Statistical analysis plan for the Healthy Hands Project; single centre cluster-randomised clinical trial of a skin care program for the prevention of contact dermatitis in health care workers

Healthcare workers (HCWs) have increased risk of developing occupational hand dermatitis (HD) due to frequent exposure to ‘wet work’ [1, 2]. Wet work, defined as unprotected exposure to humid environments and water; high frequency of hand-washing procedures or prolonged glove occlusion, is believed to cause irritant contact dermatitis in a variety of occupations [2, 3].

Amongst HCWs, nurses run the highest risk of HD, with an estimated point prevalence of 12 to 30% [2, 4, 5]. Almost 60% of sick leave reported during the first year after notification of the disease is related to HD [6]. Hence, HD represents a significant burden for affected individuals as well as for society. The annual costs due to occupational skin diseases for medical care, absenteeism and disability pensions are estimated at €98 million in the Netherlands [2, 7].

In the Netherlands, the Dutch Society of Occupational Medicine (NvAB) established a guideline for the prevention of occupational hand dermatitis (OHD) in 2006 [2, 8]. This guideline emphasises the importance of the skin barrier for the prevention of HD and recommends regular use of skin care products such as emollients and ointments. Consistently, in a more recent update of contact dermatitis guidelines issued by The Netherlands Society of Dermatology and Venereology (NVDV) [9], regular use of emollients to help prevent irritant contact dermatitis has been recommended [10–12].

Several skin care programmes have been introduced in the healthcare setting to help prevent occupational skin diseases [2, 13, 14]. The effectiveness of these programs seems to depend on several factors: (1) the effectiveness of protective measures (i.e. use of skin care products), (2) the adherence to these measures (i.e. frequency of application) and (3) the effectiveness of education on preventive behaviour (by raising awareness about the risk factors for HE and the importance of protective measures) [2, 15]. To avoid skin dryness and improve skin barrier function a variety of moisturisers and emollients have widely been used. The effect of skin care products on the skin barrier has mainly been investigated in experimental irritation studies [16, 17], and randomised controlled studies in HCW are scarce. With respect to the effectiveness of education, several intervention studies have shown moderate evidence that education influences behaviour, leading to a reduction in skin symptoms [18]. Regarding adherence, despite evidence from intervention studies that skin care is effective in prevention of HD, the adherence to these preventive measure in the workplace remains low [19].

Monitoring and feedback is a widely used behavioural strategy to improve adherence [2, 20]. The effects depend largely on the way that the intervention is designed and delivered [21]. Recently, an electronic monitoring system has been developed for the continuous registration of hand cream consumption. This system enables a detailed feedback to HCWs on the frequency and moments when the cream is used. In hand hygiene studies, a similar monitoring system was shown to improve compliance by 42% [2, 21].

As previously described in the published protocol [2], the Healthy Hands Project (HHP), a single-centre cluster-randomised controlled trial (RCT) investigated whether an intervention program, based on the provision of hand creams and regular feedback on cream consumption, can improve the skin condition of the hands of nurses engaged in wet work, when compared to a ‘care as usual’ control group. The intervention comprised of the provision of hand cream dispensers on the wards, electronic monitoring of their use, and repeated feedback to the HCW. Randomisation to the intervention program or care as usual was at the ward level. The effectiveness of the program will be assessed by measuring hand skin condition in workers before and after the study period. This trial has been registered in the Netherlands Trial Register with identification number NTR5564. In this paper, we will provide an update on the status of the trial and present a detailed description of the proposed data analysis. This statistical analysis plan (SAP )was written and submitted before the researcher analysing the data (MS) was un-blinded to the treatment allocation.

The primary objective of the HHP trial was to examine whether the provision of a skin care programme for the prevention of HD improves the hand skin condition of HCWs engaged in wet work. Hand skin condition was assessed using the Hand Eczema Severity Index (HECSI) score [22]. The HECSI score grades the intensity of erythema, induration, papules, vesicles, fissures, scaling and oedema for five areas of each hand (fingertips, fingers, palms, back of hands, wrists) on a scale from 0 (not present) to 3 (severe). The extent of affected skin in each area is graded from 0 to 4. The total score is obtained by multiplying the intensity with the extent and ranges from 0 to 360 points. The primary outcome focusses on the change in HECSI score between baseline and after 12 months.

The secondary outcome is the change in levels of natural moisturising factors (NMFs) in the uppermost layers of the skin, as a marker of early signs of barrier damage [23].

The HHP trial was a single-centre, cluster-randomised, open, blinded-endpoint, parallel-group trial conducted on the wards of one University Medical Center in the Netherlands. This academic hospital has a total of 45 departments of which approximately 20 are clinical wards. Between 2 May and 14 June 2016, a total of 504 nurses on 19 wards were enrolled. The participants in this trial were randomised to the intervention or care as usual group at the ward level. Wards were randomised, two at a time, in a 1:1 ratio stratified according to wet work exposure (high or low) using fixed blocks of size 2. Wet work exposure was estimated at the ward level from internal purchases of soap in the period January 2016 to May 2016. Nurses and some investigators (SK) were not blinded to the allocated treatment group, but the primary outcome at 12 months was assessed by an investigator blinded to treatment outcome (MS).

This SAP focusses on the analysis of primary and secondary outcomes measured at baseline and 12 months. The extended follow-up period of 6 months (i.e. 12–18 months) stated in the flow diagram (Fig. 1) will be used to assess process outcomes, and will be out of the scope of the SAP described in the present article.

The inclusion criteria were: willing to give informed written consent; aged 18 to 65 years at the start of the study; having daily exposure to wet work activities during work; and being employed as a nurse or nutrition assistant on a participating ward. The only exclusion criteria was being employed on more than one ward.

The HHP included 19 wards and 504 participants in the Netherlands from 2 May 2016 to approximately July 2016. The participants were followed up for 18 months. At the time of submission, the trial data are blinded to treatment allocation and the trial has not completed follow-up of the last participant (end of study planned for January 2018). The data will be cleaned and checked for completeness and internal consistency, blinded to treatment allocation. The database will only be locked after this SAP has been submitted for publication.

The background to, and methods for, the HHP single-centre, cluster-RCT have been previously described in the published protocol [2]. In this paper, we have provided an update on the status of the trial and presented a comprehensive SAP for the resulting data. In addition, we have clarified differences between the previously published protocol and the proposed actual statistical analyses. The principal difference from the methods presented in the protocol is the actual inclusion of 19 rather than 34 wards and 504 rather than 544 participants. Consequently, the power has fallen substantially. This paper supports transparency in reporting in the HHP by clarifying differences between the previously published protocol and the proposed actual statistical analyses.