The HubBLe Trial: haemorrhoidal artery ligation (HAL) versus rubber band ligation (RBL) for symptomatic second- and third-degree haemorrhoids: a multicentre randomised controlled trial and health-economic evaluation

Important changes to methods after trial commencement

Recruitment commenced on 12 November 2012, and, following this, in response to early observations, a number of changes were made to the protocol (see Appendix 1) and trial methods. Between the initial Research Ethics Committee (REC) approval and study commencement, there were two substantial amendments clarifying the serious adverse event (SAE) reporting and the expected adverse events. The protocol was published in BMC Gastroenterology in October 2012. 47

In October 2012 (substantial amendment 3; protocol version 4.0), a change was made to the eligibility criteria to exclude patients with hypercoagulability disorders [in addition to those on warfarin or clopidogrel bisulfate (clopidogrel)] following a request from one of the Principal Investigators during a site set-up visit.

In January 2013 (substantial amendment 6; protocol version 5.0), the baseline data collection was changed to the day of surgery, rather than at randomisation. Sites and treatment arms differed in the time between randomisation and treatment, and this change was made to standardise the data collection as much as possible. This also standardised the time between the baseline and the 12-month follow-up, as the follow-up time point was based on the date of trial treatment.

In March 2013 (substantial amendment 7; protocol version 6.0), we removed one inclusion criterion and replaced it with three exclusion criteria for clarification. The inclusion criterion ‘Either presenting for the first time or after failure of one banding’ at sites had caused some confusion at sites and it was also decided during the review of this criterion that patients with some historical treatments for haemorrhoids could be included.

The study team, the sponsor and the Trial Steering Committee (TSC) discussed the windows in which treatment is clinically relevant, and agreed that there should be a difference between surgical and non-surgical treatment for haemorrhoids (and that dietary advice would not exclude patients). The amended criteria clarified that all previous surgery for haemorrhoids and more than one injection treatment or banding in the previous 3 years would lead to exclusion. For non-surgical treatments, > 3 years was considered a historical event and therefore not relevant to the current research. As one previous banding was allowed within the protocol, this was also the case for injection treatments, as they were both non-surgical; one previous course of injection treatment would not mean that patients were excluded.

In April 2013 (substantial amendment 8; protocol version 7.0), we added ‘patients will have at least 24 hours to decide whether to take part’ to the protocol, in order to provide clarification, as it was not explicitly stated previously. The 24-hour period was an appropriate time frame for the study population and intervention.

A pre-randomisation questionnaire was introduced in July 2013 (substantial amendment 9; protocol version 8.0). The Data Monitoring and Ethics Committee (DMEC) suggested that patient-reported outcomes can be affected by the knowledge of their allocation. As baseline data collection took place on the day of surgery (following Substantial Amendment 6), the majority of patients knew their allocation by this point; the concern was that perceived pain and quality of life (QoL) may differ between the groups due to expectation bias at this time,48 even though no procedure had yet taken place. The senior trial statistician reported that the early data did indeed support this hypothesis, in particular with higher self-reported symptoms in the HAL arm. As a result of this, the protocol was amended to incorporate a questionnaire to be completed before randomisation.

To improve the questionnaire completion rates at the 12-month follow-up, we introduced a prize draw for participants who returned the questionnaires in February 2014 (substantial amendment 7; no change to protocol). We completed three draws of £50 each.

Owing to the long waiting times for the HAL procedure, the dropout rate prior to the procedure was higher than expected. To account for this attrition, in April 2014 (substantial amendment 12; no change to protocol) the recruitment target was increased to 370 in order to achieve the sample size of 350 prior to treatment.

Participants and eligibility criteria

The trial was coordinated from the Clinical Trials Research Unit (CTRU) in the Sheffield School of Health and Related Research (ScHARR). Delegated study staff located at individual centres identified and consented potential participants.

The target population was patients referred to collaborating centres for treatment of haemorrhoids.

Potential participants fell into three groups:

1. Patients presenting to the surgical outpatient clinic (SOPC) with symptomatic haemorrhoids that did not require further tests: this group was identified by the clinical team from the general practitioner (GP) referral letter and a patient information sheet was sent to them prior to their clinic appointment. If they were willing to participate then they were consented and randomised when they attended the appointment.

2. Patients presenting to the SOPC with symptomatic haemorrhoids that required further tests to exclude other diagnoses: this group were identified by the clinician at the clinic appointment and given a patient information sheet. They underwent the necessary outpatient tests (usually endoscopy) and, if negative (i.e. the symptoms are due to haemorrhoids), they were contacted by the research nurse prior to attending their follow-up clinic appointment. They were then randomised and consented when they reattended the clinic.

3. Patients who returned to SOPC following one unsuccessful RBL: they were identified by the clinician at their first clinic appointment (when they have RBL) and given a patient information sheet. They were contacted prior to a follow-up appointment (usually 6 weeks after treatment) by a research nurse. If they remained symptomatic and were willing to participate, they were consented and randomised when they reattended.

All potential participants had a minimum of 24 hours in which to decide whether or not they wished to take part. Patients with investigations excluding pathologies other than haemorrhoids, and all of those who had undergone RBL, were contacted by the research nurse before the planned follow-up clinic to ascertain whether or not they met entry criteria and were interested in entering the trial. They would then be seen by the consultant and research nurse in clinic where recruitment and randomisation took place.

Potential participants were invited to take part in the trial if they were aged ≥ 18 years and had symptomatic second- or third-degree haemorrhoids.

Patients were excluded if they met any of the following criteria:

• have had previous surgery for haemorrhoids (at any time)

• have had more than one injection treatment for haemorrhoids in the past 3 years

• have had more than one RBL procedure in the past 3 years

• with known perianal sepsis, inflammatory bowel disease, colorectal malignancy, pre-existing sphincter injury

• with an immunodeficiency

• unable to have general or spinal anaesthetic

• currently taking warfarin, clopidogrel or have any other hypercoagulability condition

• pregnant women

• unable to give full informed consent (this may be due to mental capacity or language barriers)

• previously randomised to this trial.

Settings and locations where the data were collected

The CTRU coordinated the follow-up and data collection in collaboration with centres. Participant study data were collected and recorded on study-specific case report forms (CRFs) by centre research staff, and then entered on to a remote web-based data capture system at site. To improve the rate of completion, research nurses could complete the participant questionnaires with the participants at the visit or over the telephone, or the participants could return the questionnaires by post to the centres at all time points. Patient questionnaires, at 12 months, could also be completed online and could be returned to the CTRU by participants for data entry.

Interventions

Participants were randomised to receive either RBL or HAL. Both interventions are established and well-documented procedures, which are considered standard care by NICE. 40,49

Conventional RBL uses a device that allows a rubber band to be applied to each haemorrhoid via a proctoscope. The device used across sites was a suction device (various manufacturers), but bands can be applied using a forceps ligator. This rubber band constricts the blood supply causing it to become ischaemic before being sloughed approximately 1–2 weeks later. The resultant fibrosis reduces any element of haemorrhoidal prolapse that may have been present. This is a very commonly performed procedure in all SOPCs; figures from an audit of current practice at Sheffield Teaching Hospitals (STH) highlighted that prior to the trial over 20 such procedures were carried out every week. The procedure is a basic surgical skill with which all senior staff are familiar and competent in performing. All surgeons involved in the study performed this procedure routinely.

Haemorrhoidal artery ligation uses a proctoscope that is modified to incorporate a Doppler transducer. There are two types of equipment in common use: the HALO (haemorrhoidal artery ligation operation) device [Agency for Medical Innovations (AMI) HAL Doppler system, CJ Medical, Truro, UK] and the THD (transanal haemorrhoidal dearterialisation) device (THD Lab, Correggio, Italy). Both devices operate on the same principle, essentially enabling accurate detection of the haemorrhoidal arteries feeding the haemorrhoidal cushions. Targeted ligation of the vessels with a suture reduces haemorrhoidal engorgement. When combined with a ‘pexy’ suture, both bleeding and haemorrhoidal prolapse are addressed. All of the surgeons participating in the trial ensured that the need for a pexy suture was routinely assessed and recorded. The procedure is simple, uses existing surgical skills and has a short learning curve, with the manufacturers recommending at least five mentored cases before independently practising. All of the surgeons involved in the study had completed this training prior to the start and, in addition, had carried out more than five procedures prior to delivering the study treatment.

Trial procedures were carried out as soon as possible after randomisation. RBL is a brief procedure, which, in many cases, can be carried out at the initial clinic visit. By contrast, HAL is an invasive procedure, usually performed under general anaesthetic, and requires a theatre admission, which may be some considerable time after assessment. The HubBle trial defined ‘baseline’ as being the date of procedure, and ‘recurrence at 1 year’ as being from this date onwards.

Outcomes

Primary outcome measure

The main research question of the study was to ascertain the 1-year incidence of recurrence following HAL or RBL. The primary outcome was defined as the proportion of patients with recurrent haemorrhoids at 12 months post procedure, as derived from the patient’s self-reported assessment in combination with GP and hospital records.

The trial was a pragmatic design with a dichotomous outcome. As no validated patient-reported symptom score exists, we based our definition of recurrence on Shanmugam et al. ’s systematic review12 definition:

1. Cured or improved: Symptom free or mild residual symptoms but not requiring further treatment at the end of study period; or, 2. Unchanged or worse: No symptom improvement and requiring further intervention or suffered complication or deterioration of symptoms.

This study simplified Shanmugam’s criteria12 into the following question, asked at 12 months by a research nurse: ‘At the moment, do you feel your symptoms from your haemorrhoids are (1) cured or improved compared with before starting treatment or (2) unchanged or worse compared with before starting treatment?’

Patients were considered to have recurrent haemorrhoids when any of the following were recorded (as shown in Figure 1):

1. ‘Unchanged or worse compared with before starting treatment’ at 12 months, as reported by the patient, or

2. Any subsequent procedure (RBL, HAL, THD, haemorrhoidectomy, haemorrhoidopexy, haemorrhoidal injection or other relevant procedure) over the 12 months, or

3. Presence of any symptoms or events that strongly indicated recurrent haemorrhoids [among patients not meeting (i) or (ii), as adjudicated by two trial investigators (JT, SB) who were blinded to allocated treatment].

FIGURE 1.

Decision tree for recurrence at 1 year.

Sample size

Assuming that the proportion of patients who experience recurrence following RBL is 30% and following HAL is 15%, the sample size calculated to detect a difference in recurrence rates with an odds ratio (OR) of 2 with 80% power and 5% significance was 121 individuals per group. In order to account for any between-surgeon variation and loss to follow-up, this was increased to 175 per group.

This increase was based on the conservative assumption that there would be 14 surgeons in the trial (one per centre, although the number of sites was increased) and intraclass correlation (ICC) of 2.5% in keeping with typical ICCs observed by Ukoumunne et al. 50 However, it was considered likely that each site would have a minimum of two surgeons, in which case the power to detect this difference was 85% (90% power if there was no between-surgeon variation). Because the surgical procedure was well developed and standardised, ICC was expected to be virtually zero and the proposed sample size was hoped to have closer to 90% power.

The impact of loss to follow-up was minimal for the primary end point (haemorrhoidal recurrence at 12 months). Patients who did not complete their 12-month follow-up still had their hospital notes reviewed, and their GP was contacted to ascertain whether or not any complications or operative procedures were recorded. The only dropout expected was if the patient should die, move out of the area or have no traceable patient notes; we anticipated this would be less than the 5% that we allowed for in this patient population (a previous study of RBL that used only clinical follow-up reported a 1-year loss to follow-up of 10%51).

Explanation of any interim analyses and stopping guidelines

No formal interim analyses for efficacy were carried out. However, the safety of the treatments was assessed by an independent DMEC, the members of which convened annually to review SAEs experienced by study participants.

Randomisation and blinding

Randomisation was undertaken using the CTRU’s web-based randomisation system. After consent, participants were individually randomised, in a 1 : 1 ratio, to either HAL or RBL by a member of the research team at site. To ensure equal allocation across centres, randomisation was stratified by centre using permuted blocks of random sizes 2, 4 and 6. Allocation concealment was achieved by ensuring that the participant identifier was entered, following which the allocation was revealed; no member of the study team had access to the randomisation schedule during recruitment. The study was open label, with no blinding of participants, clinicians or research staff attempted.

Analysis populations

The primary analyses were intention-to-treat (ITT) analyses in which individuals were analysed according to the treatment to which they were randomised, regardless of whether or not they underwent their allocated surgery. Secondary analyses were undertaken on a per protocol (PP) basis, which was restricted to those individuals who complied with the protocol. The categories of non-compliance were eligibility, missed windows, consent and treatment issues, which included patients who did not receive their allocated treatment. Safety summaries were reported, based both on randomised and actual treatment where these differed.

Analysis of 1-year recurrence (primary outcome)

The primary outcome was the recurrence of haemorrhoids at 1 year, as defined above (see Primary outcome measure). Analyses were undertaken using a random intercept logistic regression model in which the covariates were treatment allocation, gender, age at surgery and history of previous intervention as fixed effects; the surgeon was included as a random effect. Further sensitivity analyses assessed whether or not other baseline characteristics [including symptom score, EQ-5D-5L and body mass index (BMI)] altered the strength of the treatment effect and/or appeared to modify the treatment effect. ‘Goodness of fit’ of the logistic regression model was assessed using the Hosmer–Lemeshow test. 52 ICC coefficients were calculated to summarise the clustering that may exist around surgeons.

Secondary outcomes

Safety outcomes

General considerations

All analyses were undertaken on an ITT basis unless otherwise stated. All confidence intervals (CIs) were two-sided 95% intervals, and all statistical hypotheses were two-sided tests. Continuous measures were summarised as mean (standard deviation, SD) or median (interquartile range, IQR) as appropriate to the distribution of the data. Categorical data were summarised as numbers and percentages.

Background

We collected data as part of the trial that allowed us to conduct a full economic evaluation. The main economic analysis focused on estimating the incremental cost per QALY of HAL compared with RBL over the 12-month follow-up period of the trial. We have also presented results in terms of the incremental cost per recurrence avoided.

Patients were asked to complete the EQ-5D-5L instrument at pre-randomisation, pre-surgery (baseline) and subsequent time points following the treatment. The UK population tariffs were then used to calculate QALYs for each patient. 54 EQ-5D has been applied in previous studies in this area55 and appears to be sensitive to changes in patient outcomes. Pain was likely to be one of the main symptoms in which we might expect the treatments to differ and this is well reflected in the EQ-5D-5L instrument.

Overview

Resource use

Unit costs

Outcomes

Recurrence is the primary outcome for this study (please refer to Primary Outcome Measure, above, for details). For the economic evaluation, the difference in the number of recurrence between the intervention group (HAL) and the control group (RBL) was used as an outcome for assessing the cost-effectiveness in terms of cost per recurrence avoided.

Analysis

The base-case analysis was based on imputed data, whereas complete-case analysis was conducted as a sensitivity analysis. The economic analysis involves CUA as a primary analysis, and a CEA was performed as a secondary analysis. Both analyses involve the estimation of differential costs and differential outcomes.

A subgroup analysis was performed for patients with new haemorrhoids and patients with recurrence following RBL before randomisation. Different sensitivity analyses were performed to address uncertainty associated with the estimates from our base-case analysis.

Long-term cost-effectiveness

A long-term extrapolation analysis was undertaken to assess the cost-effectiveness of HAL compared with RBL beyond the trial time horizon. The cost and utility data reported at 1-year follow-up within the trial were used in combination with external data to estimate longer-term costs and QALYs. A time horizon of 3 years beyond the trial was chosen for this analysis. This choice of time horizon was driven by data from external studies for which recurrence data were available over this follow-up time. We constructed a three-health states Markov model for extrapolating within-trial cost–utility analyses to long-term cost-effectiveness. To maintain consistency with the trial analyses, health states were chosen based on the primary outcome of the trial – recurrence. Heath states included were new haemorrhoids, recurrence and no recurrence. The extrapolation model structure is shown in Figure 2.

FIGURE 2.

Extrapolation model structure.

Regression analysis was used to estimate costs and utilities from the trial data using a similar SUR model described earlier, with recurrence at baseline added as a covariate. This allowed us to estimate costs and utilities by health state. We assumed that costs and outcomes follow the same pattern observed within the trial for the purpose of these extrapolation analyses. For instance, the mean costs and QALYs for patients with recurrence following RBL after 1 year and treated with further procedures were assumed to be same. Costs and health benefits (utilities) occur at different time points within the long-term time horizon of this analysis. We used the UK Treasury discount rate of 3.5% per year for discounting all future costs and QALYs as recommended by NICE. 49

The annual probability of recurrence for years 1–3 beyond the trial were estimated from two studies3,37 that have reported long-term recurrence estimates. For RBL, the annual probability of recurrence was calculated from a retrospective study3 that reported 4 years’ follow-up (n = 805). For HAL, the probability of recurrence was calculated at 5-year follow-up in a study37 that assessed the long-term success rates for treating patients with grade II and III haemorrhoids following Doppler-guided HAL (n = 100). 37 The mean total costs over four years, per patient, within each group (HAL and RBL) were calculated based on the annual probabilities of transition between health states. The mean total QALYs over the same period were similarly estimated and adjusted for recurrence. The long-term ICER was then estimated from the model and is reported below (see Long-term cost-effectiveness).

A probabilistic sensitivity analysis (PSA) was performed to address the uncertainty around parameter estimates from external sources (i.e. long-term recurrence rates). The PSA was run on 1000 Monte Carlo simulations. The distribution of costs and QALYs used in the PSA simulations were estimated from the regression analysis on the trial data.

Participants who were randomly assigned, received intended treatment and were analysed for the primary outcome

Recruitment to the trial is represented in Figure 3. A total of 372 participants were randomly assigned to receive RBL or HAL; 187 patients were allocated to receive RBL, and 185 were allocated to receive HAL. Two of these participants (both randomised to RBL) were removed from the trial completely, as they were ineligible at the time of consent; therefore, a total of 370 participants were entered into the trial. In total, 969 patients were screened for entry into the trial and the reasons for non-recruitment are shown in Table 5.

FIGURE 3.

Recruitment graph.

Patient preference was the main reason for non-consent by patients, with 128 patients opting for RBL and 70 patients opting for HAL.

Losses and exclusions after randomisation

A total of 340 participants received treatment as part of the trial and the participant flow is shown in Figure 4; reasons for withdrawal during the trial are provided in Table 6. Primary outcome data were collected for 337 participants (161 HAL and 176 RBL): 256 patient questionnaires were returned at 12 months; 236 GP forms were returned at 12 months, and 337 consultant forms were completed at 12 months. There were 183 participants where all 3 of the 12 months forms were completed/returned.

FIGURE 4.

Flow chart summarising the numbers of each type of questionnaire completed throughout the trial.

Dates defining the periods of recruitment and follow-up

The first site to open to recruitment was STH NHS Foundation Trust on 9 September 2012, and recruitment finished at all sites on 6 May 2014. Follow-up was due to end 1 year after the end of recruitment but to allow for the delay in receiving the trial treatment we extended this period and the follow-up was completed at sites on 28 August 2015.

Table 7 shows the duration between randomisation and treatment (waiting time) by treatment group for the trial. These data do not include individuals who withdrew prior to treatment; 24 participants withdrew prior to the procedure in the HAL group, compared with six (eligible) participants in the RBL group. In the RBL group, 114 of 179 (63.4%) participants received treatment on the same day as they were randomised (0 days), whereas none of the HAL group participants was treated on the same day. Although the maximum waiting time is greater for RBL, the mean and median shows that waiting times are higher for HAL.

Recurrence (primary outcome)

The overall proportion of participants with a recurrence at 12 months was 49% in the RBL arm compared with 30% in the HAL arm (absolute difference 19.6%; adjusted OR 2.23, 95% CI 1.42 to 3.51; p = 0.0005). The breakdown of recurrences, overall and by criterion, is presented in Table 9. Self-reported recurrence rates were almost identical, with 29% of respondents in both arms stating that they believed that their haemorrhoids had either improved or been cured (adjusted OR for self-reported recurrence = 1.06, 95% CI 0.60 to 1.85; p = 0.85). The increased recurrence associated with RBL is mainly attributable to the high rate of additional procedures undertaken following the initial procedure [33%, compared with 14% in the HAL group (adjusted OR for further procedure = 2.86, 95% CI 1.65 to 4.93; p < 0.001)]. A further three (2%) participants in the RBL arm were considered to have symptoms that were consistent with recurrent haemorrhoids following blind review. In two cases the participants were recorded as possibly requiring further treatment at their 6-week visit but were subsequently lost to follow-up; a third patient had been hospitalised twice for excessive bleeding, but had not undergone treatment.

Sensitivity analyses were undertaken in which alternative covariates (EQ-5D-5L at randomisation or preoperative, BMI at randomisation, and grade of haemorrhoids at randomisation) were adjusted for; doing so yielded reasonable consistency in the ORs (range 2.05–2.81), all of which remained statistically significant. The PP analysis again provided a similar OR (2.21, 95% CI 1.38 to 3.54; p = 0.001).

Of the baseline covariates assessed, none had a statistically significant association with recurrence. Recurrences were more common, however, among participants who were male, had grade III haemorrhoids, had undergone previous treatment and had higher symptom scores.

Among the 80 participants who required a further intervention, the majority of participants underwent a single procedure. In most cases this was RBL, as described in Table 9, although some variation was noted across centres as described in Table 10: as the primary interest is to document second-line treatment, the treatment groups here refer to treatment, as received, as opposed to ‘as randomised’, which was reported previously. As RBL is a brief procedure with (relatively) minimal inconvenience to the patient, it could be argued that a second-line RBL is not itself indicative of a recurrence because the initial haemorrhoids remain incompletely treated. Consequently, an additional (and unplanned) analysis investigated the extent to which recurrence differed if follow-up RBL were not considered a recurrence. In total, 45 participants (31 in the RBL arm, 11 in HAL) underwent RBL as follow-up procedure. Of the 31 patients who underwent repeat RBL, six considered their haemorrhoids to be unchanged or worse at 1 year; three underwent further procedures; and one was considered a recurrence based on blind review. In the HAL arm, 8 of the 11 participants who were undergoing subsequent RBL considered their haemorrhoids to be unchanged or worse at 1 year and two participants underwent further procedures. Thus, if subsequent RBL were not considered a recurrence, the number with recurrent haemorrhoids is 66 (37.5%) in the RBL arm and 44 (27.3%) in the HAL arm (adjusted OR 1.53, 95% CI 0.96 to 2.44; p = 0.071). If a single HAL is compared with multiple RBLs the number with recurrent haemorrhoids is 66 (37.5%) in the RBL arm and 48 (30%) in the HAL arm (adjusted OR 1.35, 95% CI 0.85–2.15; p = 0.20).

Persistent significant symptoms at 6 weeks

At 6 weeks, data were available for 150 participants in the RBL arm and 143 in the HAL arm: 43 (24%) participants in the RBL arm reported their haemorrhoids as being unchanged or worse compared with 12 (7%) in the HAL arm; additionally, one participant in each arm had subsequently undergone RBL, thus resulting in the overall number of patients with persistent symptoms being 44 (29%) compared with 13 (9%) (adjusted OR 4.35, 95% CI 2.19 to 8.65; p < 0.001).

Haemorrhoid symptom severity score

The haemorrhoid symptom severity scores are summarised in Table 11 and displayed graphically in Figure 5: higher scores indicate more severe symptoms. At 6 weeks, haemorrhoids symptom severity scores were higher in the RBL group, indicating that short-term symptoms were less pronounced following HAL. The mean (SD) scores were 4.0 (3.5) in the RBL group and 3.0 (3.1) in the HAL group, with an adjusted mean difference of 1.0 (95% CI 0.3 to 1.8; p = 0.010). No difference was apparent at 1 year, with the mean (SD) being 3.6 (3.2) for RBL and 3.6 (3.3) for HAL (adjusted difference = 0.0, 95% CI –0.8 to 0.8; p = 0.98).

FIGURE 5.

Haemorrhoid symptom severity scores. Note that the area of data points is proportional to the number of participants with that value. Vertical lines represent median and IQR.

A further (post hoc) analysis looked at the proportion of participants whose symptom score was either ‘0’ or ‘1’, as this corresponds to the definition used by Nyström et al. 42 These numbers are shown in Table 12. The proportions are consistent with the previous analysis, with a greater proportion of participants in the HAL arm reporting either a score of 0 or 1 at 6 weeks.

European Quality of Life-5 Dimensions (5-level version)

The EQ-5D-5L health tariff is summarised in Table 13 and Figure 6: higher figures indicate a better health state. Prior to procedure, the mean health utility was around 0.9 in both groups but declined at days 1 and 7 in the HAL group. For RBL the mean (SD) at 1 day was 0.84 (0.19) and at 7 days 0.92 (0.15); in other words, health state was reduced for the first day but had reverted back at 1 week. By contrast, the mean health state for HAL had not returned to baseline values by 7 days, with the mean (SD) being 0.76 (0.22) and 0.83 (0.18) at 1 and 7 days, respectively. The adjusted mean differences were 0.08 (95% CI 0.04 to 0.13; p < 0.001) at 1 day and 0.08 (95% CI 0.05 to 0.12; p < 0.001) at 7 days. The two arms were nearly similar with no statistical differences (and above baseline values) at all time points from day 21 onwards.

The EQ-5D-5L inventory was also used to derive the health state used in QALYs that are reported within the health-economic analysis in the following section.

FIGURE 6.

The EQ-5D-5L. Note that the area of data points is proportional to the number of participants with that value. Vertical lines represent median and IQR.

Vaizey faecal incontinence score

Vaizey faecal incontinence scores are provided in Table 14 and Figure 7; higher scores indicate more severe incontinence. No between–group differences were noted in the Vaizey faecal incontinence scores. An improvement of around 1 unit was noted in both arms at 6 weeks, with a difference between arms of –0.1 (95% CI –1.3 to 1.0; p = 0.86). The improvement was maintained at 1 year, with a difference of –0.5 (95% CI –1.8 to 0.7; p = 0.38). A summary of these findings is presented in Table 14 and Figure 7.

FIGURE 7.

Vaizey faecal incontinence score. Note that the area of data points is proportional to the number of participants with that value. Vertical lines represent median and IQR.

Pain

Haemorrhoidal pain was assessed by asking the patient to rate his/her pain related to haemorrhoids as of today and over the last week. Pain today was asked at baseline and again at days 1, 7 and 21, and, finally, at 6 weeks. Pain over the last week was asked at days 7 and 21, and, finally, at 6 weeks. Both of these are summarised in Tables 15 and 16 and Figures 8 and 9; a score of ‘0’ = ’no pain’, whereas a score of ‘10’ = ’worst imaginable pain’. The change in pain as recorded by VAS is shown in Table 17. Pain was increased in the HAL group at one and seven days after the procedure, but the groups were similar at day 21 and at 6 weeks.

FIGURE 8.

Visual analogue scale pain scores: pain today.

FIGURE 9.

Visual analogue scale: pain over the last 7 days. Note that the area of data points is proportional to the number of participants with that value. Vertical lines represent median and IQR.

Clinical appearance of haemorrhoids at proctoscopy

Proctoscopy was conducted at the 6-week visit only if the patient reported continuing symptoms; proctoscopy was conducted more often in the RBL group. The data in Table 18 show that, in the majority of patients in whom proctoscopy was conducted, the grade had reduced at 6 weeks, with a one-grade increase observed in only four patients in the RBL group and none in the HAL group.

Pre-randomisation questionnaire

As previously mentioned, the baseline was originally measured prior to procedure as opposed to at randomisation. Following a recommendation from the DMEC, the protocol was subsequently amended to allow collection at both time points. The rationale for so doing was that self-reported severity may differ once the allocated treatment is known, thereby introducing an expectation bias. Specifically, participants randomised to the more intensive HAL surgery may rationalise their planned treatment by reporting higher levels of pain and discomfort than those randomised to receive RBL. Furthermore, it is possible that symptoms may change over time, which would also cause imbalance in pre-treatment values, as waiting time for HAL was longer than for RBL.

The agreement between pre-randomisation and pre-treatment values were therefore investigated among patients for whom data were available at both time points. The following are presented for each self-reported measure:

• The Bland–Altman plot for agreement within each treatment arm, together with the overall mean difference and 95% limits of agreement (also known as a 95% reference interval). The plot is the difference between pre-treatment and pre-randomisation (i.e. the change between the two) plotted against the average of the two. The 95% limits of agreement (also known as reference limits) for the difference were derived based on a normal distribution and quantify how closely the pre-randomisation and pre-treatment value agree. The average is plotted on the horizontal axis, which allows assessment of whether or not (dis)agreement varies according to severity.

• The difference between pre-treatment and pre-randomisation values against time elapsed between the two. The reference limits were the same as those produced for the Bland–Altman plot. Plotting the difference against elapsed time between the two measures allows an assessment of whether or not (dis)agreement increases with the time elapsed between the two measurements.

The agreement is displayed visually for each method in Figures 10–13. The difference and 95% limits of agreement are also summarised in Table 20, along with the difference between the two means and the ratio of their variances.

Overall, the extent of agreement was generally similar regardless of the severity. For the VAS pain score, there was some indication that the pre-randomisation and pre-procedure time points were in closer agreement in the HAL arm than the RBL arm, as demonstrated by the ratio of the two variances (2.63; F-test p < 0.001). The same was also true for incontinence, as estimated by the Vaizey tool, where the difference between randomisation and procedure again tended to be greater in the HAL arm than the RBL arm (ratio of variances 3.14; p < 0.001). On the other hand, the two values differed less in the HAL arm for the EQ-5D ‘thermometer’ health state (ratio of variances 0.48; p = 0.004), although this may reflect a handful of participants with outlying values. There was no indication of any systematic change (i.e. no consistent deterioration or spontaneous resolution between randomisation and procedure) on any measure.

FIGURE 10.

Agreement of haemorrhoid symptom severity scores at randomisation and pre-procedure. Bland–Altman graphs are given for (a) HAL and (b) RBL. Difference between the measures in relation to elapsed time between the two is given for (c) HAL and (d) RBL.

FIGURE 11.

Agreement of EQ-5D-5L scores at randomisation and pre-procedure. Bland–Altman graphs are given for (a) HAL and (b) RBL. Difference between the measures in relation to elapsed time between the two is given for (c) HAL and (d) RBL.

FIGURE 12.

Agreement of Vaizey scores at randomisation and pre-procedure. Bland–Altman graphs are given for (a) HAL and (b) RBL. Difference between the measures in relation to elapsed time between the two is given for (c) HAL and (d) RBL.

FIGURE 13.

Agreement of VAS pain scores at randomisation and pre-procedure. Bland–Altman graphs are given for (a) HAL and (b) RBL. Difference between the measures in relation to elapsed time between the two is given for (c) HAL and (d) RBL.

Haemorrhoidal artery ligation device

Among participants undergoing HAL, the preferred surgical device appeared to be determined by site rather than participant characteristics. Five sites (n = 73) used the AMI HALO device on all participants, with the remaining 12 sites using the THD device in 87 cases and the HALO device in two. There was no significant difference in outcomes between the types of device, although fewer recurrences were seen in those in the THD arm (26% vs. 35%; adjusted OR 0.65, 95% CI 0.33 to 1.29; p = 0.22). Symptoms and pain were also similar for the two procedures, data are shown in Table 21. The Vaizey faecal incontinence score was higher for THD than HALO at 6 weeks (mean scores 5.2 vs. 3.2) and 1 year (means 5.6 vs. 3.5), but also prior to procedure (means scores 6.3 vs. 4.8). As a consequence, the magnitude of the difference at 6 weeks (mean difference = 2.0, 95% CI 0.5 to 3.6; p = 0.01) was greater than the change from baseline (mean difference incorporating an adjustment for baseline = 1.3, 95% CI –0.3 to 2.8; p = 0.11).

Primary outcome: recurrence

Haemorrhoidal disease is a benign condition and, as such, treatment is primarily aimed at addressing the patient’s symptoms. In the absence of a validated symptom scoring system, we felt that the most important determinant of treatment success was patient-reported outcome. This was supported by a robust patient and public involvement (PPI) exercise carried out during the design of the trial. Clinical experience tells us that the physical appearance of the anal cushions post-haemorrhoid treatment correlates poorly with patients’ symptoms, meaning that anorectal visualisation is not a reliable surrogate of success. We therefore used a simple, dichotomised definition of recurrence as described by Shanmugam et al. 6,51 and measured this at 12 months post intervention. Using this patient-reported outcome alone, there was no difference in recurrence between the groups at 12 months. However, a number of patients (particularly those who underwent RBL) had undergone a further intervention for haemorrhoid symptoms within the 12 months, indicating that any symptomatic relief had been achieved only by further corrective procedures. Classing these patients as recurrence, the rate for HAL was significantly lower than for RBL (30% vs. 49%; p = 0.001) at 12 months.

The overall level of recurrence following HAL was higher than the pooled rates reported in two systematic reviews, which were 11%19 and 17.5%. 5 However, it should be noted that the recurrence rate among the constituent studies varied considerably and ranged up to 60%. 64 Similarly, reported recurrence rates for RBL also vary widely in the literature, ranging from 11% to > 50%. 6–11 This range can be attributed to many factors, but the most pertinent is probably the difficulty in defining exactly what constitutes recurrence. Some studies rely on clinical examination and patient-reported symptoms,22,25,27,51,64 whereas others rely solely on symptomatology. 24,32 When considering symptoms only, some investigators consider a patient to be recurrence free when they have no perianal symptoms at all, whereas others consider this simply an improvement. A classic example of variation dependent on definition is provided by one study64 reporting a very high recurrence rate of 60% at 30 months after a HAL procedure. However, the investigators report that, at the same time point, 86% of patients described a significant improvement in symptoms.

We adopted the more pragmatic approach6 described above in which the patient’s perception of symptomatic improvement was combined with the requirement that no further procedures were required in order to achieve a non-recurrence at 12 months. There were some practical hurdles to overcome when obtaining these data: a patient may have indicated that he/she was cured but could have forgotten about a consult or further treatment in the 12-month period since initial surgery, or a patient may have indicated recurrence when the ‘recurrence’ related to a residual skin tag65 or an alternative diagnosis clearly documented by GP or consultant records. To address these issues, we used a combination of patient-reported assessment with GP and hospital consultant perception. A decision tree was developed and followed independently by two assessors who were blinded to the intervention, with consensus reached on any areas of disagreement. This ensured fidelity and we felt this resulted in the most accurate assessment of true recurrence that was possible.

Other factors that influence the recurrence rate include the duration of follow-up and the pre-treatment haemorrhoidal grade. We chose to assess recurrence at 12 months because published data suggest that most HAL and RBL recurrences occur within 1 year. 26 Several studies have shown that using HAL for higher-grade haemorrhoids or significant prolapse adversely affects the recurrence rate;19,37,41,66,67 they also conclude that HAL is perhaps most effective for grade II and early grade III disease. We therefore limited our trial to this most clinically relevant patient population. As RBL is also most effective for this stage of disease, it was chosen as our comparator (see below).

Linked with recurrence is the need for further treatment. This was required in 31% of the RBL group and 15% of the HAL group. Eighteen per cent of the RBL group required a required further banding, the vast majority needing only one further procedure. It is common practice to repeat banding as part of a continuum of treatment in some patients with persistent symptoms, and it could be considered that these patients were actually treatment successes if by the end of the year they were cured or improved. Given that RBL is a quick, minimally invasive office procedure in comparison with HAL, it seems clinically relevant and reflective of actual practice to consider it as a ‘course’ of treatment as opposed to a single intervention. Including these patients as treatment success resulted in a reduction of the recurrence rate for the RBL arm from 49% to 37.5%, which was not statistically different from the HAL arm.

Secondary outcomes

Generalisability

Among the strengths of this study was the pragmatic, multicentre design using a mix of NHS district general hospitals and teaching hospitals across the UK, which ensures that the results are generalisable to all patients seeking treatment for grade II/III haemorrhoids. The number of subjects recruited was such that there can be considerable confidence in the conclusions drawn from the data.

Definition of recurrence

The difficulties with definition of recurrence have already been highlighted earlier and clearly alluded to in the literature. 90 We feel that the significant lengths to which we have gone to carefully define recurrence should be regarded as a strength. Our data provide a clinically meaningful result of the true incidence of recurrence that is primarily patient reported, but has taken into account clinician-derived data of further consultations and procedures. Clinician reporting alone is likely to be misleading, but at the same time must be taken into account when it comes to differentiating the symptoms related to haemorrhoids and other conditions that may co-exist (skin tags, fissures, etc.) that may lead to a false diagnosis of recurrence.

Changes from baseline data

The original design of our trial incorporated a baseline assessment of symptoms at the date of randomisation. It became clear soon after the start of the trial that there was a discrepancy between the two arms of the trial, in that many patients randomised to the RBL group had their procedure almost immediately after randomisation whereas the HAL group often had to wait a significant period on a waiting list before surgery could be carried out. It is possible that symptomatology would have changed during that period. Certainly, deterioration could have occurred, but also improvement. Recognition of the disease issue and the impending thought of surgery may have resulted in an improved lifestyle, which, in turn, may have improved symptoms (see above). The issue of surgical waiting time delay has been noted before in surgical trials,91 with a suggestion that the delay does not definitely alter data. To ensure that baseline data were collected consistently at a time close to intervention, we recollected the baseline data in those patients who were randomised to the HAL arm. Interestingly, although individual scores changed between randomisation and the actual procedure, the population average remained similar, suggesting that either could have been used as a plausible baseline value.

Justification of eligibility criteria

There are numerous therapies available for treatment of haemorrhoids, ranging from simple conservative therapies through to outpatient treatment and various surgical options. A consensus is emerging that a tailored approach is required for treatment of haemorrhoids based on grade. 92–94 Early-grade haemorrhoids can be effectively treated conservatively with diet,95,96 alteration in lifestyle94 and, perhaps, medical therapy,38,97,98 whereas at the other end of the spectrum, grade IV circumferentially prolapsed piles may respond only to haemorrhoidectomy or SH. The eTHoS trial99 hopes to provide data concerning the optimal procedure for the more advanced haemorrhoids.

Our main aim was to assess the effectiveness of HAL in the most clinically relevant scenario and compare with the most appropriate control. HAL appears to be most effective for treatment of grade II and early grade III piles92 for which non-surgical treatment is also often utilised. These non-surgical methods include RBL, injection sclerotherapy, cryotherapy, infrared coagulation, laser therapy and diathermy coagulation. Of all of these non-surgical procedures, RBL seems to be the best in terms of compliance, long-term efficacy and side effects. 100 Therefore, participants with grade II or early grade III haemorrhoids were selected as investigation group and RBL was selected as our comparator.

The consensus concerning a tailored approach regarding the treatment of haemorrhoids is difficult to confirm from the evidence available, as most published meta-analyses involve standard pairwise trials comparing only two surgical treatments directly rather than all available treatments at once. A novel method of network analysis published by Simillis et al. 101 allows simultaneous comparison of all surgical treatments. It is unfortunate that this meta-analysis has focused on treatment of only grade III and IV haemorrhoids. The addition of data from the HubBLe trial to the existing data may allow a similar network meta-analysis for treatment of earlier-grade haemorrhoids.

Haemorrhoidal grading systems

We used the Goligher grading system for haemorrhoids,102 as it is the most widely used grading system. This system has been criticised by many, particularly with regard to the definition of grade IV haemorrhoids. Goligher’s original description of grade IV haemorrhoids102 could include either skin tags or an external haemorrhoid component alone or in addition to a mucosal prolapse that needs manual positioning. There is less ambiguity with regard to grade II haemorrhoids, and grade III haemorrhoids are patient reported (i.e. the need to manually reduce). However, we have to accept that the definition of early grade III haemorrhoids has an element of ambiguity. From a pragmatic point of view, the eligibility of a patient for the trial was decided by the surgeon, who had to decide if the grade III piles were potentially treatable by the interventions prescribed.

Learning curve

The HAL procedure appears simple and easy to learn, and most investigators have assumed this to be the case. 22,103,104 However, there are no good data on the learning curve and the steps that are required to achieve competence. The existence of a learning curve has been alluded to. Pucher et al. 19 pointed out that many case series include the particular centre’s initial experience and describe a poorer outcome in the early patients. 66,105 Our trial required that surgeons had to have carried out at least five mentored cases and an additional five procedures prior to delivering the study intervention. This was based on manufacturers’ recommendations, with no data to indicate that this was adequate to have reached the top of the learning curve for competence. HAL was performed by a clinician who was accredited for independent practice or supervised by a consultant in 94% of the cases in the trial, and all procedures were completed by clinicians who met the above criteria.

If we were to be more robust in our requirement for a satisfactory level of competence we would have had to assess reproducibility. Possible assessments could include placing the stitch adequately to achieve absence of the Doppler signal. However, Denoya et al. 25 describe the Doppler signal not changing in almost half of the vessels ligated in their study. Reproducibility could relate to detection of all branches of the superior haemorrhoidal artery. Ratto et al. 106 and Denoya et al. 25 identified at least six arteries in all patients. Six ligations were certainly not carried out in all of our participants. An interesting analysis would be to see if the number of ligations correlates with the incidence of recurrence.

To add to the difficulties in assessment of reproducibility that is dependent on efficacy of detection and effective ligation of all branches of the superior haemorrhoidal artery, recent scepticism has focused on the validity of de-arterialisation. Autopsy data indicate that additional branches of the superior rectal artery are not detected even with a Doppler probe. 107 Two RCTs29,32 have randomised patients to HAL with or without Doppler probe guidance; both suggested similar results at 1-year follow-up.

Variation in intervention

Symptom severity score

At the time of design of this trial, there were no validated scoring systems for haemorrhoids. The nearest we could come to such a system was that designed by Nyström et al. ,42 who altered a validated scoring system used for assessing colopouch function after rectal resection. The score has allowed a numerical assessment of the effect of both interventions, and results have complimented the dichotomous primary outcome. However, the score is not validated. For trials such as this, for which there is a poly-symptomatic condition and treatment effect must be guided by all of the patient’s symptoms and their respective influence on QoL, there is a need for a validated score. Subsequent to our study starting, such a score, known as the Sodergren score, has been developed and validated using a small population of patients. 113 Further assessment of validity is required and data from our trial may allow further development in this area.

Length of follow-up

We chose ‘1 year’ as the time for our primary outcome, as evidence suggests that most recurrences occur within the first year. However, there is definite evidence of continued recurrences with time for both procedures3,27 and from our own data, with the striking deterioration seen at 1 year assessment compared with 6 weeks. For RBL, a further 20% may recur at 3 years, even if repeat RBL is carried out. The numbers may be similar for HAL. 37 The need for a longer-term follow-up for this subgroup is essential for further information regarding effectiveness and health economics. Modelling to assess ongoing costs and to see if HAL becomes cost-effective with time is planned but requires at least a further year of data to give more reliable predicted outcomes.

One other aspect related to follow-up is the issue of repeat banding. Our original protocol did not incorporate the analysis of HAL compared with repeat banding. The timing of this repeat banding was therefore not recorded. It is likely that most repeat procedures were carried out at, or soon after, the 6-week follow-up. However, a proportion of these were carried out later. For these patients the subsequent review at 1 year from the index procedure may not be long enough from the last banding episode to allow for confidence in the true level of recurrence.

Rubber band ligation as a course of therapy

In retrospect, an alternative design could have been considered in which HAL is compared with a ‘course’ of RBL. Such a course is difficult to define. Is it simply two episodes of RBL? Is it RBL, repeated only if there has been some response? Is it RBL ad infinitum? Is it RBL until the patient or clinician calls a halt and alternative procedures are requested? Design of such a trial is simple if the course is defined as two episodes of RBL, perhaps 6 weeks apart, with follow-up for 1 year after the last RBL. Designing a trial to evaluate alternative treatment strategies is difficult. The current study has allowed the post hoc analysis of repeat RBL to be assessed while maintaining a feasible and pragmatic design.

Trial Management Group

Mr Steven Brown (Sheffield Teaching Hospitals NHS Foundation Trust); Mr Jim Tiernan (Leeds Teaching Hospitals NHS Trust); Ms Katie Biggs (Sheffield CTRU, University of Sheffield); Dr Daniel Hind (Sheffield CTRU, University of Sheffield); Mr Mike Bradburn (Sheffield CTRU, University of Sheffield); and Mr Neil Shephard (Sheffield CTRU, University of Sheffield). Representatives from each participating NHS Trust (either the Principal Investigator or Research Nurse) were invited.

Trial Steering Committee

Professor Jonathan Lund (Derby Hospitals NHS Foundation Trust); Mr Patrick Dobbs (Sheffield Teaching Hospitals NHS Foundation Trust); Mr Jeff Garner (The Rotherham NHS Foundation Trust); and Mr Jeff Williams (Patient Representative).

Data Monitoring and Ethics Committee

Dr Jonathan Cook (University of Oxford); Mr Dermot Burke (Leeds Teaching Hospitals NHS Trust); and Mr Baljit Singh (University Hospitals Leicester).

Co-applicants

Mr Jim Tiernan (Leeds Teaching Hospitals NHS Trust); Dr Daniel Hind (Sheffield CTRU, University of Sheffield); Professor Angus Watson (Raigmore Hospital, Inverness); Professor Allan Wailoo (ScHARR, University of Sheffield); Mr Mike Bradburn (Sheffield CTRU, University of Sheffield); and Mr Neil Shephard (Sheffield CTRU, University of Sheffield).

Local principal investigators

Mr Steven Brown (Sheffield Teaching Hospitals NHS Foundation Trust); Mr Simon Radley (University Hospitals Birmingham NHS Foundation Trust); Mr Oliver Jones (Oxford University Hospitals NHS Trust); Mr Paul Skaife (Aintree University Hospital NHS Foundation Trust); Mr Anil Agarwal (North Tees and Hartlepool NHS Foundation Trust); Mr Pasquale Giordano (Barts Health NHS Trust); Mr Marc Lamah (Brighton and Sussex University Hospitals NHS Trust); Mr Mark Cartmell (Northern Devon Healthcare NHS Trust); Mr Justin Davies (Cambridge University Hospitals NHS Foundation Trust); Mr Omar Faiz (North West London Hospitals NHS Trust); Ms Karen Nugent (University Hospital Southampton NHS Foundation Trust); Professor Angus Watson (NHS Highland); Mr Andrew Clarke (Poole Hospital NHS Foundation Trust); Professor Angus MacDonald (NHS Lanarkshire); Mr Philip Conaghan (Royal Berkshire NHS Foundation Trust); Mr Paul Ziprin (Imperial College Healthcare NHS Trust); Mr Rohit Makhija (Peterborough and Stamford Hospitals NHS Foundation Trust); and Mr Jonathan Trickett (Ashford and St Peter’s Hospitals NHS Foundation Trust).