Negative-pressure wound therapy compared with standard dressings following surgical treatment of major trauma to the lower limb: the WHiST RCT

Internal pilot summary

The internal pilot took place at six recruitment centres over a period of 6 months using the same eligibility criteria and methods as the main study. The aim of this initial phase was to determine whether or not the number of eligible and recruited patients in the trauma recruitment centres over the course of 6 months would enable the delivery of a successful full trial.

The trial successfully recruited during the pilot phase and therefore continued to the main trial without any pause. Participants from the internal pilot were included in the final analysis.

Main randomised controlled trial summary

All adult patients presenting to hospital within 72 hours of sustaining major trauma and who required a surgical incision to treat a fractured lower limb were potentially eligible for inclusion. Randomisation, stratified by trial recruitment centre, open or closed fracture at presentation, and Injury Severity Score (ISS) of ≤ 15 versus ≥ 16 was generated and administered via a secure web-based service using minimisation in a 1 : 1 ratio. Participants were randomly allocated to either standard wound management or incisional NPWT.

The participants had clinical follow-up at the local fracture clinic for a minimum of 6 months, as per standard NHS practice after these injuries. Photographs of the wound were taken at 30 days, and a validated patient-reported questionnaire21 was used to assess the surgical scar. Functional and quality-of-life outcome data were collected using the DRI and the EuroQol-5 Dimensions (EQ-5D) questionnaire at 3 months and 6 months post surgery. Postoperative complications, including self-reported chronic pain [measured using the Douleur Neuropathique 4 Questionnaire (DN4)] and any further surgery related to the injury or the wound, were collected at the same time points, along with a resource use questionnaire. Completed case report forms (CRFs) were received centrally by a data administrator at the University of Oxford who entered the information into a secure password-protected database.

Inclusion criteria

Patients were eligible for the WHiST trial if they:

• were aged ≥ 16 years

• presented to hospital within 72 hours of injury

• had a major trauma injury and/or injury defined by eligibility for the UK TARN database

• had a lower-limb fracture requiring a surgical incision.

Some patients had major trauma affecting just one limb (e.g. heel, pilon and tibial plateau fractures). As the wounds associated with these injuries are always at risk of infection, we included these injuries even if the patient was subsequently not included in the TARN database.

Exclusion criteria

Patients were excluded from participation in the WHiST trial if:

• they had an open fracture of the lower limb that could not be closed primarily. Patients with open fractures that cannot be closed at the first surgery are at the highest risk of surgical site infection but incisional NPWT cannot be applied to these wounds

• there was evidence that the patient would be unable to adhere to trial procedures or complete questionnaires.

Patients who sustained injuries to other areas of the body as well as the lower limbs that could affect the primary outcome measure had their other injuries documented but were still included. For patients with more than one lower-limb injury, only the most severe wound was included as the ‘WHiST’ wound in the trial. It was at the surgeon’s discretion which injury was the most severe.

Consent

The consent procedure for this trial reflected that of the surgery, with the attending clinician assessing capacity before taking consent for the surgical procedure; this capacity assessment was then used to inform the appropriate approach to consenting to the WHiST study. A process approved by the National Research Ethics Service was used to gain consent from the patient or agreement from an appropriate consultee by an appropriately delegated member of the research team.

Conducting research in this ‘emergency setting’ is regulated by the Mental Capacity Act 2005 (MCA). 22 As patients may have lacked capacity, and the urgent nature of the treatment may have limited access to appropriate discussion with personal consultees, action was taken in accordance with section 32, subsection 9b of the MCA. This involved the clinical team making an assessment of capacity as per their usual procedures for obtaining consent for a surgical procedure. The clinical team then provided guidance to the research team as to whether the patient had capacity to consent prospectively or if consultee agreement needed to be sought.

Throughout the trial, best efforts were made to involve participants who, temporarily or permanently, lacked the capacity to decide to be involved in the trial. The clinical team made a judgement about the amount and complexity of the information that the participant was able to understand and retain. Appropriate information was communicated to the participant and updated as their understanding changed. At all times, the trial team acted in accordance with the participants’ best interests.

When the clinical team advised that prospective patient consent was appropriate, this was sought by the research team. If the clinical team advised that prospective patient consent was not appropriate, the research team approached an appropriate consultee. The main responsibility of a consultee was to advise the research team as to whether or not they thought that the participant would be happy to take part in the trial if they had capacity to consent. When a personal consultee was available, they were provided with the trial information. The personal consultee was someone who had a personal relationship with the patient, such as a family member, carer or friend. The personal consultee was given the opportunity to ask questions and discuss the trial, after which their agreement for the patient’s inclusion in the trial was recorded. When a personal consultee was not available, then a nominated consultee was identified to advise the research team. In most cases, a patient’s senior treating surgeon acted as the nominated consultee. If that surgeon was a member of the research team, another independent surgeon was identified. The nominated consultee was asked, after reviewing the trial documentation, to agree that the patient participated fully in the trial and all trial procedures; this was recorded during the electronic randomisation process.

Data collection, including linkage to routine NHS data sets, commenced as soon as consent or agreement by personal/nominated consultee had been obtained.

Patients who were able to consent before their operation were always approached. For patients who did not consent prior to surgery, the research associate provided them with all of the trial information at the first appropriate time when the patient had regained capacity. The patients were given the opportunity to ask questions and discuss the trial with their family and friends. They were then asked to provide written consent for continuation in the trial. Patients who did not consent prior to surgery and preferred not to be actively involved in the trial follow-up were asked if they were willing to consent to the research team using their routinely collected NHS data for the trial.

Patients were asked to consent to long-term follow-up (reported separately) and data linkage to routine NHS data sets. For patients who did not prospectively consent or who had a nominated consultee give prospective agreement and still lacked capacity after their surgery, every effort was made to contact a personal consultee to advise the research team about the patient’s continued participation in the trial.

If the consultee was present, they were asked to sign a consultee agreement form. When the consultee was not present at the agreement discussion (e.g. when they were being contacted via telephone), verbal agreement was recorded by the research associate on an informed agreement checklist. Personal consultees who preferred not to be actively involved in the trial follow-up were asked if they were willing to agree to the research team using the patient’s routinely collected NHS data for the trial. If no personal consultee was identified, the participant remained in the trial under the nominated consultee’s agreement provided at the time of enrolment.

Agreement for a participant to be involved in the trial was recorded in a patient’s notes. All original signed consent forms were kept in the investigator site file. Three copies of the consent forms were made: one was held in a patient’s medical notes, one was for the participant and one copy was for the trial team.

Responsibility for recording and dating both oral and written informed consent or agreement was with the investigator, or persons designated by the investigator, who conducted the informed consent discussion. Designated responsibility was recorded on the recruitment centre delegation log. Permission was sought to inform the participant’s general practitioner (GP) of their participation in the trial.

Allocation of treatment

All modern operating theatres include a computer with internet access, so a secure, 24-hour, web-based randomisation system was used to generate the treatment allocation intraoperatively. After the confirmation of randomisation and treatment allocation was received electronically by the surgical team, the allocated treatment could be administered immediately.

Blinding

As the wound dressings and topical devices were clearly visible, the treating surgeon and trial participants could not be blinded to treatment allocation. However, the treating surgeons were not involved in trial follow-up assessments or data collection for the trial. Data from clinical reporting forms were entered into a central database administered by a data clerk in the trial central office. Wound photographs taken at an outpatient clinic at approximately 30 days post surgery were reviewed independently by two experienced assessors (tissue viability specialists) blinded to the treatment allocation.

Standard dressing

The standard dressing for a surgical wound comprises a non-adhesive layer applied directly to the wound, which is then covered by a sealed dressing or bandage. The standard dressing does not use ‘negative pressure’. The exact details of the materials used were left to the discretion of the treating surgeon as per their routine practice, but the details of each dressing applied were recorded.

Incisional negative-pressure wound therapy

This uses a silicone contact layer with a silicon-based adhesive, an airlock layer, a superabsorbent layer and a polyurethane (semipermeable) layer on the top, which makes the system waterproof while allowing water vapour to pass. A sealed tube connects the dressing to a built-in mini-pump that creates a partial vacuum (–80 mmHg of negative pressure) over the wound.

It was applied to the wound at the end of the operation as per the treating surgeon’s normal practice and according to the dressing manufacturer’s instructions. The wound could be redressed again on the ward at the discretion of the clinical team; any further wound dressing was recorded and followed the allocated treatment unless otherwise clinically indicated.

Post-randomisation decline to consent and exclusions

Participants could decline to continue to take part in the trial at any time without prejudice. A decision to decline consent or withdraw did not affect the standard of care the patient received.

Participants had two options for withdrawal:

1. Participants could withdraw from completing any further questionnaires but allow the trial team to still view and retain, anonymously, any relevant hospital data that were recorded as part of normal standard of care, for example radiographs and further surgery information.

2. Participants could withdraw completely from the trial but data obtained up until the point of withdrawal were included in the final analysis of the trial; thereafter, no further data were collected for that participant.

Once withdrawn, a patient was advised to discuss their further care plan with their surgeon.

Participants were excluded in the post-randomisation phase if it was later established that they were ineligible for the trial (e.g. if they did not fulfil the criteria for ‘major trauma’), if they had a fracture that was not part of the lower limb (e.g. the spine), if they had an incision for which incisional NPWT could not be applied (e.g. with an external fixator) or if they were unable to adhere to trial procedures or complete questionnaires. In the context of major trauma, patients are sometimes taken directly to the operating theatre before their past medical history and full extent of their injuries can be assessed.

Disability Rating Index and general health-related quality of life

The DRI is measured using a self-administered, 12-item visual analogue scale (VAS) questionnaire assessing the participants’ own rating of their disability (range 0–100 in which higher scores indicate more disability). 25 This measure was chosen as it addresses gross body movements rather than specific joints or body segments. Therefore, it captured function and disability associated with different fractures and injuries of the lower limbs.

The EQ-5D is a validated measure of health-related quality of life, consisting of a five-dimension health status classification system and a separate VAS (range –0.594 to 1, with 0 equating to death and a higher score relating to better quality of life). 26 An updated version of the EQ-5D, the EuroQol-5 Dimensions, five-level version, (EQ-5D-5L) has been developed to enhance the responsiveness of the instrument to changes in patient health. 27

Responses to the health status classification system were converted into multiattribute utility (MAU) scores using tariffs developed for England. 28 These MAU scores were combined with survival data to generate quality-adjusted life-year (QALY) profiles for the purposes of the economic evaluation. The EQ-5D has been validated to be completed by a patient’s proxy in case of continued impaired capacity.

Table 1 shows the collection times for all of the trial outcome questionnaires.

Complications: wound healing and scar

The quality of wound healing was self-reported using the patient scale from the Patient and Observer Scar Assessment Scale (POSAS). 21 This consists of six questions regarding different aspects of the scar, as well as an overall assessment, to provide a subjective patient assessment of wound healing complications. An objective assessment of wound healing was also recorded using a standardised photograph of the surgical site taken at the 30-day review. The photograph was evaluated by two independent wound specialists who were blinded to the treatment allocation.

Complications: chronic pain

Chronic pain after surgery and trauma is common and disabling, but no previous studies have assessed the prevalence of persistent painful neuropathic characteristics after lower-limb fracture. Shortly after the start of the trial, and with the permission of the Research Ethics Committee, we added such an assessment. The proportion of participants reporting chronic pain with neuropathic characteristics post surgery was measured using the DN4. 29 The DN4 is a short validated neuropathic pain screening tool. The full tool includes an assessment by a clinician (10 questions, with a score of ≥ 4 indicating neuropathic pain). For the purposes of this trial, we used the patient self-reported component of the DN4 comprising seven questions, with a score of ≥ 3 indicating the presence of neuropathic pain. This screening tool is recommended for use by the International Association for the Study of Pain. 30

Complications: further surgical interventions and complications

Participants were also asked to self-report (or a consultee on their behalf, in the case of continued impaired capacity) at each of the follow-up points regarding any medical/surgical intervention they received related to their surgical wound. Any self-report of surgical treatment for infection was cross-referenced with the participant’s medical record. This allowed us to report deep infection at later time points, for example at 90 days. All other postoperative complications and surgical interventions related to the index wound were recorded in the CRFs completed in clinic or by the participant.

Health resource use

Health resource use data were collected at 3 and 6 months post surgery via patient-reported (or consultee-reported) questionnaires, which have been shown to be accurate in terms of the use of different services. 31 Unit cost data were obtained from the latest available national databases including the British National Formulary (BNF),32 the Personal Social Services Research Unit’s (PSSRU’s) Unit Costs of Health and Social Care,33 NHS Reference Costs34 and NHS Supply Chain Catalogue. 35

Statistical analysis

Analysis plan

Collection of health resource data

As randomisation of the trial occurred after the surgery, only health and social service resources used after randomisation were included in the economic evaluation. In other words, resource use data related to the fracture fixation or concurrent surgery (e.g. head, chest, abdomen, pelvis or spine), labour (e.g. surgeon) or wound closure (e.g. skin clips, sutures, glue) were not collected.

Direct medical costs were separated into two groups: costs associated with the intervention and costs incurred for other reasons attributable to the intervention. The cost of the intervention was captured by the trial CRFs and consisted of the costs of wound management, inpatient care (i.e. hospitalisation and further treatment procedures) and antibiotics. Other health-care costs, such as inpatient care, outpatient care, primary and community care, and medications, were determined by means of health resource use questionnaires as filled out by the participants (or their consultees). Direct non-medical costs, such as aids and adaptations and PSS, were captured in the same patient-reported health resource use questionnaire. Other direct non-medical costs such as travel, child care and help with housework, as well as indirect costs (lost productivity) that would be used in the sensitivity analysis, were also obtained from the same patient-reported health resource use questionnaire. The health resource use questionnaires were administered at 3 and 6 months, with a recall period of 3 months.

Free-text responses (applicable to all the ‘other’ options) were reclassified in the appropriate cost category, were removed if deemed unrelated/irrelevant to the trial by clinical experts (e.g. cardiology, renal management) or were analysed collectively as ‘other’ in the descriptive analysis and excluded in the cost analysis. Items not listed as one of the prespecified options were not included in the cost analysis because the most frequently utilised resource item in each cost category would have been listed as one of the options in our questionnaire, so the exclusion of such miscellaneous items will not materially affect findings.

Collection of unit cost

Unit direct medical costs associated with the intervention were obtained from the NHS Supply Chain Catalogue 2018/201935 (see Appendix 1, Table 37). The unit cost of standard dressing was assumed to be the mean cost of permeable or semipermeable and non-permeable film/soft polymer dressings [e.g. OpSite (Smith & Nephew plc, London, UK), Mepore (Mölnlycke Health Care, Gothenburg, Sweden), Leukomed (BSN Medical, Hull, UK), Tegaderm (3M Health Care Ltd, Loughborough, UK), Cosmopor E (Paul Hartmann Ltd, Heywood, UK), Softpore (Richardson Healthcare, Elstree, UK), Mepitel (Mölnlycke Health Care), Hydrofilm (Paul Hartmann Ltd)]. The costs of orthotic cast (i.e. backslab cast, full cast and air cast/boot) was included as part of the intervention cost when there was a clinical need. Likewise, an additional component of intervention costs was the cost associated with dressing change, which occurred in both groups when there was a clinical need. This cost was estimated from the number of dressing changes and the time taken for a band 5 hospital nurse to replace the dressing. The number and type of dressing changes were captured in the CRFs; the time taken to change a dressing was estimated to be 5 minutes per change for both types of dressing. The cost per working hour of the nurse was obtained from the PSSRU 2018. 39

The cost of inpatient care consisted of two components: the cost of hospitalisation after the initial operation and the cost of further procedures that were related to the trial (e.g. debridement, metalwork removal and revision of internal fixation). These costs were derived using the NHS Digital HRG4+ 2017/18 Reference Costs Grouper40 and the NHS Reference Costs 2017/18. 34

Unit costs of medical items other than those directly attributable to the intervention (e.g. subsequent inpatient care, outpatient care or primary and community care utilised by a patient post surgery) were sourced from the NHS Reference Costs 2017/18. 34 Medication costs were sourced from the BNF;32 classes of medications deemed related to the trial by clinical experts included analgesic, antibiotic, anticoagulant, antidepressant, bisphosphonate, corticosteroid, hypnotic and anxiolytic, nausea, supplements and vitamins.

Unit costs for direct non-medical cost items such as PSS were obtained from PSSRU, whereas the costs of aids and adaptations were obtained from the NHS Supply Chain Catalogue35 (see Appendix 1, Table 1). The total cost per patient for additional (private) cost items incurred by patients and their next of kin, such as travel expenditure, child care and help with housework, were obtained from the patients directly via the patient-reported questionnaires. To estimate indirect costs, the daily median wage was obtained from the Office for National Statistics41 to compute the cost of absenteeism.

Cost per participant

Costs were calculated by multiplying resource use by the unit cost per resource and were expressed in 2017/18 Great British pounds. Unit costs were adjusted to 2017/18 prices using the NHS Hospital and Community Health Services (HCHS) index39 for health service resources when required. As the HCHS index has been revised in 2018 and the new HCHS index was available only from 2014/15 onwards, unit costs that were earlier than 2014/15 were inflated to 2014/25 levels using the older version of the HCHS (as the HCHS index for 2014/15 in the previous and current versions are the same), and then inflated to 2017/18 levels using the new HCHS index. No discounting of costs was applied because cost-effectiveness was determined within a time horizon of < 1 year (i.e. 6 months).

Medication costs over the trial period were computed using the cost per dose for each product and the mean quantity taken per day during the reported number of days. All medications were assumed to be in tablet form unless stated otherwise. If the dose of the medication was not recorded, the defined daily dose for each medication was taken from the World Health Organization website using the relevant anatomical therapeutic chemical code. 42 For the base-case analysis, which is from the NHS and PSS perspective, only medications that were prescribed were included, as we assumed that patients bought the medications out of pocket if it was used without a prescription.

Cost of absenteeism was computed using the human capital approach in which the daily median wage was multiplied by the number of days taken off work that were attributable to the injury.

Health utilities

Responses to the EQ-5D-5L were converted into MAU scores using the algorithm developed to reflect societal preferences in England. 28,43 Cross-walking algorithms developed by van Hout et al. 44 were employed to generate supplementary utility values comparable with those derived from the EQ-5D-3L instrument. QALYs were calculated as the area under the baseline-adjusted utility curve of EuroQol-5 Dimensions, three-level version (EQ-5D-3L) utility scores from baseline, 3- and 6-month data using the trapezoidal rule. 45 As the time horizon was < 1 year, no discounting was required for health utilities.

Data analysis

All analysis was based on ITT. Means and SDs of resource use and cost values for each cost category, at each time point, within each trial allocation were calculated. Differences between health resource utilisation, the means of costs and utility scores were calculated and tested for statistically significant differences from zero using t-tests. For differences in mean costs, the bootstrap 95% CI was computed based on 1000 replications. Differences in the proportion of resource use between treatment groups were examined using chi-squared tests.

Cost-effectiveness analysis

An incremental cost-effectiveness analysis comparing the cost-effectiveness of standard dressing with that of incisional NPWT, expressed in terms of incremental cost-per-QALY gained, was performed from the NHS and PSS perspective for the base-case analysis. Results were presented using incremental cost-effectiveness ratios (ICERs) and cost-effectiveness acceptability curves (CEACs) generated via non-parametric bootstrapping with 1000 replicas. This accommodates sampling (or stochastic) uncertainty and varying levels of willingness to pay for an additional QALY. The ICER was compared with willingness-to-pay thresholds of £20,000 and £30,000 per QALY, which are commonly assumed in the UK by bodies such as the National Institute for Health and Care Excellence. 46 An additional £15,000 cost-effectiveness threshold was also included to reflect recent trends in health-care decision-making. 47 The net monetary benefit (NMB) of standard dressing versus incisional NPWT was also computed and presented in a graph across different cost-effectiveness thresholds, for which a positive incremental NMB indicated that incisional NPWT is cost-effective compared with standard dressing at the given cost-effectiveness threshold.

Missing data

Under the MAR assumption, mean imputation was used for missing baseline covariate, whereas multiple imputation by chained equations (MICE) was implemented for missing cost and QALYs in order to produce unbiased estimates of costs and health outcomes. This assumption was tested using logistic regressions of missingness of costs and QALYs against baseline covariates. Inverse probability weighting was not adopted as our data were non-monotonic (i.e. patients who did not complete the 3-month questionnaire could have completed the 6-month questionnaire).

Mean imputation was used to fill in each missing value of the baseline EQ-5D utility score with the mean of the observed values, as it ensured that imputation was done in an arm-independent way. 48

Multiple imputation for QALYs was done at the score level, whereas costs were imputed at the total cost level in each follow-up time point using the Stata^® command mi impute chained. Independent variables included in the imputation model consisted of baseline EQ-5D score, whether the fracture was open or closed at presentation, age at randomisation and smoking, education and employment status. The imputation was run 60 times in line with a ‘rule of thumb’, suggesting that the number of imputations should be similar to the percentage of incomplete cases. 49 A seemingly unrelated regression model was fitted to the imputed data to estimate total costs and total QALYs in each treatment group over the 6-month follow-up period. This approach allows for correlation between costs and outcomes and estimates the two regression equations jointly, potentially improving the precision of the estimates. Estimates obtained from each imputed data set were then combined using Rubin’s rules48 to obtain an overall mean estimate of the costs or QALYs.

Sensitivity analysis

Several sensitivity analyses were executed to explore the effects of alternative perspectives or scenarios on the cost-effectiveness results. First, a societal perspective that included medications bought out of pocket and additional (private) costs incurred by patients and next of kin, as well as the cost of absenteeism, was considered. Second, a complete-case analysis in which only patients with completed data on all cost and outcome data at all follow-up time points were included, after adjusting for the covariates, was performed.

Ethics approval and monitoring

Trial Management Group

The day-to-day trial management was the responsibility of the trial manager, based at the Oxford Trauma Unit/Oxford Clinical Trials Research Unit of the University of Oxford and supported by administrative staff. The Trial Management Group (TMG) met monthly to assess overall trial progress. It was also the responsibility of the trial manager to train the research associates at each of the trial recruitment centres.

Trial Steering Committee

A TSC was appointed and was responsible for oversight, monitoring and supervising trial progress. The TSC consisted of two independent experts, a lay member and the chief investigator. Membership is listed in the Acknowledgements.

Data Safety and Monitoring Committee

The trial DSMC adopted a DAMOCLES charter, which defines its terms of reference and operation in relation to oversight of the trial. The DSMC was tasked with monitoring ethics, safety and data integrity. The trial statistician provided data and analyses requested by the DSMC so that they could review accruing data and summaries of the data presented by treatment group, and assess the screening algorithm against the eligibility criteria. They also considered emerging evidence from other related trials or research and reviewed related SAEs that had been reported. Membership of the DSMC is listed in the Acknowledgements.

Patient and public involvement

Prior to the commencement of the trial, the research question under investigation was refined through discussions with members of a patient and public involvement group. Throughout the trial, a patient with direct experience of sustaining a lower-limb fracture and the subsequent recovery path reviewed participant documents prior to submission to the sponsor and the ethics committee. Advice was sought from this lay collaborator during management meetings, specifically when issues directly related to participant engagement were discussed. Independent lay representation was present on the TSC.

Recruitment by centre

Participants were recruited from 24 recruitment centres in England and Wales, representing the UK Major Trauma Network. Table 4 shows the number of participants recruited per recruitment centre and details on the participant’s sex, wound at presentation (open or closed), ISS (≤ 15 or ≥ 16) and type of consent at randomisation.

For those patients who lacked capacity to consent pre surgery (45%), consent for continuation in the trial was made at the first appropriate time point in the postoperative period. Table 5 shows the final type of consent/agreement obtained for all the participants recruited.

The planned overall required recruitment rate for the WHiST trial was approximately six participants per recruitment centre per month, based on 1540 participants recruited and consented over 22 months at 24 recruitment centres. Overall recruitment across recruitment centres was 4.2 participants per month. This was lower than the planned rate based on the original recruitment period; therefore, the trial recruitment was extended by 3 months to the end of April of 2018, to reach the target of 1540.

Minimisation factors by intervention group

The minimisation factors (recruitment centre, ISS and open or closed fracture at presentation) are summarised by treatment group for all randomised participants (Table 6). These factors were well balanced across treatment groups.

Post-consent eligibility errors

After consent, 27 participants were found to be ineligible for the trial. These patients were followed up as per ITT but were excluded from the PP population. Table 7 shows the reasons for ineligibility post consent.

Consented and non-consented participants

Of the 1548 participants randomised and consented, one withdrew immediately after surgery, requesting not to provide any of the baseline data. Therefore, this participant was excluded from the ITT population. Figure 1 shows the study CONSORT flow diagram.

FIGURE 1.

The WHiST trial CONSORT flow diagram. a, One participant who prospectively declined and was randomised in error; b, three participants who prospectively declined and were randomised in error.

Treatment allocation

Participants who did and those who did not receive their allocated intervention are summarised in Table 8. There were 100 crossovers in total, 92 from incisional NPWT to standard dressing and eight from standard dressing to incisional NPWT. Reasons for crossovers are also summarised in Table 8. Most crossovers in the standard dressing group [n = 7 (87.5%)] were due to surgeon’s choice, whereas most crossovers in the incisional NPWT group [n = 47 (51.1%)] were due to ‘other’ reasons: the surgical team forgot to apply the allocated treatment, or the allocated treatment was not communicated to the surgical team.

Details of treatment compliance following application of the dressings are summarised in Table 9. Incisional NPWT was in place for a median of 7 days (as recommended by the manufacturer). Participants were classed as having changed their treatment if they had a different dressing applied after < 7 days.

Available data

Primary outcome data for 197 participants (13.0% of the ITT population) were completed retrospectively by checking medical records, as these participants did not attend their follow-up appointment and could not be contacted by telephone. A total of 92 participants declined to have data collected after the primary outcome point of 30 days; therefore, data obtained from the 3- and 6-month follow-up points could be collected for a maximum of only 1456 participants (711 and 745 participants in the standard and incisional NPWT groups, respectively). Table 10 shows the follow-up rate at 3 and 6 months post surgery.

Details of the completion of the primary and key secondary outcome measures (DRI and EQ-5D utility scores) at each time point are summarised by intervention group in Table 11. Outcomes were expected to be completed for all participants who had not withdrawn or died by each time point. For participants who answered some, but not all, parts of the DRI, pro-rated scores were calculated if more than half of the questions (i.e. at least 7) had been answered. Pro-rated scores were an average across all answered questions. Eleven pre-injury DRI scores, 40 3-month DRI scores and 33 6-month DRI scores were imputed in this way.

All withdrawals up to 6 months after randomisation are summarised by intervention group in Table 12. The median time to withdrawal and the reasons for withdrawals are summarised. The proportion of withdrawals was slightly higher in the incisional NPWT group (7.4%) than in the standard dressing group (5.9%); however, the median time to withdrawal was longer in the incisional NPWT group. The main reason for withdrawal was that participants (or their consultee) did not want to complete questionnaires. Withdrawals > 6 months after randomisation will be reported separately as part of the long-term follow-up of the WHiST trial.

Baseline participant characteristics

The descriptive characteristics of the participants included in the ITT population are summarised by intervention group and overall in Table 13. These values are presented as numbers and percentages for categorical factors and means and SDs or medians and IQRs, as appropriate, for continuous variables. These variables all appear well balanced across the two treatment groups. The distribution of participant ages at enrolment is shown in Figure 2. This distribution has two peaks (i.e. it is bimodal): one in younger adults and one in older adults.

FIGURE 2.

Distribution of participant age in years at randomisation.

Operative procedures

Table 14 summarises operative procedure details for the ITT population by group and overall. These details are well balanced across the two treatment groups.

Patient-reported outcome measures at baseline

Baseline values of patient-reported outcome measures (DRI and EQ-5D) are summarised by intervention group using medians and IQRs in Table 15. For the EQ-5D variables, both immediate pre- and post-injury values are summarised. These are all similar across the two treatment groups.

Analysis of primary outcome

The number and percentage of participants with a deep infection at up to 30 days is reported by intervention group in Table 16. The deep infection rate in the control group (6.7%, 50/749) was substantially lower than anticipated: the sample size calculation for this study was based on a control group deep infection rate of 15%. The deep infection rate within 30 days in the intervention group was 5.8% (45/770).

The logistic regression analysis indicated that there was no statistically significant difference between the intervention groups in terms of deep infection rate (OR 0.87, 95% CI 0.57 to 1.33; see Table 16). The adjusted absolute risk difference between the control and intervention groups was –0.77% (95% CI –3.19% to 1.66%). This analysis was also repeated for the PP population to assess the impact of deviations from the planned trial protocol; it indicated no statistically significant difference between the two groups.

The rate of deep infection was lower than anticipated (6.3% across the whole trial) and the number of missing primary outcome data was very small [n = 28 (< 2%)]; therefore, MI was considered to be the most robust sensitivity analysis for best versus worst case.

Potential prognostic factors were investigated, and wound location (above or below knee) was included in the MICE model, along with all variables from the fitted model. Ten imputed data sets were created and combined using Rubin’s rules. No significant differences between groups were identified by this analysis (see Table 16).

Secondary analyses of the primary outcome

The number and percentage of participants with a deep infection at up to 90 days is reported in Table 17. The deep infection rate was higher than the rate reported at 30 days (13.2% in the control group and 11.4% in the intervention group) and similar to that expected in the sample size calculation. The rate at 90 days may be slightly overestimated because those with a deep infection before 30 days were counted as a deep infection before 90 days regardless of the availability of 3-month data for these individuals, whereas individuals were counted as not having a deep infection at 90 days only if they did not have a deep infection at either 30 days or 90 days. A mixed-effects logistic regression model was again used to compare the two trial groups for the ITT population using available cases. There was no significant difference in the rates of deep infection at up to 90 days between the two intervention groups (OR 0.86, 95% CI 0.61 to 1.23).

Secondary outcomes

Continuous outcomes (repeated-measures models)

For each continuous outcome (DRI, EQ-5D utility and VAS, POSAS total score and overall opinion), mean scores with SDs are reported for each intervention group at each time point (Table 20). Figure 3 provides plots of each of the continuous variables over time by intervention group. The DRI and EQ-5D variables demonstrate a trend of improvement over time, whereas the POSAS variables appear relatively constant over time. Adjusted differences between the two treatment groups at each time point were calculated using multilevel mixed-effects linear regression models. These analyses were performed for the ITT population using the available-case data set, and the results are presented in Table 20. No significant differences between the groups on any of the outcomes were identified.

TABLE 20 - Analysis of patient-reported continuous secondary outcomes at 3 and 6 months (ITT population, available-case data set)

Note

DRI scores range from 0 to 100, with higher scores indicating more disability; EQ-5D utility scores range from –0.594 to 1, with higher scores indicating better quality of life, 0 = death; EQ-5D VAS scores range from 0 to 100, with higher scores indicating better quality of life; POSAS total scores range from 6 to 60, with lower scores indicating better scars; POSAS overall opinions range from 1 to 10, with lower scores indicating a better opinion.

Outcome	Time point	Standard dressing		NPWT		Adjusted difference (95% CI)	p-value
		Mean (SD)	Total (N)	Mean (SD)	Total (N)
DRI	3 months	51.1 (23.92)	456	51.6 (23.46)	507	–0.01 (–2.79 to 2.78)	1.00
	6 months	40.2 (26.73)	432	40.6 (24.98)	469	0.03 (–2.82 to 2.88)	0.98
EQ-5D (utility)	3 months	0.5 (0.30)	470	0.5 (0.29)	528	0.00 (–0.03 to 0.04)	0.84
	6 months	0.6 (0.29)	446	0.6 (0.28)	486	0.00 (–0.03 to 0.04)	0.86
EQ-5D (VAS)	3 months	64.7 (22.78)	478	64.1 (22.24)	531	–0.73 (–3.30 to 1.84)	0.58
	6 months	69.4 (21.76)	449	69.7 (21.15)	489	0.08 (–2.57 to 2.74)	0.95
Scar assessment total score	30 days	22.9 (11.65)	607	21.4 (11.38)	648	–1.16 (–2.40 to 0.08)	0.07
	3 months	23.4 (12.40)	452	23.1 (12.87)	513	–0.36 (–1.73 to 1.01)	0.61
	6 months	21.2 (11.84)	432	21.4 (12.47)	476	0.15 (–1.26 to 1.55)	0.84
Scar assessment overall opinion	30 days	4.6 (2.65)	616	4.4 (2.65)	657	–0.18 (–0.46 to 0.10)	0.22
	3 months	4.9 (2.73)	470	4.7 (2.77)	523	–0.11 (–0.41 to 0.20)	0.51
	6 months	4.5 (2.69)	437	4.6 (2.80)	483	0.11 (–0.21 to 0.42)	0.52

FIGURE 3.

Summary plots of continuous outcome measures over time by intervention group. (a) DRI scores from pre injury to 6 months. Scores range from 0 to 100, with higher scores indicating more disability. (b) EQ-5D utility scores from pre injury to 6 months. Scores range from –0.594 to 1, with higher scores indicating better quality of life; 0 is equivalent to death. (c) EQ-5D VAS scores from pre injury to 6 months. Scores range from 0 to 100, with higher scores indicating better quality of life. (d) POSAS total scores from 30 days to 6 months. Scores range from 6 to 60 with lower scores indicating a better-healed scar. (e) POSAS overall opinion scores from 30 days to 6 months. Scores range from 1 to 10 with lower scores indicating a better opinion of the scar.

Residuals from each fitted model were used to confirm that the assumption of approximate normality was appropriate.

A sensitivity analysis of the DRI data using MI under the MAR assumption was performed; this did not change the results.

Primary outcome

This trial showed no evidence of a difference between standard dressings and incisional NPWT in the management of major trauma surgical wounds of the lower limb. The rate of deep infection at 30 days was 6.7% (50/749) in the control group and 5.8% (45/770) in the incisional NPWT group (ITT OR 0.87, 95% CI 0.57 to 1.33; p = 0.52).

Similarly, there was no evidence of a difference in the PP analysis (analysis by treatment received). The number of missing data at the primary end point was very low (< 2%); therefore, a sensitivity analysis using MI for missing data also showed no evidence of a difference between the two groups in terms of the rate of deep infection.

Secondary outcomes

After the start of the trial, the CDC changed their definition of deep infection to include any SSI up to 90 days after surgery rather than the original 30 days. To facilitate future evidence synthesis, we prespecified a secondary analysis using this 90-day time point. The rate of deep infection reported at 90 days was much closer to that anticipated at the beginning of the trial (in keeping with the existing literature for major trauma), with 13.2% (78/590) in the control group and 11.4% (72/629) in the intervention group. However, there was no evidence of a difference between groups (OR 0.84, 95% CI 0.59 to 1.19; p = 0.32).

As part of the trial data set, we recorded details of any other wound healing complications that did not fulfil the CDC criteria for a deep SSI. There was no evidence of difference in any of these complications. Interestingly, despite the wounds not fulfilling the criteria for a deep infection, 3.7% of patients were treated with antibiotics. This number is low in comparison with other related studies,54 but still suggests that clinicians are failing to follow guidance regarding the empirical use of antibiotics for surgical wounds.

In keeping with the primary analysis of deep infection, this trial found no evidence of a difference in the patients’ self-reported DRI nor in their health-related quality of life at either 3 months or 6 months. Other trials51,54 of serious lower-limb injuries have found that patients continue to report serious disability and poor quality of life even 6 months after the surgical fixation of their fractures. Similarly, the WHiST trial found that patients reported a 40-point disability score (in which 100 points represents complete disability) and similar loss of quality of life at 6 months. This is further powerful evidence of the very severe and lasting effects of these injuries on patients.

Chronic pain has previously been described as a substantial problem for patients after major trauma injuries to the lower limb,55 and this was also the case in the WHiST trial. Visual analogue pain scores were still 3/10 (in which higher scores indicate more pain) at 6 months in both groups of patients. Neuropathic pain is particularly difficult to manage with standard analgesia. The proportion of patients reporting neuropathic pain according to the DN4 was 32% (117/367) in the standard dressing group and 28% (117/414) in the incisional NPWT group (p = 0.27).

The trial found no evidence of difference in the rate of deep-vein thrombosis between the two groups of patients, nor in the rate of other local complications. There was a statistically significant difference in the rate of further surgery to the broken bone (not related to the wound), with more patients requiring further surgery in the incisional NPWT group, but the OR was close to 1 (95% CI 1.01 to 2.10).

There was no evidence of a difference in the subjective POSAS at any time during the 6 months of follow-up.

The WHiST trial also offered the opportunity to evaluate the use of photographs as an objective assessment of wound healing. Of the 1548 participants, 1123 had a photograph of their wound taken at, or shortly after, 30 days. These were assessed by independent tissue viability specialists who were blinded to the treatment allocation. There was no evidence of a difference between the two groups in terms of the proportion of wounds deemed ‘fully healed’: 84% in the standard dressing group and 87% in the incisional NPWT group. Nor was there a difference in the proportion of wounds showing signs of infection: 13.6% in the standard dressing group versus 11.4% in the incisional NPWT group. Although the agreement between the primary outcome measure of infection using the CDC criteria and this photographic assessment was significantly higher than would be expected by chance (p < 0.001), this estimate of infection using photographs (12%) was twice as high as that reported using the CDC criteria (6%). Photographic assessment is an attractive method, in that the images can be assessed in a standardised objective fashion, but it has several drawbacks. The tissue viability specialists commented that their ability to assess for signs of infection was very dependent on the quality of the photographs. In some cases, it was difficult to distinguish between ‘slough’ on the wound surface and signs of infection, which may have led to an overestimation of the rate of infection. The photographs provide only a two-dimensional assessment, which makes it difficult to appreciate swelling and fluctuance around the wound. Also, the photographs clearly provide only a single moment in time. It is possible that the patient had a clear deep infection in week 1, which, if treated successfully, may have all but resolved at 30 days. In summary, photographs provide an objective means of assessing aspects of wound healing but are probably not sufficiently reliable to replace clinical assessment in the diagnosis of infection.

Infection audit

The WHiST trial DSMC recommended that the trial team audit the medical records of patients with a CDC diagnosis of infection in the trial data set and a random sample of those who did not have a diagnosis of infection. The audit took place when the first 1000 patients had been recruited to the trial and reached their 30-day assessment.

For the large majority of patients, there was agreement between the trial data set and the routine hospital medical record. At the first review of the medical records, four further potential cases of deep infection were highlighted that had not been captured in the trial data set. However, a further detailed review of the medical records indicated that two of these patients did not fulfil the criteria for deep infection. Conversely, 49 patients were identified as having met the criteria for a deep SSI according to the trial data set, but were not identified in the routine medical record. A further detailed review of these medical records by the local clinicians showed that half (n = 24) of these participants had at least one documented sign of deep infection in the medical record but did not have documentation that fulfilled all the criteria for the CDC definition. For example, there was documentation of gaping at the wound edges but not of associated pain and/or fever. The other 25 patients had no documentation of a deep infection in the medical record at all, despite having a clear diagnosis of infection in the trial data set, for example, pus leaking from the wound at the 30-day follow-up appointment.

In summary, this audit found 49 cases in which signs and symptoms of infection were poorly documented, or indeed completely absent, from the routine medical record. These cases would have been missing from the trial data set if the medical record alone was used to collect evidence of infection. In the trial data set, the research associates reviewed the medical records of each participant and also spoke to and assessed the patient directly regarding their wound healing. A higher rate of infection was therefore documented in the trial data set. Even so, two cases of deep infection that were identified from the routine medical record were ‘missed’ in the trial data set.

Clearly, there is no perfect system for collecting evidence of infection in surgical wounds. A degree of caution is warranted in interpreting the audit data as the interpretation was not part of the pre-planned analysis of the WHiST trial but was implemented during the trial on the advice of the DSMC. However, the audit does suggest that a review of the routine medical record alone is likely to lead to an underestimation of the rate of deep infection in surgical wounds.

This has implications for the interpretation of reports of infection using routinely collected data. In the context of a randomised trial, it could be argued that any difference between interventions would still be detected using routine data, if the trial was large enough, the assumption being that any underestimation of the rate of deep infection would be equally balanced in the two groups of participants. However, in the case of observational studies, the use of routine data in isolation is likely to lead to an underestimation of the incidence of important complications such as infection.

Trial management team

Professor Matthew Costa, chief investigator; Dr Ruth Knight and Mr Karan Vadher, trial statisticians; Mrs Susan Dutton, senior trial statistician; Dr Juul Achten, senior research fellow; Dr Julie Bruce, principal research fellow; Mrs Louise Spoors and Dr Marta Campolier, trial managers; Mr Damian Haywood, senior trial manager; Mr James Masters, specialist registrar; Dr May Ee Png and Dr Melina Dritsaki, trial health economists; Professor Jagdeep Nanchahal, consultant reconstructive plastic surgeon; Dr Jason Madan, senior health economist; Mrs Suzanne Jones, public member; Miss Bojana Selinšek, Dr Scott Parsons and Miss Cheridan Morgan, trial administrators; and Mrs Kylea Draper, Miss Alice Brealy, Miss Kinzah Abbasi, Mr Hugo Strachwitz and Miss Catherine Thompsett, data clerks.

Tissue viability specialists

Ms Ria Betteridge and Ms Julie Brown.

Principal investigators

Peter Hull, Simon Scott, Sharon Scott, David Melling, Javed Salim, William Eardley, Peter Giannoudis, Jitendra Mangwani, Andrew Ridick, Paul Harnett, Edward Mills, Mike Reed, Ben Ollivere, Xavier Griffin, Mark Brinsden, Ravichandran Karthikeyan, Peter Bates, Benedict Rogers, Haroon Majeed, Damian McClelland, Sharad Bhatnagar, Caroline Hing, Rajarshi Bhattacharya, Usman Butt, George Cox and Khitish Mohanty.

Associate principal investigators

Majid Chowdhry, Sharon Scott, David Melling, Hermant Sharma, Alex Sims, Paul Baker, Ruth Varrall, Edward Massa, Ines Reichert, George Kleftouris, Hany Saleeb, Paul Harwood, Brett Rocos, Edward Hollowry, Ian Baxter, Mich Pennison, Michael Petzle, Paul Brewer, Tanvir Khan, Tim Coughlin, Matthew Baldwin, Paul Guyuer, Siddhant Kadoor, Nima Heidari, David Crone, Richard Freeman, Andreas Fontalis, John Dabis, Michalis Kaminaris, Graham Lawton, Majed Nayjar, Richard Unsworth, Tanya Tordoff, Tom Collins, Jonathan Young, Robert Jordan, Tarek Boutefnouchet, Bob Jennings, Christopher Jordan, Laura Beddard, Gareth Roberts, James Lewin, Lars Tiesl and Manhal Yasseen.

Research associates

Leila Dehghani, Simone Hargreaves, John Moorehead, Kim Dearnley, Lisa Wilson, Alanna Milne, Charlotte Bousted, Juliet James, Abdulkerim Gokturk, Marta Santamaria, Bernadette Cook, Mehvish Hussain, Tracey Lee, Tracy Bear, Elisabeth Barnett, Katherine Coates, Lizzy Shaw, Lucille Hooper, Ruth Halliday, Steven Barnfield, Carol Peel, Helen Ruth Foot, Joanne Badlok, Julie Walker, Rachel Sellars, Chris Herriott, Christine Dobb, Deborah Bunn, Elizabeth Corbishley, Helen Shirley, Kerry Third, Charlotte Bentley Jess Nightingale, Doreen Muller, Kathryn Lewis, Maria Mestre, Martin Austin, Sangeetha Prasath, Yuhan Zhang, Zoey Warnok, Rosalyn Squire, Laura Mee, Liesl Despy, Catherine Hilton, Peter May, Shanaz Ahmad, Sophie Young, William Stephens, Holly Maguire, Jonathan Sweetman, Francis Alepo, Holly Maguire, Ida Marje Ponce, Joanne Hiden, Nenette Abano, Racquel Carpio, Susan Lyjko, Donna Simpson, Karen Smith, Stuart Watson, Alexandra Ball, Elizabeth Cruddas, Isabel Carballo Horton, Jocelyn Nocerino, Kameka Angulo, Sherrin Gotru, Isabella Wilkes, Jo-Anna Conyngham, Karolina Zimmerman, Louise Young, Anne Keen, Zin Nwe Naing, Greg Scott, Helen Sharples, Joanna Teuke, Kerri McGown, Sylvia Turner, Charlie Smith, David Baker, Dima Ivanova, Maria Letts, Heather Javis and Matthew Williams.