Workstream 4: a pilot, feasibility randomised controlled trial of negative-pressure wound therapy compared with usual care

The results of workstream 2 indicated that NPWT was, with little uncertainty, not clinically effective or cost-effective. This conclusion was, however, based solely on observational data, which, even with extensive adjustment, may be subject to unresolved confounding, leading to a lack of confidence in the finding. A RCT may therefore be justified, to enable definitive, high-level evidence regarding the clinical effectiveness and cost-effectiveness of NPWT to be generated.

Given the difficulties in recruitment identified in previous studies of NPWT,54,55 we conducted a pilot, feasibility study to assess the feasibility of conducting a larger RCT of this treatment. The study was registered as ISRCTN12761776 and the protocol has been published (see Appendix 12).56 The results of this study have been published (see Appendix 13)57 and the findings relating to nurse participation in the pilot, feasibility study have been published (see Appendix 14).

Objectives

• To test the methods and feasibility of a full RCT of NPWT compared with usual care (no NPWT) for SWHSI.
• Specifically, to determine:
  • likely recruitment rate, including potential participants’ willingness to be randomised and if this depends on wound location and other factors
  • clinicians’ willingness and ability to recruit and randomise participants
  • testing of inclusion and exclusion criteria
  • fitness for purpose of data collection methods, including across and between care settings
  • ability of sites and clinicians to supply NPWT to participants in a timely fashion, irrespective of care settings, and assess any training requirements
  • ability of community staff to manage participants randomised to NPWT
  • suitability of method of outcome ascertainment
  • adequacy of duration of follow-up
  • rates of withdrawal from treatment, response rates to questionnaires, attrition from the trial and likely rates of missing data for outcomes
  • other methodological issues, including feasibility of blinded outcome assessment and whether or not trial documentation is acceptable to NHS nurses collecting data
  • feasibility of collecting data for and conducting of economic analyses in a larger trial (through exploratory analyses).

Methods

The pilot, feasibility RCT was conducted over a 10-month period (20 November 2015 to 30 September 2016: recruitment 20 November 2015 to 30 June 2016 and follow-up 20 February 2016 to 30 September 2016) in acute and community settings at three study sites in Leeds and Hull, Yorkshire, UK. Prior to study activity commencing, approval was obtained from Yorkshire and the Humber – Leeds East Research Ethics Committee (reference 15-YH-0307) and from the associated NHS trusts.

Patients were eligible for inclusion if they were aged ≥ 18 years and able to give full informed consent; were receiving care from Hull and East Yorkshire Hospitals NHS Trust (now Hull University Teaching Hospitals NHS Trust), Leeds Teaching Hospitals NHS Trust or Leeds Community Healthcare NHS Trust; had a SWHSI that could be reasonably treated with NPWT and wound dressings; had a SWHSI that was considered ready for NPWT (i.e. minimum 80% viable tissue or thin layer of slough requiring no further debridement); and were receiving adequate nutrition (as assessed by the senior nurse responsible for nursing care). Patients were ineligible for inclusion if they had limited life expectancy; active systemic infection; inadequate haemostasis or risk of bleeding; chronic wounds (e.g. pressure or foot ulcers) that were non-surgical in origin but had been surgically debrided; or had wound characteristics that precluded the use of NPWT (e.g. unclear undermining, necrotic or malignant tissue, eschar, exposed blood vessels, organs, anastomotic sites or nerves). Patients were also excluded if they had previously been or were currently receiving NPWT on their SWHSIs (including applications while in theatre for the surgery resulting in the SWHSI), if the wound was located where a vacuum seal could not be obtained, if they had participated in a research study in the previous 4 weeks, or if they were unwilling to have wound photographs taken.

Health-care professionals initially identified and screened patients for eligibility. Those meeting eligibility criteria were approached with further details of the study by a clinical or research nurse (subject to patient consent to approach), prior to study consent being obtained. As the main aim of this study was to determine the feasibility of conducting a larger trial, rather than to detect a treatment effect, a formal sample size calculation was not required. We aimed to randomise 50 patients over a 7-month recruitment period.

Baseline data were collected, including participant, surgery and wound information, wound dimensions (measurements and wound tracing) and a photograph, pain assessment [using a visual analogue scale (VAS) and the BPI]23 and quality-of-life data (SF-12 and EQ-5D-3L).21,22 As this was a feasibility trial, a primary outcome was not defined; rather, a range of feasibility (e.g. recruitment rate, time to intervention delivery) and clinical (e.g. time to wound healing) outcomes were collected to inform a future definitive trial. Participants were followed up for a maximum of 3 months, unless they withdrew consent, died or became lost to follow-up. Clinical outcomes were collected prospectively on a weekly to fortnightly basis at follow-up assessment visits conducted by the research nurses, including wound dimensions and photographs, healing status [defined as ‘complete epithelial cover in the absence of a scab (eschar)’], infection, hospital admission, return to operating theatre and treatment changes. Participants also completed self-report postal questionnaires (EQ-5D-3L,22 BPI,23 SF-1221 and resource use and pain via a VAS) at 2 weeks, 1 month and 3 months post randomisation. When consent was provided, participants were also contacted by text message on a weekly basis and asked to provide a numeric rating of their wound pain in the previous week. When participants failed to respond, one further message was sent before text message data collection discontinued. When participants’ wounds were reported to have healed during the study, text messaging data collection ceased at the point of wound healing.

Information concerning acceptability of study participation was collected from participants using Likert scales. Data relating to acceptability of documentation and training requirements for NPWT were initially collected from research nurses, using a Likert scale. Qualitative interviews were then completed with the research nurses to ascertain additional information regarding acceptability of the study processes, documentation and delivery.

Participants were allocated (1 : 1) to the intervention (NPWT) or usual care (no NPWT), using random permuted blocks, with stratification by wound area (< 28 cm², ≥ 28 cm²). Allocations were concealed through a central randomisation service at the University of York.

The study intervention (NPWT) was delivered using NPWT devices that were in use in the acute and community services in Hull and Leeds, which included V.A.C^® (KCI Medical Asia Pte Ltd, Singapore), Renasys^® (Smith & Nephew, London, UK) and PICO^® (Smith & Nephew, London, UK) devices. The NPWT treatment delivery (e.g. device type, pressure cycles, dressing change and canister emptying frequency, application and removal of the device and treatment duration) was as per the standard practice at the time to ensure that the pragmatic nature of the pilot feasibility study was maximised. This was therefore decided by the clinical care team in conjunction with the participant and nurse, with the only stipulation for NPWT being that use must be clinically appropriate.

The control group participants received usual care (wound dressings, no NPWT), with dressing type and frequency of change left to the discretion of the clinical care team. As there is no evidence to suggest that any one dressing is more clinically effective or cost-effective than another,12 types of dressings were not stipulated for the trial.

Owing to the nature of the intervention, it was not possible to blind participants or health-care professionals to trial treatment; however, the feasibility of blinded outcome assessment of healing was assessed within the trial using study photographs. Initially, it was proposed that blinded review of healing would be completed for every participant photograph obtained, but this was reduced during the study to review of only the final photograph, due to time and resource constraints. Assessment of blinding to treatment allocation was conducted through review of photographs taken at the week 1 follow-up.

Data analysis was conducted using Stata, using the principles of intention to treat and two-sided statistical tests at the 5% significance level, when appropriate. Summary statistics for all variables were generated by trial arm and overall. Time to healing was defined as the number of days from randomisation to date of healing. Data were right censored at the date of the final assessment visit for participants not deemed to have healed. Time to healing was assessed using Kaplan–Meier curves, recognising that the study was not powered to detect a treatment effect. A Cox proportional hazards regression was used to investigate the impact of covariates (contamination level of surgery, wound infection, wound size, SWHSI history and location). Cost-effectiveness was explored using total mean resource use costs and mean QALYs per trial arm.

To understand improvements to trial processes, to support research nurses in conducting a larger RCT in this area, qualitative interviews with eight research nurses were completed. Interviews were audio-recorded and fully transcribed before being analysed for thematic content using a ‘framework’ approach.41,58 Interpretation of the findings enabled linkage between individual nurses and their responses to account for variation between the type of study site (acute or community).

Prior to analysis, wound tracings were measured using Mouseyes software (R Taylor, Manchester, UK) to obtain perimeter and area estimates.59 Both adjusted (length × width × Π)60 and unadjusted (length × width) wound area (cm²) measurements were calculated using ruler measurements obtained at clinical follow-up visits.

Results

Recruitment and participation

Recruitment to the study commenced on 20 November 2015 and continued until 30 June 2016. Delays in study set up, beyond the control of the research team, had an impact on recruitment rate in the early stages of the trial; there was only one recruiting site active until the two remaining sites were opened to recruitment in February and March 2016, respectively. The overall target recruitment rate for the study was seven participants per month and, following successful opening of all three recruitment sites, the study achieved this target in the final 4 months of recruitment (Figure 4). The flow of participants through the trial is shown in Figure 5.

FIGURE 4 Cumulative target and actual participant recruitment. (Figure not reproduced here; see the original article.)

FIGURE 5 Pilot feasibility RCT: Consolidated Standards of Reporting Trials (CONSORT) flow diagram. a, Reasons not mutually exclusive. (Figure not reproduced here; see the original article.)

Of 248 patients screened for eligibility, 75.0% (n = 186) were ineligible (including one participant found to be randomised in error). The most common reason for ineligibility was that the patient was receiving or had previously received NPWT on their current SWHSI (45/248, 18.1%), had NPWT applied while in theatre (n = 7/248, 2.8%) or both (24/248, 9.7%). Twenty-two participants were eligible but did not wish to consent to study participation; the most common reason for declination was not wishing to receive NPWT (5/22, 22.7%). Reasons for ineligibility were largely distributed across the observed wound locations, with the exception of ‘receiving inadequate nutrition’, which was only reported for five patients with abdominal wounds, and ‘presence of exposed vessels or organs’, which was reported for 11 abdominal and two leg wounds only.

Forty-one of the 248 screened patients (16.5%) were randomised to the pilot feasibility RCT, with 40 participants included in the study analyses (one patient was found to be ineligible following randomisation and so was withdrawn from study data collection and analysis). Of those correctly randomised, 19 out of 40 patients (47.5%) were allocated to NPWT and 21 out of 40 patients (52.5%) were allocated to usual care.

The mean age of participants was 57.8 (SD 14.4) years and just over half (n = 22, 55.0%) were male. Participants were, on average, overweight [mean BMI 29.1 (SD 5.7) kg/m²], 60% (n = 24) were diabetic and 10 (25.0%) reported being a current smoker. The most common location of the wounds in recruited participants was the foot (n = 24, 60.0%), of which 19 out of 24 (79.2%) were due to toe or foot amputation, two (8.3%) were due to incision and drainage, two (8.3%) occurred following debridement (of amputation site, n = 1; of non-chronic ulcer, n = 1) and the cause of the remaining foot wound was unknown. Usual-care participants had their wound for a median of 12 days and over one-third (8/21, 38.1%) had a wound infection at baseline. NPWT participants had their wound for a median of 7 days and 26.3% (5/19) of participants had a wound infection at baseline. Baseline characteristics are summarised in Table 18.

Data collection methods

Twenty-six participants (65.0%) consented to data collection via text messaging; 20 participants (76.9%) subsequently provided at least one response (NPWT, n = 10; usual care, n = 10). Response rates to text messaging varied throughout the trial, ranging from 57.7% of those receiving a text in the week following randomisation to 100% of those receiving a text in week 12. Participants who responded to text messaging tended to be younger than those who did not respond (mean age 51 years vs. 60 years, respectively) and non-responders were predominantly male (85.7%). All non-responders had a foot wound, had at least one comorbidity and over half of those who did not respond to text messaging had been allocated to the control arm (57.1%). A range of wound locations were observed in those participants who did respond to text messages (e.g. foot wounds accounted for 36.8%) and 78.9% of participants had one or more comorbidities. Participants who responded to text messaging were almost evenly split between the intervention and control groups (52.6% vs. 47.4%, respectively).

Wound area (cm²) was collected by ruler measurement and wound tracing at each follow-up visit. Ruler measurements were found to overestimate the true area as obtained using wound tracings (mean difference 7.0 cm², SD 9.1 cm²), even when an adjustment of ‘Π/4’ was used. Wound area decreased in both treatment groups throughout the trial.

The movement of patients from acute to community settings did not substantially affect follow-up completion. In the majority of cases, follow-up continued as planned; however, qualitative interviews with the study nurses showed that, on occasion, it was impossible to assess the wound at follow-up visits. This was primarily due to community health-care professionals (e.g. district, community or general practice nurses) either having already attended and redressed the wound or not attending at the agreed time for the visit, meaning that research data could not be collected in full.

Negative-pressure wound therapy supply

Nineteen participants (of 40, 47.5%) were allocated to receive NPWT; however, two participants (of 19, 10.5%) did not receive the allocated intervention (one because NPWT would have caused delay in discharge and the other because all NPWT devices were with other patients in use in the community at that time). Approximately three-quarters of allocated participants (14/19, 73.7%) received NPWT within 48 hours of randomisation. Seven participants (of 19, 36.9%) did not receive NPWT within 24 hours of randomisation, with the primary reason for delay being lack of device availability (6/7, 85.7%). Participants received NPWT for a median of 18 (range 0–72) days, with eight participants discontinuing NPWT due to wound improvement. One participant recommenced NPWT following their initial cessation. Two participants (of 19, 10.5%) ceased NPWT therapy on the day they received it due to failure to maintain a seal. Twenty-one participants (of 40, 52.5%) were allocated to usual care, of whom more than half were initially treated with a hydrocolloid dressing (11/21, 52.4%), nine with a silver-containing (n = 3, 14.3%), iodine-containing (n = 3, 14.3%) or basic dressing (n = 3, 14.3%), and one with a soft polymer dressing (4.8%). Five usual-care participants (23.8%) switched to NPWT during the trial at a median of 4 (range 0–17) days after randomisation and subsequently remained on NPWT for a median of 31 (range 19–36) days. The most common reason for treatment crossover was wound deterioration (2/5, 40.0%).

Withdrawal and retention

During the trial, two participants (of 40, 5.0%) were fully withdrawn from the study due to amputation of the limb on which the reference wound was situated. The study protocol was revised following these withdrawals to enable continued data collection from participants after amputation, provided that they continued to give consent for this study. However, in the event, no further participants underwent amputation.

Participant questionnaire response rates were positive, achieving ≥ 77.5% at all time points, and 291 follow-up clinic visits were completed (NPWT, median 8; usual care, median 7) at intervals of between 4 and 33 days after randomisation (NPWT, median 7.5 days; usual care, median 9 days).

Blinded outcome assessment

In total, 308 photographs were taken as part of the study (38 at baseline, 270 at follow-up visits). Assessments of the final photographs for healing indicated that there was agreement between two reviewers for 29 (of 40, 72.5%) participants (NPWT, n = 15, 79.0%; usual care, n = 14, 66.7%), increasing to 38 participants (95.0%) following a third review. When healing status was recorded by one or more reviewers as ‘unsure’, this was due to poor photograph quality in 50.0% (4/8) of cases. For two participants (of 40, 5.0%), blinded outcome assessment of healing provided no consensus, with each reviewer recording a different healing status. The blinded assessors classed eight (of 40, 20.0%) wounds as healed (NPWT, n = 3, 15.8%; usual care, n = 5, 23.8%), of which seven (87.5%) were also recorded as healed by the unblinded investigators. Blinded assessors were unsure of healing status, or there was discordance between reviewers, for a further three wounds judged to have healed by the unblinded investigators.

The feasibility of blinding to treatment allocation was assessed using photographs taken at week 1, as the final photograph would not necessarily reflect the originally allocated treatment (given that NPWT is not necessarily used to the point of wound healing). In terms of blinding success, the assessors correctly identified treatment allocation for 15 participants (of 40, 37.5%): nine participants allocated to NPWT and six participants allocated to usual care. Allocation to NPWT was incorrectly identified in six cases (15.0%).

Documentation acceptability: nurses

Half of the eight research nurses who responded (n = 4, 50.0%) indicated that they found that the questionnaires asked all relevant questions and were straightforward to complete. There were mixed opinions on the ease of participant identification, with three nurses agreeing (agree or strongly agree), two nurses remaining impartial (neither agree nor disagree) and three strongly disagreeing. Half of the nurses (n = 4, 50%) indicated that trial processes were clear and the randomisation process was straightforward and just over half (n = 5, 62.5%) found that the assessments were straightforward to complete.

Qualitative interviews were subsequently conducted to ascertain further information on responses provided. Five face-to-face and three telephone interviews were conducted with research nurses across the three study sites, which identified key points of feedback on the trial procedures and processes:

• Trial documentation was found to be acceptable, straightforward and in keeping with expectations based on nurses’ prior research experience. Amendments to clarify consistency in interpretation of eligibility criteria, processes for recording of wound depth and general minor amendments to study documentation were, however, suggested. The nurses acknowledged that the majority of issues encountered with the implementation of this pilot, feasibility RCT were not dissimilar to other new studies.
• Nurses based at the secondary care sites highlighted issues with engagement of consultants, due to concerns that randomisation could interfere with their preferred wound management plan. Nurses suggested that direct principal investigator involvement enabled more patients to be successfully screened at the preoperative stage. Nurses based within the community settings noted that the majority of patients were identified in the acute settings. Randomisation was regarded as an easy process, although availability of a 24-hour randomisation service would have been preferred.
• Given the geographical spread of participants at each study site, arranging visits to coincide with a patient’s clinic appointment was the most efficient approach, although this was not always viable and often time was wasted waiting for district, community or general practice nurses at home visits. The frequency of follow-up visits was noted as having an impact on time available to recruit and the value of continued follow-up beyond wound healing (up to three visits completed post healing) was queried. Access to equipment was generally better in the acute setting than in the community setting, although there were occasions when equipment needed to be borrowed by the trial team across both acute and community settings to facilitate patient treatment.
• There was mixed feedback regarding nurses and consultants indirectly involved with the study. At one site, there was a perception that if NPWT was involved, some nurses would avoid a study due to a lack of confidence or training in its application and also the potential for additional work involved. At another site, it was reported that negative pressure is widely used and therefore there was a potential bias towards use of negative pressure. It was suggested that additional principal investigator involvement to promote the study, particularly during the set-up phase, would be helpful.
• All research nurses received training in the application of NPWT but, as a result of randomisation, not all nurses had the opportunity to apply it. Generally, there was confidence in the use of NPWT; however, staff turnover and levels of junior staffing were noted to highlight gaps in training. Nurses reported that for patient safety, patient lifestyle, in particular mobility requirements, should be taken into account during recruitment.

Study acceptability: patients

Feedback from participants indicated that taking part in the study had matched expectations (30/31, 96.8%), and that completing questionnaires had been straightforward (28/31, 90.3%) and manageable (29/31, 93.6%).

Clinical outcomes

Healing

Participants were followed for a median of 84 days after randomisation (range 13 to 105 days) to assess wound healing. During the study, 10 wounds (of 40, 25.0%) were deemed to have healed based on assessment at weekly follow-up [six in the NPWT group (31.8%) and four in the usual-care group (19.1%)]. Time to healing was summarised using Kaplan–Meier curves (Figure 6). Adjusted Cox proportional hazards models indicated that wound size (p = 0.03) and duration of wound (p < 0.001) were predictors of time to wound healing, with smaller and longer duration of wounds associated with increased time to healing (see Appendix 13).

FIGURE 6 Pilot feasibility RCT: Kaplan–Meier curve – time from randomisation to healing. (Figure not reproduced here; see the original article.)

Infection

During the trial, 17 participants (NPWT, n = 6, 31.6%; usual care, n = 11, 52.4%) experienced an infection. Of these, nine experienced infection at baseline only (NPWT, n = 3, 15.8%; usual care, n = 6, 28.6%), four during follow-up only (NPWT, n = 1, 5.3%; usual care, n = 3, 14.3%) and four both at baseline and during follow-up (NPWT, n = 2, 10.5%; usual care, n = 2, 9.5%).

Hospitalisation

Participants in the NPWT group spent a median of 29 nights in hospital during the trial and participants in the usual-care group spent a median of 10 nights in hospital.

There were five reported hospitalisations during the trial, all of which were associated with serious adverse events (as Adverse events). In four instances (of five, 80.0%) the participants were subsequently discharged and in one instance (20.0%) the participant was withdrawn from the study (due to amputation) and subsequently died, and so further information regarding hospital discharge was not collected.

Patient-reported outcomes

Participant-completed questionnaires indicated that pain severity (BPI),23 interference and VAS wound pain reduced (improved outcome) in the NPWT group between baseline and month 1, but increased again (worse outcome) slightly at month 3 of follow-up. In the usual-care group, pain interference followed the same pattern as in the NPWT group, but VAS wound pain reduced consistently over time and BPI interference fluctuated between the follow-up time points. There was lower reported pain severity (improved outcome), interference scores (BPI)23 and wound pain (VAS) for usual-care participants at all time points relative to the NPWT group (Table 19). A total of 20 participants provided at least one weekly text message providing a pain score (NPWT, n = 10; usual care, n = 10). Mean scores reduced between week 1 and month 3 (week 1 4.0, SD 3.0; week 12 1.6, SD 2.0); however, there was some fluctuation observed in scores during this time (see Appendix 15).

There was improvement (i.e. better health status) in the mean SF-12 PCS over time in both groups. MCSs reduced in the NPWT group over time, but improved initially in the usual-care group (between baseline and 1 month) before declining slightly at the month 3 follow-up visit. Further detail regarding these results is presented in Table 20 (see also Appendix 16, Figures 7 and 8).

Adverse events

Adverse event data collection throughout the study identified 28 non-serious and five serious adverse events. Two serious adverse events were reported in the NPWT group from one participant and three in the usual care group from two participants. All five events were hospitalisations and only one (usual-care group) was classed as related to the reference wound. Three events (of five, 60.0%) were said to be possibly or probably related to treatment, all of which were recorded as expected. There were no suspected unexpected serious adverse reactions. The majority of non-serious adverse events related to the reference wound (25/28, 89.3%) and were reported as definitely related to treatment (19/28, 67.9%). The most common event type was ‘other’ (12/28, 42.9%), which included events such as skin irritation (3/12, 25.0%) and dressings falling off (3/12, 25.0%), followed by ‘pump failure’ (8/12, 28.6%).

Economic outcomes

Usual-care participants reported that they received a mean of 8.0 (SD 14.8) nurse visits at home and had a mean of 12.41 (SD 15.04) visits at the GP surgery for their SWHSIs. NPWT participants reported a limited number of nurse visits at the GP surgery (mean 1.14, SD 2.24), but additional nurse visits at home (mean 6.78, SD 9.82) and substantial mean inpatient hospital stays (mean 17.5, SD 19.8) and GP visits (mean 1.35, SD 2.70) (Table 21).

Costs associated with NHS and health-care resource use were calculated on a per-contact basis, using NHS Reference Costs61 and Unit Costs of Health and Social Care62 (see Appendix 17). The associated NHS and health-care costs were higher for the NPWT group: the mean total cost for the NPWT intervention was £9490 (SD £7346), compared with £1153 (SD £1806) for patients receiving usual care (Table 22).

Negative-pressure wound therapy participants reported a mean baseline EQ-5D-3L score of 0.34 (SD 0.10) and usual-care participants reported a mean baseline score of 0.54 (SD 0.08). At 3 months, following adjustment for baseline utility, mean scores had improved over time [NPWT participants 0.49 (SD 0.35) and usual-care participants 0.77 (SD 0.23)] (Table 23, see also Appendices 18 and 19).

The results show that NPWT had higher total NHS and health-care costs and also achieved lower QALY gains than usual care (complete-case analysis –0.007, 95% CI –0.04 to 0.02; multiple imputation analysis –0.004, 95% CI –0.03 to –0.02).

During the study, 1135 dressing changes were completed (NPWT, n = 518; usual care, n = 617). The average number of dressing changes completed was 27.26 for NPWT participants and 29.38 for usual-care participants. The majority of dressing changes were completed by district, community or general practice nurses (NPWT, n = 253; usual care, n = 272) or the participant’s clinic nurse (NPWT, n = 107; usual care, n = 167). The distribution of dressing changes by group corresponds to the distribution by group of nursing visits received or attended by participants: nursing visits at home or at the GP clinic were higher for usual-care participants than those allocated to NPWT.

Conclusions

This pilot, feasibility RCT provides invaluable information on the feasibility of conducting a large-scale trial of NPWT for treatment of SWHSIs and identifies areas of key consideration in the planning of future research.

Recruitment, participation and retention

The consistent rate of recruitment observed following the opening of all sites to recruitment, demonstrates that it is possible to recruit participants to a trial of NPWT compared with usual care. A full trial would, however, need to ensure that an adequate number of recruiting sites are identified. On the basis of the recruitment rate observed (two patients per site per month), to recruit 400 participants to a similar trial over an 18-month recruitment period, 11 study sites would need to be opened to recruitment. This example is based on hypothetical detail and would therefore be subject to revision following full sample size calculations for a larger trial. A larger trial should consider the impact of, and account for in associated sample size calculations, the rates of participants randomised to, but subsequently not receiving, NPWT (10.5% of participants in the pilot feasibility RCT) and rates of crossover to NPWT (23.8% of usual-care participants in the pilot feasibility RCT).

Qualitative interviews indicated that nurses perceived the majority of SWHSI patients to be identified within acute care settings, highlighting that the primary source of recruitment in a future RCT may be best achieved in acute care settings. This distribution may, however, differ between individual localities and so work would be required prior to recruitment to ensure that appropriate and effective recruitment strategies and pathways were implemented at individual sites.

Having previously received, or currently receiving, NPWT for their SWHSI was found to be the most frequent reason for patients’ ineligibility to take part in the study. Prior to undertaking a full RCT, it would be necessary to undertake some promotional work with clinical colleagues and research groups in order to fully engage surgeons and nurses within the study site, and to ensure consistent and continued referrals into the study. In addition, a future trial should recruit study sites where there is sufficient equipoise with regard to treatment of patients with SWHSIs, thus thorough evaluation of potential trial sites will be essential.

Work to engage clinical colleagues and research groups prior to recruitment commencement would also be crucial to allay resistance to trial recruitment from certain clinical groups. Anecdotal evidence from the study sites indicated some resistance to trial recruitment participation from colorectal surgeons, either because NPWT was primarily used for wound management rather than healing or because a surgeon’s prior experience of NPWT had led them to feel that this treatment was inappropriate for use in this patient group. Given the small number of sites involved in this pilot feasibility study, it is impossible to draw firm conclusions on the likely resistance among specific clinical groups, which may be observed in a larger RCT. Promotional work with clinicians across potential research sites in advance of study commencement should, however, help to minimise any resistance to recruitment in a larger RCT. Where necessary, promotional work may need to be targeted to ensure that clinicians from a cross-section of surgical disciplines are aware of, and open to, recruitment for a larger trial.

Participant-reported questionnaire response rates in excess of 75% and the limited number of withdrawals observed also demonstrate that it is possible to retain participants in a RCT of NPWT compared with usual care. During the study, only two (5.0%) participants were withdrawn from the trial due to amputation of the limb on which the reference wound was situated. Data collection did not continue for these participants. The trial team subsequently amended the study protocol to reflect that clinical data collection would cease at the point of amputation, but participant-reported data could continue to be collected, subject to continued consent. A future trial should ensure that the process for continued data collection following amputation is discussed and is clearly defined prior to study recruitment commencing.

Data collection methods

Assessment of wound measurements indicated that ruler measurements, both with and without adjustment, overestimate wound size, compared with wound tracings. Wound tracings should therefore be chosen over ruler measurement in any future RCT.

We have shown how wound pain data can be successfully collected on a weekly basis using text messaging; however, there were limited numbers of participants who responded to this method of data collection and rates of response varied considerably throughout the study. Exploration of participant characteristics in relation to text message respondents and non-respondents indicated that collection of data using this method may be most effective in younger populations; those who routinely responded were, on average, almost 10 years younger than those who did not respond routinely. The use of text messaging data collection in a larger study should therefore be considered carefully, before implementation, to ensure that use of this method will provide sufficient data for analysis.

Negative-pressure wound therapy supply

Negative-pressure wound therapy delivery was largely completed within 48 hours; however, three participants did not receive NPWT within this time frame. Furthermore, two participants did not receive NPWT at all, due to delays to discharge and lack of intervention availability. A future RCT would therefore need to establish the availability of NPWT devices with sites from the outset, to ensure timely intervention delivery.

Nurses involved in the study reported that they had received sufficient training within their individual institutions to confidently apply NPWT. It was, however, noted that nursing colleagues were not always confident in using NPWT, which corresponds with findings from workstream 3 of this research programme. A future trial would therefore need to invest sufficient resource to ensure that NPWT training was provided to all staff who may be providing trial treatment to a participant.

Blinded outcome assessment

This pilot feasibility trial explored the feasibility of blinded outcome assessment in the context of NPWT for SWHSI healing. Blinded and unblinded investigator assessment in relation to wound healing agreed 87.5% of the time and so blinded outcome assessment can be deemed to be feasible and appropriate within a future RCT. Blinded assessment of healing was, however, often hampered by the quality of the wound photograph. A robust photography protocol would therefore be required for a full RCT to facilitate consistency of imaging throughout the study and accurate blinded healing assessment.

It was not possible within this study to blindly assess time to healing due to difficulties in transferring the necessary numbers of photographs required to complete this activity. Dependent on the outcomes included in a larger study, this may warrant further exploration and a robust and tested process for transfer of photographs to blinded assessors for review should be established prior to the study commencing.

It is important to note that assessors correctly identified treatment allocation in approximately one-third (37.5%) of cases when reviewing photographs taken in the first week of the study. Therefore, assessment of time to healing in a larger study may become biased through identification of treatment allocation and so inclusion of this assessment should be carefully considered before implementation.

Documentation acceptability

Participant questionnaires were largely well received and so limited changes would be required for data collection in a larger RCT. The variation in opinions from research nurses involved in the study regarding study documentation and study processes, especially frequency and method of follow-up visits, indicates that further work should be conducted during the design of a larger trial to develop and test data collection processes to ensure fitness for purpose.

Clinical events

Given the limited number of participants involved in this study, there were insufficient data to enable any true treatment effect of NPWT to be identified; and this was not the purpose of this study. A full RCT is therefore required to fully assess the clinical effectiveness and cost-effectiveness of NPWT as a treatment for patients with SWHSIs.

The cohort study undertaken before this pilot, feasibility RCT (workstream 1) estimated that a follow-up duration of 90 days should be sufficient to observe wound healing. However, in this pilot feasibility study, only 25% of participants healed, which meant that a median time to healing could not be calculated. As a result, a 3-month follow-up may be inadequate to reliably investigate and compare healing rate; therefore, a longer follow-up should be considered for future trials. The difference in time to healing between this pilot feasibility study and our cohort study may be explained by the greater proportion of participants with foot and leg wounds in the pilot feasibility study (usually secondary to diabetes mellitus). The complexities associated with wound healing in patients with diabetes mellitus63 may explain these lower healing rates.

Regional arrangements for the supply of NPWT in community settings may have resulted in between-group differences in relation to nights spent as an inpatient. It is, however, worth noting that this may be due to specific regional arrangements present in the participating sites, rather than being an issue that would manifest in a larger national study. To ensure that inpatient stay is not distorted by regional arrangements, it is imperative to assess provisions for treatment delivery (e.g. local discharge arrangements and availability of NPWT within community settings) prior to commencing research activity at any study site. Associated with NPWT supply is the aforementioned requirement for training. Nurses noted that clinical colleagues were not always confident in using NPWT and so sufficient resource would be required to ensure full coverage of training across all those who may provide treatment to a participant.

This pilot feasibility study has clearly demonstrated the feasibility of completing a full RCT to provide definitive evidence in relation to NPWT effectiveness as a treatment for patients with SWHSIs. The pilot feasibility study has also identified key elements of, and recommendations for, recruitment, data collection, outcome measurement, trial design and safety reporting, which would need to be considered during the design of a larger RCT to investigate the clinical effectiveness and cost-effectiveness of NPWT compared with usual care.