Hughes abdominal closure versus standard mass closure to reduce incisional hernias following surgery for colorectal cancer: the HART RCT

Article history

The research reported in this issue of the journal was funded by the HTA programme as project number 12/35/29. The contractual start date was in February 2014. The draft report began editorial review in June 2020 and was accepted for publication in August 2021. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The HTA editors and publisher have tried to ensure the accuracy of the authors’ report and would like to thank the reviewers for their constructive comments on the draft document. However, they do not accept liability for damages or losses arising from material published in this report.

Copyright statement

Copyright © 2022 O’Connell et al. This work was produced by O’Connell et al. under the terms of a commissioning contract issued by the Secretary of State for Health and Social Care. This is an Open Access publication distributed under the terms of the Creative Commons Attribution CC BY 4.0 licence, which permits unrestricted use, distribution, reproduction and adaption in any medium and for any purpose provided that it is properly attributed. See: https://creativecommons.org/licenses/by/4.0/. For attribution the title, original author(s), the publication source – NIHR Journals Library, and the DOI of the publication must be cited.

2022 O’Connell et al.

Background

An incisional hernia (IH) is a common and potentially serious complication following abdominal surgery. An IH is a bulge in the abdomen through or close to a previously made incision caused by the patient’s intestines, organs and/or other tissue protruding through a weakening in the abdominal muscles as a result of surgery. The reported rates of IH range from 8.6% to 39.9% following open colorectal surgery and from 4.7% to 24.3% following laparoscopic surgery. 3–7 In a systematic review8 that included 14,618 patients, the incidence of IH was 12.8% at 2 years’ follow-up, with incidence rates as high as 35.6% among patients who had received a midline incision during surgery. In patients with colorectal cancer, the rate of IH has been reported to be as high as 39.9%. 5

A number of potential risk factors for IH have been identified, including male sex, increased age, increased BMI, history of chronic obstructive pulmonary disease, history of smoking and certain medications. Surgeon-modifiable risk factors include surgical technique and suture type for abdominal closure; however, although a number of studies have been conducted investigating different surgical methods, uncertainty remains around the impact of such surgeon-modifiable factors on IH rates, with several studies reporting conflicting results. For example, three meta-analyses9–11 concluded that non-absorbable stitches reduce the risk of IH, one meta-analysis12 reported that absorbable stitches are associated with a lower risk of IH and one meta-analysis8 reported no difference in IH rates when comparing absorbable and non-absorbable stitches.

Recent cost analyses have found that the treatment and repair of IH places a considerable strain on already-stretched health-care resources. Direct per-patient cost estimates range from €3497 to €16,367 in European countries13,14 and from US$6530 to US$16,889 in the USA,15–17 with hospitalisation and surgery costs, as well as complications, adverse events and recurrences, identified as the main cost drivers.

An IH can be diagnosed as a result of patient-reported symptoms, such as a lump, abdominal pain and symptoms of obstruction. If the hernia has become incarcerated or strangulated, then this can also lead to tissue necrosis.

Treatment can vary depending on the size and anatomy of the hernia, the general health of the patient and the desired level of physical activity post repair. However, it will generally require one of two types of surgery: an open or, sometimes, a laparoscopic hernia repair. In attempting to reduce the risk of hernia recurrence, there is an increasing reliance on the use of synthetic or biologic mesh to facilitate the repair.

Rationale

More than 30,000 patients are diagnosed with colorectal cancer in the UK each year. 18 Most of these patients will undergo surgery as part of their treatment, and the incidence of complications following surgery is high. 19 One common complication after abdominal surgery is the occurrence of an IH following the closure of the midline incision.

An IH may have a negative impact on a patient’s quality of life (QoL) and their overall experience. 20 The outcomes for patients with IHs are poor, and many will suffer with chronic pain or suffer a repeat hernia even after the first repair. This can, in turn, lead to increased NHS resource use as a result of additional or longer hospital stays. It is, therefore, important to identify surgical procedures and strategies that can reduce the risk of IHs in patients who undergo abdominal surgery.

The aim of this study was to assess the potential of Hughes abdominal repair, an alternative wound closure method, to prevent IHs. The study recruited colorectal cancer patients who were due to receive surgical treatment for their cancer. The care pathway followed its standard course, except at point of abdominal wall closure, when the patient was randomised to either Hughes abdominal closure or standard mass closure.

Literature update

A review of the literature was conducted in key databases, including MEDLINE, EMBASE, Cochrane Database of Systematic Reviews and Cochrane Register of Controlled Trials. The review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA),21 and the PRISMA flow chart (Figure 1) details the amount of evidence sifted and excluded at each stage. Search strategies were developed in MEDLINE and adapted for other databases (see Appendix 1, Table 29).

FIGURE 1.

The PRISMA flow diagram. Reproduced with permission from Moher et al. 21 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: https://creativecommons.org/licenses/by/4.0/. The figure includes minor additions and formatting changes to the original figure.

Reproduced with permission from Moher et al. 21 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: https://creativecommons.org/licenses/by/4.0/. The figure includes minor additions and formatting changes to the original figure.

The study team was aware of a published systematic review and meta-regression that reported on factors affecting rates of midline IH in 14,618 patients from 56 individual studies, as well as the prevalence of IH at the 2-year follow-up. 8 The quality of the studies included was variable. All study types were considered for inclusion in the meta-analysis if they described a population of adult patients undergoing primary suture closure of a midline laparotomy wound. The review included a literature search up to March 2013. For this reason, it was decided to update the literature searches from this point. Initial database searches were carried out for the period from January 2013 to November 2018, with update searches carried out in September 2019 and again in April 2020 to ensure completeness. Following the removal of duplicate records, searches identified 3417 potentially relevant records (see Figure 1).

During the initial sifting of the updated literature search results, a Cochrane review published in 2017 was identified, entitled ‘Closure methods for laparotomy incisions for preventing incisional hernias and other wound complications’. 22 Assessment of the Cochrane review indicated that the searches were comprehensive and directly relevant to this study. As this provided a thorough review of relevant evidence up to February 2017, the authors decided that a full literature review was not required at this time. The decision was made to limit the evidence review for this report to new evidence and to include only relevant evidence published after the date of the searches in the Cochrane review.

Although an earlier systematic review8 had included all study types, the decision was also made to limit the searches to randomised controlled trials (RCTs) reporting the primary outcome, incidence of IH following abdominal surgery, in line with the Cochrane review,22 as it was considered that there were sufficient RCTs available.

As this was not a full systematic review of the literature, it was not registered on the PROSPERO database.

In addition, a review of clinical trial databases was conducted to identify any relevant ongoing clinical trials. Five additional relevant ongoing trials were identified, three of which have corresponding trial protocols published. 23–25 Ongoing trials include:

• the ESTOIH study,23 investigating the influence of stitch length, using an elastic, extra-long-term absorbable monofilament suture, on the long-term clinical outcome of abdominal wall closure

• the HULC trial,24 investigating whether or not a combination of small stitched fascial closure and onlay mesh augmentation after elective midline laparotomies reduces the risk of IH

• the CONTINT trial,25 comparing continuous slowly absorbable sutures with interrupted rapidly absorbable sutures for abdominal wall closure after midline incisions for emergency laparotomy

• the E-STITCH trial,26 comparing the small-tissue-bite technique with the large-bite technique for the closure of emergency midline laparotomy

• the Rein4CeTo1 trial,27 comparing the IH incidence 1 year after planned colorectal cancer surgery performed through a midline incision that is closed either by a standardised small stitch 4 : 1 technique or with the same technique plus a reinforced tension-line suture.

Full details of these additional studies can be found in Appendix 2.

Based on the adjusted criteria for inclusion in the review, only two additional studies not covered by the Cochrane review were identified as relevant: the STITCH trial28 and the HART feasibility trial. 1

In total, three studies are included in the literature review. 1,22,28 The methods and results of each of the studies are detailed in the following sections.

Hypothesis

The null hypothesis states that, in patients having midline abdominal wall closure following elective or emergency colorectal cancer surgery, there is no difference in the rate of IH over 1 year between those undergoing Hughes repair and those undergoing standard mass closure.

The alternative hypothesis states that, in patients having midline abdominal wall closure following elective or emergency colorectal cancer surgery, Hughes repair alters the incidence of IH over 1 year when compared with standard mass closure.

Patient and public involvement

Patient and public involvement (PPI) was an integral part of the HART study. A minimum of two PPI representatives were involved at any given time throughout the study, one of whom had experience of colorectal surgery for a colorectal cancer. PPI representatives had previous experience of working with research groups and sat on the Trial Management Group for the study. In addition, a PPI representative sat on the Steering Committee for the trial. PPI began at the protocol development stage and continued right through to the interpretation, discussion and dissemination of results.

Patient and public involvement representatives were paid honoraria and out-of-pocket expenses in line with Health Care Research Wales and INVOLVE guidelines for attending meetings.

Although the impact of PPI was not an outcome of the study, feedback was sought from the patient representatives at all stages of the study and their experience of being part of a clinical study is reported.

Trial design

This was a multicentre, single-blinded randomised controlled trial. The patients were randomised 1 : 1 to enable the comparison of two suture techniques for the closure of the midline abdominal wound following surgery for colorectal cancer.

Changes to trial design

The study was split into three phases: feasibility, pilot and main. The feasibility phase assessed recruitment, randomisation, deliverability and early safety of the surgical technique. Following a successful feasibility phase, the trial was approved by the independent Data Monitoring Committee (DMC) and progressed to the pilot and main phases. No changes were adopted in the trial design specified during the feasibility phase and the study finished its primary end point as per the published study protocol. 1,2

Conduct of the study: approvals and trial registration

The trial was conducted in compliance with the protocol, the Declaration of Helsinki as currently revised29 and the principles of Good Clinical Practice (GCP),30 and in accordance with all applicable regulatory guidance.

The study protocol and all subsequent amendments were reviewed and approved by the Wales 3 Research Ethics Committee (REC) (MREC 12/WA/0374) and the research and development offices of the participating NHS sites. A list of key protocol amendments can be found in Appendix 3.

Annual progress and safety reports were submitted to the REC.

The trial was registered in the ISRCTN registry and the trial registration number is ISRCTN25616490.

Participants

The study identified patients who were due to undergo abdominal surgery for the treatment of colorectal cancer. Patients undergoing emergency surgical treatment and patients receiving elective surgical treatment were screened. Patients were not excluded if they had undergone previous abdominal surgery for conditions other than the new colorectal cancer. To be eligible for inclusion, patients had to be considered suitable for either Hughes repair or standard mass closure. Patients had to have a midline incision at least 5 cm in length or they were excluded from the study.

Surgery was carried out as per standard surgical procedure in a general surgical unit within the NHS across the UK. Prior to commencing closure and provided that the patient still met the inclusion criteria, the closing surgeon accessed the telephone randomisation system and the patient was allocated to either Hughes repair or standard mass closure.

An initial feasibility phase in which 30 participants were randomised to Hughes repair or standard mass closure was conducted at University Hospital of Wales, Cardiff, only. Following the feasibility phase, which demonstrated that a RCT comparing Hughes repair with standard mass closure would be acceptable to patients,28 the study moved into the pilot and main phases. Patients were randomised to the pilot (n = 80) and main (n = 722) phases of this study (Hughes repair, n = 401; standard mass closure, n = 401) over a period of approximately 3 years and 5 months. No changes were made to the inclusion and exclusion criteria from the feasibility study to the full trial. Data from the pilot and main phases of the study were combined and are referred to as the main study phase in the rest of this report.

Inclusion and exclusion

Consent

Patients identified as eligible were approached by a member of the research team to discuss taking part in the study. Interested patients were given a patient information pack containing a letter of invitation, a consent form and a patient information sheet for the study. Patients were given time to review the information and ask questions. Patients who agreed to take part in the study were consented prior to their surgery (Figure 2). In 2018, an updated consent statement was implemented to meet with GDPR (General Data Protection Regulation) requirements. This was sent to each site for their records.

FIGURE 2.

Flow diagram of consent process. PIS, participant information sheet.

The mean time from signing the consent form to randomisation (surgery) was 2.5 days [standard deviation (SD) 7.2 days]. The maximum time was 91 days and the minimum time was 0 days. For patients undergoing elective surgery, the mean time from consent to randomisation (surgery) was 2.7 days, and for patients undergoing emergency surgery, the mean time from consent to randomisation was 0.4 days.

Interventions

The Hughes abdominal repair technique involves the use of ‘near and far’ sutures to close the abdominal wall. The technique combines a standard mass closure [two-loop 1-polydioxanone (PDS) sutures] with a series of horizontal and two vertical mattress sutures within a single suture (1 nylon), theoretically distributing the load along the incision length as well as across it (Figure 3). Surgeons use loop 1-PDS for the mass closure element of the Hughes repair, with the multiple nylon sutures used for the ‘near and far’ sutures. The principles of the technique are to ensure that only sound normal tissues are used for repair, to use graduated tension for easy approximation and to use a monofilament nylon suture, which has the advantage of slipping easily through tissues to create a pulley system. 2 This was a standardised intervention arm with no variation permitted.

FIGURE 3.

Hughes repair. Reproduced with permission from Cornish et al. 2 © 2016 The Author(s). Open Access. This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: https://creativecommons.org/licenses/by/4.0/. The figure includes minor additions and formatting changes to the original figure.

Reproduced with permission from Cornish et al. 2 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: https://creativecommons.org/licenses/by/4.0/. The figure includes minor additions and formatting changes to the original figure.

Standard care (control arm) comprised standard mass closure. This involves closing all layers of the abdominal wall (excluding the skin), usually using non-absorbable sutures, although ‘slow-resorbing’ sutures, such as PDS, are also widely used. Variation was allowed in this arm according to surgeon preference, but the technique used was recorded on the patient case report form (CRF).

All patients were given the same postoperative rehabilitation advice regardless of study arm.

Primary outcome

The primary outcome measure was the incidence of IH at the 1-year clinical examination.

The clinical presence of a hernia was assessed by an independent surgeon who was blinded to the closure technique wherever possible. In some centres, this examination was carried out by a nurse specialist who had appropriate accreditation for patient examination and was also blinded to the closure technique. The CRF included explicit details/instructions on how to carry out the examination. The presence of a hernia was detected as a reducible, palpable mass, usually with a cough impulse, which may cause pain or discomfort.

Secondary outcomes

• Quality of life (QoL) measured using SF-12 and FACT-C. Questionnaires were administered at baseline and at 30 days, 6 months, 1 year and 2 years following randomisation. QoL was compared between arms (Hughes repair vs. standard mass closure) and between patients who developed IH and patients who did not develop IH.

• Cost-effectiveness of Hughes repair compared with standard mass closure in the first year from the perspective of the NHS. Information regarding health-care resource use (including surgery-specific resources) was collected using a Client Service Receipt Inventory (CSRI) questionnaire at baseline, 6 months and 1 year.

• Postoperative full thickness abdominal wall dehiscence (burst abdomen) within 30 days of surgery, as well as details of any repair surgery and closing sutures used.

• Identification and characterisation of patient and surgical factors associated with an increased risk of developing an IH.

• Prevalence of hernia at 1-year clinical examination.

Tertiary outcomes

• Prevalence of clinically detectable IH at 5 years from surgery.

• Effect of Hughes repair and standard mass closure over the 5 years from surgery.

• Cost-effectiveness of Hughes repair compared with standard mass closure over 5 years from the perspective of health and social care.

• Sensitivity and specificity of CT image identification over 2 years compared with clinical examination over 2–5 years from surgery.

• Quality of life between patients with IH and patients with no IH in both arms over 5 years.

Sample size

A clinically important difference between the study arms was deemed to be a reduction in the IH rate from 30% in the standard mass closure arm to 20% in the Hughes repair arm. To detect this difference, it was calculated that a total of 640 patients would be required, providing 80% statistical power with a 5% level of significance. Assuming loss to follow-up of about 20% at 1 year, as seen in similar trials, HART aimed to recruit a total of 800 patients. A completed sample of 640 participants was calculated to yield 80% power of detecting (with a 5% significance level) a standardised difference (differences in means scaled by SD) of 0.225 in QoL (the principal secondary outcome).

Interim analyses and stopping guidelines

No interim analysis was planned in this study. Primary analysis was performed when all 1-year visits had been completed, data had been collated and the database had been locked. A separate analysis was carried out on data collected in year 2 of follow-up. The statistician received unblinded data after the database was locked for the final data analysis at year 1 and again at year 2.

The study followed the principle of allowing for early stopping were there to be a safety concern. During the study, the DMC and the study sponsor were responsible for stopping the study early had continuation of the trial been considered not in the patients’ best interest. Study data were reviewed by the DMC approximately every 6 months and reports were made to the Trial Steering Committee (TSC). The sponsor was ultimately responsible for trial progression after the consideration of recommendations by the DMC and TSC. The TSC also monitored recruitment and study progress to inform any decision to halt the study were it to be considered that the study had failed to deliver its objective as a result of delayed recruitment or lack of data besides any safety concern for the patients.

Randomisation and treatment allocation

We used an adaptive randomisation design to allocate the eligible patients to arms of similar sizes using an allocation ratio of 1 : 1. The sequence was created by a computerised random number generator and the allocation of participants was balanced by controlling the stratification variables. This dynamic randomisation stratified by operation type (elective or emergency) and site. 1 The principles of the randomisation design can be found in Russell et al. 32 The customised randomisation process was hosted by Sealed Envelope^TM (London, UK; www.sealedenvelope.com; accessed March 2021), an independent company registered with the Information Commissioner’s Office, which provided a validated and fully automated 24-hour access telephone service for this dynamic randomisation. Randomisation was performed by the closing surgeon and took place as close as possible to commencement of closure. HART study surgeons were provided with training on the use of the telephone randomisation system.

Blinding

Study participants were informed prior to surgery that they would be randomised to receive one of the two closing techniques. Both the study participants and the post-surgical clinical assessors were blinded to the closure techniques. To attempt to maintain blinding, the method of closure was not documented in the operating notes and/or the clinical assessor was asked to complete the hernia examination (CRF) before reviewing the patient notes. The surgeon could not be blinded to the arm allocation because they were to undertake the closure as per randomisation. The data entry staff, the trial manager and the data manager were not blinded because of their central role in data collection and collation. The trial statisticians were blinded until the point of data lock during the final analysis stage.

Data

All data, including those on screening, eligibility, randomisation, surgery and follow-up, were collected on a patient CRF and managed using the MACRO database (Elsevier, Amsterdam, the Netherlands; www.elsevier.com/en-gb/solutions/veridata/macro; accessed March 2021), a secure electronic data capture system recording the information collected on CRFs and allowing rapid data extraction.

Radiology data (CT scans and related data) were managed in REDCap (Research Electronic Data Capture; www.project-redcap.org; accessed March 2021), a secure web application for data collection.

The data collection comprised pre-surgery data, intraoperative data and post-surgery follow-up data. For detailed information of the data collected in the CRF, see Report Supplementary Material 1.

Patients were assessed for eligibility using the study inclusion and exclusion criteria, and eligible patients who gave informed consent were recruited to the trial. Once they were recruited, and prior to surgery, patients’ baseline demographic information, medical history, surgical history, hernia status and QoL were collected. Intraoperative data were collected immediately post surgery and included the surgery type, grade of surgeon, details on the materials used and surgical outcomes.

Quality-of-life data were collected using the validated SF-1233 and FACT-C questionnaires. 34 Questionnaires were given to patients to complete while they were waiting for follow-up visits or, where necessary, were sent by post. The SF-12 is a short version of the SF-36 item health survey. It is a general health questionnaire capturing information on both physical and mental health across eight domains. FACT-C is a 37-item colorectal cancer-specific tool that adds a subset of 10 colorectal cancer-specific items to the original 27-item FACT-G (which is used for any cancer population). FACT-C consists of five subscales: physical well-being, social and family well-being, emotional well-being, functional well-being and the Colorectal Cancer Subscale. Both SF-12 and FACT-C can be self-administered by the patient or completed in an interview with the patient. Both tools are used widely in research, and a review of available generic and colorectal cancer-specific patient-reported outcome measure tools35 suggests that the SF-12 and FACT-C were appropriate choices for this study.

The MACRO database was locked for year 1 data collection in March 2019 and for year 2 data collection in April 2020. The REDCap database was locked for year 1 data in September 2019. The later datalock for REDCap was to allow all CT scan images and data to be transferred from individual sites via the Picture Archiving and Communications System (PACS) and entered into the database to be reviewed independently by two radiologists.

Following data lock, data were checked for completeness. Any outstanding queries were raised with individual sites and resolved, and SAE coding was completed and reviewed.

Year 1 data analysis included data collected at 12 months ± 2 months (i.e. 10–14 months) and year 2 data analysis similarly included a 2-month window either side (24 months ± 2 months; range 22–26 months). If a patient’s treatment course and clinical requirements meant that no CT was undertaken during this window, the CT scan closest to the 1-year time point was used. In total, 86.5% of CT scans were recorded within 14 months, with 95% recorded within 16 months.

Follow-up

Patients consented to be followed up at 30 days, 6 months and 1 year and annually thereafter. Postoperative follow-up data included data from clinical examination, CT scanning, patient-completed SF-12 and FACT-C quality-of-life forms and patient diaries. Full details of the data collected at individual follow-up points are reported in Appendix 4, Table 31.

Safety

As part of the monitoring of adverse events, information related to surgical site infection (SSI) and postoperative burst abdomen was collected. Colorectal cancer stage information was collected within 30 days post surgery.

Statistical methods

The trial analysis was carried out in accordance with the statistical analysis plan (SAP) using treatment allocated [intention to treat (ITT)], with participants in the arm allocated at randomisation. QoL data were collected using recognised and validated patient-reported outcome measure tools, namely the generic SF-12 and the condition-specific FACT-C.

Continuous variables that follow an approximately normal distribution were summarised using n (non-missing sample size), means, SDs, minimums and maximums. Skewed continuous variables were summarised using medians and interquartile ranges. Categorical variables were summarised using frequencies and percentages. All hypothesis testing was planned to be two-tailed with a 5% significance level and no adjustment for multiple testing.

Binary logistic regression analysis for the outcome variable IH was adjusted for all of the important baseline covariates. As this was a variable selection method, we employed a stepwise backward selection search starting with the full model, considering all of the adjusting covariates and removing the least significant, and repeating until only statistically significant covariates and the arm indicator remained in the model. The initial list of variables was arm indicator, age, gender, ethnicity, BMI, diabetes, any chemotherapy, radiotherapy, history of high alcohol use, history of chronic obstructive pulmonary disease (COPD), any IH present clinically, American Society of Anesthesiologists (ASA) class, whether the patient was from a high (≥ 50 enrolled participants) or a low (< 50 enrolled participants) recruiting site, baseline QoL measures [i.e. the SF-12 Physical Component Summary (PCS) and Mental Component Summary (MCS) and FACT-C score] and baseline Physiological and Operative Severity Score for understanding Mortality and Morbidity (POSSUM) score.

Data processing and analyses were performed using Stata^® version 16 (StataCorp LP, College Station, TX, USA).

Baseline data

Baseline characteristics, including demographics, medical history, specific conditions (e.g. COPD and diabetes), chemotherapy, radiotherapy, abdominal aortic aneurysm, abdominal surgery history, current hernia status (incisional, non-incisional), baseline POSSUM score and QoL measures (FACT-C and SF-12), were summarised by treatment arm using appropriate descriptive methods for all randomised participants. Full details can be found in the statistical and health economic analysis plan (see Report Supplementary Material 2).

Primary end point

The primary outcome was the incidence of IH, and we tested the null hypothesis that there is no difference between the two surgical procedures (i.e. Hughes repair and standard mass closure) at 1 year (the primary end point). Patient and surgical factors associated with a risk of IH were identified via binary logistic regression models.

Secondary end points

Patient and public involvement

Two PPI representatives were involved in this study at any given point. Both representatives had previous experience of working with research groups and sat on the Trial Management Group for the project. A PPI representative sat also on the Steering Committee for the trial.

The PPI representatives were involved throughout this study from protocol development stage to study conclusion.

Patient and public representatives co-participated with the study team in areas such as designing patient materials and promoting the study to patients. PPI representatives worked with study team members to produce a patient information pack introducing the trial and explaining the aims of the research. The PPI representatives were involved with the development of the patient information sheets, patient questionnaires and contributed extensively to the writing of the Plain English summary for the study.

Patient and public representatives were in attendance at all Trial Management Group meetings and actively contributed to any discussions and decision-making processes. Over the course of the study, PPI representatives joined the clinical team at a number of public events giving presentations about public involvement in research to nurses and medical students. They were also involved with providing feedback to trial participants throughout the study and keeping sites up to date with trial progress via a study newsletter and designed a thank-you card giving information about the trial and details of where the results would be posted, which was sent to all surviving participants following the completion of the trial.

The PPI representatives were paid honoraria and out-of-pocket expenses in line with Health Care Research Wales and INVOLVE guidelines for attending meetings.

Patient representatives will have further input into the dissemination of the final results.

Participant flow

Recruitment

The first study site opened for recruitment in August 2014 and the last site opened in December 2016. A total of 28 sites opened for recruitment; one proposed site did not open and one site, although open for recruitment, did not recruit any patients to the study. Recruitment was planned to be completed by March 2017. However, owing to slower than anticipated recruitment, the recruitment phase of the study was extended and recruitment was completed in January 2018 (Figure 4). Detailed recruitment by site is reported in Appendix 3, Table 30.

FIGURE 4.

Cumulative recruitment.

Screening and eligibility

Screening logs showed that 3086 patients across 28 sites were screened for eligibility, with 1771 patients deemed to be eligible for study inclusion. Primary reasons for non-eligibility included not having a midline incision, not having colorectal cancer and declining to take part in the study. In a large number of cases, screening logs did not report a reason for non-eligibility (Figure 5).

FIGURE 5.

The CONSORT flow diagram of consented and randomised patients. a, Some sites included only eligible patients on their screening logs; therefore, the number screened is likely to be larger; b, a small number of patients (Hughes repair, n = 15; standard mass closure, n = 13) reported no year 1 data and could not be included in year 1 analysis, as no study discontinuation form had been completed. Of these 28 patients, eight were seen at year 2 (four in each arm), with five having no IH; c, 58 patients in Hughes repair arm and 62 patients in standard mass closure arm reported no year 2 data and could not be included in year 2 analysis. As no study discontinuation form had been completed, these patients could not be considered lost to follow-up in year 2.

Missing data

The study team made every effort to capture a full data set. The frequency of missing data is summarised for each variable. The statistical analysis plan and study protocol necessitated that if data were not missing completely at random, then the trial statistician and chief investigator would discuss the findings and agree on an approach to handle the missing data. The trial statistician identified no reason to indicate that data were not missing completely at random; therefore, protocol procedures for handling data not missing completely at random were not needed. For the key primary outcome, participants with no follow-up data were automatically excluded from the analysis. We scored the QoL measures (i.e. SF-12 and FACT-C) using their scoring algorithms. After this scoring, 20% of data were missing in the outcome scores of these QoL measures at baseline. For those missing, we adopted a multiple imputation method to impute the outcome scores outcome scores.

Baseline data

A summary of the baseline data for the study cohort is presented in Table 3 and a detailed summary of intraoperative characteristics is given in Table 4. The two arms were well balanced, with no differences observed between the arms in any of the baseline characteristics.

Primary outcome

Secondary outcomes

Quality of life

All of the QoL measures (i.e. SF-12 PCS, SF-12 MCS and FACT-C scores) were well balanced between the arms at baseline. No significant difference in score was observed between the arms for either SF-12 or FACT-C at any time point (Tables 7 and 8).

TABLE 7 - Summary of QoL (SF-12) PCS and MCS scores at time points from baseline to 2 years

Notes

Missing n (%) = number missing (mentioned only when some are missing).

MCS, higher score indicates better mental health; PCS, higher scores indicates better physical activity.

QoL measure	Hughes repair arm	Standard mass closure arm	Total	Mean difference (I – C) (95% CI)	p-value (t-test)
SF-12: PCS
Baseline
Mean (SD)	43.6 (5.5)	43.7 (5.5)	43.7 (5.5)	–0.09 (–0.93 to 0.76)	0.8
Missing, n (%)	82 (20.5)	78 (19.5)	160 (20.0)
30 days
Mean (SD)	41.5 (5.7)	40.8 (5.8)	41.2 (5.7)	0.73 (–0.19 to 1.64)	0.1
Missing, n (%)	108 (26.9)	108 (26.9)	216 (26.9)
6 months
Mean (SD)	42.8 (5.0)	42.9 (4.8)	42.8 (4.9)	–0.08 (–0.88 to 0.72)	0.8
Missing, n (%)	101 (25.2)	121 (30.2)	222 (27.7)
1 year
Mean (SD)	43.1 (4.4)	43.7 (4.5)	43.4 (4.4)	–0.56 (–1.30 to 0.18)	0.1
Missing, n (%)	112 (27.9)	134 (33.4)	246 (30.7)
2 years
Mean (SD)	43.0 (4.3)	43.2 (4.4)	43.1 (4.3)	–0.13 (–0.95 to 0.68)	0.75
Missing, n (%)	178 (44.4)	187 (46.6)	365 (45.5)
SF-12: MCS
Baseline
Mean (SD)	52.5 (11.9)	52.9 (12.1)	52.7 (12.0)	–0.41 (–2.2 to 1.50)	0.7
Missing, n (%)	81 (20.2)	78 (19.4)	159 (19.8)
30 days
Mean (SD)	46.5 (13.2)	48.2 (12.8)	47.30 (13.0)	–1.77 (–3.87 to 0.33)	0.1
Missing, n (%)	107 (26.7)	106 (26.4)	213 (26.6)
6 months
Mean (SD)	51.50 (13.3)	51.7 (12.3)	51.6 (12.8)	–0.15 (–2.23 to 1.94)	0.9
Missing, n (%)	101 (25.2)	121 (30.2)	222 (27.7)
1 year
Mean (SD)	53.9 (12.1)	53.7 (11.7)	53.8 (11.9)	0.21 (–1.77 to 2.18)	0.8
Missing, n (%)	112 (27.9)	133 (33.2)	245 (30.5)
2 years
Mean (SD)	54.0 (12.0)	54.0 (12.8)	54 (12.4)	0.08 (–2.25 to 2.42)	0.95
Missing, n (%)	178 (44.4)	187 (46.6)	365 (45.5)

TABLE 8 - Summary of QoL FACT-C at time points from baseline to 2 years

FACT-C score, higher is better health (score 0–136).

Time point	Hughes repair arm	Standard mass closure arm	Total	Mean difference (I – C) (95% CI)	p-value (t-test)
Baseline
Mean (SD)	70.6 (9.9)	71.7 (10.0)	71.08 (10.0)	–1.21 (–2.78 to 0.35)	0.13
Missing, n (%)	88 (21.9)	90 (22.4)	178 (22.2)
30 days
Mean (SD)	66.8 (8.9)	65.9 (8.8)	66.3 (8.9)	0.86 (–0.59 to 2.32)	0.25
Missing, n (%)	116 (28.9)	116 (28.9)	232 (28.9)
6 months
Mean (SD)	66.3 (9.9)	66.6 (9.3)	66.5 (9.7)	–0.33 (–1.92 to 1.25)	0.68
Missing, n (%)	107 (26.7)	125 (31.2)	232 (28.9)
1 year
Mean (SD)	67.6 (8.7)	67.3 (9.7)	67.4 (9.2)	0.34 (–1.21 to 1.89)	0.66
Missing, n (%)	124 (30.9)	133 (33.2)	257 (32.0)
2 years
Mean (SD)	67.0 (8.9)	68.2 (8.9)	67.6 (8.9)	–1.14 (–2.83 to 0.54)	0.18
Missing, n (%)	171 (42.6)	195 (48.6)	366 (45.6)

Tables 9 and 10 outline the QoL scores in patients with and patients without IH (regardless of their allocated arm). Patients with an IH at year 1 had significantly lower mean PCS scores than patients without an IH (p = 0.03) at baseline. However, this difference was not observed at other time points. No statistically significant differences were observed in mean SF-12 MCS (see Table 7) or FACT-C (see Table 8) scores when comparing patients with and patients without an IH at any time point.

TABLE 9 - Summary of QoL (SF-12) PCS and MCS scores in patients with and without IH at year 1

MCS, higher score indicates better mental health; PCS, higher score indicates better physical activity.

QoL measure	Patients with IH (N = 107)	Patients without IH (N = 565)	Total (N = 672)	Mean difference (IH – no IH) (95% CI); p-value (t-test)
SF-12: PCS
Baseline
Mean (SD)	42.6 (5.6)	44.0 (5.4)	43.8 (5.4)	–1.40 (–2.70 to –0.11); 0.03
Missing, n (%)	27 (25.2)	96 (17.0)	123 (18.3)
30 days
Mean (SD)	41.2 (5.9)	41.1 (5.6)	41.1 (5.7)	0.13 (–1.18 to 1.44); 0.8
Missing, n (%)	21 (19.6)	117 (20.7)	138 (20.5)
6 months
Mean (SD)	42.4 (4.9)	43.0 (4.9)	42.9 (4.9)	–0.59 (–1.73 to 0.54); 0.3
Missing, n (%)	23 (21.5)	111 (19.7)	134 (19.9)
1 year
Mean (SD)	43.8 (4.6)	43.3 (4.4)	43.4 (4.4)	0.51 (–0.50 to 1.52); 0.3
Missing, n (%)	19 (17.8)	108 (19.1)	127 (18.90)
SF-12: MCS
Baseline
Mean (SD)	53.7 (11.4)	53.2 (11.6)	53.3 (11.6)	0.48 (–2.30 to 3.23); 0.7
Missing, n (%)	27 (25.2)	95 (16.8)	122 (18.1)
30 days
Mean (SD)	47.3 (12.2)	47.7 (12.9)	47.6 (12.8)	–0.46 (3.42 to 2.50); 0.8
Missing, n (%)	21 (19.6)	114 (20.2)	135 (20.1)
6 months
Mean (SD)	52.1 (12.7)	52.2 (12.4)	52.2 (12.4)	–0.11 (–3.01 to 2.78); 0.9
Missing, n (%)	23 (21.5)	111 (19.6)	134 (19.9)
1 year
Mean (SD)	52.8 (12.1)	53.9 (11.8)	53.7 (11.9)	–1.01 (–3.73 to 1.70); 0.5
Missing, n (%)	19 (17.8)	107 (18.9)	126 (18.8)

TABLE 10 - Summary of QoL (FACT-C) over 1 year in patients with and without IH at 1 year

FACT-C score, higher is better health (score range 0–136).

Time point	Patients with IH (N = 107)	Patients without IH (N = 565)	Total (N = 672)	Mean difference (I – C), 95% CI; p-value (t-test)
Baseline
Mean (SD)	70.9 (8.5)	71.6 (9.7)	71.5 (9.5)	–0.74 (–30.2 to 1.55); 0.5
Missing, n (%)	29 (27.1)	110 (19.5)	139 (20.7)
30 days
Mean (SD)	66.7 (7.7)	66.2 (9.1)	66.3 (8.9)	0.42 (–1.65 to 2.50); 0.7
Missing, n (%)	22 (20.6)	132 (23.4)	154 (22.9)
6 months
Mean (SD)	65.9 (9.8)	66.7 (9.6)	66.6 (9.6)	–0.85 (–3.13 to 1.42); 0.5
Missing, n (%)	25 (23.4)	119 (21.1)	144 (21.4)
1 year
Mean (SD)	67.5 (7.8)	67.4 (9.5)	67.4 (9.2)	0.08 (–2.10 to 2.25); 0.9
Missing, n (%)	24 (22.4)	113 (20.0)	137 (20.4)

The results of the mixed-model repeated-measures analysis of the PCS scores showed no statistically significant differences in the between-group difference mean scores at any time point (Table 11 and Figure 6). Similar results were observed from the mixed-model repeated-measures analysis of the MCS scores (Table 12 and Figure 7). Following an initial sharp decline in MCS score in the month immediately following their index operation in both arms, patients reported improvement over time, with similar scores in both arms from 6 months onwards. In terms of FACT-C scores, the between-arm differences of the mean change from baseline to first month and to year 1 are statistically significant. Despite that the actual difference in the FACT-C score is not clinically relevant, the score in the standard mass closure arm seems to decrease steeply when compared with the Hughes repair arm at these two time points. The detailed mixed-model repeated analysis for the FACT-C score can be found in Appendix 5, Table 41 and Figure 13.

TABLE 11 - Results of the mixed-model repeated measures of the QoL (PCS) over 2 years: mean change

Mean change from baseline and between-arm differences are predicted means and 95% CIs, estimated from mixed-effect models and adjusted by age, gender, ethnicity, BMI, COPD, any chemotherapy or radiotherapy and other baseline characteristics. Significant covariates were age, sex (female), any chemotherapy or radiotherapy (reference: no therapy) and visit time (30 days, 6 months; reference: baseline).

Note

PCS, higher score indicates better physical activity.

Variable	Mean baseline PCS score (SD)	Mean change in PCS score from baseline (95% CI);a p-value
		Month 1	Month 6	Year 1	Year 2
Standard mass closure	43.7 (5.5)	–2.9 (–3.65 to –2.21); 0.00	–0.8 (–1.51 to –0.04); 0.04	0.0 (–0.76 to 0.73); 0.9	0.5 (–1.31 to 0.29); 0.2
Hughes repair	43.6 (5.5)	–2.2 (–2.91 to –1.46); 0.00	–0.8 (–1.53 to –0.10); 0.03	–0.6 (–1.35 to 0.10); 0.09	–0.6 (–1.40 to 0.17); 0.13
Between-arm difference in means		0.8 (–0.27 to 1.77); 0.15	–0.0 (–1.07 to 0.99); 0.94	–0.6 (–1.64 to 0.43); 0.25	–0.1 (–1.22 to 1.01); 0.86

FIGURE 6.

Mean PCS scores over time by arm.

TABLE 12 - Results of the mixed-model repeated measures of the QoL (MCS) over 2 years: mean change

Mean change from baseline and between-arm differences are predicted means and 95% CIs, estimated from mixed-effect models and adjusted by age, gender, ethnicity, BMI, COPD, any chemotherapy/radiotherapy and other baseline characteristics. Significant covariates were gender (female), BMI, ethnicity: white (reference: non-white); ASA class: ≥ 3 (reference: class 1); both chemotherapy and radiotherapy (reference: no therapy) and visit time (30 days, 6 months; reference: baseline).

Note

MCS, higher score indicates better mental health.

Variable	Mean baseline MCS score (SD)	Mean change in MCS score from baseline (95% CI);a p-value
		Month 1	Month 6	Year 1	Year 2
Standard mass closure	52.9 (12.1)	–5.2 (–6.58 to –3.87); 0.00	–1.5 (–2.97 to –0.19); 0.03	0.0 (–1.4 to 1.4); 0.99	0.0 (–1.5 to 1.51); 0.99
Hughes repair	52.5 (11.9)	–6.2 (–7.57 to –4.86); 0.00	–0.9 (–2.25 to 0.45); 0.2	0.8 (–0.54 to 2.2); 0.24	0.9 (–0.56 to 2.4); 0.22
Between-arm difference in means		–1.0 (–2.91 to 0.93); 0.31	0.7 (–1.25 to 2.62); 0.49	0.8 (–1.13 to 2.78); 0.41	0.9 (–1.19 to 3.02); 0.40

FIGURE 7.

Mean MCS scores over time by arm.

Sensitivity and specificity of computed tomography scanning compared with clinical examination for identification of incisional hernia at 1 year

The rate that IH was diagnosed by clinical examination or by CT scan at 1 year is reported in Table 13. Clinical examination was considered the reference test, and the results from this were considered the ‘true’ results. In total, 630 patients had both a 1-year clinical examination result and a CT scan available for analysis and are included in the analysis for sensitivity and specificity.

TABLE 13 - Incidence of IH at 1 year by clinical examination and/or CT scan

CE, clinical examination.

	Hughes repair arm, n	Standard mass closure arm, n	Total study (both arms), n
IH on CE only	7	9	16
IH on CT only	114	103	217
IH on both CE and CT	37	47	84
Total IH	158	159	317
Negative for IH on both	156	157	313
Total	314	316	630

In total, across both study arms, 317 IHs were diagnosed by clinical examination and/or CT scan, of which 84 were identified by both modalities: 37 in the Hughes repair arm and 47 in the standard mass closure arm. Clinical examination identified 44 IHs in the Hughes repair arm and 56 IHs in the standard mass closure arm, that is clinical examination identified a further seven IHs in the Hughes repair arm and nine in the standard mass closure arm that were not detected on the patient’s CT scan. Conversely, CT scan alone identified an additional 114 IHs in the Hughes repair arm and an additional 103 IHs in the standard mass closure arm (see Table 13).

From the 2 × 2 tables (Table 14) and the analysis of these results, it appears that CT scanning has high sensitivity and moderate specificity when compared with clinical examination. The sensitivity of CT scanning was 84%, with 301 scans indicating the presence of an IH, and the specificity was 59%, with 313 patients without a hernia having a negative CT scan across both arms (Table 15). Of the 301 patients with a positive CT scan, only 84 had an IH on clinical examination, indicating that CT scanning had a low positive predictive value (27.9%). Similarly, of 329 patients with a negative CT scan, 313 did not have an IH on clinical examination, indicating that CT scanning had a very high negative predictive value of 95% (see Table 15). The corresponding ROC curve (Figure 8) or area under the curve was 72%.

TABLE 14 - The 2 × 2 table of IH diagnosis by clinical examination (reference standard) vs. CT scan

Note

A total of 681 patients with radiology data are available. Out of these 681 patients, 630 have clinical examination data available, leaving 51 (7.5%) patients as missing.

IH by clinical examination	IH by CT scan, n (%)
	Positive	Negative	Total
All patients
Positive	84 (84.0)	16 (16.0)	100
Negative	217 (40.9)	313 (59.1)	530
Total	301	329	630
Hughes repair arm
Positive	37 (84.1)	7 (15.9)	44
Negative	114 (42.2)	156 (57.8)	270
Total	151	163	314
Standard mass closure arm
Positive	47 (83.9)	9 (16.1)	56
Negative	103 (39.6)	157 (60.4)	260
Total	150	166	316

TABLE 15 - Sensitivity and specificity of CT scanning (clinical examination as reference standard)

	Hughes repair	Standard mass closure	All patients
Sensitivity (%)	84.1	83.9	84
Specificity (%)	57.8	60.4	59.1
Positive predictive value (%)	24.5	31.3	27.9
Negative predictive value (%)	95.7	94.6	95.1
ROC areas (95% CI)	0.71 (0.65 to 0.77)	0.72 (0.67 to 0.78)	0.72 (0.67 to 0.76)

FIGURE 8.

The ROC curve (reference test: clinical examination). Area under ROC curve = 0.7153.

Methods of the health economic evaluation

Before the analysis commenced, a health economic analysis plan was produced, reviewed by the trial team and incorporated into the SAP. The health economic team followed this analysis plan during the conduct of the economic evaluation without deviation.

Total costs

By adding up the implementation costs of Hughes repair (in the intervention group only), post-surgical inpatient costs (including adverse events such as SSI and burst abdomen) and subsequent primary, secondary and social care for all patients in both trial groups, the total and mean costs per patient (including 95% CIs) of the two closure methods were calculated to derive the incremental costs (or cost savings) of the intervention at the 12-month follow-up.

Health outcomes

Clinical examination identified 50 (14.8%) IHs in the Hughes repair arm and 57 (17.1%) IHs in the standard mass closure arm (see Chapter 3).

Quality-adjusted life-years (QALYs) required for the cost–utility analysis were derived from SF-12 responses. We used the SPSS syntax algorithm for SF-12 version 2 to map SF-12 responses to SF-6D scores and patient utilities based on UK value sets for baseline and the 1-, 6- and 12-month follow-up points. 47 During this procedure, missing SF-12 items were replaced with the trial population mean of each item if at least half of the items were complete. 48 When more than six items were missing, the questionnaire was considered ‘not completed’ and no utility value was calculated. We reported utility values for all available cases, which were defined as those with at least a complete baseline questionnaire and at least a utility score at 12 months. Utilities for participants who died during the follow-up period were set to 0 between the point of death and the end of follow-up. The cost–utility analysis used the ITT population, with missing utility values imputed as the mean of each time point, in the base case. Sensitivity analyses used last observation carried forward/backward (LOCF) imputation and available cases (available data for all patients at each time point) to test the impact of missingness on the analysis results. Once the analysis populations were defined, and imputation performed where required, QALYs for each individual patient were calculated according to an area-under-the-curve approach and linear interpolation. This used all four time points to estimate the overall QALYs as a combined measure of patients’ QoL over 12 months.

Cost-effectiveness analysis

The CEA compared the incremental costs with the primary outcome (number of IHs) at the 1-year follow-up to calculate the incremental cost-effectiveness ratio (ICER) for Hughes repair reported as cost per IH avoided.

The base-case analysis was based on the ITT population for the primary outcome of number of IHs identified at clinical examination. The total costs at 12 months for the ITT population (after mean imputation) only were considered. The results of the comparative analysis of incremental costs and effects were summarised in terms of incremental cost-effectiveness ratios (ICERs). An ICER can be represented as:

where C₁ and E₁ are the costs and effects of the intervention arm and C₀ and E₀ are the costs and effects of the control arm, with ΔC and ΔE the incremental costs and effects of the intervention compared with the control.

The ICER is reported to determine the cost-effectiveness of the intervention compared with competing alternatives and to aid decision-making. No established willingness-to-pay threshold for the cost-effectiveness in reduction of IHs is available. Furthermore, cost-effectiveness is a spectrum rather than a dichotomy, with the maximum threshold increasing depending on the circumstances. The reported ICERs from our analysis are presented to assist the decision-making process and are not an absolute statement on whether or not the intervention can be deemed to be cost-effective.

Cost–utility analysis

A within-trial CUA was undertaken to assess the incremental costs per QALY gained as a result of the use of Hughes repair at 12 months. The total costs at 12 months and QALYs for the ITT population were used to calculate the incremental cost per QALY gained.

Quality-adjusted life-years incorporate quantity of life (additional life-years) and QoL in one measure. Thus, by dividing the difference in costs by the difference in QALYs, a cost per QALY can be calculated for each comparison to establish whether the new intervention is less or more cost-effective than, or similarly cost-effective to, the comparator treatment. The ICER resulting from the CUA was compared with the willingness-to-pay threshold of £20,000 per QALY gained as standardised by NICE. No conditions for non-inferiority were applied in this analysis. The results are reported as ICERs showing the extra cost of producing one extra QALY or the extra savings achieved by sacrificing one additional QALY.

Sensitivity analysis

Sensitivity analyses were undertaken to test the robustness of the results of both CEA and CUA considering the uncertainty in input parameters, such as costs and outcomes, and in different scenarios. Deterministic, univariate sensitivity analysis changed surgery, complication and health-care costs and outcomes individually within plausible ranges (using 10%, 20% and 30% of the mean value). Scenario analyses tested different assumptions and recalculated the ICER. Furthermore, probabilistic sensitivity analysis used non-parametric bootstrapping to address joint parameter uncertainty and assess the impact on the ICER during 1000 simulations that were undertaken using random sampling of the distributions of costs and outcomes with results presented on cost-effectiveness planes and as cost-effectiveness acceptability curves (CEACs). The cost-effectiveness plane is a scatterplot of the point estimates obtained as a result of the 1000 simulations depicted in four quadrants representing the probability of the intervention being more or less costly and more or less effective than the control. A CEAC is a curve that describes the probability that the intervention will be cost-effective at different willingness-to-pay-thresholds based on the probabilistic sensitivity analysis. We also undertook scenario analyses to test for the potential impact of baseline imbalances, training costs and imputation method.

Cost–consequences analysis

A cost–consequences table was created to present all of the relevant primary and secondary outcomes alongside the costs in tabular form (without combining them into ICERs) to leave decision-makers the option to form their own view of relative importance.

Results of the health economic evaluation

Total health-care costs

The total post-surgery health-care costs (based on available cases) included the cost of the index hospital stay post surgery, the cost of all primary, secondary and social care in the 12 months following discharge from hospital after the index operation, and the cost of any additional cancer treatment. Patients in the Hughes repair arm accrued mean per-patient post-surgery health-care costs of £15,295.24 (SD £9343.15), compared with £14,650.33 (SD £9092.01) in the standard mass closure arm. The incremental cost of Hughes repair was £644.91 (95% CI –£633.01 to 1922.83). The higher cost was caused by the longer index hospital stay and higher inpatient costs in the first 6 months following surgery, but the difference was not statistically significant (p = 0.322).

Adding the intervention mean per-patient implementation costs of £106.26 increases the cost difference to £751.17 (95% CI –£526.75 to £2029.10; p = 0.249) for the Hughes repair arm in the available-case population.

Total costs of the intention-to-treat population

Consistent with the statistical analysis, the cost-effectiveness and cost–utility analyses were based on the ITT population. Including all patients for whom primary outcome data were available at the 12-month follow-up point resulted in 339 patients in the Hughes repair arm and 333 patients in the standard mass closure arm.

Although these patients had complete primary outcome data, not all cost data were recorded at all time points for every patient. Missing data were spread relatively evenly between both trial arms, with 9.3% of cases with missing data in the standard mass closure arm and 7.7% in the Hughes repair arm. It was noted that 4.8% of patients in the standard mass closure arm and 3.2% of patients in the Hughes repair arm had systematically missing data because the first 37 patients entered into the trial had not been surveyed for health resource utilisation data at baseline. Therefore, as the data were not missing at random and the remaining missing data affected < 5% of the sample, the population mean health-care cost was used for imputation of missing data.

The mean total cost per patient for the ITT population was £15,490.32 (SD £8688.03) in the Hughes repair arm and £14,873.87 (SD £8685.73) in the standard mass closure arm, with a mean difference of £616.45 (95% CI –£699.56 to £1932.47; p = 0.358). This difference is made up of higher costs for the index hospital stay post surgery (mean £213.10, 95% CI –£527.05 to £953.26; p = 0.572) and subsequent health-care costs (mean £453.07, 95% CI –£498.64 to £1404.79; p = 0.350), lower costs for cancer treatment and adjuvant chemotherapy (mean –£155.98, 95% CI –£510.49 to £198.54; p = 0.388) plus an incremental intervention cost of £106.26 for all patients who underwent Hughes abdominal repair (Figure 9).

FIGURE 9.

Mean cost differences between Hughes repair and standard mass closure patients in the ITT population.

There was a statistically non-significant baseline imbalance in health-care costs for the ITT population. In the 6 months prior to randomisation, reported health-care resource use amounted to a mean total cost of £2065.30 (SD £3350.06) per patient in the Hughes repair arm (based on data for 256/339 patients available) compared with £1756.59 (SD £2623.74) in the standard mass closure arm (based on data for 247/333 patients), with a mean difference of £308.71 (95% CI –£217.47 to £834.10; p = 0.250) per patient. The difference was mainly attributed to the higher secondary care costs incurred by patients in the Hughes repair arm prior to surgery and fell slightly when mean imputation was applied to all missing cases to £231.02 (95% CI –£163.30 to £625.33; p = 250).

Health outcomes

The incidence of IH at 1-year clinical examination was 14.75% in the Hughes repair arm compared with 17.12% in the standard mass closure arm (OR 0.84, 95% CI 0.55 to 1.27; p = 0.40).

Using the ITT population (with mean imputation), there was no difference in QALYs between the arms at 12 months (p = 0.979). In the 12 months post randomisation, patients accumulated, on average, 0.6902 (SD 0.0725) QALYs in the Hughes repair arm compared with 0.6901 (SD 0.0673) QALYs in the standard mass closure arm. The QALY difference was 0.0001 (95% CI –0.0105 to 0.0107). The QALY outcomes for available cases and the LOCF imputation of missing data are summarised in Appendix 6.

Cost-effectiveness analysis

The unadjusted mean incremental cost for the ITT population (n = 672) at the 12-month follow-up point was £616.45 higher in the Hughes repair arm. As the number of patients who experienced IH following surgery was 57 (17.12%) in the standard mass closure arm, compared with 50 (14.75%) in the Hughes repair arm (see Table 6), the mean ICER is £26,034 per clinical hernia avoided.

Sensitivity analyses

One-way deterministic sensitivity analyses showed that results are sensitive to changes in costs and effects, with ICERs ranging from £7784 to £81,538 per hernia avoided. The ICERs are most affected by changes in health-care costs and the between-arm difference in hernias.

The majority of results of the 1000 bootstrap iterations during probabilistic sensitivity analysis indicate that Hughes repair is more costly but also more effective than standard mass closure in preventing hernias. However, the results are distributed across all sectors, with cases in which Hughes repair both dominates and is dominated by standard mass closure (Figure 10). This is a result of the small differences in cost and effect.

FIGURE 10.

Cost-effectiveness plane (ITT analysis) for the base case (incremental cost per hernia avoided).

No willingness-to-pay threshold exists for cost-effectiveness analysis. However, the incremental cost of a hernia could be used as a proxy. During the trial, 107 patients experienced an IH and 565 did not. The mean total cost per hernia patient was £15,968.60 (SD £10,411.67). Patients who did not experience IH had a mean total cost of £15,036.42 (SD £8321.39). This means that the costs of patients with hernias were, on average, £932.18 (95% CI –£1175.71 to £3040.08; p = 0.383) higher than those of patients without hernias.

Removing Hughes repair training costs from the total costs, as part of scenario analysis, did not alter the ICER. However, after adjusting for the baseline imbalance in the health-care costs (£231.02 higher in the Hughes repair arm), the ICER was reduced to £16,278 per hernia avoided.

Cost–utility analysis

Based on the unadjusted incremental cost of Hughes repair compared with standard mass closure and the marginal QALY increase, the base-case ICER is £4,359,353 per QALY gained.

Sensitivity analyses

The direction of the results does not change when inputs are altered during deterministic sensitivity analysis (Table 27). Owing to the small between-arm differences in QALY gain, the results are sensitive to changes in this variable without the cost-effectiveness conclusions being affected.

TABLE 27 - Results of the one-way sensitivity analyses

N/A, not applicable.

Parameter	Change	Most cost-effective option
N/A	N/A	Standard mass closure
Costs	100% registrar or 100% consultant at training	Standard mass closure
Costs	Training costs removed	Standard mass closure
Costs	–10%	Standard mass closure
Costs	–20%	Standard mass closure
Costs	–30%	Standard mass closure
Costs	+ 10%	Standard mass closure
Costs	+ 20%	Standard mass closure
Costs	+ 30%	Standard mass closure
Costs	Long-course radiotherapy (25 fractions)	Standard mass closure
Costs	Baseline imbalance adjusted	Standard mass closure
Costs	Patients who withdrew removed instead of censoring at withdrawal date	Standard mass closure
Utilities	–10%	Standard mass closure
Utilities	–20%	Standard mass closure
Utilities	–30%	Standard mass closure
Utilities	+ 10%	Standard mass closure
Utilities	+ 20%	Standard mass closure
Utilities	+ 30%	Standard mass closure
All parameters	Last observation carried forward	Standard mass closure
All parameters	All available cases used	Standard mass closure

Probabilistic sensitivity analysis shows that results are distributed across all quadrants of the CE plane when all analysis inputs are altered simultaneously within predefined ranges. Possible scenarios range from Hughes repair being dominated to it dominating standard mass closure, as a result of the small differences in costs and QALYs between the arms (Figure 11). Overall, the probability that Hughes repair is more cost-effective at a willingness-to-pay threshold of £20,000 per QALY gained is 18.9% (Figure 12).

FIGURE 11.

Cost-effectiveness plane (ITT analysis) for the base-case CUA (incremental cost per QALY gained).

FIGURE 12.

Cost-effectiveness acceptability curve (ITT analysis) for the base-case CUA (incremental cost per QALY gained).

Removing all staff training costs as part of scenario analysis did not affect the ICER. Using LOCF as imputation method instead of mean imputation in the base case gives a mean QALY gain over 12 months of 0.6856 for Hughes repair patients and 0.6844 for standard mass closure patients. The slightly higher QALY difference of 0.0012 (95% CI –0.0119 to 0.0143) reduces the ICER to £516,421 per QALY gained.

Any utility imbalances at baseline were adjusted automatically as QALYs are calculated as relative gains. However, when adjusting for the baseline imbalance in health-care costs, the ICER was reduced to £2,725,658 per QALY gained, with a probability of 29.4% that Hughes repair is cost-effective.

Cost–consequences analysis

Table 28 summarises the results of the cost–consequences analysis.

Limitations

The pragmatic nature of the study, chosen to maximise study engagement by surgeons and sites, and patient recruitment nonetheless led to a wider variation in the incidence of IH in the standard mass closure arm than had been anticipated and may have affected the results. Shortly after the HART study commenced, the STITCH trial28 was published, showing reduced IH rates using the small-bites technique. This led to the significant promotion of this technique, which coincided with the early part of the HART study. This introduced two issues likely to have had an impact on our findings. First, surgeons were learning a second new technique (small stitch) in a non-controlled fashion, which potentially confounded the control arm. Second, it led to some surgeons losing equipoise, as they may have considered the issue of IH prevention resolved and felt that there was no role for the Hughes technique.

Intraoperative randomisation also highlighted the issue of equipoise. Hughes repair took longer to perform, with a mean time for fascial closure of 22 minutes in the Hughes repair arm, compared with 13.2 minutes in the standard mass closure arm. It is possible that at the end of difficult procedures surgeons chose not to randomise and so this important cohort of patients were lost to the study. This is hinted at by considering those who were randomised but did not receive their allocated treatment. A total of 15 patients fell into this category, 13 of whom were in the Hughes arm. Of these 13 patients, three had an IH at 1 year by clinical examination (i.e. 23%, compared with 14.8% in the Hughes arm) (analysed on an ITT basis). In addition, among patients who had consented but were subsequently not randomised, reasons for non-randomisation included ‘ran out of time’. These arguments around equipoise may also explain the large numbers of patients who were screened but did not enter the trial. In addition, screening logs do not report the route (emergency or elective) through which patients considered eligible but not consented were identified. Only 59 patients (standard mass closure arm, n = 29; Hughes repair arm, n = 30) underwent emergency surgery, representing 7.3% of the study population. In the UK, up to 20% of patients with colorectal cancer present in the emergency setting,18 suggesting that a number of potentially eligible patients may not have been identified. The mean time from consent to randomisation was only 0.4 days for emergency patients (compared with 2.7 days for elective patients), suggesting an increased urgency in treating patients in the emergency setting, and so surgeons may have had concerns around factors such as giving eligible patients enough time to decide to take part in the study.

One further limitation of this trial is considered to be the lack of follow-up time. European Hernia Society guidelines published after the commencement of the HART study recommend a minimum of 2-year follow-up and also imaging as a method of detection rather than clinical examination alone. 52 The 1-year follow-up suggests that rates of IH do not differ significantly between the two methods of closure. Further analysis of 2-year data similarly shows no significant difference in the incidence of IH detected by clinical examination. The lack of long-term follow-up in clinical trials is a common limitation, and outcomes such as IH can develop at any time after surgery, lending support to the idea that a minimum follow-up for trials should be considered to ensure that there is enough follow-up time for primary outcomes to be observed.

A limitation of the health economic evaluation is the use of mean imputation rather than multiple imputation to account for missing data in line with the analysis plan and protocol. Although this will cause CIs to be narrower than when using multiple imputation, considering that the number of missing cases was small, and also the overwhelming evidence pointing to the intervention being less cost-effective, it is unlikely that the approach to address missing data and the missingness will have had an impact on the findings. In addition, the health economic evaluation uses independent samples t-tests to derive incremental costs and QALYs. This approach does not account for any potential imbalances at baseline or the likely skewed nature of the data, particularly costs, which are often gamma or log-normal distributed. However, although the simple analysis does not adjust for baseline covariates, no significant baseline imbalances were found and undertaking regression modelling would not change the direction of the results. The results of the cost–utility analysis are sensitive to changes in patient QoL, which is caused by the small differences in utility scores between the two arms. Considering that disease-specific questionnaires have shown significantly reduced QoL in patients with IH after colonic cancer resection,53 this could be caused by an insensitivity and lack of responsiveness of the generic SF-12 instrument to small changes in health status with specific conditions. However, the SF-36 questionnaire, a longer version of the SF-12, was found to be suitable for picking up changes in QoL in hernia patients,54 and no significant difference was recorded in this trial in FACT-C scores. Therefore, the difference in hernia incidence of seven cases in year 1 could be too small to affect the mean QoL of the trial population.

Generalisability

The HART study was consciously designed as a pragmatic study to ensure that the results were as generalisable as possible across the UK and beyond in similar health-care systems. With its simple design and the addressing of a common clinical problem with a low-tech solution, it had the potential to be practice changing.

All patients with colorectal cancer were eligible, including elective and emergency patients. This was important because up to 20% of patients with colorectal cancer present as emergencies. A relatively small number of emergency patients was recruited in the end, reflecting the difficulties of trial recruitment in that setting. However, the broad demographic picture is representative of the colorectal cancer population in age, sex, ethnicity and comorbidity, allowing the results to be extrapolated across this group. The wide range of sites, from large university hospitals to smaller district general hospitals, also enhances this. In relation to ethnicity specifically, statistics published in 2009 from the National Cancer Intelligence Network55 report that age-standardised rates for colorectal cancer in England are lower in all ethnic groups, including black, Asian and Chinese ethnic groups, for both male and female patients than in the white ethnic group, and the authors consider this to be borne out in the lower proportion of patients recruited to the trial from these ethnic groups. The authors acknowledge, however, that understanding when risk may differ for different ethnic groups and ensuring that these groups are adequately represented in clinical trials is crucial to generating results that are of benefit to those most at risk.

This trial included only colorectal cancer patients and it is unclear whether patients undergoing surgery for colorectal cancer are different in any way from patients undergoing midline incisional surgery for other reasons. It is certainly appreciated that patients with abdominal aortic aneurysms are at very high risk for IH following surgery and, from the CT scan results presented here, it is clear that colorectal cancer patients are also a vulnerable group.

Patient and public involvement

Patient and public involvement representatives understood that involvement in the HART study was a long-term commitment. The study lasted over 7 years from inception to the 1-year results point and, as a consequence, there was some turnover in the representatives involved at various points in the study. The HART study has been very fortunate to have had PPI representatives who have been actively involved from the very beginning and remain involved in working on the dissemination of the study results. PPI representatives were asked to attend monthly trial management meetings and took part in a number of public events, giving presentations and providing information about the HART study. This long-term commitment to an individual study has proven rewarding and given our PPI members a sense of ownership.

Patient and public involvement in the HART study resulted in a number of important developments that helped the clinical team to engage with study participants in a positive way. For example, one PPI representative suggested that trial packs be produced that could be given patients who were recruited to the study. Later in the process, the PPI representative also suggested that a thank-you card could be sent to all patients who had been recruited to the study to acknowledge the importance of the role of patients who agree to take part in clinical studies. Some study participants took the time to e-mail the study team to say how much they appreciated that gesture. It is perhaps important for study teams to remember that, although patients may be happy to take part in clinical trials and may well have a personal interest in taking part, it is nonetheless important to acknowledge the role that their participation plays, and small gestures can mean a lot.

Listening to the PPI representatives can provide a number of learning opportunities for clinical team members. Although the PPI representatives reported that they were treated with respect and felt that the study team always listened to their point of view and involved them in decision-making processes throughout the study, the patient representative did also indicate they felt a little out of their depth at the start of the process and conscious of the time taken by study team members in helping them to understand their role in the study. It is, therefore, important that study team members maintain an awareness of the needs of their PPI representatives, provide the relevant support and ensure that the representatives understand the value of their contribution. As a result of their positive experience of being involved in the HART study, one PPI representative stated they had gained so much confidence that they are now involved in training others, including research staff, in PPI. Another PPI representative has gone on to become the lead lay research partner for the Wales Cancer Research Centre and participated in the NICE Colorectal Guideline Committee. One patient representative reported that, although they found it stimulating to be part of the HART study, they had concerns about the possible benefit that they could bring, as they were not a colorectal cancer patient. The other patient representative had undergone major surgery for colorectal cancer but had not had an IH, and it is possible that having a patient representative with experience of both colorectal cancer and IH would have brought additional benefit and insight. In future studies, consideration could be given to the inclusion of a patient representative with direct experience of the outcome under investigation (IH) as well as a patient representative with more broadly relevant experiences.

Lessons learned

Data collection and management for such a large, multicentre study posed some logistical issues over the course of the study. Making CT images available to the study radiologists for review presented a particular issue. The study protocol stated that CT images should be transferred to the lead site over the course of the study, but this did not happen. Image files could not be stored in the study databases as part of the patient CRF; therefore, all CT images had to be transferred to the lead site via PACS at the end of the data collection period. In addition, once transferred to the lead site, the CT images remained on the system for only a short time before they were automatically deleted, presenting an additional difficulty. Not resolving this issue until close to the end of the study did create a time pressure that might have been avoided and, for future studies that require images, due consideration should be given to the most efficient way for the study team to access data, including any image files that might not easily be transferred between sites or made available on study databases.

Although all sites kept screening logs, some inconsistencies were noted in the way that sites recorded information. Some sites kept detailed screening logs only from the point at which patients were considered eligible, meaning that the number of people actually screened for eligibility might have been larger than the 3098 reported. Sites were given standard templates to record screening information; however, the use of the template was not mandatory and, as a result, there was variation from site to site in the level of detail recorded.

According to the study screening logs, 700 patients who had been initially considered eligible for inclusion in the trial were not consented to the study. In 259 cases, the reason given was that the patient declined to take part. No further information is available about the reasons for declining and, although patients are not required to give any reason why they do not want to take part in a study, it may be useful, where possible and a patient is willing, to ask for a reason so that researchers might better understand why people do not want to take part in studies. In the case of the HART study, the follow-up requirements may have been considered burdensome, with patients required to consent to 5 years of follow-up along with commitments to completing questionnaires and attending clinic visits at numerous points throughout the 5 years.

Although the study recruited participants from both the elective and the emergency settings, the number of patients recruited from the emergency setting was very small. This might be reflected in the fact that the study screening logs reported that in 288 cases the reason for not consenting an eligible patient was the surgeon’s decision, suggesting that consideration could be given to whether specific recruitment and consent processes may be required that account for the difference between elective and emergency routes. This might help to ensure that the proportion of patients recruited from the emergency setting reflects the current UK situation.

The HART study was a large trial conducted over 7 years. Changes in the make-up of the research team both at individual study sites and within trial management had an inevitable impact on the day-to-day running of the study. Changes in personnel at study sites can have an impact on recruitment rates and follow-up data collection, and changes in the trial management team can result in delays in identifying and responding to problems.

Conclusions

The HART study has provided a rich data set to enhance the understanding of the common and frustrating problem of IH following colorectal surgery. This will continue to yield more insights going forward. In considering a novel closure method using interrupted sutures to reinforce the abdominal wall (the Hughes technique), the trial has failed to show that this confers a statistically significant benefit. It has provided the most definitive evidence for the high incidence of IH after colorectal cancer surgery that will become the reference point in the surgical literature and will encourage research investment in future prevention methods, including the use of mesh.

The HART study has identified new risk factors for IH formation, including preoperative radiotherapy, that will inform future research.

The study has demonstrated the important clinical finding that IH cannot be confidently excluded by clinical examination alone, and this will inform future clinical practice.

The finding of a lower PCS score in patients who go on to develop IH strongly supports the concept of prehabilitation that is being widely promoted to improve postoperative outcomes.

Finally, the HART study has demonstrated that it is possible to carry out high-quality multicentre interventional surgical research in the UK across a wide variety of hospital settings, fostering collaboration and engagement with surgeons and research departments and, crucially, patients and the public and their representatives.

Contributions of authors

Susan O’Connell (https://orcid.org/0000-0002-5887-2771) (Healthcare Research Scientist/Triallist) led the writing of the Health Technology Assessment report and contributed to the interpretation of the trial findings. She contributed to the day-to-day trial management.

Saiful Islam (https://orcid.org/0000-0003-3182-8487) (Trial Statistician) co-wrote the statistical and health economic analysis plan and carried out all of the study data analyses. He contributed to the interpretations of the outcomes and drafting of the report.

Bernadette Sewell (https://orcid.org/0000-0001-5471-922X) (Health Economist) designed the health economic evaluation, co-wrote the statistical and health economic analysis plan, carried out the analysis and wrote the health economics sections of the report. She reviewed the final report for consistency.

Angela Farr (https://orcid.org/0000-0002-2087-9310) (Health Economics Analyst) was involved in the health economic analysis and drafting and checking the health economics sections of the report.

Laura Knight (https://orcid.org/0000-0003-2726-8026) (Healthcare Research Scientist/Trial Manager) was responsible for the day-to-day trial management from May 2018. She was responsible for the co-ordination and checking of safety data and contributed to the literature review and writing of the report.

Nadim Bashir (https://orcid.org/0000-0003-0501-1342) (Senior Data Manager at the Swansea Trials Unit) built and maintained the MACRO database for the collection of the data, developed checks and corresponded with sites to ensure data quality and completeness and assisted the statistician with preparation of analytical data sets. He reviewed the final report for accuracy.

Rhiannon Harries (https://orcid.org/0000-0001-7095-3673) (MD FRCS) (Consultant Colorectal Surgeon based at Swansea Bay University Health Board) was a co-investigator for the HART study and was involved in the grant application and the pilot and feasibility studies. She undertook most of the surgeon training and contributed to the writing of the final report.

Sian Jones (Public and Patient Involvement) contributed to the design of the trial. She contributed substantially to the PPI sections of the report and reviewed the final report for accuracy.

Andrew Cleves (https://orcid.org/0000-0003-0431-3138) (Healthcare Research Scientist/Triallist) contributed to the day-to-day trial management. He was responsible for the acquisition of radiology data from individual sites. He reviewed the final report for accuracy.

Greg Fegan (https://orcid.org/0000-0002-2663-2765) (Director of the Swansea Trials Unit) was involved with the statistical and health economic analysis plan, was a conduit between the data manager and statistician to maintain the latter’s blinding prior to data lock, assisted the trial statistician with the analysis, interpreted the trial findings and supported the drafting of the report.

Alan Watkins (https://orcid.org/0000-0003-3804-1943) (Professor, Health Data Science) contributed to study management and design modification, represented the study team on its oversight committees, co-wrote the statistical and health economic analysis plan, and supported the clinical effectiveness analysis. He contributed to the drafting of the report.

Jared Torkington (https://orcid.org/0000-0002-3218-0574) (Professor of Colorectal Surgery) was the chief investigator. He was involved with the original grant application, co-designed the trial, co-wrote the statistical and health economic analysis plan, interpreted the trial findings and supported the drafting of the report.

Publications

Cornish J, Harries RL, Bosanquet D, Rees B, Ansell J, Frewer N, et al. Hughes Abdominal Repair Trial (HART) – abdominal wall closure techniques to reduce the incidence of incisional hernias: study protocol for a randomised controlled trial. Trials 2016;17:454.

HART Collaborative. Incisional hernia following colorectal cancer surgery according to suture technique: the Hughes Abdominal Repair Randomised Trial (HART). Br J Surg 2022; in press.

Data-sharing statement

All data requests should be submitted to the corresponding author for consideration. Access to anonymised data may be granted following review.

Patient data

This work uses data provided by patients and collected by the NHS as part of their care and support. Using patient data is vital to improve health and care for everyone. There is huge potential to make better use of information from people’s patient records, to understand more about disease, develop new treatments, monitor safety, and plan NHS services. Patient data should be kept safe and secure, to protect everyone’s privacy, and it’s important that there are safeguards to make sure that it is stored and used responsibly. Everyone should be able to find out about how patient data are used. #datasaveslives You can find out more about the background to this citation here: https://understandingpatientdata.org.uk/data-citation.

Disclaimers

This report presents independent research funded by the National Institute for Health and Care Research (NIHR). The views and opinions expressed by authors in this publication are those of the authors and do not necessarily reflect those of the NHS, the NIHR, the HTA programme or the Department of Health and Social Care. If there are verbatim quotations included in this publication the views and opinions expressed by the interviewees are those of the interviewees and do not necessarily reflect those of the authors, those of the NHS, the NIHR, the HTA programme or the Department of Health and Social Care.

Amendments to the study protocol

The HART study protocol was amended on six occasions. A list of the protocol sections impacted by changes made at each amendment is given below. Full details of specific changes are recorded in change logs that are held in the study archive.

Approval of amendments to the study protocol

The amendments were reviewed and approved by the sponsor, the chief investigator, the REC and the research and development offices of the NHS trusts at which recruitment of patients took place.