Randomised controlled trial of silk therapeutic garments for the management of atopic eczema in children: the CLOTHES trial

Primary objective

• To assess whether or not the addition of silk therapeutic garments to standard AE care reduces AE severity in children with moderate to severe disease over a period of 6 months.

Secondary objectives

• To estimate the within-trial cost-effectiveness of silk therapeutic garments from a NHS and wider (family and employer) perspective.

• To explore parent/guardian and child views on and experiences of using silk garments and factors that might influence the use of these garments in everyday life.

• To examine prescribers’ and commissioners’ views on the use of silk garments for the management of AE.

Informed consent

Written informed consent was obtained from the parent/guardian of each participant at the baseline visit, prior to any trial procedures being carried out. In addition, assent was provided by the children if they wanted to. Consent to take part in the genetic study (FLG genotyping), and for samples to be stored and used for potential future research, was included as optional.

Silk garments

The silk garments used in the study are licensed as a medical device with a Conformité Européenne (CE) mark for use in AE, denoting that they comply with European Union legislation and safety requirements. They are 100% silk garments made from antimicrobially protected, knitted, sericin-free silk. Sericin is removed from the silk fibres during manufacturing because it is a protein that coats the outside of the fibres and has the potential to cause allergic reactions.

Two products were chosen for inclusion in the trial (DermaSilk and DreamSkin), as these were the two brands available on prescription in the UK at the time of trial design. Distribution of the intervention to participants was handled from the co-ordinating centre, where a stock of garments across a range of sizes in both brands was maintained. Participants received three sets of garments (long-sleeved vest and leggings or long-sleeved body suits and leggings, depending on the age of the child) and were instructed to wear the garments as often as possible during the day and at night, either as underwear or as pyjamas (Figure 2). Three sets were provided to allow for the washing and rotation of garments. The child’s height at randomisation was used to determine the correct size of garments, which were posted out to participants as soon as possible after randomisation. On receipt of the garments, participants were instructed to try on one set to check that they fitted correctly and then confirm this with the co-ordinating centre. Standardised care instructions were provided on a paper insert included in the garment package (Box 1) and instructions were also replicated in the participant diary.

FIGURE 2.

Garments being worn.

Garments were replaced as required during the 6-month RCT (if they were worn out, were lost or no longer fitted the child). If replacement garments were required, the participants returned the worn garments to the co-ordinating centre with a completed garment request form and new garments were sent out. After the 6-month RCT period was over, garments were not replaced.

Standard care

All participants continued with their standard AE care in line with National Institute for Health and Care Excellence (NICE) guidance,5 including regular emollient use, avoidance of irritants and topical corticosteroids (or calcineurin inhibitors) for controlling inflammation. Participants were asked not to change their standard AE treatment for the duration of the trial unless medically warranted.

Participants who frequently used wet- or dry-wrap dressings for their AE were excluded, but occasional use of wet or dry wraps was monitored but not prohibited.

If a research nurse suspected that the AE had become infected, participants were advised to contact their normal medical team for confirmation of diagnosis and subsequent treatment.

Primary outcome

The primary outcome of AE severity measured using the objective Eczema Area and Severity Index (EASI)31 was assessed at baseline and at 2, 4 and 6 months. Baseline EASI score was used as a covariate in the analysis model. EASI score was assessed by trained research nurses who were blinded to treatment allocation. EASI score was chosen as the primary outcome as it is a validated scale recommended as the core outcome instrument for AE signs. 32 EASI involves an evaluation of four AE signs [erythema (redness), excoriation (scratching), oedema/papulation (swelling and fluid in the skin) and lichenification (thickening of the skin)] and an assessment of percentage area affected by eczema in four body regions (head and neck, upper limbs, trunk and lower limbs). EASI score ranges from 0 to 72, with higher scores representing more severe disease.

All research nurses received training in the use of EASI (using standardised training photographs and assessment of patients with AE by two independent assessors until concordance was reached). Resources were provided to assist in assessing the signs and body surface area (see Appendix 7).

Participants were assessed by the same research nurse at all time points in order to minimise interobserver variability.

Secondary outcomes

• Global assessment of AE by research nurses [Investigator Global Assessment (IGA)] and by participants [Patient Global Assessment (PGA)] at baseline and at 2, 4 and 6 months, using a six-point scale (clear, almost clear, mild, moderate, severe and very severe).

• Self-reported AE symptoms using the recommended core outcome instrument,33 the Patient Oriented Eczema Measure (POEM), which captures frequency of itch, sleep loss, bleeding, weeping/oozing, cracking, flaking and dryness. 34 It has a range from 0 to 28, with higher scores representing more severe disease. POEM scores were collected weekly using an online questionnaire for the first 6 months and once again at 8 months. Obtaining self-reported eczema severity every week for 6 months was used to capture long-term control of flares as well as self-reported eczema symptoms.

• Three-Item Severity (TIS) scale35 at baseline and at 2, 4 and 6 months, assessed by research nurses at a single representative body site (defined as the most bothersome patch of AE that was covered by the garments). The selected representative body site did not have to be the same at each visit. The TIS measures three clinical signs (erythema, oedema/papulation and excoriation) and the total score ranges from 0 to 9, with higher scores representing more severe disease. Given the importance of an objective measure to capture eczema severity in this observer-blind trial, it was felt that a second validated eczema severity scale was warranted.

• Use of AE treatments: number of days of use of topical steroids, topical calcineurin inhibitors, emollients and wet/dry wrapping, assessed weekly throughout the trial. At each visit, research nurses assessed change in AE treatment regimen and categorised as no change, neutral change, reduction or escalation.

• Health-related quality of life at baseline and at 6 months from the perspectives of the family [Dermatitis Family Impact (DFI)],36 the main carer [EuroQoL-5 Dimensions-3 Levels (EQ-5D-3L)]37 and the child [Atopic Dermatitis Quality of Life (ADQoL) preference-based index38 and Child Health Utility-9 Dimensions (CHU-9D)39 in those aged ≥ 5 years].

• Durability of the garments, adherence and acceptability of use (as assessed by children and parents/carers). Adherence was collected weekly, and information on durability and acceptability was captured at 6 and 8 months in the participant questionnaires. Sticker charts were provided for children to record how many days/nights the garments had been worn for the intervention group and how many days/nights they had been in the study for the standard care group (see Appendix 8). These were intended to help keep children engaged in the study and to assist in completing the adherence data in the weekly questionnaires.

• Health resource use for treatment of AE throughout the trial: health-care visits, inpatient stays, medications, tests, personal items for AE and time off work or school.

Safety outcomes

Skin infections requiring antibiotic or antiviral treatment self-reported by parents and serious adverse events related to AE (hospitalisation as a result of AE) were recorded.

Tertiary outcomes

Although it was assumed that the different brands of garments were similar, the effects of receiving different brands of garments were also explored. Another additional exploratory analysis was conducted based on AE severity scores in areas covered by the garments (body and limbs) compared with areas uncovered by the garments (head and neck). All tertiary analyses were considered exploratory.

Data collection

Trial data generated by all centres were entered by research nurses directly into a web-based MACRO database (MACRO 4 version 3800, Elsevier, London, UK), maintained by the Nottingham Clinical Trials Unit (CTU). Access to the trial database was controlled by user logins and research nurses could enter/edit data for their site only. Paper worksheets were provided for research nurses to record data during the clinic visit (see Appendix 9) and were transcribed after the visit. Participant questionnaires completed at clinic visits were transcribed by the research nurses into the trial database. Data entry was checked against the paper record for 100% of the primary outcome and for a 10% sample of all data.

Participants were provided with a diary booklet in which they were encouraged to record all health-care visits for eczema, eczema prescriptions, purchases for AE and time off work/school because of AE (see Appendix 10). The diary was reviewed by the research nurse at each clinic appointment and used as an aide memoir to complete the relevant sections of the trial database.

Missing and/or ambiguous data were queried with research nurses and resolved whenever possible.

Weekly questionnaires were completed by the participant online or in paper format (see Appendix 11) and sent to the Nottingham CTU for data entry on the bespoke in-house system. The preference for paper or online questionnaires was recorded at baseline. Participants completing online questionnaires were emailed a unique web link to the questionnaire each week on the day completion was due. A further reminder e-mail was sent at the beginning of day 3 if the questionnaire had not been completed. Links remained active until the end of day 3, after which time the week’s entry was classed as missing. Participants who failed to complete the weekly questionnaire for ≥ 3 weeks in a row were contacted by the Nottingham CTU and encouraged to complete the questionnaires.

For the week 24 (6-month) (see Appendix 12) and week 32 (8-month) (see Appendix 13) questionnaires, online submission remained open for 14 and 7 days, respectively, in order to ensure maximum data completion at the primary end point and end of trial. For these time points, non-responders were contacted by telephone and a paper copy of the questionnaire was sent by post if required.

Preliminary analyses

Descriptive statistics of demographic and clinical measures were used to examine balance between the randomised arms at baseline.

Primary outcome

The primary analysis used a multilevel model with observations at 2, 4 and 6 months, nested within participants, and included participants in whom EASI was assessed at least once at follow-up. The model assumed that missing EASI scores were missing at random given the observed data. The model used a random intercept and slope at the participant level, with an unstructured covariance matrix for these random effects. Diagnostic plots to check the normality of the residuals from the fixed part of the model, homogeneity of the variance of the residuals and the normality of the random effects when the model was initially fitted indicated that the assumptions for the multilevel model were not met. The score was therefore log-transformed for analysis and the effect of the trial garments is presented as a ratio of geometric means. 42,43 This ratio was back-transformed to the original EASI scale to facilitate interpretation of findings.

The effect of trial garments on AE severity changing over the study period was explored by including an interaction term between group and time point in the model. There was no evidence of a differential effect over time, so a single treatment effect has been reported that averages the treatment effect over all time points.

Sensitivity analyses for the primary outcome were performed:

• To adjust for variables that had an observed imbalance between the groups at baseline.

• Using multiple imputation (by chained equations) for missing outcome data.

• To explore the impact of adherence in wearing the garments on the primary outcome by estimating the complier average causal effect (CACE) at 6 months using instrumental variable regression methods. This analysis aims to provide an unbiased estimate of the treatment effect among compliers, defined as participants who would comply with their allocation regardless of the treatment arm to which they were randomised. Estimates are presented for two measures of compliance:

  1. binary compliance, defined as participants who wore the trial garments for at least 50% of the days or 50% of the nights

  2. continuous compliance, defined as each additional 10% of time that the garments were worn. This was calculated by summing the number of days and nights that the trial garments were reported to be worn, then dividing by the total number of days and nights in questionnaires completed about garment wear.

A planned subgroup analysis based on the presence or absence of loss-of-function mutations in FLG was conducted by adding an interaction term between allocated treatment and FLG genotype (none, one or two FLG null mutations) to the primary analysis model.

Secondary outcomes

The global assessment scores (IGA and PGA) were dichotomised into ‘clear, almost clear or mild AE’ versus ‘moderate, severe or very severe AE’, and analysed using generalised estimating equations to allow estimation of risk difference and risk ratio. The TIS score was analysed using the multilevel model framework as outlined above for the primary outcome (not transformed). For the global assessment scores and the TIS score, the effect of the trial garments changing over the study period (2-, 4- and 6-month visits) was explored by including an interaction term between group and time point in the models. There was no evidence of a differential effect over time for any outcomes, so a single treatment effect per outcome has been reported that averages the treatment effect over all time points.

For each participant from the weekly questionnaire data, the mean of their weekly POEM scores between week 1 and week 24 and the percentage of days that topical treatments were used were calculated. The participant mean POEM scores and percentage of days that topical steroids were used were analysed using a linear model weighted according to the number of weekly questionnaires completed.

Quality-of-life outcomes at 6 months were analysed using linear models. Changes to treatment regimen were based on whether or not a participant had reported any treatment escalation over the 6-month RCT period and analysed using a generalised linear model. Skin infections were analysed using negative binomial regression.

Adherence to wearing the trial garments was summarised using the percentage of days and nights that the study garments were worn. Participants were classified as adherent if they wore the trial garments for at least 50% of the days or 50% of the nights. This was done for participants who completed at least half (12/24) of the weekly questionnaires. Sensitivity analyses explored adherence for all participants by making different assumptions about garment wear during periods in which the questionnaire was not completed. Adherence to wearing the trial garments was explored descriptively according to age group and baseline eczema severity.

Serious adverse events, durability and acceptability of use of the garments and information from the follow-up questionnaire at 8 months were summarised descriptively.

Tertiary outcomes

The primary analysis assumed that the effect of the different brands of garments was similar, but the impact of garment brand on AE severity was explored in a tertiary analysis. AE severity according to brand was explored by adding a term for garment brand to the primary analysis model for the EASI described above. AE symptoms according to brand were explored by comparing POEM scores after 2 months of garment wear (baseline and 2 months for the intervention group and 6 and 8 months for the standard care group).

During the study there was a supply problem with one of the garment suppliers (DreamSkin), which meant that the randomised schedule was not followed during this time and participants received the alternative brand (DermaSilk). Any participants randomised to the intervention group during a time period that DreamSkin garments of the required size were out of stock were not included in the tertiary analysis by brand of garments. Similarly, any participants in the standard care group who completed their 6-month visit during a period when DreamSkin garments of the required size were out of stock were not included.

Adherence and acceptability of the garments at 6 and 8 months were summarised descriptively by allocated group and allocated garment brand.

Additional analyses

On completion of the pre-planned analyses, and following concerns that the baseline EASI scores appeared lower than might be expected for children with moderate to severe eczema, an additional post hoc analysis was conducted to explore the interaction between baseline severity and treatment group. This was conducted by adding an interaction term between allocated group and baseline EASI score (log-transformed and continuous) to the primary analysis model.

Participants

The mean age of participants was 5 years (SD 3.6); 58% were male and 79% were of white ethnicity (Table 4). The majority (72%) had previously been treated in secondary care for their AE, 72% had moderate or severe AE (based on IGA scores at baseline) (Table 5) and 37% were reported to use a potent or very potent steroid as their main steroid (Table 6).

The demographic and AE characteristics were well balanced at baseline, although there were slightly more boys in the intervention group than in the standard care group, and parent-reported history of asthma and food allergy was higher in the standard care group than in the intervention group (see Table 4). The mean EASI score was slightly higher in the intervention group than in the standard care group, as more children had a baseline EASI score of > 30 (intervention, 14 participants; standard care, 4 participants; Figure 6); however, the median and interquartile ranges were similar between the groups (see Table 5). Other AE severity measures were similar between the two groups apart from the PGA scores [a greater proportion of participants in the standard care group rated their AE as moderate, severe or very severe than the intervention group (75% vs. 66%) (see Table 5)]. Health-related quality of life in the two groups was similar at baseline (Table 7).

FIGURE 6.

Histogram of baseline EASI scores by group. (a) Standard care; and (b) intervention group.

FLG genotype

Table 8 shows the FLG genotyping results for participants of white ethnicity. Samples from 219 participants were tested for the six most prevalent FLG loss-of-function mutations in the white European population: R501X, 2282del4, R2447X, S3247X, 3702delG and 3673delC. Genotyping methods for the four most prevalent genotypes (R501X, 2282del4, R2447X and S3247X) were largely successful (n = 217 were included in the analysis), but the genotyping methods used for 3703delG and 3673delC were unsuccessful for 24 participants (11% of samples tested) owing to the suboptimal quality and quantity of DNA obtained from paediatric saliva samples.

In total, 217 participants of white European ethnicity were categorised as FLG wild type (individuals in whom none of the prevalent mutations was identified), FLG heterozygotes (carrying one FLG null mutation) and FLG homozygotes or compound heterozygotes (individuals carrying two FLG null mutations) for the four most prevalent mutations (R501X, 2282del4, R2447X and S3247X). Of these, 74 participants had at least one mutation.

Contamination

Only six participants in the standard care group reported wearing silk clothing during the 6-month study period.

Durability of the garments

Information about the number of garments and replacement garments sent out by the co-ordinating centre during the trial is presented in Chapter 4. This section presents information reported by parents on the 6-month questionnaire about the condition of the trial garments. Just over half of parents reported that at least one garment (top or leggings) could no longer be worn (Table 11). Children aged 4 years or under were more likely than older children to require replacement garments, as they outgrew the garments over the 6-month study period. Just over one-third of responders at 6 months reported that garments could no longer be worn as they had worn out or torn.

Acceptability of use of silk clothing

At 6 months, 85 out of 121 (70%) participants reported being satisfied or very satisfied with the garments (95% CI 61% to 78%) and 89 out of 121 (74%) participants were either happy or very happy to wear the garments (95% CI 64% to 81%).

Blinding

Blinding appeared to be successful. Research nurses remained blinded to treatment allocation for 289 out of 300 (96%) participants. Unblinding occurred for three participants in the standard care group and eight participants in the intervention group. This unblinding was first reported at 2 months for one participant in the standard care group and seven participants in the intervention group and at the 4-month visit for all other participants.

Unblinding mainly occurred as a result of the child or parents saying that they had or had not received the garments. Unblinding occurred for two participants because they wore the garments to the assessment visit.

Primary analysis

Mean AE severity based on EASI scores improved in both groups during the 6-month follow-up period; however, there was no clinically important difference between the groups in the nurse-assessed EASI scores (Table 12 and Figure 8). Averaged over the 2-, 4- and 6-month follow-up visits, the ratio of geometric mean EASI scores was 0.95 (95% CI 0.85 to 1.07; p = 0.43). This CI is equivalent to a difference of approximately 1.5-point improvement to 0.5 points worse for the intervention group, compared with the standard care group in the original EASI scale units.

FIGURE 8.

Primary outcome: geometric mean EASI scores with 95% CIs.

Arithmetic means of the EASI scores and log-transformed EASI scores for each group and time point are given in Appendix 14 (see Table 52).

Sensitivity analyses for the primary outcome

Sex, history of food allergy and history of asthma were added as covariates into the analysis model because of baseline imbalance. This estimate of the ratio of the geometric mean was the same as for the primary analysis (0.95, 95% CI 0.85 to 1.07).

Multiple imputation of missing EASI data, assuming that scores were missing at random, gave a very similar result to the primary analysis (Table 13). When it was assumed that missing scores were systematically worse in the standard care group, the 95% CI for the geometric mean was equivalent to scores of 2 points lower to 0.1 points higher for the intervention group than for the standard care group.

Further exploratory analysis of the EASI scores for areas covered by the garments (body and limbs) and areas uncovered by the garments (head and neck) can be found in Appendix 16.

Causal effect of adherence with wearing trial garments on primary outcome

The CACE was estimated using the EASI scores at 6 months for participants who completed 12 or more questionnaires (standard care, n = 127; intervention, n = 124). Table 14 presents the CACE estimate based on a binary definition of adherence of wearing the trial garments for at least 50% of the days or 50% of the nights and the CACE estimate for each additional 10% of the time the garments were worn. The intention-to-treat estimate for participants included in the CACE analysis is also presented using the EASI scores at 6 months for comparison (analysis according to randomised group).

The intention-to-treat and CACE estimate based on wearing the garments for at least 50% of the days or 50% of the nights are similar (see Table 14) as a result of 82% of intervention participants satisfying this definition (see Table 9). The ratio of geometric means for all comparisons is greater than 1, favouring the standard care group. The CACE estimate for each additional 10% of time that garments were worn suggests that eczema severity (EASI scores) did not improve with greater amounts of garment wear. A further summary table using all data at 6 months is shown in Appendix 17.

Subgroup analysis for primary outcome according to FLG status

Eczema Area and Severity Index scores according to group and FLG status (none, one or two FLG null mutations) for participants of white ethnicity are shown in Table 15 and Figure 9. Participants with FLG gene mutations were no more likely to benefit from the silk clothing than participants without a mutation (p-value for interaction effect 0.47).

FIGURE 9.

Geometric mean EASI scores by group and FLG status. FLG +/+ denotes no mutations; FLG +/– denotes one FLG null mutation; and FLG –/– denotes two FLG null mutations.

Post hoc subgroup analysis for primary outcome according to baseline eczema severity

Eczema Area and Severity Index scores according to group and baseline eczema severity (almost clear or mild EASI scores and moderate or severe EASI scores)44 are shown in Table 16. There was no evidence that the clothing was more or less effective depending on the severity of eczema at baseline.

Global assessment of atopic eczema

The proportion of participants with a nurse-assessed IGA of AE of moderate severity or worse decreased in both groups during the follow-up period, but there was no difference between the two groups: relative risk 0.98 (95% CI 0.82 to 1.12; p = 0.63; Table 17).

In contrast, for the participant-rated IGA, fewer participants rated their AE as moderately severe or worse in the intervention group than standard care: relative risk 0.83 (95% CI 0.70 to 0.98; p = 0.03; see Table 17).

Self-reported atopic eczema symptoms using the Patient Oriented Eczema Measure

Mean weekly POEM scores by group are shown in Figure 10. The mean of the participants’ mean weekly POEM scores over the 6-month study was 2.8 points lower in the intervention group than in the standard care group (95% CI –3.9 to –1.8; p < 0.001; Table 18). There was a more obvious separation of the groups in the first 3 months of the trial than in the final 3 months.

FIGURE 10.

Mean weekly patient reported symptoms (POEM scores) with 95% CI. Baseline POEM scores (16.6 standard care; 17.3 intervention): data not shown on graph as scores were collected in clinic rather than by online questionnaire.

Three-Item Severity scale

The mean TIS scores improved in both groups during the follow-up period. No between-group differences were observed: difference in means 0.09 (95% CI –0.22 to 0.40; p = 0.57; Table 19).

Use of atopic eczema treatments

The percentage of days during the study that emollients, topical corticosteroids, calcineurin inhibitors and wet/dry wraps were used is shown in Table 20.

The mean percentage of topical corticosteroid use was slightly less in the intervention group than the standard care group, equivalent to using topical corticosteroids on 6 days fewer over the 24 weeks (95% CI equivalent to using steroids for between 16 days fewer and 4 days more). The mean frequency of usage was similar in the two groups for the other topical treatments. Details of the amount of topical corticosteroid prescribed over the 6-month trial are summarised in Chapter 4, Table 30.

The potency of participants’ main topical corticosteroid was similar in the two groups at 6 months (Figure 11).

FIGURE 11.

Potency of main steroid at baseline and follow-up.

Changes in AE treatments during the trial are shown in Table 21. There were no differences between the groups in the percentage of participants who escalated their AE treatment between baseline and 6 months, although participants in the standard care group were more likely to have escalated treatment within the first 2 months of the study.

Health-related quality of life

Health-related quality-of-life outcomes for the DFI, EQ-5D-3L, ADQoL and CHU-9D are shown in Table 22. There were no differences between any of these quality-of-life outcomes between the two groups. The difference in means were all close to 0 and favoured the intervention group for DFI, EQ-5D-3L and ADQoL, and favoured the standard care group for CHU-9D.

Brand of garments

Figure 12 shows the numbers of participants randomised to the two brands of clothing. Garments were required during a DreamSkin out-of-stock period for 19 participants randomised to the intervention group (eight randomised to DermaSilk and 11 to DreamSkin) and 26 participants at 6 months who were allocated to the standard care group (nine randomised to DermaSilk and 17 to DreamSkin). These participants are not included in the analysis according to brand of clothing.

FIGURE 12.

Randomisation to brands of garments. a, Four participants in the intervention group randomised to DreamSkin initially received DreamSkin garments but were issued DermaSilk replacements as the required size of DreamSkin clothing was out of stock when the replacements were required; b, 12 participants in the standard care group were not sent any clothing at 6 months because they did not attend the 6-month visit; however, two of these participants completed the questionnaire at 8 months; c, clothing supply at 6 months was delayed for five of these participants so that they could receive DreamSkin garments.

The EASI scores during the 6-month RCT were similar according to garment brand (Figure 13). There was no differential effect of the garments according to brand (ratio of geometric means for DermaSilk vs. DreamSkin was 0.98, 95% CI 0.83 to 1.16). Adherence, satisfaction and the child being happy to wear the clothing were all similar at 6 months regardless of brand (Table 27). Further comparisons of the brands using data from the 8-month questionnaires can be found in Appendix 18.

FIGURE 13.

Geometric mean EASI scores by brand of garment.

Aims and perspective

The aim of the economic evaluation was to estimate the within-trial cost-effectiveness of silk therapeutic clothing plus standard care compared with standard care alone from a NHS perspective in the base case, and from a NHS and wider (family and employer) perspective in secondary analyses. Personal Social Services costs were not explicitly asked about, as the clinical team felt that these were unlikely to be relevant to those with childhood eczema.

Two forms of economic evaluation were used, cost–utility analysis and cost-effectiveness analysis, in order to enable comparisons with non-eczema interventions and other eczema interventions, respectively.

The economic evaluation adhered to published and well-accepted guidelines for the economic evaluation of health-care interventions, as appropriate. 52–54

Resource use: identification, measurement and valuation

The range of resource use and costs captured was in keeping with the chosen perspective. Three categories of resource use were identified as important to capture. These were intervention resource use, other health resource use and wider family/employer costs. All resources were costed at 2014/15 price year levels in UK pounds sterling.

Measurement of outcomes and quality-adjusted life-years

The main economic analysis is the cost–utility analysis, whereby effectiveness is measured in terms of QALYs for the child. This will be presented as the base case because it enables decision-makers to compare the value for money afforded by this intervention with that for other conditions. It also enables us to clearly differentiate the results of the cost–utility analysis based on the ADQoL from the other cost–utility analyses based on the CHU-9D and the parental EQ-5D-3L.

In the base-case analysis, utility was measured in all children using the disease-specific ADQoL. 38 The ADQoL consists of four binary choice questions,38 in answer to which respondents were asked to indicate whether they agreed more with the statement on the left or on the right:

1. you cannot join in some activities with other children/you are not limited in joining in activities with other children

2. you are very moody/you are not very moody

3. you cannot be comforted/you are quite settled

4. you sleep badly most nights/generally, you sleep very well.

The ADQoL was completed by parental proxy for children aged < 7 years and by self-report in those aged ≥ 7 years. The developer of the ADQoL used standard gamble methods to estimate utility values for the 16 health states described by the instrument. This involved asking adult participants to imagine that they were 10 years old and would be in the health state described for the rest of their life. The utility values generated ranged from 0.356 to 0.841. 38

The primary measure of effectiveness for the cost-effectiveness analysis was the difference in the proportions achieving treatment success at 6 months – defined as those with at least a 50% improvement compared with baseline on the primary outcome measure, EASI. 31 Secondary analysis was conducted using continuous data from the DFI, where scores range from 0 (no impact on family life, best score) to 30 (maximum impact on family score, worst score). 36

In addition, the main carer was asked to record their own utility using the EQ-5D-3L in order to see if the intervention impacted on parental quality of life (e.g. through sleep loss). The EQ-5D-3L is a generic preference-based health-related quality-of-life instrument with five dimensions, each of which has three levels. 37 The UK tariff was valued using time trade-off methods with a sample of the general adult population. Utility values on the EQ-5D-3L range from –0.594 to 1. 63 As there is no guidance to the best method for analysing adult utility data in combination with data on the child’s utility,64 we present a cost per QALY for the child as the base case and a cost per QALY per carer in secondary analyses, such that we do not combine the two sets of utility in a single analysis. It should be noted that the cost per QALY analysis for the main carer excludes any QALYs gained by the child with eczema and also probably underestimates the impact on the wider family in cases where a child with eczema lives with more than one adult, and in cases where they have siblings.

All utility instruments were measured at baseline and 6 months, and used to estimate QALYs for the trial period by using linear interpolation and area under the curve with and without baseline adjustment,65 and adjustment by centre and age. The total area under the curve (without baseline adjustment) was measured as:

to reflect the 6-month time frame. The primary cost–utility analysis reports the incremental cost per QALY based on the ADQoL because children of all ages, or their main carer, were asked to complete this instrument. Secondary analyses will report the cost per QALY based on the main carer’s EQ-5D-3L values separately. Previous work has not explored the ability of the EQ-5D-3L to detect impacts on carers’ quality of life for this condition.

Statistical analysis and analysis of uncertainty

Neither costs nor benefits were discounted,53 reflecting the 6-month time horizon of the analysis. A complete-case analysis approach was undertaken, with participants included only if they had complete cost and effect data at each time point.

In line with the statistical analyses, the primary analysis was conducted using the principles of intention to treat: all participants with data at baseline and follow-up were included, regardless of adherence to the allocated intervention. If > 10% of participants had missing data, then imputation of missing values was to be conducted as a sensitivity analysis.

The economic evaluation is a ‘within-trial analysis’. This means that costs and benefits were only evaluated for the trial follow-up period (6 months). Costs and outcomes in both arms of the study were estimated using the methods described in Measurement of outcomes and quality-adjusted life-years. This information on costs and benefits was used to conduct a complete-case incremental economic analysis comparing the silk garments in addition to standard care with standard care alone. This was completed for both the cost-effectiveness and cost–utility analyses. Conclusions are based on the estimated results. ICERs were calculated using accepted methodology. 52,53

The statistical analysis estimated both the unadjusted and adjusted estimates, where the latter controlled for any differences in baseline characteristics (i.e. the cost regression adjusted for baseline costs, age and recruiting centre, whereas the QALY regression adjusted for baseline utility, age and recruiting centre). All adjusted analyses used a regression-based approach (seemingly unrelated regression equations)66 to estimate incremental costs and QALYs with the exception of the cost-effectiveness analysis for EASI. As EASI was analysed as a binary variable (coded 1 for treatment success and 0 otherwise), generalised linear models, assuming costs and effects were independent, were employed. The cost regression used the Poisson family and identify link function, whereas the effects regression used the binomial family and identify link function.

As cost data were skewed, we used non-parametric bootstrapping to estimate adjusted mean (95% CI) incremental cost and mean (95% CI) incremental QALY estimates. Bootstrapping was also used to estimate cost-effectiveness acceptability curves;67,68 these show the probability that each of the intervention groups is the most cost-effective option at different monetary valuations of the outcome variable. A range of ceiling ratio (or willingness to pay per QALY) values were tested, including the £20,000 and £30,000 per QALY thresholds used by NICE in cost–utility calculations. 69

Sensitivity analysis tests the robustness of results in the face of any uncertainties. 53 It also improves the generalisability of results by indicating what could happen with different values of a parameter. Two areas of uncertainty were considered worth exploring. The first was the cost of the silk therapeutic garments, because 46% of total costs from a NHS perspective were accounted for by the cost of the silk garments in the study. Instead of using the Health and Social Care Information Centre Prescription Cost Analysis net ingredient cost per item, we reran the analysis using unit costs for silk garments based on the NHS Business Services Authority actual cost formula to estimate the actual cost to the NHS:56

In particular, this approach takes account of the average discount enjoyed by the NHS when purchasing prescription items. As payment for consumables and out-of-pocket expenses is not relevant for silk garments, we did not include these but did include the 90p fee pharmacists receive for dispensing. We did not do this for all prescription items as medication costs were not significantly different between the two treatment arms.

Second, as there is uncertainty about how best to capture utility in children, we included a second generic preference-based instrument, the CHU-9D,39 at baseline and 6 months in order to compare the results with those gained using the disease-specific ADQoL. 38 The generic health-related quality-of-life instrument, CHU-9D,39 was used only in children aged ≥ 5 years at baseline. The CHU-9D consists of nine dimensions (worry, sadness, pain, tired, annoyed, school work/homework, sleep, daily routine, ability to join in activities) using a recall period of today or last night, dependent on the question; each dimension has five levels. 39 The wording of the dimensions and levels resulted from qualitative work with children and young people. It can be self-completed by children and young people aged 7–17 years with a proxy version for parents to complete for children aged < 7 years. At the time the CLOTHES trial was designed, the CHU-9D had not been used with the under-fives and so we did not use it with those aged < 5 years. Additional guidance now exists to help parental proxies complete the CHU-9D for this age range (personal communication with K Stevens, University of Sheffield, 2014). The CHU-9D was valued by the UK general adult population using standard gamble methods and utility using this instrument can range between 0.33 and 1. 39 In this study, the CHU-9D was self-completed by children and young people aged ≥ 7 years and parental proxy completed for 5- and 6-year-olds. QALYs for the trial period based on the CHU-9D were estimated using the same methods described for estimating QALYs based on the ADQoL and main carer EQ-5D-3L (see Measurement of outcomes and quality-adjusted life-years).

Some of the subgroup and sensitivity analyses included in the original health economic analysis plan (see Appendix 19) were not conducted for the following reasons:

• imputation of missing values, as missing values were < 10%

• per-protocol analysis adjusting for adherence; sensitivity analyses in Chapter 3, Causal effect of adherence with wearing trial garments on primary outcome showed no evidence of a causal effect of adherence with wearing trial garments on the primary outcome.

• subgroup analysis based on impairment in skin barrier function (FLG genotype), as subgroup analysis of the primary outcome suggested no differential effect

• resource use data collected in the observational period (6–8 months) were not included, as there was no overall effect at 6 months.

All statistical analysis was undertaken in Stata 14.

Resource use and costs

Base-case cost–utility analysis

Base-case cost–utility analysis from a NHS perspective

The adjusted mean difference in QALYs per participant was 0.0064 (95% CI –0.0004 to 0.0133) (see Table 29 for the key findings from the base-case analysis). Combined with adjusted mean cost, the adjusted mean incremental cost per QALY was £56,811 (Figures 15 and 16; see also Table 29), suggesting that silk garments for moderate to severe eczema are not cost-effective within currently accepted thresholds. At a willingness to pay of £30,000 per QALY, the probability of silk garments being cost-effective was 12.13%.

FIGURE 15.

Incremental cost-effectiveness plane for the base-case analysis (ADQoL).

FIGURE 16.

Cost-effectiveness acceptability curves for the intervention (solid line) and standard care group (dashed line): base-case adjusted costs and ADQoL utility scores. NICE threshold for willingness to pay per QALY = £20,000–30,000. 69

Cost-effectiveness analysis

Cost–utility analysis for main carer EuroQoL-5 Dimensions-3 Levels

The main results for the cost–utility analysis based on main carer quality of life, as measured using the EQ-5D-3L, are shown in Table 32. The adjusted mean incremental cost was £369.76 (95% CI £216.02 to £523.51) and the adjusted mean incremental QALY gain was 0.0029 (95% CI –0.0045 to 0.0102) (Figure 17), giving an ICER of £251,849 per QALY. At a willingness to pay of £30,000 per QALY, the probability of silk garments being cost-effective was 1.31%.

FIGURE 17.

Incremental cost-effectiveness plane (main carer EQ-5D-3L).

Uncertainty and sensitivity analysis

Alternative outcome measure: Child Health Utility-9 Dimensions analysis

For the analysis using the CHU-9D instead of the ADQoL to estimate child utility scores, standard care dominates as silk garments were both more expensive and less effective (in terms of QALYs measured using the CHU-9D for utility) (Figure 18 and see Table 32) than standard care. The probability of silk garments being cost-effective at a willingness-to-pay threshold of £30,000 per QALY was 0.06%.

FIGURE 18.

Incremental cost-effectiveness plane (CHU-9D).

Main findings

This is the first economic evaluation of silk therapeutic garment use in children with moderate to severe eczema. The economic analysis extends the clinical analysis to show that silk garments in addition to standard care are unlikely to represent value for money for the NHS. In terms of the costs, the additional costs of providing silk garments to intervention participants were not recouped through cost savings from lower use of wider health-care resource items nor from families in terms of reduced out-of-pocket costs or from employers in terms of reduced time off work by parents caring for their child with eczema. In terms of outcomes (when looking at patient-assessed health-related quality of life), although the ADQoL showed a very small, non-significant, positive benefit in terms of QALYs in favour of silk garments, this was not sufficient to outweigh the higher costs of providing silk garments. The CHU-9D showed a very small non-significant decrease in QALYs (suggesting worse health) in the intervention arm. When coupled with the higher costs, this suggests that silk garments do not represent value for money for the NHS. The small, non-statistically significant, differences found when estimating QALYs (using the different instruments) suggest that the difference between intervention and usual care groups is negligible and may have been due to chance.

The result of this within-trial cost-effectiveness study also provides an indication of the cost of treating moderate to severe eczema to the NHS. Over a 6-month period, health-care costs were in the region of £345 (SD £647.40) per child (using 2014/15 costs and excluding the cost of silk garments). In addition, the reported personal costs and time lost from work/school as a result of eczema were considerable and, provide important data to inform studies on the societal impact of this condition.

Strengths and weaknesses

Given the clinical result, it could have been argued that an economic evaluation was unnecessary. However, resource use and quality-of-life data were collected alongside the trial, and these results provide useful data to inform future studies and decisions regarding health commissioning.

The study did not explicitly ask families to record how much time they spent applying treatments for their child’s eczema. Were a treatment to be effective at reducing disease severity, it could conceivably reduce the amount of time a parent or child spends applying/taking medications and it might be important to capture this effect from a family perspective in future studies.

The study attempted to measure QALYs for the main carer in addition to the child. Our approach had limitations because children may have also lived with another parent or carer in addition to siblings, and we did not seek to capture the health-related quality-of-life effects of these extended family members. It is unclear how best to capture the wider impacts of eczema within a family in an economic evaluation. Further research looking at whose QALYs to capture, how to aggregate QALYs estimated for patients, carers and siblings, and how to present such results is needed. 64 However, given that this study found no difference in effect between treatment arms, the results are unlikely to be sensitive to the assumptions made in relation to such ‘beyond the patient’ effects. 64

Data collection

Interviews and focus groups with children were completed from February 2015 to May 2015. Children were recruited using a convenience sampling strategy76,77 via the study research nurses. They represented all of the recruiting centres and had all completed the trial before participating. Ten semistructured, audio-taped face-to-face or telephone interviews were conducted by the researcher (EW). The children in the interview group ranged in age from 9 to 15 years. There were three focus groups in total, comprising two groups of 7- and 8-year-olds, each with two participants, and a further group of 5- and 6-year-olds, with four participants. Parents and siblings accompanied the children during data collection according to age and personal preference. The demographic details of the children are presented in Appendix 20 (see Table 59). The researcher (EW) used a wide range of developmentally appropriate, child-friendly techniques, such as drawing, collage, photography, storytelling, stickers and a puppet. The activities were adapted to the needs and interests of individual children. The resources chosen were sex and ethnicity neutral. The activities were selected with sensitivity; for example, the use of stickers and foam people allowed children to illustrate their point without the added pressure of having to be a ‘good drawer’. Care was taken to enable children to express their views through whatever medium they preferred. A broad topic guide was used as a basis for interactions with the children (see Appendix 20). This was shared with the children so that they understood what would happen next. At the end of each interview and focus group the researcher summed up her understanding of the responses and checked with each child that this was correct.

All data collection focused on finding out from children what it is like to have eczema, what they thought of the silk garments and of being in the trial, and whether or not other children with eczema should be given the garments. All interviews were audio-recorded and transcribed in full. Photographs were taken of artefacts produced during the sessions, as most children wanted to take these home. The children enjoyed taking photographs of what they had made and while they were doing this the researcher used the opportunity to verify her understanding of their thoughts. This was documented alongside the photographs in preparation for data analysis.

Data analysis

The analysis of the data required a sound understanding of children’s physical, cognitive, emotional and social development. This contributed to ensuring that the analysis reflected the children’s voices as faithfully as possible. Interpretation of drawings, collages and craft work was undertaken with the children at each stage.

The data from the interviews and focus groups were analysed using the three methods of holistic, selective and detailed data analysis of Van Manen:78 (1) data were viewed as a whole, (2) phrases or illustrations that seemed to represent the experience under study were identified and (3) written data were reviewed line by line in order to identify themes. 79 The aim was to be attentive to the voices of children and recognise subtle significations. 80 This was achieved through prolonged and intensive engagement with the totality of the data.

Data analysis yielded six key themes: (1) living with eczema, (2) expectations of the garments, (3) wearing ‘silks’ (a term often used by the children to describe their silk garments), (4) did they help?, (5) thoughts about the garments and (6) being part of the study. A description and analysis of each theme, together with supporting illustrations, is provided with a tabulation of each theme containing exemplar data extracts.

Theme 1: living with eczema

This theme comprises two subthemes: (1) health aspects of living with eczema and (2) social and emotional implications of living with eczema (Table 33).

TABLE 33 - Views of children: theme 1 – living with eczema

Health	Social and emotional
Itchy and hot, hands and feet sore. Eczema keeps me awake at night	It makes me feel sad and grumpy
My skin on my back feels like fire, like rocket fire	It’s . . . like poo. I itch all the time and it’s just not fair
It hurts a lot . . . when water touches it, it burns	I beat myself up for it because like I really want to get rid of it
	You can't do as many things as you want to and even the sports you want to do, you can't do it because it irritates or makes it worse
	People say I am ugly

Health aspects of living with eczema

All children portrayed a negative view of living with eczema; they particularly suffered with itching, heat, and flaking and dry skin, all in different areas of their bodies and with differing severity. A picture by an 8-year-old boy depicts how burning, red and sore the eczema on his back felt (Figure 19). Many said that eczema affected their sleep at night and that this impacted on their daytime activities, particularly school and play.

FIGURE 19.

Photograph showing depiction of eczema by an 8-year-old boy.

Social and emotional implications of eczema

Several children spoke of the emotional effect that eczema had on their lives, including frustration with the condition and how it affected their general well-being. The children spoke of how eczema affects their ability to fit in with their peer group and the limiting effects the condition could have; for example, most did not like exposing their skin to others and did not enjoy sporting activities for this reason.

Theme 3: wearing silk garments

This theme recounts when the children chose to wear the silk garments. There are two subthemes: (1) day versus night and (2) school versus home, leisure and play (Table 35).

TABLE 35 - Views of children: theme 3 – wearing silk garments

Day vs. night	School vs. home, leisure and play
Always wear them at night	I wore them to school but not PE [physical education] days . . . people would laugh
During the day we use the silk suits, put the cream on underneath, then put all the suits on . . . and that sort of stops my clothes rubbing against my skin. Wearing the tights day and night	Poor fit trousers so not worn to school/don’t fit under tights
I prefer to wear them during the day rather than the night because the silk is really comforting on my skin	One of the problems was that at school – because they’re quite strict on uniform – that’s a problem
Wearing them at night because some people ask like what’s that and it’s a bit annoying	Yes I wore them to school but it was a bit weird when like everyone was like ‘What is that?’
I wear them mainly at night	I did the upper body one once at school . . . but everyone kept on asking about it and I just didn’t like it
	I usually just wear them at home
	I wouldn’t wear them at a sleepover just because I would find that really embarrassing, because no offence, it didn’t really look that good

Day versus night

Figure 20 is an example of how a child conveyed the times at which he wore the garments. Few children of any age wore their silks during the day; a majority of the children preferred to wear them at night and only when at home.

FIGURE 20.

Photograph showing a child’s depiction of times garments were worn.

School versus home, leisure and play

Children reported not wanting peers to see the garments for fear of attracting unwelcome questions or comments. There was a seasonal element to garment wear, particularly not wanting the garments to be visible when wearing shorts and t-shirts in the summer. There were many issues with being able to wear them for school, ranging from fit, uniform-requirements, physical education issues and the reactions of other school children. Peer groups and friendships were very important for all children; only those with very secure friendships felt that they could openly speak about their eczema and the garments.

Theme 4: did they help?

This theme has three subthemes: (1) getting better, (2) getting worse and (3) no difference but liked them anyway. There was a real mixture of perceived success of the garments, ranging from a perceived complete cessation of eczema to no effect at all, to worsening of symptoms (Table 36).

TABLE 36 - Views of children: theme 4 – did they help?

Getting better	Getting worse	No difference but liked them anyway
It made it feel better but the crusty is still there	I stopped wearing them as soon as I realised they were making [my] eczema worse. It helped the cream absorb better but it did make it all inflamed and more itchy	It was comfy. Top and bottoms were smooth
Better, the itch goes away	It kind of got worse. It just got more dry and really red	Feels nice on my skin
At night I got to sleep through for the first time. It made me feel good inside	[At] night, it wakes you up because it is scraping at my skin and because you have got loose bits of dead skin . . . it catches and it rips it	They feel nice, I like wearing them. They . . . helped me be less annoyed
It’s kind of helping the eczema go away. Less sore

Getting better

Some children reported that the silks had improved their sleep, whereas others felt that the silks had made them more comfortable in bed but had not improved their eczema per se. Improved sleep was linked with better well-being at school and at home the following day. A few children stated that the clothing improved their mood.

Getting worse

Other children were disappointed in the effect of the garments, which affected their self-esteem. Some children suggested that it was their own fault the garments had not worked for them.

No difference but liked them anyway

Some children liked wearing the garments, even if they had no perceptible effect on their eczema. This was predominantly because they were comfortable, although some parents reported that the garments made their child feel special. Figure 21 shows what an 8-year-old girl thought other children should know about what was ‘good’ and ‘bad’ about the silks. The blue-stamped areas represent where the garments were hot and uncomfortable.

FIGURE 21.

Photograph of an 8-year-old girl’s thoughts on garments.

Theme 5: thoughts about the garments

This theme comprises three subthemes: (1) quality and cost, (2) fit and fabric, and (3) design and appearance (Table 37).

TABLE 37 - Views of children: theme 5 – thoughts about the garments

Quality and cost	Fit and fabric	Design and appearance
Looked like a trial clothing rather than a purchase	‘Random sizing’ poor fit. Uncomfortable – too tight	[Didn’t like] that you could see through them
I’ve got massive holes gaping wide	I would make them a bit tighter, problem with the elastic, and they bunch at the bottom	It’s not . . . that private to be wearing something basically see-through . . . like wearing cling film
Get hot . . . and they can tear	I just found them really, really uncomfortable. They were tight. There wasn’t the give in them	Gap between trouser and top untreated and elastic waistband itchy
I think the best thing about them is that they are 100% silk and they don’t have any . . . elastic	Kind of transparent – needed to wear pants under	No good for summer – long sleeves
We didn’t think they were going to be as expensive		They look really grubby
		They are not very pretty

Quality and cost

Children were remarkably aware of cost and quality issues. Many thought that the garments were quite poor quality in fit, appearance, design or the fabric itself, especially given the cost. They were disappointed in the shape, size and colour. The cost was an issue for further purchase for a few, with older children (and their parents) feeling that the garments should be available on prescription from a GP.

Fit and fabric

Many felt that the sizing and fit, especially of the trousers, were poor. Children were aware that the garments did not always wash well, turning baggy and grey, and affecting fit over time as illustrated in Figure 22. Most children used the garments with creams, which made the silks sticky and oily, and, in some cases, smelly. Some felt the texture was rough and actually irritated their skin more; some also felt that the silks actually made them hotter.

FIGURE 22.

Child’s illustration of garments over time.

Design and appearance

The fact that the silks were see-through was an issue for many; this limited their use to the home, and many children had to wear another layer of clothing over the silks. At night, the garments tended to be worn under pyjamas or onesies. Issues of fit, appearance and smell compounded children’s sense of being ‘different’ from their peers. Children commented that, although three sets of garments were provided, these did not reflect sex differences in relation to puberty nor the normal varied physical growth spurts experienced at this time.

Children suggested a range of improvements to the garments, including using smoother and more closely woven fabric, resembling ‘proper’ silk. Some wanted the garments to be tighter, but others preferred the looser fit. A few thought that different colour options would be good, especially the youngest and oldest children. In terms of design, a ‘onesie’ was seen as a positive option, as were short-sleeved tops and shorts for summer. Others felt that additional protection for hands and feet would be beneficial, as these areas were not covered by the trial garments (except for those aged < 2 years) and were often troublesome. Children suggested that a few changes would make the garments more wearable in the daytime provided that this was under normal clothing, including school uniform, in the colder winter months. All except one participant felt that they would recommend other children to try the garments, even if they had not been particularly successful personally. Many of the children recognised that everyone’s experience of eczema was different and that they would respond differently to different treatments, and they seemed keen to ensure that other children did not miss any opportunity to improve their eczema.

Theme 6: being part of the study

This theme comprised two subthemes: (1) helping others and (2) the research process (Table 38).

TABLE 38 - Views of children: theme 6 – being part of the study

Helping others	The research process
Happy to know I am helping everyone in the world, who has eczema probably in the world	It felt quite good because I was doing it in a trial
It felt like I was helping the other people who had it as well so then they got some research. It felt quite good	[The focus group was] lots of fun it gives you a chance to talk to others about your feelings . . . you get to do lots of fun stuff at the end!
I didn’t mind wearing them because I knew it was for a trial to see if it could help people with eczema

Helping others

Many of the children were really pleased to be taking part in the trial. Indeed, the majority of children demonstrated a significant level of altruism, wanting to be part of the trial particularly if it would help other children.

The research process

All children relished the opportunity to be with others with eczema and to talk about their experiences. Many were disappointed that the data collection was a one-off event. Most enjoyed the whole research process, and younger children valued the small gifts they received and enjoyed charting their progress. A few of the older children reported that they would ‘give the garments another try’, having had the opportunity to talk about them with the researcher.

Children engaged readily in the interviews and focus groups. Older children appreciated the chance to discuss their experience of eczema and the garment trial. In addition to the activities outlined in previous sections, all children were encouraged to feed back to the researcher throughout the process. All children in focus groups and face-to-face interviews chose to ‘mark’ the researcher’s work in a similar way to their experience of their schoolwork being marked by a teacher. The researcher’s understanding and performance, including whether or not she had asked all of the pertinent questions and understood their answers, were assessed by the children ‘ticking’ the work with carefully selected coloured pens.

Children took home all their artwork and checked the final photographs for use by the researcher, which gave them a sense of achievement and accomplishment. Children were asked how the findings of the project would best be communicated: most advocated either posters or a website, which they felt should be fun and incorporate activities. Figure 23 shows a dissemination poster designed in one of the focus groups.

FIGURE 23.

Dissemination poster designed by a child.

Child participants presented a range of views about the silk garments. All had high hopes when they entered the trial and, with few exceptions, their expectations were not realised. Some found the garments comfortable but many did not like the look, feel and fit of them. Even those children of a relatively young age were aware that the garments were expensive. Patterns of use varied, but most children reported wearing the garments fairly regularly at night. Although the children tended to report a limited effect on their skin condition, they almost universally recommended that other children should ‘give the garments a try’, perhaps exemplifying their experience of trying different treatment options and recognition of individual differences and needs.

Data collection

A series of in-person focus groups and telephone interviews with the parents of children in the trial was conducted between November 2014 and August 2015. A convenience sample of parents was recruited via trial information leaflets and invitations, and with the support of the research nurses. 76,77 All participants had reached the end of the trial prior to participating and they represented all recruiting centres. In total, 28 mothers and five fathers participated, representing 11% of all trial participants. The age range of their children was 2–14 years and the sex split was 19 boys and 13 girls; 17 of the children were allocated to the standard care group and 15 were allocated to the intervention group. The demographic details are summarised in Appendix 20 (see Table 60).

The researcher (FC) conducted the semistructured focus groups and telephone interviews using a prepared topic guide (see Appendix 20). The four focus groups had between two and four participants and lasted between 42 and 95 minutes. Telephone interviews lasted between 18 and 50 minutes. All data were audio-taped and transcribed verbatim.

Data analysis

Data were analysed using the five-stage Framework Analysis process. 81 The five stages comprise (1) familiarisation with the data through reading full transcripts; (2) development of a theoretical framework through identification of recurring and important themes; (3) indexing and pilot charting; (4) summarising data in an analytical framework; and (5) synthesising data by mapping and interpreting. In line with usual qualitative research practice, the researcher completed a reflective account and this was used to inform the analysis process. Stages 1–3 of the framework are illustrated in detail in Appendix 20 and stages 4 and 5 are documented below. The analysis process yielded four key themes: (1) despair and hope, (2) fit, durability and care, (3) perceived impact of the garments and (4) engaging in the trial. A total of 13 subthemes were identified. A description of each theme is provided below, and this includes a tabulation of each theme and subtheme with exemplar data extracts.

Data collection

Interviews with clinicians and commissioners in England were completed from June 2014 to January 2015. A purposive76,77 sample of participants was recruited via advertisements on health professional and dermatology websites, and snowballing. This approach yielded a range of dermatology specialist and primary care generalist participants from across the country. Some had dual roles, for example GP and commissioner, so they have been categorised by their stated primary role, and included dermatology specialist nurses (n = 9), dermatologists (n = 4), GPs (n = 3), pharmacists (n = 3) and health-care commissioners (n = 2). All clinicians had a minimum of 5 years’ experience of caring for people with eczema. Demographic information about the 21 participants is summarised in Appendix 20 (see Table 63). Semistructured telephone interviews using an interview guide (see Appendix 20) were conducted by the researcher (FC) lasting from 9 to 24 minutes. Interviews were audio-recorded and transcribed verbatim.

Data analysis

Data were analysed using the five-stage Framework Analysis process, described in Data analysis, and the researcher’s reflective account was used to inform the analysis process. Stages 1–3 of the framework are illustrated in detail in Appendix 20 and stages 4 and 5 are documented below. The analysis process yielded four key themes: (1) knowledge base, (2) reasons to use silk garments, (3) reasons for not using silk garments and (4) outcome measures. A total of 14 subthemes were identified. A description and analysis of each theme is provided below, and this includes a tabulation of each theme and subtheme with exemplar data extracts.

Findings

Implications for interpretation of trial results

• Children’s responses were particularly helpful in providing additional detail on possible reasons for non-adherence in wearing the garments.

• Parents and children portrayed mixed views on the garments. Few, if any, reported the longed-for ‘miracle cure’ and there was a significant sense of disappointment in relation to effectiveness and the quality, fit and durability of garments.

• Some parents found the weekly questionnaires useful in prompting more regular use of their child’s usual eczema care.

• Clinicians and commissioners were generally equivocal about the use of silk garments. Most wanted a robust evidence base to inform treatment decisions and were eager to see the results of the CLOTHES trial.

• Results of the qualitative studies were in line with the quantitative trial data, suggesting that important differences between the groups had not been missed.

Added value of qualitative studies within trials

• The added value of the qualitative work is that it has provided a deeper, richer and more detailed understanding of the ‘what’, ‘why’ and ‘how come’ underpinning participant’s beliefs and behaviours. It has uncovered understandings, which help to inform interpretation of the results of the study.

• Parents who participated were universally positive about having an opportunity to talk with a researcher who was interested in their child’s condition, but, perhaps more importantly, focus group participants relished meeting other parents in the same situation.

• Likewise, children wanted to talk with the researcher and were very able to express their views with clarity, given age-appropriate means of communication.

Lessons for conduct of nested qualitative studies

• Recruitment to qualitative studies nested within RCTs can be challenging. Research nurses who had personal contact with participants were crucial to successful recruitment. Participants in some centres were geographically widespread and the ages of children completing the trial at a time suitable for involvement in the focus groups meant that many participants were unable to take part.

• Early awareness of the forthcoming focus groups may help to boost recruitment.

• Child data added a valuable dimension to our understanding of childhood eczema and the use of silk garments. However, these were time-consuming to collect as there was a need for pre- and post ‘playtime’. Recruitment was particularly challenging, as the groups were age banded to ensure the appropriateness of data collection methods.

Contextual factors relating to patient and public information involvement

Many factors contributed to ensuring strong PPI in the CLOTHES trial.

Stages of research and opportunities for patient and public information impact

A summary of the stages of research that provided opportunities for PPI engagement are summarised in Table 48, along with details of the PPI activities and their impact.

Trial delivery and recruitment

Following publication of a press release announcing the start of recruitment into the CLOTHES trial, the trial team received considerable interest from media outlets including television, radio, newspapers and magazines. Having access to PPI partners who were willing to give interviews about the impact that AE has on their lives ensured that the news stories were widely distributed and engaging. In total, three parents, two children with AE and a representative from the National Eczema Society gave media interviews (Figure 24).

FIGURE 24.

Patient and public involvement co-applicant Amina being interviewed with her son Tahmid about the trial for BBC local news. BBC, British Broadcasting Corporation.

As a result of this media coverage, the trial team received 492 expressions of interest from potential participants in the space of 3 months. This resulted in a surge in recruitment at the start of the trial and meant that the trial was consistently ahead of recruitment targets throughout the study (Figure 25).

FIGURE 25.

Recruitment graph for the CLOTHES trial showing impact of initial media interest.

Reflections from our patient and public involvement partners

As a member of the CEBD [Centre for Evidence Based Dermatology] panel I was part of the initial priority-setting research which led to this trial and have felt privileged to be given the opportunity to play such an active role in this research, which is close to my heart.

Early mainstream media interest in this trial has made me to feel more involved than I may have had in other projects, making it a pleasant and rewarding experience. Throughout the research process I feel my contribution has been valued and treated with professionalism and the team have made a conscious decision to drive inclusive behaviours through the use of ‘simple’ plain English versus technical jargon.

Amina (PPI co-applicant)

As an attendee at the original Centre of Evidence Based Dermatology Patient Panel where the concept for the CLOTHES trial was first aired, I have been delighted to be involved at various stages of this well-run trial. I know the result will be of use and interest to the followers of the Nottingham Support Group for Carers of Children with which I am involved.

Amanda (member of CEBD patient panel and main contact for Nottingham Support Group for Carers of Children with Eczema)

As the lay member of the Trial Steering Committee I did not have as active a role as the other PPI members, but took part in all the meetings and kept up to date with the progress of the trial. I also liaised with the National Eczema Society and took part in the CLOTHES Collaborators meeting in June 2015. I am impressed with the interest in non-pharmacological interventions for the treatment of AE shown by the patients and carers in this trial and with the PPI input at all stages of the trial.

Rosemary (PPI member of Trial Steering Committee)

Trial logistics

Both clothing suppliers were supportive and were responsive in addressing questions or concerns about use of the silk garments. They worked closely with the trial co-ordinators at the Nottingham CTU to ensure efficient supply of the garments and development of an audit trail.

The central team at the CTU acted as the distribution hub for all trial garments to increase efficiency of stock supply, to maintain blinding of local site staff and to act as an intermediary between participants and the clothing companies (in the event that questions arose regarding the clothing). A detailed inventory of stock was kept so that supplies could be ordered weekly, providing accountability of garment usage and serving as a reordering system. Any returned stock (because of ill fit, excess wear or child growth) was returned to the manufacturers for them to see how the clothing held up with regular use.

Although three sets of garments were supplied to all participants, the need for replacement garments during each participant’s 6 months’ involvement in the trial was underestimated during our planning phase. This presented an increased financial burden on the companies, which remained supportive and supplied the additional garments after discussions with the trial team about why they were required. In this regard, it was helpful to work with two companies, so that the financial burden could be shared.

Having two companies involved in the study was also particularly beneficial when one of the companies experienced supply problems partway through the recruitment period. Each supplier remained supportive of the trial, doing all they could to allow the trial to continue. The company that experienced stock shortages kept back reserves of their dwindling stock for the trial, leaving them with fewer to sell on the market. The other company provided cover for their commercial competitor, providing more garments to the trial than originally intended.

Independence from the trial team

As outlined in Table 50, we attempted to replicate the model of engagement with commercial companies through an independent third party, as was successfully achieved in our previous NIHR HTA-funded SWET. To achieve this, we asked a representative of the lead NHS trust and the local lead for the comprehensive research network to be independent arbiters between the companies and the Trial Management Group. However, despite initial agreement from all parties, this approach did not work well in reality. This was largely because the nature of the ordering and stock control system meant that it was necessary for the trial management team to have weekly contact with representatives from the clothing suppliers. As a good working relationship developed, the majority of trial communication inevitably involved the trial manager. Nevertheless, this situation ensured that any concerns were resolved quickly, which was beneficial for delivery of the trial, and direct contact with the rest of the Trial Management Group (including the chief investigator) was kept to a minimum.

Training in trial design and interpretation

Given the importance of this trial to the financial concerns of both companies, we strove to ensure that the rationale for our trial design was clearly explained before, during and after the trial, so that the clothing suppliers had a good understanding of the trial design and conduct. In particular, we explained our approach to controlling for bias and the implications of running a pragmatic RCT.

As outlined in Table 50, we conducted face-to-face meetings to discuss these issues prior to start of recruitment, and again prior to data analysis, to ensure that the clothing suppliers understood the reasoning behind trial design decisions. In particular, these meetings focused on explaining the following key aspects:

• what a pragmatic trial is and implications for the design of the trial

• importance of the primary outcome and the need for an objective primary outcome

• interpreting unblinded trials

• understanding p-values and CIs

• individual versus group mean effects

• the need for transparent reporting of results

• timelines and availability of the findings

• ownership of the data and write-up of study results.

Child interview and focus group topic guide

• Grand tour question:

  • Tell me a bit about your eczema, what it’s like living with it?

• Mini tour questions:

  • How have you got on with the special clothing?

  • How much did you wear the clothing (day/night/away from home)?

  • What was it like wearing the clothing (skin condition, comments from others)?

• Example questions:

  • Can you tell me about any differences you have noticed (skin condition/well-being)?

• Experience questions:

  • Were there particular things you liked or did not like about using the special garments?

Parent interview and focus group topic guide

• Grand tour question:

  • Tell me a bit about the eczema, what it’s like living with it?

• Mini tour questions:

  • How have you got on with the special clothing?

  • How much did you/your child wear the clothing (day/night/away from home)?

  • What was it like wearing the clothing (skin condition, comments from others)?

  • How did you get on with looking after the garments (washing, size)?

• Example questions:

  • Can you tell me about any differences you have noticed (skin condition/behaviour/well-being)?

• Experience questions:

  • Were there particular things you or your child liked or did not like about using the silk garments?

  • How would you feel about continuing to use the special clothing (what things might make you continue or make you stop, barriers to use)?

  • What makes you think that the garments have worked or not worked?

Stages 1–3 of parent data analysis