Emollients for prevention of atopic dermatitis: 5- year findings from the BEEP randomized trial

Background: The effectiveness of emollients for preventing atopic dermatitis/eczema is controversial. The Barrier Enhancement for Eczema Prevention trial evaluated the effects of daily emollients during the first year of life on atopic dermatitis and atopic conditions to age 5 years. Methods: 1394 term infants with a family history of atopic disease were randomized (1:1) to daily emollient plus standard skin- care advice (693 emollient group) or standard skin- care advice alone (701 controls). Long- term follow- up at ages 3, 4 and 5 years


| BACKG ROU N D
Atopic dermatitis (syn, atopic eczema, eczema) is a global problem affecting around 1 in 5 children 1 and 1 in 20 adults. 2,3 The prevalence of atopic dermatitis (AD) seems to be increasing especially in cities undergoing rapid demographic development. 4 Genetic factors such as genes coding for skin barrier proteins and immunological responses appear to be important, 5 but the increased prevalence over time, increased risk in smaller families, and migrant studies suggest that environmental factors also play a role. 6 While many effective topical and systemic treatments are available for established AD, 7 prevention of AD has remained elusive. 8,9 Most previous preventive strategies focused on allergen reduction during pregnancy and during infancy with little evidence of benefit. 10 Some evidence exists for a possible role of probiotics, 11 but the exact combination of bacterial strains and timing is still unclear and issues such as selective reporting may have impacted the evidence base. Interest in the role of a defective skin barrier preceding AD development led to the hypothesis that enhancement of the skin barrier from birth might prevent a chain of events resulting in skin was via parental questionnaires. Main outcomes were parental report of a clinical diagnosis of atopic dermatitis and food allergy.
Results: Parents reported more frequent moisturizer application in the emollient group through to 5 years. A clinical diagnosis of atopic dermatitis between 12 and 60 months was reported for 188/608 (31%) in the emollient group and 178/631 (28%) in the control group (adjusted relative risk 1.10, 95% confidence interval 0.93 to 1.30).
Although more parents in the emollient group reported food reactions in the previous year at 3 and 4 years, cumulative incidence of doctor-diagnosed food allergy by 5 years was similar between groups (92/609 [15%] emollients and 87/632 [14%] controls, adjusted relative risk 1.11, 95% confidence interval 0.84 to 1.45). Findings were similar for cumulative incidence of asthma and hay fever.
Conclusions: Daily emollient application during the first year of life does not prevent atopic dermatitis, food allergy, asthma or hay fever.

K E Y W O R D S
asthma, atopic dermatitis, food allergy, prevention, rhinitis

G R A P H I C A L A B S T R A C T
Graphical Abstract1394 newborns at high risk of atopic dermatitis randomized to either daily emollients (Doublebase Gel [Dermal Laboratories] or Diprobase Cream [Bayer]) for 12 months or standard skin care advice. Regular emollients did not prevent atopic dermatitis, food allergy, asthma or allergic rhinitis during the first 5 years of life. Severity of atopic dermatitis similar in the two groups at 5 years. Abbreviation: AD, atopic dermatitis

Funding information
Health Technology Assessment (HTA) 12/67/12; National Institute for Health Research inflammation and establishment of AD. 12 The risk of atopic dermatitis is strongly associated with mutations in the gene encoding filaggrin -a protein that contributes to skin barrier integrity that suggests an impaired skin barrier as a critical defect in the development of AD. 13 Dysfunction in the skin barrier starts soon after birth, making enhancement of the skin barrier a possible target for AD prevention by reducing inflammation from irritants and sensitization through the skin. 14 The "outside-in" hypotheses suggests that there is a complex interplay between epithelial barriers, environmental factors and the immune system in the development of systemic allergic diseases such as AD. 15 Food sensitization may be initiated through an impaired skin barrier, especially in those with AD, so prevention of AD may also prevent the development of subsequent food allergy 16,17 .
Furthermore, if associated conditions such as asthma and allergic rhino-conjunctivitis truly follow-on from AD in predisposed individuals in the so-called 'allergic march', 18 then it might also be possible to prevent such co-morbidities by preventing earlyonset AD with emollients. 16 Two pilot studies had suggested an efficacy signal for preventing AD using such an approach. 19,20 The rationale for the follow-on BEEP (Barrier Enhancement for Eczema Prevention) study was to conduct a definitive large randomized controlled trial to evaluate whether whole-body daily emollient application for the first year of life could prevent AD in high-risk children, compared with standard skin care. 21 Results for the 2-year primary outcome of AD did not show any protective effect of daily emollient on AD development (adjusted relative risk 0.95 [95% confidence interval (CI) 0.78 to 1.16], P = .61). 22 Secondary outcomes for AD were consistent with the primary outcome.
Parental reported skin infections were more common in the emollient group during the first year (adjusted incidence rate ratio 1.55, 95% CI 1.15 to 2.09). There was also no evidence that emollient reduced the risk of food allergy (adjusted relative risk 1.47, 95% CI 0.93 to 2.33). Other studies have found similar results on risk of AD 23 but findings are controversial, with some small studies and systematic reviews reporting positive effects. 24 The purpose of the five-year follow-up of children in the BEEP trial was to evaluate the longer-term effects of daily emollient application during infancy on AD and other atopic outcomes up to 5 years of age. 21 2 | ME THODS

| Study design and participants
BEEP was a multicentre, 2-arm, parallel-group randomized controlled trial which recruited participants from 12 hospitals and four general practice sites in the UK. The trial was approved by the West Midlands Ethics Committee, UK (14/WM/0162). The protocol 21 and results for the primary outcome at 2 years have been published. 22 Briefly, between November 2014 and November 2016 after informed consent from the parent/guardian, term newborns (≥37 weeks gestation) at high-risk of developing AD (at least one first-degree relative with parent-reported doctor-diagnosed AD, allergic rhinitis or asthma) were randomized (1:1) to apply emollient all over the body daily for the first year plus standard skin-care advice (emollient) or standard skin-care advice only (control). Standard general skin care advice was provided in booklet and video format at the time of randomization and included guidance to use mild cleansers and shampoos specifically formulated for infants, and to avoid soap, bubble bath, and baby wipes. 22 Randomization was stratified by recruiting centre and number of first-degree relatives with atopic disease (1, 2, or >2). Participating families were not blinded to group allocation. Parents whose children were allocated to the emollient group were initially sent both Doublebase Gel (Dermal Laboratories, Herts, UK) and Diprobase Cream (Bayer, Berks, UK) and specified which emollient they wanted when reordering. No emollients were supplied after the child reached 1 year of age. Adherence was assessed by asking parents about emollient use at 3, 6, and 12 months and was deemed satisfactory if emollients were applied at least 3-4 times per week to most of the child's body. We used a similar definition for contamination in the control group.
The sample size for the trial was calculated for the primary outcome of a diagnosis of AD in the last year as defined by the UK working party refinement of the Hanifin and Rajka diagnostic criteria for eczema at age 2 years 25 assessed by research nurses blinded to treatment allocation. The original target sample size of 1282 was based on a relative reduction of 30% in the primary AD outcome at the 5% significance level (two-sided) with 90% power assuming 30% of children would have AD in the control group, and 20% dropout.

| Outcomes
Long-term follow-up outcomes (defined as tertiary in the protocol) were: • Presence of AD in the previous year at 3, 4 and 5 years based on parental report of a clinical diagnosis of AD.
• Any parental report that in their opinion their child had AD at 3, 6, 12, 18 months, 2, 3, 4 and 5 years. • Presence of other atopic diseases: • Parental reported wheezing, allergic rhinitis and food allergy symptoms at 3, 4 and 5 years.
• Parental report of a clinical diagnosis of asthma or allergic rhinitis by 5 years. These outcomes were added to capture the lifetime experience and fluctuating nature of AD and food allergy. The first 12 months were not included for AD 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 AD. 28 Health-related quality of life and health economic long-term outcomes will be reported separately. No additional long-term safety data was recorded between years 2 and 5 of follow-up.

| Statistical analysis
Details of the analyses of the long-term outcomes were added to the SAP 29 by the trial statistician after the analysis of the primary and secondary outcomes at which point, the investigators, trial management, data management, statisticians and participants were aware of the results. Full details of definitions and derivations of the longterm tertiary outcomes are given in version 2.0 of the SAP, which was finalized prior to the database lock for the analysis of the longterm outcomes at 60 months. All analyses were carried out using Stata 17.0 (StataCorp LP, College Station, TX, USA).
Analysis was according to randomized group regardless of adherence with the allocation in the first year. The main analyses made the assumption that missing outcomes were missing at random, that is, did not depend on the unobserved outcomes given the observed data. All analyses adjusted for randomization stratification variables using a fixed effect for number of immediate family members with atopic disease and a random effect for the recruiting centre.
Analysis of binary long-term outcomes at 3, 4 and 5 years used mixed effects logistic regression models including the outcome collected at earlier time points (i.e. 12 and 24 months where applicable) with 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. Adjusted risk differences and risk ratios along with corresponding 95% confidence intervals (CI) were obtained using Stata's Margins command with standard errors computed using the delta method. 30 Multiple imputation was used to impute missing outcomes collected at 5 years only and the cumulative incidence outcomes.
Between-group effects in each imputed dataset were estimated using mixed effects logistic regression. Adjusted risk differences and risk ratios were obtained, as described above, and combined using Rubin rules for multiply imputed data. Further details of the multiple imputation model and sensitivity analyses are in the Supporting Information and SAP. Exploratory subgroup analyses for FLG genotype was done by including an interaction term in the analysis model for the parental report of clinical diagnosis of AD from the age of 12 months to 60 months. The baseline characteristics of infants in whom the 5-year questionnaire was completed were similar in the two groups ( Table 1).

Families of infants in both groups in whom the 5-year question-
naire was not completed were more likely to have joined the study after the birth of their baby rather than consenting antenatally, had slightly younger mothers on average, were more likely to be of nonwhite ethnicity, were more likely to be in a household with other children, lived in areas on average with lower deciles of the Index of Multiple Deprivation and were less likely to have a first degree relative with a history of AD at randomization ( Table 1).

| Moisturizer use during follow-up
At 3 years, parent-reported application of a moisturizer at least 3 times per week over all or most of the child's body in the past year was still increased in the emollient group (139/449, 31%) compared with the control group (94/471, 20%), and differences remained at 4 years (25% vs 18%) and 5 years (22% vs 16%). In both groups and at all time points, this frequent whole-body moisturizer use was more common in children with reported AD.

| AD outcomes
Diagnosis of AD at 3, 4 and 5 years was consistently slightly higher in the emollient group when compared to the control group, but adjusted differences were small, and none were statistically significant. The lack of difference between emollient and control groups for AD diagnosis was consistent for different methods of defining AD in the last year, including parental report of a clinical diagnosis, UK Working Party Diagnostic Criteria for AD ( Table 2) and parental opinion of whether their child had developed AD (Table S2). AD of moderate severity or worse as measured by parent-reported symptoms on the POEM was also very similar between the groups at 3, 4 and 5 years ( Table 2).  Table 3). Parental report of immediate reactions to foods containing cow's milk, egg or nuts and of a clinical diagnosis of food allergy in the previous year were also slightly higher in the emollient group than in the control group at 3 and 4 years ( Table 3). At 5 years, all outcomes relating to food allergy were similar between the two groups ( Table 3).   [4,9] 4 [3,7] 6 [4,9] 4 [2,7] Note: Other ethnicities include mixed ethnicity, Middle Eastern and south American.

| Wheezing and allergic rhinitis outcomes
( Table 4). Parental report of symptoms of allergic rhinitis were similar between groups at 3, 4 and 5 years, with approximately a quarter of parents in each group reporting such symptoms in the previous year ( Table 4).

| Cumulative incidence outcomes
There were no differences between the two groups in the cumulative incidence of a parental report of a clinical diagnosis of AD, food allergy, asthma or allergic rhinitis by 5 years ( ing the impact of a worse outcome in those with missing data were consistent with the main analyses (see Table S3). Subgroup analyses according to FLG genotype found no evidence of an interaction (see Table S4). Although safety data was not specifically recorded in the 3-to 5-year follow-up period, no safety concerns such as serious infections or slippages were spontaneously reported during that period. Note: Adjusted relative risk/difference in risk estimated using a mixed effects logistic regression model using all available outcome data (including time points prior to 3 years) adjusting for randomization stratification variables and including a random effect for participants. The number of participants and observations included in each analysis model are shown in Table S1. Note: Adjusted relative risk/difference in risk estimated using a mixed effects logistic regression model using all available outcome data (including time points prior to 3 years) adjusting for randomization stratification variables and including a random effect for participants. The number of participants and observations included in each analysis model are shown in Table S1. a Immediate defined as reaction within 2 h of eating the food.

| Main findings
This study presents the first long-term follow-up data from an emollient for AD prevention trial documenting AD and other atopic outcomes to 5-years. Consistent with earlier findings from the BEEP trial, we found no evidence for an effect of daily emollient application during the first year of life on longer-term AD risk.
Our data also show no clear evidence for an effect of regular emollient application during infancy on risk of other atopic outcomes during the first 5 years of life. Some food allergy outcomes were increased in the emollient group, consistent with findings at age 2 years. Food allergy findings were however inconsistent and imprecise with no effect seen in cumulative incidence of parent-reported food allergy diagnosis by age 5 years. Similar to AD outcomes, we can be reasonably confident that daily emollient during infancy did not reduce food allergy risk. There was also no evidence of a protective effect of emollients for the development of parentally reported wheeze or doctor-diagnosed asthma or allergic rhinitis -perhaps now best considered as co-morbidities rather than sequential development of similar diseases.

| Interpretation in context with other studies
Our findings are consistent with another large clinical trial 34 and with the recent individual patient data (IPD) meta-analysis of emollient prevention studies. 35 The IPD included 10 trials of 5154 participants and showed that skincare interventions did not change the risk of AD by the age 1-3 years (RR 1.03, 95% CI 0.81 to 1.31; I2 = 41%; moderate certainty; 3075 participants, 7 trials).
One single-centre study 36 has reported a 30% reduction in AD at 12 months following early initiation of daily specialized emollient use until 2 months of age. Other studies using more sophisticated emollients containing ceramides have not shown any benefit for AD prevention. 35 Not all emollients are the same in terms of their effects on the skin barrier. 37 It is still possible that some emollients could reduce or delay AD development as the role of epithelial barrier disruption in the development of allergic disorders is quite convincing. 38 Perhaps barrier enhancement would work in a low-risk rather than high-risk population or perhaps only when combined with enhanced skin care such as reduced bathing and soap avoidance, but the evidence for benefit so far has been disappointing.
The alternative conclusion is that emollient application in early life does not work in terms of preventing AD and that the strongest influences on AD development in high-risk children are genetic and in utero programming. Although data on food allergy from BEEP is inconclusive, data from the Enquiring About Tolerance (EAT) trial showed a significant dose-response relationship between parentreported moisturization frequency at 3 months of age and the subsequent development of food allergy raising the possibility that that regular application of moisturizers in early life could paradoxically promote food allergy development through transcutaneous sensitization. 39 Note: Adjusted relative risk/difference in risk estimated using a mixed effects logistic regression model using all available outcome data (including time points prior to 3 years) adjusting for randomization stratification variables and including a random effect for participants. The number of participants and observations included in each analysis model are shown in Table S1.

| Strengths and limitations of this study
that a parent report of AD in their child after 2 years would vary according to their allocation status. Questionnaire completion in the control group was very slightly higher and it is unclear whether this was due to chance or some other factor. Non-responders to longterm follow-up differed slightly from responders as listed above, but sensitivity analyses assuming non-responders were more likely to

| Implications for research
Since application of simple emollients does not appear to prevent AD or associated atopic conditions in high-risk families, we suggest that the value of further emollient prevention studies needs to be carefully considered, with a priority given to novel approaches to in-  Note: Analysis used multiple imputation for missing outcomes and included all randomized participants (693 emollient and 701 control). See Supporting Information for further details of the multiple imputation model. Adjusted relative risk/difference in risk estimated in each imputed dataset for food allergy, asthma and allergic rhinitis outcomes using a mixed effects logistic regression model adjusting for randomization stratification variables (using fixed effect for of number of immediate family members with atopic disease and a random effect for the recruiting centre) and for the AD outcome, due to convergence problems with the mixed effects logistics regression models in some of the imputed datasets, using generalized estimating equations with the Binomial family and log/identity link respectively, with an exchangeable correlation matrix to account for randomization being stratified by centre and number of immediate family members with atopic disease (1, 2, or more than 2) included as a covariate. Estimates were combined using Rubin's rules. a Outcome derived from responses to questionnaires at 12 (food allergy only), 18 (AD only), 24, 36, 48 and 60 months.
increased risk of food allergy. 39 Concerns regarding increased skin infections and food allergy are therefore both additional reasons why emollient use for AD prevention should not be recommended.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request. Access to the data will be subject to review of a data sharing and use request (available from ctu@nottingham.ac.uk) by a committee including the CI and sponsor, and will only be granted upon receipt of a data sharing and use agreement. Any data shared will be pseudoanonymised which may impact on the reproducibility of published analyses. The study protocol, statistical analysis plan and health economics analysis plan are available on the trial website: https://www.nottingham.ac.uk/ research/groups/cebd/projects/1eczema/beep-maintrial.aspx.

TR I A L R EG I S TR ATI O N
ISRCTN21528841.