Chen et al.

Trials (2017) 18:231

DOI 10.1186/s13063-017-1976-6

UPDATE Open Access

The role of anti-IgE (omalizumab/Xolair) in

the management of severe recalcitrant
paediatric atopic eczema (ADAPT):
statistical analysis plan
Tao Chen1, Susan Chan2,3, Gideon Lack3, Suzie Cro4 and Victoria R. Cornelius4*

Abstract
Background: The Atopic Dermatitis Anti-IgE Paediatric Trial (ADAPT) is a trial to determine the clinical efficacy and
safety of omalizumab for children with severe atopic eczema. This article describes the detailed statistical analysis
plan for the ADAPT as an update to the published protocol and is submitted prior to knowing all outcomes.
Method and design: The ADAPT is a randomised, double-blind, placebo-controlled trial with a primary objective to
determine whether anti-IgE reduces eczema severity as assessed by the validated eczema score (objective SCORAD)
after 24 weeks of treatment in children with severe eczema. This articles outline the overall analysis principles
including considerations on sample definition in each analysis, missing data, and adjusted covariates. Comparability
and representativeness of the randomised groups, primary and sensitivity analyses of the primary and secondary
outcomes as well as subgroup analysis are described.
Results: This prespecified statistical analysis plan has been developed to comply with international guidelines which
will increase the transparency of the data analysis for the ADAPT.
Trial registration: ISRCTN, identifier: ISRCTN15090567. Registered on 3 December 2014;
EU Clinical Trials Register, EudraCT Number: 2010-020841-29. Registered on 14 May 2010. The first participant was
enrolled on 15 January 2015.
Keyword: Statistical analysis plan, Eczema, Paediatric, Atopic dermatitis, Anti-IgE, Omalizumab, Randomised controlled
trial, Xolair

Background safety of omalizumab (Xolair, Novartis) for children with

Eczema is a chronic inflammatory skin disorder with a life- severe atopic eczema. Fuller details on the rationale and
time risk of up to 22% of children by the age of 12–14 years. the design for the trial are given in the study protocol
The available literature suggests that anti-immunoglobulin E [5]; the prespecified statistical analysis plan (SAP) has
(anti-IgE) may be of benefit in the treatment of eczema from been developed and finalised without knowing the treat-
at least the age of 7 years [1–4]. Studies and case reports to ment allocation or treatment-related study results.
date have had small numbers of participants, have not been This paper describes important features of the trial de-
randomised or placebo controlled, or have included a sign and the statistical method and procedures which
heterogenous mix of participants of different ages. need to be adhered to and performed by the statistician
The Atopic Dermatitis Anti-IgE Paediatric Trial responsible for this study. It should be read in conjunc-
(ADAPT) is a trial to determine the clinical efficacy and tion with the protocol.

* Correspondence: v.cornelius@imperial.ac.uk Method and design

4
Imperial Clinical Trials Unit, School of Public Health, Imperial College
London, Stadium House, 68 Wood Lane, London W12 7RH, UK The ADAPT is a randomised, double-blind, placebo-
Full list of author information is available at the end of the article controlled trial of anti-IgE therapy in children with
© The Author(s). 2017 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0
International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and
reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to
the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver
(http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.
Chen et al. Trials (2017) 18:231 Page 2 of 8

severe eczema who have failed topical therapy. In total, Minimum Clinically Important Difference (MCID)
62 children aged 4–19 years are planned to be recruited The study was powered to detect a minimum important
within 18 months. Participants are eligible if they have treatment effect of a 13.5-point absolute change in ob-
severe eczema defined as an objective SCORing Atopic jective SCORAD score taking into account the patient
Dermatitis (SCORAD) score of over 40 at assessment burden and high treatment cost. The MCID is the smal-
(detailed exclusion and inclusion criteria are in the pub- lest difference in an outcome measure that represents a
lished protocol). Participants will be individually rando- clinically relevant outcome to the patient, regardless of
mised in a 1:1 ratio to two treatment arms (omalizumab cost and burden. There is no verified MCID for
and matched placebo) using minimisation to ensure the SCORAD score in this severely affected paediatric popu-
balance of total IgE (≤1500 and >1500) and age (<12 and lation. In order to determine MCID, published studies
≥12 years). The allocation will be performed by an on- have recommended the use of both anchor- and
line randomisation system hosted at the King’s Clinical distribution-based methods [7]. A study by Schram et al.
Trials Unit (CTU). [8], which adopts an anchor-based approach, suggests
that a MCID for the objective SCORAD score is 8.2.
However, this is based on data from three RCTs on
Primary objective treatments for atopic eczema which included adults. The
The primary objective of this study is to assess whether MCID reported by Schram et al. for children only, based
omalizumab will reduce eczema severity as assessed by on a subsample of n = 25, with an average age of
the validated eczema score (SCORAD) after 24 weeks of 9.4 years, is 9.0. Since the patients included in the study
treatment in children with severe eczema. by Schram et al. also had a milder baseline severity we
employed a distribution-based method using data col-
Secondary objectives lected from the trial to calculate a MCID. Using the data
The study will examine the influence of the study inter- from the first 47 ADAPT patients who completed week-
vention on the rate of treatment failure, rate of alterna- 24 assessments (75% of total sample size) adopting 0.7
tive systemic therapy, quality of life, eczema severity as SD of the change in score from baseline gives a MCID
assessed by the Eczema Area and Severity Index (EASI), of 8.5. These MCIDs will be used to guide interpretation
effect on co-existing allergic disease, number of eczema of the results from the primary analysis.
exacerbations, infective episodes of eczema, change in
reactivity to food and aeroallergens and change in General statistical principles
allergen-specific IgE. Detailed descriptions of the pri- The principle of intention-to-treat (ITT) will be the
mary and secondary outcomes can be found in Table 1 main strategy of the analysis adopted for the primary
as well as in the study protocol [5]. outcome and all secondary outcomes. That is, all ran-
domised participants will be analysed in the group ran-
domised regardless of whether the allocated study
Sample size treatments were received, or whether other interventions
Omalizumab is administered via subcutaneous injec- were received and regardless of any protocol deviations or
tions which require fortnightly or monthly attendance violations [9]. A safety set (SS) population will consist of
at clinic to receive them. It is available from the man- participants who receive at least one dose of allocated
ufacturers at an undisclosed cost in the United King- treatment, regardless of their eligibility for the study. The
dom. In order for omalizumab to be adopted into harm analyses will compare the harm outcomes between
practice, a treatment effect that would make an im- the two treatment groups in the safety population.
portant impact on the children’s quality of life would All regression analyses will include the minimisation
be required. Through discussion and consultation variables (IgE (≤1500, >1500) and age (<10 or ≥10 years)
with the funder and clinicians, a relative reduction of as covariates. This is because adjustment for these strati-
around 33% in symptoms was selected to be the fication factors in the randomisation process will main-
minimum important treatment effect to detect. Given tain correct type I error rates [10]. Additionally, for
the inclusion criteria the mean baseline SCORAD continuous outcomes, the outcomes measured at base-
score is anticipated to be 45 and we aim to detect a line will be included in regression analysis to increase
change in SCORAD score of 13.5 points between the power [11].
treatment arms. Based upon a study by Hindley [6] Any examination of subgroups, not specifically identi-
we assume that the standard deviation (SD) is 15, fied in the SAP, will be considered exploratory in nature
using a significance level of 5% with 90% power, and and will be clearly identified. All p values will be two-
including a 15% dropout rate we aim to recruit 62 sided and the significance level is set at 5% unless other-
participants (31 each to each arm). wise stated.
Chen et al. Trials (2017) 18:231 Page 3 of 8

Table 1 Outcomes and analysis models

Outcome Endpoints Category Analysis model
Primary
Objective SCORAD Difference in the objective SCORAD in both Continuous outcome Mixed-effects linear regressions
groups after 24 weeks of treatment measures (primary)
Instrumental variable regression
Secondary
Treatment failure Participants who have persistent severe Binary outcome Logistic regression model
eczema despite 2 courses of rescue therapy measures
(0.5 to 1 mg/kg/day of orally administered
prednisolone for a week at a maximum dose
of 40 mg/day, followed by a week at 50% of
this dose)
Alternative systemic therapy Requirement for alternative systemic therapy Binary outcome Logistic regression model
measures
Eczema quality of life • POEM Continuous outcome ANCOVA
• (C)DLQI measures
Eczema severity • Subjective SCORAD Continuous outcome ANCOVA
• EASI score measures
Effect on co-existing allergic PADQLQ Continuous outcome ANCOVA
disease measures
Number of eczema exacerbationsc • Clinician-diagnosed exacerbation of eczema Numerical outcome Poisson regression
or measures Negative binomial regression
• Increase on SCORAD by 15 points from last model
recorded SCORAD with participant/parent Zero-inflated Poisson regression
perception of worsening eczema model (as appropriate)
Infective episodes of eczemac Clinician-diagnosed and -treated infective Numerical outcome Poisson regression
episode of eczema, or clinically apparent, measures Negative binomial regression
culture-positive infective exacerbations model
Zero-inflated Poisson regression
model(as appropriate)
Allergen-specific IgEa Change in allergen-specific IgE Continuous outcome ANCOVA
measures
Reactivity to food and aeroallergensa Change in skin-prick test reactivity to food Numerical outcome Poisson regression
and aeroallergens measures Negative binomial regression
model
Zero-inflated Poisson regression
model
Safety
Adverse eventsb Spontaneously reported AE will be collected Binary outcome Descriptive analysis
throughout the follow-up period
Urea and electrolytes, creatinine, Surveillance tests where abnormal ranges are Binary outcome Descriptive analysis
FBC, eosinophils, LFT, IgE, vitamin D, defined using the ranges specified by the
iron level, bone profile processing laboratory
a
Only collected at screening and 24 weeks of treatment. The remaining outcomes are collected at baseline, 4-weekly during the 24 weeks of treatment, 36 weeks
and 48 weeks
b
Blood test and urine samples will be collected at baseline, 24 weeks, 36 weeks and 48 weeks. Clinical observations will be examined at every visit
c
Chi‐square goodness‐of‐fit tests will be used to select the suitable model
ANCOVA analysis of covariance, IgE immunoglobulin E, SCORAD SCORing Atopic Dermatitis, PADQLQ Paediatric Allergic Disease Quality of Life Questionnaire,
(C)DLQI (Children’s) Dermatology Life Quality Index, EASI Eczema Area and Severity Index, AE adverse events, POEM Patient-oriented Eczema Measure, FBC full
blood count, LFT liver function test

Missing data assumption). Sensitivity analysis will explore departures

The number and proportion of participants missing ob- from the main MAR analysis assumption for all patients
jective SCORAD values by visit number will be tabu- on the primary outcome using a pattern-mixture mul-
lated. The primary analysis includes all observed data tiple imputation (MI) approach [12].
and assumes the probability that missing data is not Imputation under MAR will initially be performed
dependent on the values of the unobserved data itself, separately within each arm following the guidance sug-
conditional on the observed values of the variables in- gested by White et al. [13]. The variables in the imput-
cluded in the analysis model (missing at random (MAR) ation model will be the same as those in the analysis
Chen et al. Trials (2017) 18:231 Page 4 of 8

model without including more auxiliary variables (e.g. be calculated from the non-missing values for the base-
predictors of missingness) after taking into account the line variable using pooled data from both treatment
relatively small sample size of this study [14]. Imputa- groups [15]. With reference to those categorical vari-
tions will then be modified to reflect departures from ables, the imputed mean will be rounded up to nearest
the MAR assumption. category level. This is justifiable because randomisation
We will investigate the impact of a better or poorer re- ensures that baseline scores are independent of treat-
sponse than that predicted by MAR (lower/higher ob- ment group and imputation keeps the statistical effi-
jective SCORAD scores) for patients with missing data. ciency in the estimation of the treatment effect.
Specifically, we define δ as the postulated mean differ- For those missing items within questionnaires, we
ence in the rate of change of the objective SCORAD firstly use the missing value guidance provided for
score between the observed and unobserved cases over questionnaires. If no guidance is provided we will then
24 weeks, conditional on the variables in the imputation impute the missing values using the mean of the ob-
model. For each patient we then modify the MAR im- served items within the same subscale if 20% or fewer
puted observations accordingly by δ. Imputed data sets items are missing. The scale score will be calculated
will be analysed using the primary analysis model. Re- based on the complete values and these replacements
sults will be combined across imputed data sets using [16]. If more than 20% of items are missing in the ques-
Rubin’s rules. We will repeat the analysis for a range of tionnaire, multiple imputation will be used as discussed
δs corresponding to ±10, 20, 30, 40 and 50% of the rate above.
of change of the objective SCORAD score observed over
24 weeks in all patients. We will also consider the possi- Statistical analysis
bility that data is missing informatively in one arm only Trial profile
and employ the outlined imputation approach separately A Consolidated Standards of Reporting Trials (CON-
by trial arm. SORT) flow chart will be constructed (see Fig. 1). This
For baseline covariates, the amount of missing data is will include the number of eligible patients, the number
expected to be small. However, if this happens, in case of patients agreeing to enter the trial, the number of par-
of loss of power using observed data, mean values will ticipants withdrawing and lost to follow-up, the number

Fig. 1 Consolidated Standards of Reporting Trials (CONSORT) trial flow chart for the Atopic Dermatitis Anti-IgE Paediatric Trial (ADAPT)
Chen et al. Trials (2017) 18:231 Page 5 of 8

continuing through the trial, and the number included (a) Y ij ¼ β0 þ β1 T RT i þ β2 SCORAD0i þ β3 IgE i þ β4 Agei
in the analyses.
þβ5 t 12 þ β6 t 16 þ β7 t 20 þ β8 t 24 þ β9 t 12
Comparability/representativeness of randomised groups T RT i þ β10 t 16  T RT i þ β11 t 20  T RT i
All baseline descriptive variables of participants will be þβ12 t 24  T RT i þ bi þ eij
summarised by treatment arm. Continuous data will be
expressed as N/mean/SD/min/Q1 (lower quartile)/me- (b) Y ij ¼ β þ β T RT i þ β SCORAD0 þ β IgE i þ β Agei
0 1 2 i 3 4
dian/Q3 (upper quartile)/max. Tabulations of frequen-
cies for categorical data will include all possible þβ5 t 12 þ β6 t 16 þ β7 t 20 þ β8 t 24 þ β9 t 12
categories and will display the number of observations T RT i þ β10 t 16  T RT i þ β11 t 20  T RT i
in a category as well as the percentage (%) relative to
þβ12 t 24  T RT i þ b1i þ b2i t j þ eij ;
number of available values within the respective treat-
ment group unless otherwise specified. The number of
where j = time points (8, 12, 16, 20 and 24 weeks), i =
missing values is reported for both types of variables.
participants,
This will allow a visual assessment of whether the ran-
TRTi: dummy variable (TRTi = 0 or 1) of patient i,
domisation procedure succeeded in producing compar-
IgEi: dummy variable for IgE (=0 or 1) of patient i,
able arms, and tests of statistical significance will not be
SCORAD0i : baseline SCORAD score of patient i,
undertaken between arms at baseline; rather the clinical
Agei: dummy variable for age (<10 or ≥10 years) of
importance of any imbalance will be noted.
patient i,
The number of participants who receive the injection
txx: dummy variable for time (= 0 or 1) at time point
outside the planned visit window of 5 days or more will
xx weeks.
be reported by visit number and treatment arm. Also,
Where bi and b1i are random intercepts, b2i is random
the mean cumulative dosage by planned dose will be
slopes, both eij and b1i, following normal distributions.
plotted by treatment arm and separately for those receiv-
An unstructured covariance matrix will be used. Models
ing monthly and fortnightly injections.
will be fitted using residual maximum likelihood
(REML). The estimated treatment effect at 24 weeks, β1
Descriptive statistics for outcomes
+ β12, will be reported with 95% confidence intervals and
The distributions of all efficacy outcomes (in Table 1)
corresponding p value.
will be presented in histograms (continuous/count) or
Model (a) will be the primary analysis model unless
bar charts (ordinal/binary) both overall and by group at
there is strong evidence for mis-specification of the
each assessment point. A single table will be outputted
model. The random slope model is less restrictive and
with summary statistics for all outcomes by group and
possibly more realistic in its assumptions, i.e. the object-
visit point. Furthermore, summary statistics will be plot-
ive SCORAD score trajectories for each individual start-
ted by line graphs for each outcome across time by
ing from a different level and following a different trend
intervention. Only participants with a completely re-
with a different slope. The primary interest is in deter-
corded outcome will be used to calculate the summary
mining whether β1 + β12 is significant and whether this
measures.
varies between the two models (a and b).
The conclusion of the trial will be based on this
Analysis of primary efficacy outcome
analyses.
Primary analysis
A linear mixed model will be used to obtain an estimate
for the mean difference in objective SCORAD scores be- Planned sensitivity analyses
tween the two treatment groups. Participant will be in- To investigate the robustness of the results of the pri-
cluded as a random intercept (investigating adding a mary analysis we will undertake a number of sensitivity
random slope on time), and time (investigating the pos- analysis:
sibility of linearising this effect across 8, 12, 16, 20 and
24 weeks), time-by-group interaction, baseline objective 1. Subsequent adjustment for cumulative use of potent
SCORAD score, IgE (≤1500, >1500) and age (<10 or topical steroids (continuous variable) at 24 weeks,
≥10 years) as fixed effect. An overall treatment effect for alternative systemic therapy (yes/no) at 24 weeks,
objective SCORAD score at 24 weeks will be estimated. rescue medication (yes/no) at 24 weeks based on the
The response yij is the objective SCORAD score meas- primary model
urement for patient i at time tj. Both random intercept 2. An analysis of Complier Average Causal Effect
model (a) and random intercept and slope model (b) will (CACE) by a two-stage least squares instrumental
be fitted as specified below: variable regression would be performed for the
Chen et al. Trials (2017) 18:231 Page 6 of 8

primary endpoint as analysis based upon ITT may reaction, serious AE, serious adverse reaction or unex-
underestimated the effect of actually receiving the pected serious adverse reactions). These will be sum-
treatment [17]. Here, we defined ‘compliers’ as those marised over the 48-week follow-up period and, where
who complete more than 50% of injections (that is appropriate, by time of occurrence.. The numerator will
injections received relative to injections planned for indicate the number of affected participants at each time
the 24-week study period in groups). Randomisation point from the SS population. The denominators will
will be used as an instrumental variable for treatment show how many participants were in the trial at the cor-
received with the same covariates in primary analysis responding time point. If appropriate, the difference in
models. proportion (95% confidence interval) will be estimated
and time-to-event curves by treatment arm will be plot-
Analysis of secondary efficacy outcomes ted. All AE will be listed individually.
All analyses for secondary efficacy outcomes will be
based on the ITT population and defined at week 24 un- Subgroup analyses
less specified otherwise. The missing data will be tackled A subgroup analysis is planned to investigate whether
according to the strategies mentioned above. intervention effects differ between adherence, defined as
For each secondary outcome, we will adjust for the the injections received relative to the injections planned for
minimisation variables IgE (≤1500, >1500), age (<10 or the 24-week study period in groups (≤50%, >50%; ≤75%,
≥10 years) and baseline data (as appropriate). >75%; ≤90%, >90%). All subgroup analyses will be analysed
Treatment failure (binary) and alternative systemic using the same method as for the primary outcome. The
therapy (binary) will be analysed using a logistic regres- results will be displayed by means of a forest plot.
sion model. The estimated treatment effect (odds ratio)
will be reported with 95% confidence intervals and cor- Software
responding p value. Subjective SCORAD, EASI, Patient- Data management: an online data collection system for
Oriented Eczema Measure (POEM), Paediatric Allergic clinical trials (MACRO; InferMed Ltd.) will be used.
Disease Quality of Life Questionnaire (PADQLQ), (Chil- This is hosted on a dedicated server at Kings’ Clinical
dren’s) Dermatology Life Quality Index ((C)DLQI) scores Trial Unit. The CTU data manager will extract data
and allergen IgE levels will be analysis using analysis of periodically as needed and provide these in comma sepa-
covariance (ANCOVA). The estimated treatment effect rated (.csv) format.
(mean difference) will be reported with 95% confidence Statistical analysis: analysis will be performed using
intervals and corresponding p value. The number of statistical software Stata, R or SAS.
skin-prick test reactivities, infective episodes of eczema
count, and number of eczema exacerbations will be ana- Tables and figures
lysed by Poisson regression, Negative binomial regres- The SAP describes the conventions to be used for pre-
sion or zero-inflated Poisson regression models after senting results in text and in tables and figures. Those
checking the distribution of the dependent variable by conventions are based on the International Conference
Pearson chi‐square goodness‐of‐fit tests will ensure the on Harmonisation (ICH) guideline for reporting clinical
selection of the correct statistical model. The estimated trial results. The planned tables are:
treatment effect (odds ratio) will be reported with 95%
confidence intervals and corresponding p value. A sum-  Analysis population by study centre and treatment
mary of models for each of the outcomes can be found group
in Table 1.  Withdrawals, protocol deviations and violations by
treatment group
Analysis of safety outcomes  Baseline demographic and clinical characteristics by
Information on adverse events (AE) will be collected by treatment group
means of spontaneous reports from patients and carers,  Baseline blood/urine investigations by treatment
clinical observation and clinical examinations and blood group
tests. Adverse events will be coded using terms chosen  Descriptive analysis for primary efficacy outcomes
by the clinical investigators with reference to the Med- by treatment group across study visits
ical Dictionary for Regulatory Activities (MedDRA) at  Inferential analysis for primary efficacy outcomes by
the Preferred Term level. Abnormal ranges for blood treatment group
tests will be defined using the ranges specified by the la-  Sensitivity analysis for primary efficacy outcomes by
boratory processing the sample. treatment group
Adverse events will be tabulated overall by severity  Secondary efficacy outcomes by treatment group
and type (AE, adverse reaction, unexpected adverse  AE by treatment group across study visits
Chen et al. Trials (2017) 18:231 Page 7 of 8

The planned figures are: Authors’ information

Not applicable

 Flow chart of participants through the study Competing interests

 Bar chart/histogram for efficacy outcome over time The authors declare that they have no competing interests.
by overall and treatment group
Consent for publication
 Linear graph for efficacy outcome over time by Not applicable
treatment group
 Forest plot of effect of treatment on primary for all Ethics approval and consent to participate
Ethical approval has been agreed and is being overseen by London
participants and for prespecified subgroups Westminster NHS Research Ethics Committee (REC reference number: 11/
LO/0123). We will obtain informed consent from all participants in the
This article presents the SAP for the ADAPT and study.

should be read in conjunction with the trial protocol [5].

Publisher’s Note
Abbreviations Springer Nature remains neutral with regard to jurisdictional claims in
(C)DLQI: (Children’s) Dermatology Life Quality Index; AE: Adverse events; published maps and institutional affiliations.
ANCOVA: Analysis of covariance; CACE: Complier Average Causal Effect;
CTU: Clinical Trials Unit; EASI: Eczema Area and Severity Index; ITT: Intention- Author details
1
to-treat; MAR: Missing at random; PADQLQ: Paediatric Allergic Disease Department of Clinical Sciences, Liverpool School of Tropical Medicine,
Quality of Life Questionnaire; POEM: Patient-oriented Eczema Measure; Liverpool L3 5QA, UK. 2Guy’s and St. Thomas’ NHS Foundation Trust,
SAP: Statistical analysis plan; SCORAD: SCORing Atopic Dermatitis; Westminster Bridge Road, London SE1 7EH, UK. 3King’s College London,
SD: Standard deviation; SS: Safety set King’s Health Partners, Asthma-UK Centre in Allergic Mechanisms of Asthma,
Allergy and Respiratory Science, Guy’s Hospital, London SE1 9RT, UK.
4
Imperial Clinical Trials Unit, School of Public Health, Imperial College
Acknowledgements London, Stadium House, 68 Wood Lane, London W12 7RH, UK.
The authors wish to acknowledge the NIHR EME programme for funding the
study (NIHR EME 11/14/24 ADAPT), Novartis for supplying the study drug Received: 28 September 2016 Accepted: 4 May 2017
and King’s Clinical Trials Unit for providing support in the design and
conduct of the trial (Caroline Murphy, Joanna Kelly) and the trial manager
(Rahi Jahan). The Data Monitoring Committee (Professor Graham Roberts, References
Dr. Ahmed Massoud, Ms. Helen Moyses) and Trial Steering Committee 1. Lane JE, Cheyney JM, Lane TN, Kent DE, Cohen DJ. Treatment of
(Dr. Nerys Roberts (chair), Dr. Mike Coren, Dr. Helen Cox, Professor John recalcitrant atopic dermatitis with omalizumab. J Am Acad Dermatol.
Harper, Mrs. Kathryn Humphreys, Dr. Susan Chan, Dr. Emma Wedgeworth, 2006;54(1):68–72.
Professor Gideon Lack, Dr. Victoria Cornelius) for their advice and contributions 2. Belloni B, Ziai M, Lim A, Lemercier B, Sbornik M, Weidinger S, Andres C,
to the protocol and statistical analysis plan. We wish to thank Dr. Kun Liu for his Schnopp C, Ring J, Hein R, Ollert M, Mempel M. Low-dose anti-IgE therapy
important contribution to the development of the statistical methods. in patients with atopic eczema with high serum IgE levels. J Allergy Clin
The study was designed and delivered in collaboration with the UK Clinical Immunol. 2007;120:1223–5.
Research Collaboration (UKCRC)-registered King’s College London Clinical 3. Vigo PG, Girgis KR, Pfuetze BL, Critchlow ME, Fisher J, Hussain I. Efficacy of
Trials Unit. The study is supported by the NIHR Biomedical Research Centre anti-IgE therapy in patients with atopic dermatitis. J Am Acad Dermatol.
which is based at Guy’s and St. Thomas’ NHS Foundation Trust and King’s 2006;55:168–70.
College London, and by a National Institute for Health Research comprehensive 4. Sheinkopf LE, Rafi AW, Do LT, Katz RM, Klaustermeyer WB. Efficacy of
Biomedical Research Centre award to Guy’s and St. Thomas’ NHS Foundation omalizumab in the treatment of atopic dermatitis: a pilot study. Allergy
Trust and King’s College London. Asthma Proc. 2008;29:530–7.
5. Chan S, Cornelius VR, Chen T, Raulovic S, Wan M, Jahan R, Lack G. Atopic
Funding Dermatitis Anti-IgE Paediatric Trial (ADAPT): the role of anti-IgE in severe
This investigator-led trial is funded by the National Institute for Health paediatric eczema: study protocol for a randomized controlled trial. Trials.
Research Efficacy and Efficacy and Mechanism Evaluation Programme (NIHR 2017;16:136.
EME 11/14/24) and the Guy’s and St. Thomas’ Charity. The study drug and 6. Hindley D, Galloway G, Murray J, Gardener L. A randomised study of ‘wet
placebo were supplied by Novartis. Novartis had no role in the design of the wraps’ versus conventional treatment for atopic eczema. Arch Dis Child.
study or writing of the protocol or statistical analysis plan. 2006;91:164–8.
7. Howard R, et al. Determining the minimum clinically important differences for
outcomes in the DOMINO trial. Int J Geriatr Psychiatry. 2011;26(8):812–7.
Availability of data and materials 8. Schram ME, Spuls PI, Leeflang MM, Lindeboom R, Bos JD, Schmitt J. EASI,
The protocol and statistical analysis plan can be obtained by contacting the (objective) SCORAD and POEM for atopic eczema: responsiveness and
corresponding author. The study team will retain the exclusive use of data minimal clinically important difference. Allergy. 2012;67(1):99–106.
until publication of major outputs has been completed. Funding will be 9. White IR, Horton NJ, Carpenter J, Pocock SJ. Strategy for intention to treat
sought to make the data publically available. analysis in randomised trials with missing outcome data. BMJ. 2011;342:d40.
10. Kahan BC, Morris TP. Reporting and analysis of trials using stratified
Authors’ contributions randomisation in leading medical journals: review and reanalysis. The BMJ.
TC participated in development of the trial protocol and statistical analysis 2012;345:e5840.
plan and was responsible for preparing the manuscript. VC obtained grant 11. Turner EL, Perel P, Clayton T, Edwards P, Hernandez AV, Roberts I, Shakur H,
funding, participated in development of the study design, trial protocol, Steyerberg EW, CRASH trial collaborators. Covariate adjustment increased
statistical analysis plan and manuscript. SChan is the chief investigator, power in randomized controlled trials: an example in traumatic brain injury.
obtained grant funding, contributed to the study design and protocol and J Clin Epidemiol. 2012;65:474–81.
reviewed the manuscript. GL conceived the study, obtained grant funding, 12. Carpenter J, Kenward M. Missing data in randomised controlled trials—a
contributed to the study design and protocol and reviewed the manuscript. practical guide. Birmingham: National Institute for Health Research. 2008;
SCro participated in development of the trial statistical analysis plan, and vol. Publication RM03/JH17/MK.
contributed to, and reviewed, the manuscript. All authors read and approved 13. White IR, Royston P, Wood AM. Multiple imputation using chained
the final manuscript. equations: issues and guidance for practice. Stat Med. 2011;30:377–99.
Chen et al. Trials (2017) 18:231 Page 8 of 8

14. Hardt J, Herke M, Leonhart R. Auxiliary variables in multiple imputation in

regression with missing X: a warning against including too many in small
sample research. BMC Med Res Methodol. 2012;12:184.
15. White IR, Thompson SG. Adjusting for partially missing baseline
measurements in randomized trials. Stat Med. 2005;24:993–1007.
16. Fayers PM, Curran D, Machin D. Incomplete quality of life data in
randomized trials: missing items. Stat Med. 1998;17:679–96.
17. Sussman JB, Hayward RA. An IV for the RCT: using instrumental variables to
adjust for treatment contamination in randomised controlled trials. The
BMJ. 2010;340:c2073.

Submit your next manuscript to BioMed Central

and we will help you at every step:
• We accept pre-submission inquiries
• Our selector tool helps you to find the most relevant journal
• We provide round the clock customer support
• Convenient online submission
• Thorough peer review
• Inclusion in PubMed and all major indexing services
• Maximum visibility for your research

Submit your manuscript at

www.biomedcentral.com/submit