Effectiveness of a skin care programme for the prevention of contact dermatitis in healthcare workers (the Healthy Hands Project): A single‐centre, cluster randomized controlled trial

Background Healthcare workers (HCWs) are at risk of developing hand dermatitis (HD). Guidelines recommend moisturizers to prevent HD, but in practice their effectiveness has been poorly investigated. Objectives To assess whether an intervention aimed at improving skin care leads to a reduction in HD severity. Methods In this 1‐year randomized controlled trial, 9 wards (285 HCWs) were allocated to an intervention group (IG), and 10 wards (216 HCWs) were allocated to the control group (CG). The intervention included provision of cream dispensers with electronic monitoring of use, regularly communicated to the HCWs. The primary and secondary outcomes were change from baseline in Hand Eczema Severity Index (HECSI) score (ΔHECSI) and change in natural moisturizing factor (NMF) level (ΔNMF). Results At 12 months, the rates of loss to follow‐up were 41% and 39% in the IG and the CG, respectively. The HECSI score was reduced in the IG by −6.2 points (95%CI: −7.7 to −4.7) and in the CG by −4.2 points (95%CI: −6.0 to −2.4). There was no significant difference in ΔHECSI or ΔNMF between the groups. Relative improvement in the HECSI score was significantly higher in the IG than in the CG (56% vs 44%). In a subgroup of HCWs with mild HD, the IG showed a larger HECSI score decrease than the CG (P < 0.001). Conclusion Although there was no significant effect on the primary outcomes, the intervention showed overall positive effects on the HECSI score.

tive than other monitoring systems that track individuals' actions, which can be seen by staff as punitive or an intrusion of their privacy. 17 Recently, we reported on an electronic monitoring system developed for the continuous registration of hand cream consumption with electronic dispensers in wards. 18 This system enables detailed feedback on the frequency and time-pattern of hand cream use on wards. In the present study, we aimed to assess the effectiveness of this system in a randomized controlled trial (RCT). The intervention comprised installing hand cream dispensers on the wards, electronic monitoring of use, and repeated feedback. The effect of the intervention was measured as the change in clinical severity of HD between baseline and follow-up at 12 months, as compared with the control group (CG). Furthermore, as a secondary outcome, we determined the change in the stratum corneum (SC) levels of natural moisturizing factors (NMFs) as a potential sensitive biomarker of skin barrier damage.
Cluster randomization was used to minimize the threat of "treatment contamination." This article is focused on the primary and secondary outcomes of the intervention. We will report the effects of the intervention on cream use behaviour in a separate article.

| PATIENTS AND METHODS
The Healthy Hands Project is a single-centre, cluster-randomized, parallel-group controlled trial with blinded outcome assessments. We have previously described the design 18 and statistical methods 19 in detail.
The participants in this study were HCWs clustered in wards of the Amsterdam University Medical Centre, Amsterdam, The Netherlands. They were enrolled between 2 May and 14 June 2016. Only wards known to have substantial exposure to wet work were included. These wards were categorized as low-exposure or highexposure wards on the basis of soap purchase in the previous year.
The eligibility criteria were provision of written informed consent, being employed as an HCW or nutrition assistant on one of the wards, and being exposed to "wet work." Ethics approval to conduct the trial was granted by the Medical Ethics Review Board of the Academic Medical Centre (AMC) (reference number, NL54372.018.15; Trial registration, NTR 5564; date of registration, November 2, 2015).
Informed consent to participate in the trial was obtained from all participants.

| Intervention
The intervention comprised provision of hand cream dispensers placed at accessible locations in the wards, continuous electronic monitoring of cream use, and feedback on cream use at ward level.
The feedback was provided by means of posters reflecting the compliance of HCWs with the skin care recommendations that they were given. The HCWs were instructed to use hand cream at least two times per shift, preferably before starting a shift, following wet work activities, or after finishing a shift. The hand cream in the electronic dispensers, Stokoderma Aqua Sensitive from the Stoko-Deb Group, was a perfume-free and silicone-free soft cream, with no particular pharmacological function. The main ingredients were moisturizing compounds, such as glycerol and urea. 18 Both the intervention group (IG) and the CG received basic education on skin care and skin protection behaviour every 3 months from baseline to the end of the study.
This education took the form of a short small-group lesson lasting 5 to 10 minutes. The objective was to assess the effectiveness of the intervention in reducing the severity of HD in HCWs exposed to wet work.
The primary outcome was the difference in the Hand Eczema Severity Index (HECSI) score between baseline and 12-month follow-up (ΔHECSI). The secondary outcome was the difference in NMF levels determined in the SC samples collected at baseline and followup (ΔNMF).

| Assessment
At baseline, HCWs completed a questionnaire on self-reported atopic dermatitis, here defined as a history of flexural dermatitis on the elbows or knees, a history of HD, glove use, frequency of hand washing, and the use of hand alcohol and hand creams at work. All outcomes were evaluated at the individual level.
The clinical assessment of the hands was performed by a trained physician (M.S.) using the HECSI scoring system. 20 The HECSI grades the intensity (scale 0-3) of clinical symptoms, including erythema, vesicles, fissures, scaling, papules and oedema in combination with the extent (affected area) of disease (scale 0-4) per area of the hand (fingertips, fingers, palms, backs of hands, and wrists). To obtain the total score, the extent per location was multiplied by the intensity of all clinical symptoms and summed. 20 The HECSI score ranges from 0 (no dermatitis) to 360 (extremely severe dermatitis). The severity of disease was classified as mild (0-11 HECSI points), moderate (12)(13)(14)(15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27) points) and severe (≥28 points) dermatitis. 21 The minimal detectable change (MDC) in the HECSI score was defined according to Norman et al 22 as 0.5 times the standard deviation of the mean ΔHECSI value. 22 The MDC represents the smallest change that can be detected by a measurement instrument that corresponds to a true change rather than being attributable to measurement error.

| NMF analysis
The SC samples for the NMF analysis were collected with a tapestripping technique. Round, adhesive tapes (3.8 cm 2 , D-Squame; CuDerm, Dallas, Texas) were attached to the dorsal part of the dominant hand, and pressed on for 5 seconds with a standardized force (D500, D-Squame Pressure Instrument; CuDerm). The tape stripping from the same skin site was repeated with a new tape five consecutive times. For the NMF analysis, the fifth consecutive tape was used. 18,23 For the analysis, a slightly modified method of Dapic et al was applied. 24 Briefly, NMF was extracted from the tapes by two successive extractions with 0.5 mL of water. The concentration of NMF in the extracts was analysed with high-performance liquid chromatography with ultraviolet detection. 24 In contrast to the published protocol, the protein levels on the tape used to compensate for variable amounts of the SC being removed by a tape were not estimated from the optical density values, because visual inspection of the tapes showed that the SC was not evenly distributed over the tape, which is the prerequisite for applying this method. Instead, the protein amount on the tape was determined in the extracts with a Pierce bovine serum albumin assay (Thermo Fisher Scientific, Rockford, Illinois). 24 The NMF levels were expressed as mmol/g protein.

| Randomization and statistical analysis
We based the sample size calculation on the Osnabruck Hand Eczema Severity Index score, 25 as there were no studies available on primary prevention in which the HECSI was used, and the two scores are well correlated (r = 0.85; P < 0.001). Using a two-sided t test with a significance level of 0.05, we calculated that a study with 17 wards per treatment group with 16 HCWs per ward (a total of 544 HCWs) would have 81% power to detect a difference of 0.4 in group means, 25 on the assumption of an SD of 1.2 points and an intracluster correlation of 0.05. 18 These power calculations were performed in PASS 15 (NCSS, Kaysville, Utah).
The HCWs were randomized to the IG or the CG at the ward level. Wards (as the unit of randomization) were randomized in fixed size blocks of two, and stratified into "high" or "low" levels of exposure to "wet work." Wet work exposure was estimated at the ward level from the quantity of soap purchased in the period January to The HCWs were not blinded to treatment allocation.
The statistical analysis plan has been described in detail previously. 19 Briefly, characteristics of wards, working years, working hours, sex, self-reported atopic dermatitis, self-reported HD in the past year and baseline NMF and HECSI values are presented by the use of descriptive statistics, and no formal statistical testing was performed. We used counts and percentages to present categorical variables. For continuous variables, the mean and SD (normally distributed data) or the median and interquartile ranges (IQRs) (deviation from a normal distribution) were used. The intraclass correlation (ICC) was also calculated. Two-sided P values of <0.05 were considered to be statistically significant. Statistical uncertainty was expressed by the use of two-sided 95% confidence intervals (CIs). The analyses were performed by an investigator (M.S.) supervised by the other investigators (S.K. and J.S.) and the Clinical Research Unit of the AMC. All statistical analysis was performed with IBM SPSS statistics version 24 (IBM Corp., Armonk, New York). When writing this manuscript, we adhered to the CONSORT statement. 26,27 The main analyses of the primary and secondary outcomes was performed on a modified 'intention-to-treat' population, consisting of all participants with a HECSI score at baseline and 12 months (i.e. per protocol population). We present crude means and 95%CIs for ΔHECSI and ΔNMF for the IG and the CG. To model the absence of ΔHECSI values when HECSI scores at 12 months were not available, we used a simple joint model approach. We used a linear model for ΔHECSI and ΔNMF, and a binary model with a logit link function for the missing data. We obtained P values for the difference between the IG and the CG by using generalized estimating equations with an exchangeable working correlations matrix to account for clustering within wards. We adjusted the analysis of the primary outcome for the binary stratification factor ward level exposure to wet work in the preceding year.
We performed a sensitivity analysis by using the observed (rather than predicted) ward level exposure to wet work and, for missing data, on the primary outcome at 12 months. Furthermore, a sensitivity analysis was carried out for participants with missing HECSI score data at 12 months. For this purpose, we performed three types of sensitivity analysis: (a) assuming the best possible outcome (HECSI score of 0); (b) assuming the worst possible outcome (highest observed HECSI score); and (c) applying multiple imputation by using baseline characteristics for the difference between baseline and 12-month HECSI scores.
As some intervention studies have reported relative change in disease severity, 12 we performed a post hoc analysis of the relative change in the HECSI score from baseline. This was calculated by use of the formula (ΔHECSI/HECSI T0) × HECSI T0. Furthermore, we used the MDC as proposed by Norman et al to a define meaningful improvement in HECSI score in this trial. 22 The MDC was calculated as (0.5 × SD) = 0.5 × 10.2 = 5.1.

| RESULTS
The study flowchart is shown in Figure 1. We included 501 HCWs from 19 wards; on average, 6 to 58 HCWs per ward were included.
Nine wards with a total of 285 HCWs were allocated to the IG, and 10 wards with a total of 216 HCWs were allocated to the CG. Nine wards with a total of 169 HCWs in the IG and 10 wards with a total of 121 HCWs in the CG had high exposure to wet work ( Table 1) The baseline characteristics of the HCWs are shown in Table 1.  There was no significant difference in the change in HD severity between the IG and the CG; the decrease in HECSI scores was 2.6 points higher in the IG than in the CG, but this difference was not sig-

| Post hoc analysis: Relative decrease in HECSI scores and MDC
We performed three post hoc analyses that had not been defined in the statistical analysis plan 19 : the relative change in HECSI scores, the MDC for HECSI scores, and subgroup analysis. The relative change in HECSI scores from baseline to follow-up (percentage of the baseline value) is shown in Table 3. At follow-up, the percentage changes in HECSI scores were 56% in the IG and 44% in the CG, representing a significant difference between the two arms (P = 0.001; ICC = 0.085).
We calculated the MDC to be 5.1 HECSI points, which is a value that was only reached in the IG.
To explore whether HD severity plays a role in the effectiveness of the intervention, we performed a subgroup analysis based on HECSI scores at baseline. To define the two subgroups, we used a cut-off of HECSI score of 11 points, which was defined by Hald

| Secondary outcomes
At baseline, the NMF levels were similar between the IG and the CG. At follow-up, NMF levels decreased in both groups: −1.0 (SD 1.6) in the IG, and −1.2 (SD 1.6) in the CG. There was no significant difference between the IG and the CG. The electronically measured cream use in the IG was, on average, 0.4 application events per HCW per shift during the trial. 28

| DISCUSSION
With respect to the primary and secondary outcomes, that is, respective changes in HECSI and NMF from baseline to follow-up, this RCT could not provide evidence for effectiveness of the intervention.

| Primary outcome
Although, in the present study, we found no significant effect of the intervention on the main outcomes, the intervention did have overall positive effects on the severity of HD symptoms, especially and significantly in specific subgroups. The decrease in HECSI score was larger in the IG than in the CG (respectively, 6.2 and 4.2 points). Furthermore, in contrast to the CG, the decrease in the IG exceeded the estimated MDC of 5.1 points. As some intervention studies have reported relative changes in disease severity, 12 we performed a post hoc analysis, which showed relative improvement in HECSI score to be significantly larger in the IG than in the CG (56% vs 44%; P < 0.001). 12 To explore whether HD severity plays a role in the Intervention studies focusing on the effectiveness of skin care in high-risk occupations are scarce, and skin care is often part of a multicomponent prevention programme. 12 The effect of moisturizers vs no treatment has previously been investigated in three studies focused on the primary prevention of occupational HD; all three suggested a beneficial effect of moisturizers. 12,13,29 On the basis of these three studies, a recent Cochrane review concluded that there may indeed be a clinically important risk reduction for the development of  (12) Working years, median (IQR) (n = 498) 11 (5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18)(19)(20) 11  Hours worked per week, mean (SD) (n = 499) 31 (7) 31 (6) History of atopic dermatitis, n (%) 35 (16) 41 (14) History emphasized that the quality of evidence is low. 30 This intervention was aimed at reducing HD symptoms by improving skin care. Indeed, the intervention showed significantly higher cream use in the IG than in the CG. 28 However, looking at the electronically collected data in the IG, 28 the average frequency of 0.4 cream applications per nurse per shift was much lower than recommended in the present study and in current guidelines (two applications per shift).
In this trial, we provided, both in the IG and in the CG, basic educational lessons to achieve the same level of knowledge, and tried to avoid differences between the two arms. As previously studied, education itself could have a positive effect on the severity of HD. 2,31 This may at least partly explain why skin symptoms were improved not only in the IG but also (although to a lesser extent) in the CG during the trial.

| Secondary outcome
Interestingly, NMF levels in both groups were significantly decreased at follow-up as compared with baseline. Exposure to water and soap has been shown to reduce NMF levels by different mechanisms 32-35 ; however, this cannot explain the decrease in NMF levels, as wet work exposure did not change during the trial. One of the possible explanations might be the introduction of a new disinfectant, containing npropanol instead of the formerly used ethanol, shortly before the start of our trial. In addition to the stronger skin barrier-damaging effect of n-propanol than of ethanol, 36 our recent work also showed an NMF level-decreasing effect of n-propanol. 33 The introduction of this new disinfectant might also explain another discrepancy in the NMF results. A reduction in NMF level has been suggested to be a contributory factor for dry skin in HD patients, and one may expect that a decrease in NMF level would parallel an increase in skin symptoms; however, this was not the case in the present study. The new disinfectant formula contained glycerol, which is a frequently used humectant in skin care products. 37 Notably, glycerol was also present in the hand creams provided to the IG. It may be speculated that the addition of glycerol to the new disinfectants could have at least partly compensated for the NMF level-decreasing effect on skin hydration associated with propanol. These positive effects of glycerol might also have interfered with the primary outcome, and may explain why the HECSI score was also improved in the CG at follow-up.

| Strengths and limitations
The strengths of our study included stratified randomization, controlling for cross-contamination, and comparable groups with regard to baseline values of the primary and secondary outcomes and covariates. In contrast to studies focused on patients, we included one highrisk occupational group (HCWs) under real occupational conditions. Assessment of outcomes and statistical analysis were performed blinded to allocation. To the best of our knowledge, this is the first study to use real-time monitoring to quantitatively assess the consumption of creams. In addition, we applied an objective and validated tool (HECSI score) for clinical assessment of the severity of HD by a trained physician. A quantitative biomarker of early skin barrier damage (NMF) was included as our secondary outcome.
There are several factors that might explain the lack of effect in the present RCT. One of the main drawbacks of this trial is the substantial loss of power resulting from overestimation of the number of available clusters, a higher loss to follow-up, and higher ICC values than assumed. Driven by organizational constraints during recruitment, we had to regard several subdepartments together as one cluster. Although these wards were at different locations, they shared the same management, and cross-contamination was likely, as HCWs rotated within such large clusters. However, this aspect was not foreseen when power was calculated, leading to a considerable loss of   30 Finally, the change in hand alcohol (from an ethanol-containing to a propanol-containing formula and the addition of glycerol) on the wards just before this trial started is also considered to be a limitation.
This is relevant, as it may have influenced both primary and secondary outcomes.

| CONCLUSION
This is the first trial to report the effectiveness of a prevention programme in the healthcare environment focused on the application of creams combined with continuous monitoring and feedback on skin care performance. This intervention was reported to improve hand cream use. 28 However, the IG did not show significantly better improvement in the primary outcome (ΔHECSI) and secondary outcome (ΔNMF) than the CG. Notwithstanding this, the intervention showed overall positive effects on the severity of HD symptoms, supporting the benefits of creams in the workplace, in particular in HCWs with mild HD. As occupational health interventions tend to be complex and dependent on context, evaluation based strictly on the primary and secondary outcomes in the total group might not reflect the overall benefit of the intervention. 39 In the present study, we did not focus on the barriers to and facilitators of hand cream use; however, the fact that cream use, despite it resulting in some improvement during the trial, still remains quite low, is intriguing. To design successful HD prevention strategies in the future, further investigation of these factors is needed.