergy at 12 and 24 months.
   • Parental report of food allergy at 12 and 24 months. Parents were specifically questioned about cow’s milk, egg, peanuts and other nuts plus ‘any other food’.
   • Allergic sensitisation at 24 months to any of the following common allergens: milk, egg, peanut, cat, grass pollen, house dust mite (added in protocol version 4.0 after obtaining separate funding).
   • Confirmed diagnosis of food allergy at 24 months to milk, egg, peanut or ‘any of milk, egg or peanut’ (added in protocol version 4.0 after separate funding obtained). The diagnosis was derived from a combination of parental report, allergic sensitisation and food challenge.
6. Health-related quality of life:
   • Child Health Utility instrument-9 domains (CHU-9D)50 at 24 months in order to estimate quality-adjusted life-years (QALYs).
   • Parental quality of life measured using the EuroQol-5 Dimensions, five-level version (EQ-5D-5L) at baseline and 24 months in order to estimate change in parental QALYs, if any.
7. Health economic outcomes:
   • Healthcare resource use at 3, 6, 12, 18 and 24 months.
   • Cost-effectiveness and cost–utility at 24 months (combining health resource use and health-related quality-of-life outcomes).

Safety outcomes

1. Number of skin infection events during the first year.
2. Number of infant slippage incidents (slippage in hand and slippages to the floor) that occur within an hour of applying emollient during the first year.

Tertiary outcomes (long-term follow-up)

1. Presence of eczema in the previous year at 36, 48 and 60 months based on parental report of a clinical diagnosis of eczema.
2. Any parental report that in their opinion their child has eczema at 3, 6, 12, 18, 24, 36, 48 and 60 months.
3. Presence of eczema at 36, 48 and 60 months based on completion by parents of UKWP Diagnostic Criteria for Atopic Dermatitis.
4. Severity of eczema at 36, 48 and 60 months as measured by POEM.
5. Presence of other atopic diseases:
   • Parental-reported wheezing, allergic rhinitis and food allergy symptoms at 36, 48 and 60 months.
   • Parental report of a clinical diagnosis of asthma or allergic rhinitis by 60 months.
   • Parental report of a clinical diagnosis of food allergy at 36, 48 and 60 months.
6. Health-related quality of life:
   • CHU-9D at 36, 48 and 60 months in order to estimate QALYs.
   • Parental quality of life: EQ-5D-5L at 36, 48 and 60 months in order to estimate parental QALYs.
7. Health economic outcomes:
   • Healthcare resource use at 36, 48 and 60 months.
   • Cost–utility and cost-effectiveness at 60 months (combining health resource use and health-related quality-of-life outcomes).
8. Cumulative incidence outcomes (not specified in protocol, see below for rationale for addition).
   • Parental report of clinical diagnosis of eczema from the age of 12 to 60 months.
   • Parental report of clinical diagnosis of food allergy by 60 months.

Exploratory outcomes (long-term follow-up, not specified in protocol)

1. Parental report of reaction to egg or nuts by 60 months.
2. Parental report of immediate reaction to egg or nuts by 60 months.

The protocol initially specified the paediatric quality of life (PedsQL)51 questionnaire to estimate QALYs for the child at 24, 36, 48 and 60 months. This was changed in May 2016 to the CHU-9D questionnaire to reflect the latest research in the area. The Infant Dermatitis Quality of Life questionnaire52 was also removed as an outcome in May 2016 to reduce the questionnaire burden and also due to BEEP being a prevention trial rather than a treatment trial.

Cumulative incidence tertiary outcomes and exploratory outcomes (not in the protocol) were specified in version 2.0 of the SAP, as a better measure of lifetime experience of eczema and food allergy than single sweeps of 1-year period prevalence. Eczema is a condition that undergoes relapses and remissions – both short term over the course of weeks or months that reflect seasonal influences such as temperature, pollen and humidity (which is adequately captured by enquiring about a one period prevalence) or over the course of years, with some children having eczema which then clears and some getting early eczema which then clears and then returns at different sites. Food allergy can cause only intermittent reactions, often less than once per annum; the condition can be immunologically present for many months or years before a clinical reaction is experienced, especially for allergy to foods which are not widespread components of everyday diets such as tree nuts. The first 12 months were not included in the tertiary outcome for cumulative incidence of eczema as transient eczematous rashes are common in the first year of life and often reported by parents as ‘eczema’ but are less likely to be true AE.

Randomisation and blinding

Within 21 days of the birth of the baby, a baseline visit was conducted (either on the phone, via e-mail or face to face) with the research nurse to confirm eligibility and collect baseline data. Babies meeting the eligibility criteria were then randomised by the research nurse via a web-based randomisation system developed and maintained by Nottingham Clinical Trials Unit (NCTU).

The randomisation schedule was based on a computer-generated pseudo-random code using random permuted blocks of randomly varying size created by NCTU and held on a secure University of Nottingham server. Infants were allocated in a 1 : 1 ratio to either emollient and standard skin-care advice (intervention group) or standard skin-care advice only (control group). Randomisation was stratified by recruiting centre and number of immediate family members (parents/siblings) with atopic disease (one, two, or more than two family members with either eczema, asthma or hay fever).

Parents were informed of their child’s allocation by staff at NCTU by letter. While it was not possible to blind parents to the treatment allocation, efforts were made to minimise expectation bias by emphasising that information on the effect of emollient in addition to standard skin-care advice was limited.

Research nurses were not informed of the allocation to maintain blinding for the follow-up visit for the primary outcome at 2 years. To reduce the chance of the research nurse being unblinded by the parents at the 2-year visit prior to the skin examination for eczema, the appointment letter reminded parents not to tell the nurse which group they had been assigned to in the first year and skin assessments were conducted first during the visit.

Researchers involved in the food allergy assessment process were also blinded to treatment allocation.

The trial statisticians and health economists remained blinded to treatment allocation until after the initial database lock for the analysis of primary outcome data at 2 years.

Interventions

Skin-care advice

Both groups received advice on general skin care in booklet and video format at the time of randomisation.47 Skin advice was based on 2006 NICE guidance for postnatal care up to 8 weeks of birth available at the time (since replaced)53 and updated in 2016)54 and supplemented with expert opinion from dermatological nursing and skin barrier research expertise (SL and MC). The booklet and video advised to use mild cleansers and shampoos specifically formulated for infants, and to avoid soap, bubble bath and baby wipes. Parents were advised to seek medical advice from their GP if their baby developed any skin problems. Infants in the control group were allocated to skin-care advice alone.

Intervention group

Families in the intervention group were advised to apply emollient (Doublebase Gel® or Diprobase Cream®) at least once daily to the whole body (excluding the scalp) until the child reached 1 year, in addition to skin-care advice described above. These two emollients were chosen based on our pilot work with families from our preceding programme grant for applied research.42 We were able to establish that the liquid-paraffin-based emollients, Doublebase Gel (Dermal Laboratories Ltd) and Diprobase Cream (Bayer Plc) are popular with parents and do not have any detrimental effects on skin barrier function. Parents in the pilot study also told us that having a choice of emollients was very important to them.

They were also advised to apply emollient after every bath, even if they had already applied the emollient that day. Daily application was advised in order to encourage regular use of emollient several times a week, but since the study was designed to reflect how the intervention might be delivered in normal practice, no prompts or reminders were sent to parents. The skin- care advice booklet and video for the intervention group additionally provided guidance on how to apply emollients correctly by dotting over the skin and using gentle downward strokes rather than rubbing in. The booklet also contained warnings about the skin being slippery after application and the need to clean up spillages from the floor to avoid slipping.

Upon randomisation, the central trial pharmacy sent the families a 500 g container of each emollient by post. Families reordered their preferred emollient during the intervention period (1 year) by contacting the NCTU. Parents were advised to stop applying emollients when their child reached 1 year of age and no further emollients were supplied after this point.

Trial assessments and procedures

Parents were initially contacted by telephone by the NCTU approximately 2 weeks after randomisation to check they have received their skin-care advice pack and web link for the video, and for infants randomised to the intervention group, to check the date that the family started applying the emollient. Parents were also reminded to contact the NCTU if they had any questions or problems (to protect the research nurses from becoming unblinded).

Parents were asked to complete web-based questionnaires at 3, 6, 12, 18, 36, 48 and 60 months to collect information on any skin problems, eczema and other allergy symptoms (including diagnosis, prescriptions and health resource use), feeding, skin-care practices (including emollient use during the intervention period), skin infections, infant slippage incidents within an hour of applying skin-care products, reactions to food and quality of life as detailed in Table 2 (all questionnaires available at47). Reminders were sent by e-mail after 2 and 3 weeks of non-completion and the questionnaires could be completed up to 4 weeks after the initial e-mail invitation was sent. Paper copies of the questionnaire were sent by post (with pre-paid envelopes) if parents did not wish to complete the questionnaires online. From May 2016, staff at the NCTU also telephoned and/or sent a text message to participants who had not completed the questionnaire after the first reminder and gave participants the opportunity to complete the questionnaire on the phone.

At 24 months, a face-to-face visit with a research nurse took place either in the family home or in clinic according to parent preference. Prior to the first appointment, 28 research nurses were trained in all assessments either by attending an in-person training day or by watching a series of training videos. This included training in diagnosing eczema using the UK diagnostic criteria, the EASI, skin prick test (SPT) and anaphylaxis. After this initial training, nurses used the UKWP Diagnostic Criteria and performed an EASI assessment in clinic with three patients with the local Principal Investigator. Specialist allergy nurses accompanied research nurses at their initial 24 months visits to sign off on proficiency in SPT competency and using an EpiPen (Viatris UK HealthCare Ltd., Potters Bar, Hertfordshire, UK, for anaphylaxis).

After arranging the 24-month visit, the research nurse sent parents an information leaflet and sheet to explain that their child would be offered an allergy test (SPT) at the visit. The nurse then discussed the SPT during the visit and took consent for this if the parents were willing for the child to have the optional allergy test.

During the visit, the research nurse conducted the skin examination for the UKWP Criteria and the EASI and for parents who had given the additional optional consents, collected a saliva sample and conducted a SPT. Parents were also asked to complete the POEM, EQ-5D and CHU-9D questionnaires and another questionnaire to collect information about skin problems, reactions to food, symptoms of allergic rhinitis and wheezing, skin care/washing and other characteristics potentially associated with the development of eczema (e.g. antibiotic use since birth, furry pets, dust mite reduction measures, number of other children in the household and whether the child attends nursery or playgroup), as detailed in Table 2 (case report form for 24-month visit available at NIHR Funding and Awards47). At the end of the visit, nurses recorded if they had become aware of which group the child had been randomised to and if so whether this happened before, during or after the skin examination.

Skin prick testing was carried out following a trial-specific procedure in line with the British Society for Allergy and Clinical Immunology procedures for SPT55 (Appendix 2 of protocol v7.0 26 February 2021, see NIHR Funding and Awards47). The following allergens were tested; grass pollen mix, dust mite and cat (Allergopharma GmBH & Co., Reinbek, Schleswig-Holstein, Germany), peanut (Inmunotek, Madrid, Spain), fresh skimmed cow’s milk and fresh chicken egg. Positive (1% histamine) and negative (0.9% saline) controls were used (Allergopharma GmBH & Co.). The research nurse measured the size of the reactions, and the results were reviewed by the BEEP food allergy team.

If it was not possible to complete a face-to-face visit, data collection was attempted remotely by telephone, text, e-mail or post. If no follow-up with the family was possible at 24 months, the NCTU attempted to collect key minimal data around diagnoses of eczema and food allergy (including number of prescriptions, primary and secondary care visits) from the child’s GP.

Children with a positive SPT or history suggestive of food allergy, where further information was needed to confirm a food allergy, were invited for a supervised oral food challenge (Figure 5) in an allergy clinic at Imperial College Healthcare National Health Service (NHS) Trust or Sheffield Children’s Hospital NHS Trust. A trial-specific food challenge information sheet was sent to parents of children invited to the food challenge and if needed the food allergy team contacted parents to gather more information about the child’s food allergy history. Parents who agreed to take part gave consent for their child to have a food challenge to confirm whether their child had a food allergy. Oral food challenges were conducted by experienced allergy nurses blinded to treatment allocation following a trial-specific standard procedure using incremental doses or a single dose of the relevant food (see Appendix 3 of protocol v7.0 26 February 2021, see NIHR Funding and Awards47). Allergy nurses were able to review SPT results and allergy history as needed to inform clinical decision-making. Trained nurses supported by a consultant paediatric allergist evaluated whether there had been a clinical reaction to the food using modified practical allergy (PRACTALL) and Integrated Approaches to Food Allergen and Allergy Risk Management criteria.57,58 Clinical reactions during the food challenge were treated using standard clinical guidelines, with antihistamine, bronchodilator, intramuscular adrenaline or other medications, as needed.

FIGURE 5 Food allergy assessment process. (Figure not reproduced here; see the original article.)

For participants invited to a food challenge where this did not take place, a panel of experienced paediatric allergists agreed allergy status (see Figure 5). The panel developed and validated an algorithm to make the food allergy diagnosis.59 The panel were blinded to treatment allocation and used all available information from the questionnaires and assessments and, where necessary, from secondary care records and direct communication with participants’ parents. Following the SPT and food challenge (where applicable), parents and the child’s GP were sent a letter summarising the findings and any recommended action by the BEEP food allergy team.

Saliva samples were sent to the Centre for Dermatology and Genetic Medicine at the University of Dundee for deoxyribonucleic acid (DNA) extraction by standard techniques. DNA samples were genotyped for the four most common FLG null mutations in the white European population (2282del4, R501X, S3247X and R2447X).60

Small tokens of appreciation were sent to all participating families by the NCTU throughout the trial including BEEP branded muslin or bib at randomisation, birthday card and BEEP branded plastic cutlery set or storybook at the child’s first birthday and BEEP branded cloth shoulder bag sent at 18 months. At 24 months parents received a £10 voucher either with the invitation letter for the primary follow-up visit or at the visit according to their allocation for the nested SWAT. A £10 voucher was also sent to parents on completion of the 48- and 60-month questionnaire. Parents were also sent trial newsletters every 6 months (from January 2016). The newsletter sent to parents in February 2020 summarised the results for the primary and secondary outcomes up to 24 months.61

Trial oversight

An independent Trial Steering Committee (TSC) provided overall supervision of the trial. The TSC had an independent chair and four independent members (parent representative, statistician, consultant dermatologist and paediatric epidemiologist as detailed in Acknowledgements). The TSC met at least once a year during recruitment and follow-up for the primary outcome.

Due to the very low medical risk associated with the intervention, there was no separate data monitoring committee for the trial. Safety outcomes were monitored by the TSC during closed sessions of the meetings attended by only the independent members.

Sample size

The sample size was based on assuming 30% of children in the control group would have eczema between 1 and 2 years of age (based on previous epidemiological studies in this high-risk population) and a conservative relative reduction of 30% in the intervention group. This relative reduction was considered conservative as in the pilot study, a 50% reduction in eczema at 6 months was observed [43% developed eczema in the control group (n = 55) and 22% developed eczema in the emollient group (n = 53), 95% CI for RR 0.28 to 0.90]. The anticipated effect size was lower in the main trial due to the more pragmatic study design and the longer-term outcome assessment; however, such a reduction would still have important implications for families and health services.

A total of 1282 children were required to allow this difference to be detected (i.e. 30% of children in the control group, 30% relative reduction to –21% of children in the intervention group) at the 5% significance level (two-sided) with 90% power and allowing for 20% attrition at 24 months.

The protocol specified that the assumptions underpinning the sample size would be checked by independent members of the TSC after approximately 21 months of recruitment (i.e. by checking the percentage of children with eczema in the control group and percentage with follow-up data). However, the original target sample size of 1282 was exceeded prior to this point and the Chief Investigator requested that the sample size review therefore be brought forwards. At this point (July 2016), sites were told to stop consenting women to the trial however randomisation continued for women who consented before this date prior to the birth of their baby. In August 2016, independent members of the TSC were sent details of the follow-up questionnaire completion rate in each group and parental-reported medical diagnosis of eczema in the control group from the questionnaires by a NCTU statistician independent of BEEP. The TSC were asked to advise on whether consent and randomisation should continue. The TSC recommended that consent to the trial should be permanently terminated but randomisation should continue for any women who had consented but had not yet been randomised. The total number of children randomised was expected to be approximately 1400.

Stopping rules and discontinuation

There was no planned interim analysis of treatment efficacy for the primary outcomes. The criteria below (Box 1) were specified in the protocol in relation to recruitment and adherence to trigger discussion with the TSC and funder regarding the best course of action.

• Recruitment: If recruitment (as documented in the recruitment plan) was < 50% of the expected rate by 15 months, and strategies to overcome the identified barriers to recruitment had not been successful.
• Adherence to the intervention: If fewer than 90% of families in the intervention group had applied emollient over the majority of their child’s body at some stage and fewer than 70% were still using emollient at 6 months.
• Emollient use by the control group: If emollient use in the control group exceeded 25% of families at 6 months. This excluded the use of emollients for the treatment of eczema and only applied to emollient use that closely reflected the intervention (i.e. regular widespread use in the first year of life, defined as widespread emollient use over the majority of the child’s body at least 3 or more days per week).

Statistical methods

Analyses are detailed in the SAP. Version 1.0 of the SAP was finalised prior to database lock and release of treatment allocation codes for analysis of the primary and secondary outcomes at 24 months.47 Version 2.0 of the SAP was finalised prior to the database lock for the analysis of the tertiary outcomes at 60 months.47 All analyses were carried out using Stata 15.0 or above (StataCorp LP, College Station, TX, USA), unless otherwise specified.

The main approach for analysis was to analyse participants (children) as randomised regardless of adherence with the allocated intervention using available data (i.e. without imputation for missing data) with sensitivity analysis for key outcomes using multiple imputation for missing data. Estimates of the intervention effect are presented with 95% CIs.

Descriptive analyses

Baseline characteristics were summarised by allocated group to describe the sample recruited and to examine balance between the two groups.

Follow-up completion was summarised in the two groups and baseline characteristics of children with and without primary outcome data in each group compared descriptively. Information collected on follow-up questionnaires and at the 24-month visit on characteristics which may be associated with the development of eczema and on the timing of introduction of allergenic foods as well as recent consumption of these foods were summarised by allocated group to inform the interpretation of the results.

Compliance and contamination

Compliance in the intervention group at each questionnaire time point was defined as widespread emollient use over the majority of the child’s body at least 3 or more days per week. The ‘majority of the child’s body’ was defined as at least two of the three body areas asked about on the questionnaire (face/neck, arms/legs or trunk). Contamination at each questionnaire time point in the control group was defined as use of a moisturiser or oil at least 3 days per week over most or all of the child’s body since the last questionnaire. Compliance in the intervention group and contamination in the control group at each questionnaire time point were tabulated.

Compliance and contamination over the first year of life were described using an ordered categorical variable, as defined in Table 3.

Compliance/contamination was summarised for participants with complete data on compliance/contamination (i.e. completed questionnaires at 3, 6 and 12 months) and for all participants using the assumptions for missing data described below.

1. Participants with no reported data on emollient/moisturiser use were categorised as not compliant in the intervention arm and not contaminated in the control arm.
2. If participants missed a questionnaire(s) and went on to complete a subsequent questionnaire, missing emollient/moisturiser use was based on the next subsequent observation carried backwards. For example, if the 6-month questionnaire was missed for a participant in the intervention group and they reported being compliant at 12 months, then it was assumed that they were also compliant at the 6-month time point. The rationale for this was that it was assumed that compliance was likely to decrease over time so that this would be conservative for the intervention group.
3. If participants completed questionnaires initially and missed later questionnaires (e.g. completed at 3, did not complete 6 or 12 or completed at 3 and 6, did not complete 12), then it was assumed that there was no compliance (intervention)/contamination (control) for the later missed questionnaires.
4. Categorisation of compliance/contamination over the first 12 months following randomisation for participants with missing emollient/moisturiser use data then proceeded according to the above table.

Compliance/contamination was summarised according to (1) allocated group and (2) allocated group and whether there was a parental report of eczema diagnosis by a doctor or nurse in the first year (yes, no or unknown).

Primary outcome

The adjusted RR and difference in risk for the primary outcome were estimated using generalised estimating equations with the binomial family and log/identity link, respectively, with an exchangeable correlation matrix to account for randomisation being stratified by centre and number of immediate family members with atopic disease (one, two or more than two) included as a covariate.

Sensitivity analysis for the primary outcome

The following sensitivity analyses for the primary outcome were performed:

• Analysis repeated including diagnosis of eczema data collected from GP records for participants with missing primary outcome data.
• According to the method of collection of the primary outcome data: between-group estimates for the risk of eczema were calculated separately for outcomes collected during face-to-face visits and outcomes collected by telephone/e-mail/SMS using generalised estimating equations adjusting for stratification variables.
• Analysis repeated replacing ‘visible flexural dermatitis’ in the UKWP criteria with any visible dermatitis (EASI score of > 0) in the derivation of eczema between 1 and 2 years of age.
• Using multiple imputation for missing primary outcome data (further details described below).
• Repeating the analysis assuming that all participants in the intervention group with missing primary outcome data were eczema free or had eczema and all participants in the control group with missing data had eczema or were eczema free (i.e. best-case and worst-case scenario for intervention).

A sensitivity analysis was also planned to further adjust for any baseline characteristics with an observed imbalance between groups. However, no important differences between groups in the baseline characteristics were observed so this sensitivity analysis was not conducted.

Multiple imputation was performed using chained equations.62 The following variables were used in the imputation model:

• Allocated group.
• Randomisation stratification variables: centre, number of immediate family members with atopic disease (one, two or more than two).
• Age of mother, number of other children in household at screening (none, one, two, three or more), any furry pets in the household at screening (yes/no) and baby sex (baseline variables identified as predictive of drop-out by examination only).
• Variables used in subgroup analyses – number of FLG null mutations, the number of immediate family members with eczema, water hardness, season of birth and regular use of probiotic supplements during pregnancy.
• Summary of compliance and contamination in the first year of life as per the definitions above (Table 3).

The following outcomes were imputed:

• Diagnosis of eczema in the last year at age 2 years (primary outcome).
• Parental report of a clinical diagnosis of eczema at 2 years.
• Parental report of immediate allergy to cow’s milk, egg or peanut at 2 years.
• Allergic sensitisation to cow’s milk, egg or peanut.
• Food allergy to any of milk, egg or peanut at 2 years.

Forty data sets were imputed and the results of the analyses on the imputed data sets were combined using Rubin rules for multiply imputed data. This analysis assumed that unobserved outcomes are missing at random (MAR) and depend on observed characteristics but not the unobserved outcomes, since it was considered most plausible that missing outcomes would have been similar to observed outcomes for participants with similar characteristics.

Secondary analysis of the primary outcome

To explore the effect of application of emollient in the first year of life in parents who would comply with the allocated treatment, the complier-average causal effect (CACE) was estimated.63 CACE models were implemented as latent growth mixture models64 in MPlus (Version 5.2), with compliance/contamination status included as a training variable for estimating class membership.

Two separate odds ratios (ORs) for the CACE were estimated based on the following definition of compliance:

1. full compliance over the first year of life as per Table 3 – participants in the control group who met the criteria for full contamination were considered as always-takers
2. compliance within the first 3 months (i.e. in the full compliance or early-onset application categories in Table 3) – participants in the control group who met the criteria for full or early-onset contamination were considered as always-takers.

For children with incomplete data on compliance/contamination, the assumptions described above were used to categorise their compliance/contamination.

Subgroup analysis of the primary outcome

Exploratory subgroup analyses were conducted by including appropriate interaction terms in the regression model for the following variables:

1. Number of FLG mutations (no FLG mutation vs. one or two FLG mutations).
2. Number of immediate family members with atopic disease (one, two or more than two).
3. Number of immediate family members with eczema (zero, one, two or more).
4. Season of birth (spring – born in March, April or May, summer, autumn, winter).
5. Water hardness (dichotomised into hard/very hard and moderate/soft).
6. Parental-reported regular use of probiotic supplements during pregnancy (yes/no).

Subgroup analyses one to three were specified in the protocol and subgroup analyses four to six were specified in the SAP prior to database lock for the primary outcome.

The FLG subgroup analysis included children whose mother and father were reported to be of white European ethnicity, since the mutations tested are most prevalent in the white European population. Children who were found to have at least one mutation (regardless of ethnicity) were also included in this subgroup analysis but children of other ethnicities without FLG mutations were excluded because the genotyping may have been a false-negative result.

Secondary outcomes

For binary secondary outcomes, the adjusted RR and difference in risk were estimated using the analysis model specified for the primary outcome.

Time to onset of eczema was presented descriptively by showing the cumulative percentage of children with eczema at 3, 6, 12, 18 and 24 months in a bar graph and table and presented separately according to: first parental report of a clinical diagnosis of eczema and first topical corticosteroid and/or immunosuppressant prescription for eczema.

Additional analyses were also performed for the main food allergy outcome of confirmed diagnosis of food allergy at 24 months to any of milk, egg or peanut. Sensitivity analyses were conducted using multiple imputation for missing outcomes as described above and repeating the analysis including panel decisions of ‘unclear – possible food allergy’ as allergic and ‘unclear – food allergy unlikely’ as not allergic. The CACE was also estimated for the confirmed food allergy outcome (using methods described above). Exploratory subgroup analyses were conducted by including appropriate interaction terms in the regression model for the following variables: number of FLG mutations, number of immediate family members with atopic disease and number of immediate family members with eczema.

Safety outcomes

The adjusted incidence rate ratio for the number of skin infections reported per child was estimated using generalised estimating equations with a negative binomial family and log link with an exchangeable correlation matrix to account for randomisation being stratified by centre and number of immediate family members with atopic disease (one, two or more than two) included as a covariate. Slippage incidents were compared between allocated groups as a binary variable (any slippages reported in the first year/none) using the same analysis model as specified for the primary outcome.

For each questionnaire time point, parental-reported skin infections and slippage incidents since the last questionnaire were also presented descriptively by allocated group and parental-reported emollient/moisturiser use (none/some/widespread over the majority of the child’s body at least 3 or more days per week).

Tertiary outcomes

The analysis of the tertiary outcomes was in keeping with the analysis of the primary and secondary outcomes. Analysis was according to randomised group, regardless of adherence with allocation and estimates of the intervention effect are presented with 95% CIs. The main analysis of the tertiary outcomes assumed that missing outcomes were MAR, that, it does not depend on the unobserved outcomes given the observed data.

Analysis of binary tertiary outcomes at 36, 48 and 60 months used mixed-effects logistic regression model, which gives valid inferences when data are assumed MAR.65 The models included the outcome collected at earlier time points in the trial (i.e. 12 and 24 months where applicable) as dependent variables and adjusted for the randomisation stratification variables, using a fixed effect for the number of immediate family members with atopic disease (one, two, or more than two) and a random effect for the recruiting centre, and a random effect for participant. Models included an allocated treatment-by-time interaction to estimate the between-group difference at each follow-up time point and where technically possible an interaction between number of immediate family members with atopic disease and time. Adjusted risk differences and adjusted risk ratios along with corresponding 95% CIs were obtained using Stata’s margins command with standard errors computed using the delta method.66

Multiple imputation using chained equations was used to impute missing outcomes collected at 60 months on parental report of a clinical diagnosis of asthma and parental report of a clinical diagnosis of allergic rhinitis and the derived outcomes of parental report of a clinical diagnosis of eczema from the age of 12 to 60 months and parental report of a clinical diagnosis of food allergy by 60 months. The following variables were used in the imputation model: allocated group, randomisation stratification variables (centre, number of immediate family members with atopic disease) and baseline variables identified as predictive of drop-out (by examination only: mothers age at randomisation, number of other children in the household at randomisation, decile of index of multiple deprivation). Fifty data sets were imputed. Between-group effects in each imputed data set were estimated using a mixed-effects logistic regression model including a fixed effect for randomisation stratification variable of number of immediate family members with atopic disease and a random effect for the recruiting centre. The adjusted risk differences and adjusted risk ratios were computed in each imputed data set (computed using the delta method described above) and combined using Rubin rules for multiply imputed data.

To explore the robustness of the results to the MAR assumption, sensitivity analysis was conducted for the tertiary outcomes of parental report of clinical diagnosis of eczema from the age of 12 to 60 months and parental report of clinical diagnosis of food allergy by 60 months under a missing not at random (MNAR) assumption using controlled multiple imputation.67 Delta (δ)-based multiple imputation was used to modify the value imputed under a MAR assumption by a fixed amount to explore how the results change if participants with missing outcomes were more likely to have a worse outcome than predicted (based on the MAR assumption). A range of δ values were used in the sensitivity analysis.

Summary of changes to the protocol

A full list of all substantial amendments to the protocol can be found in Appendix 2. All amendments were reviewed by the Sponsor before submission to, and approval by, the Research Ethics Committee and/or the Health Research Authority.

Patient and public involvement

Patient and public involvement (PPI) members of the team provided a vital and valuable role throughout the trial. The aim of PPI was to enhance the design of the main trial.

We were fortunate in being able to conduct a thorough pilot trial as part of a programme grant42 in preparation for this definitive trial. The pilot RCT was similar in design to the main trial and therefore provided the opportunity for meaningful and significant input from parents who had direct experience of a similar eczema prevention trial. Their opinions and feedback during the design of the main trial were therefore largely based on experience rather than hypothetical scenarios and were key in reaching the final design of this main trial. The main areas of input from parents who had participated in the pilot were around the following issues:

1. Continue applying emollient until the child was 12 months old rather than the 6 months tested in the pilot. Parental feedback strongly suggested was that this was acceptable, and in fact, many had chosen to do so anyway in the pilot. Therefore, we felt confident that an extended intervention period of 12 months could be introduced to the main trial without high risk to intervention adherence and this was reflected in the continued high adherence rates during months 6–12 of the intervention period.
2. Restricting choice of emollient to cream/gel formulation only. Parents reported in the pilot that they liked having a choice of emollient, so it was important to retain this. However, because the cream/gel formulation was the most popular, the decision was made to offer a choice from within this emollient type. A group of parents who had participated in the pilot trial participated in a preference study which, along with mechanistic studies, supported the choice of emollient for the main trial.
3. Number of visits reduced to only screening and 24-month follow-up. The pilot trial involved multiple visits to the nurse to check the skin for signs of eczema. The feedback from parents was that although they appreciated the support, they often felt the visits were not necessary, especially when there was nothing wrong with their child’s skin. Taking this parental feedback into account and to design a large definitive trial that was more practical to deliver, the interim follow-up was conducted via online questionnaires and via contact with the co-ordinating centre over the telephone, coupled with advice to visit the GP if they had concerns about their child’s skin. This was a successful approach in terms of making it a viable trial but contacting the higher proportion of non-completers of online questionnaires than anticipated required more resource from the co-ordinating centre. Incentives to complete the questionnaires were introduced part way through the trial to help with completion rates.
4. Addition of allergic sensitisation (skin prick) tests at the 24-month visit. As these tests were not done as part of the pilot trial it was essential to get input from parents into the acceptability of these in the main trial. Parents were generally very keen for their child to receive these tests, but due to funding restrictions, SPTs were introduced part way through the trial recruitment period. As a result, and because of additional concerns raised by families participating in the main trial about the need to travel to the oral food challenge, the SPTs were introduced as optional and separate consent was sought.

Further details of how parental input shaped the main trial design are in the pilot study report.42

In the early stages of the trial set-up, the Centre of Evidence Based Dermatology (CEBD) patient panel also provided input. The panel comprises of people with lived experience of a wide range of skin diseases but the input into this trial was sought mainly from members who are parents of children with eczema along with some adults with eczema. The panel reviewed and improved the BEEP patient-facing study documentation, including the participant information sheet (PIS), consent forms and the design and content of the online follow-up questionnaires. These were refined as a result of the feedback to ensure they were suitable. As the main trial progressed through the pilot phase, further feedback was sought from both parents and an eczema support group.

Patient and public involvement representatives were involved in promoting and publicising the trial (through local television and radio) alongside the Chief Investigator. In addition, the PPI team were consulted to help modify documentation to increase up-take of the SPT at the 24-month visit and subsequent food challenges. Their advice was sought to help disseminate the important primary outcome findings. We also decided to share the 2-year primary outcome results directly with participating families61 as our first audience, that is before the main publication in The Lancet. Additional PPI colleagues with lived experience of eczema also participated in our 5-year results reveal meeting and were helpful in contextualising the key results on lack of benefit and possible signals of increased minor skin infections and skin sensitisation.

A PPI representative sat on the TSC and attended the meetings throughout the trial and provided key advice on how to balance the key messages for parents and carers.