Laparoscopic supracervical hysterectomy compared with second-generation endometrial ablation for heavy menstrual bleeding: the HEALTH RCT

Article history

The research reported in this issue of the journal was funded by the HTA programme as project number 12/35/23. The contractual start date was in January 2014. The draft report began editorial review in October 2018 and was accepted for publication in February 2019. 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.

Declared competing interests of authors

T Justin Clark reports grants and personal fees from Hologic Inc. (Santa Clara, CA, USA), outside the submitted work, and membership of the Health Technology Assessment (HTA) Prioritisation Committee. John Norrie declares grants from the University of Aberdeen and the University of Edinburgh during the conduct of the study, and membership of the following National Institute for Health Research (NIHR) boards: HTA Commissioning Board (2010–16); NIHR HTA and Efficacy and Mechanism Evaluation Editorial Board (2014–19); HTA Commissioning Sub-board (Expression of Interest) (2016–present); HTA Funding Boards Policy Group (2016–present); HTA General Board (2016–present); HTA Post-board Funding Teleconference (2016–present); the Pre-exposure Prophylaxis Impact Review Panel (2018); and the NIHR Clinical Trials Unit Standing Advisory Committee (2018–present). Siladitya Bhattacharya is the Editor-in-Chief of HROpen and an Editor for Cochrane Gynaecology and Fertility.

Copyright statement

© Queen’s Printer and Controller of HMSO 2019. This work was produced by Cooper et al. under the terms of a commissioning contract issued by the Secretary of State for Health and Social Care. This issue may be freely reproduced for the purposes of private research and study and extracts (or indeed, the full report) may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NIHR Journals Library, National Institute for Health Research, Evaluation, Trials and Studies Coordinating Centre, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

2019 Queen’s Printer and Controller of HMSO

Background

Heavy menstrual bleeding (HMB) is a common problem, affecting approximately 1.5 million women in England and Wales. It accounts for one-fifth of all gynaecology outpatient referrals and has a major impact on women’s physical, emotional, social and material quality of life (QoL). 2 The condition is initially treated in primary care by either oral medication or insertion of the levonorgestrel-releasing intrauterine system (Mirena^®; Bayer AG, Whippany, NJ, USA). If medical therapy fails or is deemed unsuitable, surgical treatment can be offered, either EA, which destroys the lining of the cavity of the uterus (endometrium), or hysterectomy (removal of the uterus). Neither medical treatment nor EA can guarantee complete resolution of symptoms and up to 59% of women on oral drugs3 and 13.5% of those using the levonorgestrel-releasing intrauterine system (Mirena^®)4 ultimately undergo surgical treatment within 2 years. Over one-third (38%) of women referred to an English NHS hospital between April 2009 and March 2012 with HMB underwent a surgical procedure, with about three-quarters of these being EAs. 2 Of those treated with EA, 19% have required a subsequent hysterectomy for relief of their symptoms. 5

Scale of the problem in the UK and use of NHS resources

Hospital Episode Statistics data indicate that a total of 136,921 hysterectomies and 128,434 EAs were performed in England and Wales for HMB between April 1997 and December 2009. 6 Current types of EA that are recommended by the National Institute for Health and Care Excellence (NICE) are second-generation (non-hysteroscopic) procedures, including bipolar radiofrequency ablation (Novasure^®; Hologic Inc., Marlborough, MA, USA) and thermal balloon EA. 7

Evidence explaining why this research is needed

The NICE guideline on HMB recommends both EA as well as hysterectomy as options for women with HMB resistant to medical treatment,7 but a significant minority of women treated with EA are likely to need repeat EA or hysterectomy. A recent individual patient data meta-analysis8 of results from randomised trials has shown that, despite the the greater invasiveness of conventional hysterectomy (removal of the uterus and the cervix) and associated longer hospital stay and prolonged recovery, fewer women are dissatisfied with it than with EA. Additionally, a health economic model based on these data showed that hysterectomy is more cost-effective. 9 The accompanying HTA monograph10 showed that a quarter of all women who undergo EA will require subsequent gynaecological surgery, with just under one-fifth requiring hysterectomy. These findings, which are consistent with those of a relevant Cochrane review,11 suggest that the optimal surgical treatment for HMB that is unresponsive to medical treatment may well be hysterectomy, but its effectiveness needs to be balanced against its invasive nature and increased short- and long-term morbidity. 5

Removal of the cervix as part of a conventional total hysterectomy can be technically challenging and result in injury to surrounding blood vessels, ureters and the bladder. In contrast, LASH is the use of keyhole surgery to remove the body or major part of the uterus, which is the part responsible for menstrual bleeding, but conserves the cervix. This approach reduces operating time as well as surgical morbidity, while conserving the uterosacral ligament support to the cervix and upper vagina. Routine cervical screening is required for women undergoing LASH. The procedure is quick, minimally invasive, relatively easy to learn and associated with a low risk of complications, short hospital stay (< 24 hours) and rapid recovery. 12,13

Before this technique is incorporated into routine clinical practice, it is important that it is subjected to robust evaluation. The authors of two small RCTs comparing LASH with a first-generation EA (endometrial resection13 or second-generation EA) thermal balloon12 suggest that LASH could lead to a better QoL, but emphasised the need for larger evaluative studies to confirm this.

The last decade has seen widespread use of laparoscopic techniques in gynaecology due to increased familiarity with the procedures, more sophisticated instruments, better training and greater laparoscopic surgical skill. As a result of this, LASH could be delivered by most general gynaecologists, with minimal morbidity to women who are currently being treated with EA. Advances in perioperative care mean that, unlike conventional hysterectomy, women treated by this procedure may not need to stay in hospital any longer than those receiving EA.

HEALTH is a multicentre RCT comparing the clinical effectiveness and cost-effectiveness of LASH with second-generation EA (the current first-line surgical treatment for HMB) in women with HMB seeking surgery. The trial is relevant and timely, as rigorous evaluation of this new surgical option will provide much needed high-quality evidence to underpin any decision to use it in routine NHS practice.

Description of the surgical procedures

Questions addressed by HEALTH

The aim of this study is to compare the clinical effectiveness and cost-effectiveness of LASH and second-generation EA in women with HMB.

The primary objective is to compare (1) patient-reported satisfaction, measured on a six-point Likert scale (from ‘totally satisfied’ to ‘totally dissatisfied’) and (2) condition-specific QoL, measured using the Menorrhagia Multi-Attribute Quality-of-Life Scale (MMAS), at 15 months post randomisation. The corresponding economic objective is to estimate the incremental cost per quality-adjusted life-year (QALY) gained for LASH compared with EA at 15 months post randomisation.

The hypothesis being tested is that LASH is superior to second-generation EA for the treatment of HMB in terms of patient satisfaction, QoL and costs.

Study design

HEALTH was a parallel-group, multicentre RCT designed to compare the clinical effectiveness and cost-effectiveness of LASH with second-generation EA in women with HMB. Further details of the study design have been described previously1 and are represented in Figure 1. All trial case report forms (CRFs) and participant-completed questionnaires are available at URL: www.journalslibrary.nihr.ac.uk/programmes/hta/123523/#/ (accessed 22 May 2019).

FIGURE 1.

HEALTH flow diagram. EQ-5D-3L, EuroQol-5 Dimensions, three-level version; NRS, Numerical Rating Scale; SF-12, Short Form questionnaire-12 items. Reproduced from Cooper et al. 1 This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

Reproduced from Cooper et al. 1 This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

Study population

Women aged < 50 years with HMB who were eligible for EA and willing to be randomised between LASH and EA.

Women were excluded from trial entry if any of the following criteria were met: they had plans to conceive; endometrial atypia; abnormal cytology; uterine cavity size > 11 cm; any fibroids > 3 cm; contraindications to laparoscopic surgery; previous EA; and inability to give informed consent or complete trial paperwork.

Recruitment

Health technologies being compared

Women were randomised to one of two surgical treatments for HMB:

1. LASH [removal of the uterine corpus (body) by means of keyhole surgery]

2. second-generation EA [destroying the endometrium (lining of the womb) by means of a silicone balloon containing hot fluid or radiofrequency energy delivered through an intrauterine mesh electrode].

Treatment allocation

Eligible and consenting women were randomised to one of the two treatment arms in a 1 : 1 allocation ratio, using the randomisation application at the trial office at the Centre for Healthcare and Randomised Trials (CHaRT). The randomisation application was available as both an interactive voice-response telephone system and as an internet-based application, and used a minimisation algorithm based on centre and age group (< 40 vs. ≥ 40 years). 21

Blinding

Baseline data were reported by women before randomisation using self-completed questionnaires. Surgeons and participants could not be blinded to the allocated procedure.

Delivery of the intervention

Following randomisation, participants were placed on the waiting list for the appropriate treatment. As per Scottish and UK government guidelines, it was anticipated that treatment would occur within 12–18 weeks of randomisation. 22–24 Surgeons used their standard practice so that the technique they normally used was not modified for the purposes of the trial. All other aspects of care were left to the discretion of the responsible surgeon.

Data collection during follow-up

Participant-reported outcomes were assessed by self-completed questionnaires at baseline (before surgery), 6 weeks and 6 months after surgery, and 15 months following randomisation (Table 1). A self-completed 14-day diary was also collected. Up to two reminders were sent to participants by post, e-mail, telephone or text message, taking into account any preferences they had for mode of communication.

Intraoperative and postoperative data were collected by the local research team at the time of the randomised procedure. A short CRF was also completed for any related hospital readmissions during the follow-up period.

Study outcome measures

The co-primary clinical outcome measures were:

• patient satisfaction, measured on a six-point scale (from ‘totally satisfied’ to ‘totally dissatisfied’), at 15 months after randomisation

• MMAS score, a condition-specific QoL outcome,25 ranging from 0 (worst possible health state) to 100 (best possible health state), based on six items, measured at 15 months after randomisation.

The primary study objective was to compare the LASH and EA groups with respect to the two co-primary outcomes. These outcomes were addressed in a hierarchy. First, the patient satisfaction outcome was considered and, if this showed a statistically significant difference at a p-value of < 0.05, then the MMAS score outcome was also considered. By specifying this hierarchy, it was not necessary to apply any adjustment for multiple statistical testing, as the overall false-positive rate is controlled at 5%.

The 15-month follow-up post-randomisation time point was chosen to accommodate the 12- to 18-week waiting time for treatment. The intention was that the primary outcome questionnaire (triggered at 15 months post randomisation) would be completed at approximately 12 months post surgery, in order to facilitate comparison with the most similar RCTs in the literature.

The primary economic outcome was the incremental cost (to the health service) per QALY gained (LASH vs. EA).

Secondary outcome measures

Other outcome measures included pain score at days 1–14 and at 6 weeks post surgery; acceptability of treatment at 6 weeks post surgery; satisfaction with treatment at 6 months post surgery; MMAS score at 6 months post surgery; menstrual outcomes at 6 months post surgery and at 15 months post randomisation; generic health-related quality of life (HRQoL) [Short Form questionnaire-12 items (SF-12) and EuroQol-5 Dimensions, three-level version (EQ-5D-3L), scores at 6 weeks and 6 months post surgery and at 15 months post randomisation]; perioperative complications, including recovery details and need for additional gynaecological surgery; cost; and cost-effectiveness.

Pathology results for all endometrial biopsies and uterine specimens were also checked.

Statistical analyses were used to compare secondary outcomes by randomised groups. These analyses, however, were regarded as hypothesis-generating and no adjustment was made for multiple statistical testing.

Safety reporting

Adverse events (AEs) were either notified to the study office by the local research team or reported by the women in their follow-up questionnaires. If an AE was suspected, it was verified by the local research team, if possible. Unrelated AEs were not recorded.

In HEALTH, ‘relatedness’ was defined as an event that occurred as a result of a procedure required by the protocol, whether or not it was either (a) the specific intervention under investigation or (b) administered outside the study as part of normal care.

The following events were also potentially expected: admission to a high-dependency unit/intensive care; emergency hysterectomy; laparotomy; port site hernia; blood transfusion; wound infection; lower urinary tract infection; endometritis; blood stained vaginal discharge; anaesthetic complications; low-grade pyrexia; blood loss; haematoma; constipation; pelvic discomfort/pain; internal bleeding or injury; deep-vein thrombosis; pulmonary embolism; injury to the wall of the uterus; bladder/bowel/ureteric injury; and voiding dysfunction.

Any serious adverse events (SAEs) related to the participants’ HMB treatment that were not further interventions (e.g. being admitted to hospital for an infection) were recorded on the SAE form. Hospital visits that were associated with further interventions due to HMB (e.g. further surgery) were recorded as an outcome measure. All deaths from any cause (related or otherwise) were recorded on the SAE form.

It was a requirement to report to the sponsor any SAEs that were deemed related and unexpected within 24 hours of receiving the signed SAE notification. Such SAEs would also be reported to the main Research Ethics Committee within 15 days of the chief investigator becoming aware of the event. All related SAEs were summarised and reported to the appropriate authorities within their regular progress reports.

Sample size

An individual participant data meta-analysis of abdominal hysterectomy compared with first-generation EA8 suggested a target difference of an odds ratio (OR) of 2.84 (95% vs. 87%) for patient satisfaction. Such an OR also equates to a medium-sized standardised effect (Cohen’s d). It was calculated that 292 participants per group would provide 90% power to detect a difference in total satisfaction rates of 8% (87% vs. 95%), using a two-sided continuity-corrected chi-squared test. This would also allow > 90% power to detect a 10-point difference in MMAS scores [assuming a standard deviation (SD) of 33 units]. Based on an expected 10% drop-out rate, the recruitment target was 648 participants in total (324 participants per group).

Owing to changes to the analysis plan during the trial, the implications for these calculations were later revisited (see Statistical analysis).

Data Monitoring Committee

The independent Data Monitoring Committee (DMC), which consisted of a methodologist, a clinician and a statistician, met on four occasions during the trial and considered interim reports of trial data by randomised groups (denoted as group 1 and group 2). On each occasion they agreed that the trial should continue as planned. At the final meeting, which took place in September 2017, after recruitment had ended, the DMC was shown the distributions of the primary outcomes using interim data.

Statistical analysis

Analyses were conducted using Stata^® version 15 (StataCorp LP, College Station, TX, USA). Categorical variables were described with the number and percentage in each category. Continuous variables were described using mean and SD or median and interquartile range (IQR), depending on the distribution of data.

Analyses were based on the intention-to-treat principle, analysing women in the groups to which they were randomised. Analyses used a two-sided 5% significance level, with corresponding 95% confidence intervals (CIs) generated as appropriate [see URL: www.journalslibrary.nihr.ac.uk/programmes/hta/123523/#/ (accessed 22 May 2019)].

Analyses of the two co-primary outcomes (patient satisfaction and MMAS at 15 months after randomisation) were conducted independently.

Patient satisfaction was collected on a six-point scale: ‘totally satisfied’, ‘generally satisfied’, ‘fairly satisfied’, ‘fairly dissatisfied’, ‘generally dissatisfied’ or ‘totally dissatisfied’. The original analysis plan specified that patient satisfaction be treated as a binary variable, but after considering the distribution of interim data (September 2017), and to make better use of the data collected, the DMC requested that this outcome be treated as ordinal with four categories (‘totally satisfied’, ‘generally satisfied’, ‘fairly satisfied’, with the remaining three categories combined into a single ‘dissatisfied’ category), and analysed using an ordinal logistic regression (OLR) model with adjustment for minimisation factors: age (< 40 vs. ≥ 40 years) and centre (treated as a random effect).

We originally intended the MMAS score to be treated as a continuous outcome and analysed using a linear regression model, adjusted for baseline MMAS scores (treated as a continuous variable) and the two minimisation factors (age and centre). However, interim data presented to the DMC were highly skewed, with over half of all trial participants reporting a maximum MMAS score of 100 at 15 months post randomisation. After discussion, the DMC recommended that the analysis of this outcome be changed to an OLR model with four categories (0–50, 51–75, 76–99, 100).

The Project Management Group (PMG) accepted the recommended changes to the analysis of the primary outcomes after viewing the distribution of these outcomes with both groups combined. The decision to change the analysis method for both co-primary outcomes to an ordered categorical analysis had implications for the power of the study. It seemed appropriate to continue to use a common OR of 2.84 in the revised power calculations. Based on the formulae in a report by Whitehead,26 assuming a common OR of 2.84 and using the expected proportions in each category from interim trial data with treatment groups combined, a total of 292 participants per group would have > 90% power to detect statistically significant differences between the randomised groups for patient satisfaction. There were no accurate data on which to base our decisions regarding the MMAS outcome, so a pragmatic decision was made to use the same OR for this outcome. Both outcomes had similar distributions after classification into four categories and were addressed in a hierarchy (MMAS score was considered only if patient satisfaction showed a statistically significant difference). Therefore, it was considered unnecessary to recruit additional participants to the trial.

The MMAS total score was calculated only if the six constituent items were completed. There does not appear to be a precedent in the literature for imputing scores for MMAS items from other items as the response category weighting varies by item. In addition, to account for the nature of the treatments being offered (i.e. the vast majority of the women in the LASH arm and 40–50% of the women in the EA arm would be expected to be amenorrhoeic following treatment), the instructions for completion of the MMAS were altered slightly before the start of the study [see URL: www.journalslibrary.nihr.ac.uk/programmes/hta/123523/#/ (accessed 22 May 2019)].

It was expected that a small number of women might have to wait longer than 12–18 weeks for their operation and therefore would receive their 6 months post-surgery and 15 months post-randomisation questionnaires simultaneously. It was therefore agreed that scores for the co-primary outcomes and the main outcome used in the economic analysis (EQ-5D-3L utility score) could be used from the 6-month post-surgery questionnaire in lieu of the 15-month data, provided these were provided within 3 months of the due date of the latter.

A sensitivity analysis using a multiple imputation approach was used to explore the impact of missing data on the robustness of the results of the analyses of co-primary outcomes. Missing values for the outcome and for one covariate (MMAS score at baseline) were imputed using multiple imputation by chained equations using the mi package in Stata. 27 Twenty imputed data sets were created. There were no missing data for the other covariates (age group and centre), as these were required for the minimisation algorithm.

A further sensitivity analysis using the results of the original analysis method for MMAS score (i.e. a linear regression model treating MMAS score as a continuous outcome) was also presented.

Exploratory subgroup analyses were performed for the following groups: uterine cavity length (≤ 8 cm vs. > 8 cm), menstrual pain (dysmenorrhoea) at baseline (severe/crippling pain vs. other categories, determined using a five-point Likert scale), patient age (< 40 years vs. ≥ 40 years) and presence or absence of fibroids. These prespecified subgroup analyses were conducted by including a treatment by subgroup interaction term in the corresponding OLR model for the co-primary outcomes (patient satisfaction and MMAS score at 15 months post randomisation). The effect sizes from these subgroup analyses were displayed graphically in a forest plot, along with results for each recruiting centre and whether or not the operator was a consultant. Centres with fewer than 50 randomised participants were included in an ‘other centre’ category.

Secondary outcomes were analysed using generalised linear models (GLMs), adjusted for the minimisation factors and, when appropriate, a baseline measure. Continuous outcomes were analysed using linear regression, binary outcomes using logistic regression and ordered categorical outcomes using OLR. As many continuous outcomes had an extreme skewed distribution, an ordered categorical analysis had to be used instead. The category thresholds were decided by the PMG after viewing the distribution of the outcome of both groups combined.

Time-to-event outcomes such as time until return to normal activities (e.g. paid work for those in employment) were analysed using Cox proportional hazards regression models.

Pain data from the participant diaries (collected in the first 14 days after the operation) were analysed using a repeated-measures model.

All AEs were described using the number and percentage within each randomised group. All expected SAEs and other SAEs were recorded in detail and the number and percentage in each randomised group reported.

Economic evaluation

The economic analysis consisted of a trial-based analysis of individual patient-level cost and effect (QALY), and a decision modelling component to inform cost-effectiveness in the longer term. See Chapters 5 and 6 for a detailed description of the methods used.

Management of the study

The trial management team, based within CHaRT, University of Aberdeen, provided day-to-day support for the recruiting centres led by a local principal investigator (PI). The PIs, supported by dedicated research nurses, were responsible for all aspects of local organisation, including recruitment of participants, delivery of the interventions and notification of any problems or unexpected developments during the study period.

The study was supervised by the PMG, which consisted of representatives from the study office and grant holders. The study was further overseen by a Trial Steering Committee (TSC), which comprised four independent members and an independent DMC (see Data Monitoring Committee).

Patient and public involvement

Study recruitment

Study design and recruitment methodology have been described previously1 (see also Chapter 2). Women with HMB who attended gynaecology outpatients and pre-assessment clinics, and who were eligible for EA, were invited to participate in HEALTH. Women were asked if they would be willing to be randomised to either LASH or second-generation EA. The centres that randomised women into HEALTH, including numbers recruited by centre, are described in Appendix 1, Table 26. The recruitment rate is illustrated in Figure 3.

FIGURE 3.

Recruitment to the trial over time.

Non-recruited women

Of the 2552 women approached, 1888 (74.0%) did not participate in the trial because they were ineligible (n = 1201, 47.1%) or declined participation (n = 589, 23.1%) or because their consultant later indicated a preference for a particular treatment (n = 98, 3.8%) (see Figure 2 and Appendix 1, Table 27). The most common reasons for women declining to take part in the study were a preference for a particular treatment [preference for EA, n = 151 (25.6%); preference for hysterectomy, n = 126 (21.4%); preference for medical management, n = 89 (15.1%)] and an unwillingness to accept randomisation (n = 54, 9.2%). Ninety-eight women (16.6%) did not give a reason for declining to take part (see Appendix 1, Table 28).

Reasons for ineligibility included ‘fibroids > 3cm’ (n = 361, 30.1%), a preference to continue with medical management (n = 244, 20.3%) and age > 50 years (n = 239, 19.9%). In addition, 98 women who were deemed eligible for HEALTH were not included for other clinical reasons. In the majority of cases, this was because a treatment pathway had already been decided prior to study entry [hysterectomy, n = 25 (25.5%); EA, n = 18 (18.4%); and ‘other’ treatment, n = 16 (16.3%)] (see Appendix 1, Table 28).

Randomised participants: baseline characteristics

The baseline characteristics of the 660 women who agreed to participate in HEALTH and who were truly eligible to take part are described in Tables 2–4 and Appendix 1, Table 29.

Flow of participants through the trial

The CONSORT flow diagram shows the number of participants providing data at each stage of the trial (Figure 4). Response rates are based on the number receiving an operation (14-day diary, 6-week and 6-month questionnaire) or the number randomised (15-month questionnaire), after accounting for withdrawals. Questionnaire completion rates ranged between 80% and 89% (exact return rates are reported in the CONSORT flow diagram; see Figure 4).

FIGURE 4.

Consolidated Standards of Reporting Trials (CONSORT) flow diagram. a, Other operations (LASH group): total hysterectomy (n = 5), EA (n = 1), hysteroscopy/polypectomy (n = 1); b, other operations (EA group): total hysterectomy (n = 5), hysterectomy/polypectomy (n = 4); c, reasons included: unwillingness to have surgery (n = 1), private treatment (n = 1), no reason given (n = 1); and d, reasons included: unwillingness to have surgery (n = 2), requested a different operation (n = 2), family illness (n = 1), moved abroad (n = 1), did not want to complete questionnaires (n = 2). PQ, participant questionnaire. Reproduced from Cooper et al. 29 © The Author(s). Published by Elsevier Ltd. This is an Open Access article under the CC BY-NC-ND 4.0 license.

Reproduced from Cooper et al. 29 © The Author(s). Published by Elsevier Ltd. This is an Open Access article under the CC BY-NC-ND 4.0 license.

Operation details and operative outcomes

Forty-four participants [21/330 (6.4%) randomised to LASH and 23/330 (7.0%) randomised to EA] did not undergo an operation. These women were not asked to complete the patient diary or questionnaires at 6 weeks and 6 months post surgery, but were sent the final questionnaire at 15 months post randomisation.

Table 5 provides details for the 616 women who received an operation. The median number of days between randomisation and treatment was higher in the LASH group [84 (IQR 57–127) days] than in the EA group [63 (IQR 41–97) days] (see Appendix 2, Figure 16). Six women across both arms waited a year or more for their operation and therefore received the 6 months post-surgery and the 15 months post-randomisation questionnaires at around the same time.

Of those undergoing treatment, 291 out of 309 (94.2%) randomised to LASH and 297 out of 307 (96.7%) randomised to EA received the allocated procedure. Twelve of those randomised to LASH actually received an EA, five underwent a total hysterectomy and one had a hysteroscopy/polypectomy. One woman randomised to EA received LASH, five had a total hysterectomy and four had a hysteroscopy/polypectomy (see Figure 4). In total, nine of the EA operations could not be completed during the first admission (one in the LASH group and eight in the EA group), three women were subsequently readmitted for LASH, two women were readmitted for the total hysterectomy and two women were readmitted for EA.

Compared with the EA group, the LASH group included higher proportions of women who were operated on by a consultant (77.3% vs. 57.3%), received thromboprophylaxis (98.4% vs. 69.2%) and received parenteral postoperative opiates (30.4% vs. 15.0%). Fewer women in the LASH group than in the EA group were noted to have a uterus free from fibroids during surgery (75.7% vs. 91.1%) and more (32.4% vs. 5.3%) stayed in hospital for > 24 hours (see Table 5).

Results for the co-primary outcomes

MMAS scores at 15 months post randomisation

Total MMAS scores were available for 262 out of 330 (79.4%) women in the LASH group and 268 out of 330 (81.2%) women in the EA group at 15 months post randomisation. This included MMAS scores for two women whose 6-month data were imputed for the 15-month time point. A further 29 women completed at least one of the six items, but were excluded from the primary analysis because a total score could not be derived.

The total MMAS score ranges from 0 (worst possible health) to 100 (best possible health). Both groups reported a considerable improvement in MMAS scores after surgery. At baseline, median scores were 28.6 (IQR 14.7–43.7) in the LASH group and 29 (IQR 15.7–47.7) in the EA group, but by 15 months post randomisation the majority of women in each group had the best possible score (MMAS score = 100) (Figure 5 and Appendix 2, Table 31).

FIGURE 5.

Total MMAS scores at (a) baseline; (b) 6 months post surgery; and (c) 15 months post randomisation.

The results for the primary analysis, OLR adjusting for age, centre and baseline MMAS, are presented in Table 6. An OR of 1.87 (95% CI 1.31 to 2.67) was obtained, suggesting that women randomised to receive a LASH had almost twice the odds of being in a more favourable MMAS category than women randomised to EA (p = 0.001). The corresponding unadjusted OR was 1.90 (95% CI 1.35 to 2.68; p < 0.001) (see Appendix 2, Table 32).

The proportional odds assumption was investigated by examining binary logistic regression models using three splits of the data. All approaches yielded similar ORs that were consistent with the main result and the Brant test was not statistically significant (p = 0.07). There was also no change in interpretation using a per-protocol analysis, when restricting analyses to those operated on by a consultant or when treating the MMAS score as a continuous outcome (see Appendix 2, Table 32).

The mean between-group difference in MMAS scores of 6.3 points was lower than the difference of 10 points specified in the original sample size calculation. However, as the distribution of scores was highly skewed, the median is a more appropriate summary to use, and this was 100 in each group. The 95% CI for the effect size for the primary analysis did not include the OR of 2.84 specified in the revised sample size calculation, but an OR of 1.87 could still be considered to have a clinically important impact on patients.

Results for the secondary outcomes

Patient diary (1–14 days)

Table 34 in Appendix 2 reports data for the patient diary, which was completed in the first 14 days following surgery. In both groups there was a reduction in self-reported levels of pain (0 = no pain, 10 = worst imaginable pain) (Figure 6) and in the proportion of women taking paracetamol or other painkillers over these 2 weeks. In addition, fewer women were using pads for vaginal bleeding or discharge. By day 14, 177 out of 256 (69.1%) women in the EA group and 34 out of 267 (12.7%) women in the LASH group were using pads.

FIGURE 6.

Mean pain during the first 14 days post surgery (0 = no pain, 10 = worst imaginable pain).

Overall, those in the LASH group had pain scores that were almost 1 point higher than those in the EA group (mean difference 0.92, 95% CI 0.59 to 1.24; p < 0.001; see Appendix 2, Table 34).

Pain and return to usual activities (6 weeks post surgery)

By 6 weeks after surgery, over half of the women in both groups reported no pain on a 10-point scale from 0, no pain, to 10, the worst pain imaginable (Table 9). After adjusting for the minimisation factors (age group and centre), an OR of 1.43 (95% CI 1.05 to 1.96; p = 0.03) was obtained, suggesting that those in the EA group had lower levels of pain at 6 weeks than those in the LASH group.

TABLE 9 - Patient outcomes and return to normal activities 6 weeks post surgery

HR, hazard ratio.

OLR using categories 0, 1–5, 6–10.

Favours EA (p < 0.05).

Cox regression.

Estimated median (95% CI) from Cox regression.

Outcome	LASH (N = 309)	EA (N = 307)	Adjusted ORa (95% CI), p-value
Level of pain today (0 = no pain, 10 = worst imaginable), median (IQR) [range], n	0 (0–1) [0–10], 241	0 (0–1) [0–10], 234	1.43 (1.05 to 1.96),b p = 0.03
Current employment, n (%)
Full time	104 (43.7)	99 (42.1)
Part time	82 (34.5)	82 (34.9)
Not working	52 (21.8)	54 (23.0)
			Adjusted HRc (95% CI), p-value
Days until return to paid work, median (95% CI)d [n]	42 (37 to 42) [186]	10 (7 to 14) [181]	0.23 (0.18 to 0.30),b p < 0.001
Days until return to unpaid work, median (95% CI)d [n]	21 (17 to 25) [255]	7 (5 to 7) [251]	0.64 (0.57 to 0.73),b p < 0.001
Days until return to sporting or social activities, median (95% CI)d [n]	42 (34 to 42) [255]	14 (14 to 18) [250]	0.48 (0.42 to 0.56),b p < 0.001

Figure 7 shows the time to return to work by randomised group for the women in full- or part-time paid employment. Those in the LASH group returned to work after a median of 42 days, whereas those in the EA group returned after a median of 10 days (see Table 9). A Cox proportional hazards regression model adjusted for age group and centre suggested a statistically significant difference in favour of EA [adjusted hazard ratio (HR) 0.23, 95% CI 0.18 to 0.30; p < 0.001].

FIGURE 7.

Time to return to (a) paid work; (b) unpaid work; and (c) sporting or social activities.

Those in the EA group returned more rapidly to both unpaid work (adjusted HR 0.64, 95% CI 0.57 to 0.73; p < 0.001) and sporting or social activities (adjusted HR 0.48, 95% CI 0.42 to 0.56; p < 0.001) (see Table 9).

Summary of the clinical effectiveness results

Table 38 in Appendix 2 provides a summary of the primary and secondary analyses, including both adjusted and unadjusted effect sizes for all outcomes as follows: mean differences for continuous outcomes; both ORs and risk differences for binary outcomes (i.e. both relative and absolute effect sizes); ORs for ordered categorical outcomes and HRs for time-to-event outcomes. Adjusted analyses include the minimisation factors, age group and centre (random effect), in the model, as well as a baseline score, if this is available.

The results of the unadjusted analyses tended to be similar to those of the adjusted analyses.

Using a threshold of a p-value of < 0.05, there was evidence that those randomised to EA had lower levels of pain in the first 6 weeks following treatment. In addition, those randomised to EA also had improved EQ-5D-3L utility scores and SF-12 PCSs at 6 weeks, but there was no evidence of a difference between groups at later time points. Women in the EA group also returned to work and usual activities sooner than those in the LASH group.

Most self-reported outcomes at the 6 months post surgery and 15 months post-randomisation time points tended to favour LASH. There was evidence that women in the LASH group had better QoL outcomes and were more satisfied with their treatment than those in the EA group.

The results of both co-primary outcomes (satisfaction and MMAS score at 15 months post randomisation), both strongly favoured those in the LASH group. The results of the secondary outcomes should be treated as exploratory because no adjustment was made for multiple statistical testing. There is, however, a pattern suggesting greater short-term benefits for EA but longer-term improvements in patient-reported outcomes for the LASH group. In particular, LASH was strongly favoured for all the questions concerning acceptability, satisfaction and recommendation to a friend.

Introduction

This chapter reports on the within-trial economic evaluation of LASH compared with second-generation EA. 1 The rationale for the economic evaluation in health care is to help inform the adoption of technologies that provide good value for money in the context of constrained health service resources. The within-trial economic analysis reported in this chapter considers the 15-month post-randomisation follow-up period only. As the full impact of the alternative interventions on resource use and individuals’ HRQoL is likely to accrue over a much longer time horizon, a Markov model was also developed to extrapolate the trial-based findings. This model-based economic analysis is reported in Chapter 6 and constitutes the primary economic analysis for HEALTH.

Objectives of the economic evaluation

The primary economic objective of the within-trial cost-effectiveness analysis was to estimate the incremental cost per QALY gained for LASH compared with EA at 15 months post randomisation. Two of the three secondary economic objectives are addressed in the current chapter: (1) to compare the costs and consequences of LASH and EA at 15 months; and (2) to assess the wider societal costs associated with changes in productivity. The third secondary economic objective of modelling the longer-term cost-effectiveness of LASH compared with EA is addressed in Chapter 6.

Methods

Results

Cost–utility analysis results

The incremental analysis was conducted using the multiple imputation data set and is presented in Table 15. The adjusted mean costs per participant were £2886 and £1282 for LASH and EA, respectively, producing an adjusted difference of £1604. The average cost difference resulting from the imputed data was slightly narrower than the unadjusted difference based on complete data (see Table 13). The mean adjusted QALYs per participant were 0.978 and 0.974 for LASH and EA, respectively, giving an adjusted QALY difference of 0.004 in favour of LASH.

TABLE 15 - Trial-based incremental cost-effectiveness analysis (base case; NHS perspective; 15 months post randomisation)

Differences adjusted by study minimisation variables.

Intervention	Total cost (£)	Incremental cost (£)a	Total QALYs	Incremental QALYsa	ICER (£/QALY gained)
EA	1282		0.974
LASH	2886	1604	0.978	0.004	458,334

Therefore, intention to treat with LASH resulted in significantly higher mean costs and a slight non-significant QALY gain compared with EA at 15 months post randomisation. The ICER for LASH compared with EA came to £458,334 per QALY gained over this relatively short time horizon. Although the EQ-5D-3L health state utility was higher in the LASH group by 15 months post randomisation, the mean QALYs accruing over 15 months remained very similar between groups owing to the earlier improvement in HRQoL with EA than with LASH.

Figure 8 shows the scatterplot of the difference in mean costs and difference in mean QALYs based on the 1000 bootstrapped iterations of the regression analysis with nested multiple imputation. LASH was clearly more costly, on average, than EA, as all the iterations resulted in mean cost differences that were greater than zero. However, the mean difference in QALYs was centred on 0.004 (favouring LASH), with a substantial proportion of the bootstrapped iterations favouring EA. Although Figure 8 shows that ≈60% of the bootstrapped iterations generated a QALY gain favouring LASH over EA, all these points lie above accepted cost-effectiveness thresholds,42 suggesting that LASH had little chance of being considered cost-effective based on its incremental cost-per-QALY ratio over a 15-month time horizon. This was to be expected, given the relatively short duration of follow-up combined with the higher initial treatment costs for LASH and the more delayed pattern of improvement in HRQoL described above.

FIGURE 8.

Trial-based incremental cost-effectiveness scatterplot for LASH vs. EA (base case; imputed data; 1000 bootstrap iterations).

Discussion

The within-trial cost–utility analysis reported in this chapter indicated that, over a 15-month post-randomisation follow-up period, intention to treat with LASH resulted in increased costs to the health service (mean difference £1604). This was mainly driven by the cost of the initial procedure, with LASH taking twice as long to perform and resulting in a longer hospital stay than EA. Costs to society were also increased following LASH (£1596), because women treated with LASH took longer to return to paid and unpaid productive activities than those treated with EA. There was very little difference in QALYs between the treatment allocation groups, assessed over 15 months from randomisation, resulting in the ICER for LASH being unfavourable at this time point. The 15-month ICER also remained unfavourable to LASH in all the sensitivity analyses and subgroup analyses performed.

Strengths of the trial-based economic analysis include the availability of randomised data on resource use and HRQoL profiles collected prospectively as part of HEALTH. This enables accurate and unbiased estimation of mean differences in costs and QALYs over the trial follow-up period. The pragmatic design and intention-to-treat principles also enhance the generalisability of the findings to routine practice in the UK NHS.

The key limitation of the trial-based economic analysis relates to the relatively short duration of follow-up. Although there was no notable difference in QALYs between the groups by 15 months post randomisation, this time horizon was insufficient for drawing any conclusions on cost-effectiveness. The follow-up data indicate that there were more readmissions to hospital among those randomised to EA (n = 27) than there were among those randomised to LASH (n = 15), including 13 hysterectomies following EA compared with only one surgical episode related to menstrual bleeding post LASH. This trend for an increased incidence of further gynaecological surgery following EA is likely to become more pronounced in the future,10 reducing the incremental cost of LASH and resulting in QALY gains favouring it. Furthermore, the lack of difference in QALYs by 15 months post randomisation belied the fact that the EQ-5D-3L health state utility curves crossed during follow-up. Those randomised to EA experienced a shorter waiting time and quicker rise in health state utility following surgery, but then a subsequent decline in their EQ-5D-3L score by 15 months. The LASH group, on the other hand, experienced continued improvement in their EQ-5D-3L score out to 15 months post randomisation, by which time the mean score was higher than the EA group. As QALYs were computed as the area under the health state utility curve, terminating follow-up at 15 months has truncated the incremental QALY gains that would be expected to accrue to LASH if the 15-month difference in health state utility was maintained or increased over time. It is therefore necessary to extrapolate the trial results over a longer time horizon to inform cost-effectiveness. This is the focus of Chapter 6.

Introduction

The purpose of this chapter is to report on the details of further modelling conducted to extrapolate the trial-based cost-effectiveness findings beyond 15 months post randomisation. Although the within-trial analysis is useful for informing differences in costs and outcomes over a relatively short time horizon, it does not capture expected differences over the medium (2–5 years) to long term (5–10 years). Eventually 20–25% of women assigned to EA are expected to require further surgical treatment,5,9,10 resulting in downstream QALY losses and higher readmission costs than with LASH. These anticipated costs and consequences should be accounted for in a full economic evaluation. 43 Furthermore, although there was no significant between-group difference in QALYs observed by 15 months post randomisation in HEALTH, the mean health state utility scores appeared to be diverging by this time point (see Chapter 5, Table 14). Although the difference was not statistically significant, the direction of the effect in favour of LASH is consistent with the reported superiority of LASH on the primary clinical outcomes and with the expectation of higher repeat surgery rates in the EA group.

To estimate longer-term economic differences, a simple Markov model was developed to extrapolate the estimated 15-month difference in utility and simulate the incidence of further gynaecological surgery over time. The key objective of the analysis was to inform the long-term cost-effectiveness of LASH compared with EA. 1 This analysis forms the primary economic analysis of HEALTH.

Methods

Model structure

The Markov model was constructed in TreeAge Pro software (TreeAge Software, Inc., Williamstown, MA, USA) and was informed by reviewing decision models identified in a recent systematic review that was undertaken to inform the NICE clinical guidelines on the assessment and management of HMB. 43 A structure similar to that of other models was developed to assess the cost-effectiveness of treatments for HMB in the UK NHS was chosen. 9,10,43 The state occupancy and estimated pay-offs are updated on a constant monthly Markov cycle.

A cohort of women with HMB enter the model in the ‘HMB’ health state and are assigned to treatment with either EA or LASH. Women are modelled to receive treatment as observed in HEALTH by intention to treat. However, for simplicity, index procedures are modelled to occur in the first cycle of the model, with the waiting time factored out.

Following the index treatment in the EA arm (Figure 9), women move to either the ‘post EA’ or ‘complication post EA’ health state. The complication health state is designed to capture the cost and utility impact of postoperative complications resulting in readmission to hospital. These events are assumed to be transitory and so the health state utility impact is modelled to last for one cycle (1 month). Following this, women transit to the ‘post EA’ health state for the subsequent model cycle unless they progress to further surgery. The transition to further surgery from the ‘post EA’ health state is modelled on a monthly basis. Although the model has the capability to include first-generation EA (rollerball) as a second-line surgical treatment, the base case assumes that hysterectomy is the treatment of choice for women who require further surgery following failed second-generation EA. Repeat EA is not explicitly recommended as a treatment option in the recently updated NICE clinical guidelines for HMB. 43 For women who transition to ‘hysterectomy post EA’, the surgical intervention occurs in the first model cycle following the transition. Following this, women either enter a temporary health state representing severe postoperative complications, or they move to ‘convalescence post total hysterectomy’. The convalescence period is modelled to last for 3 months, after which all women transition to the ‘well post total hysterectomy’ health state. Finally, ‘death’ is included as an absorbing state in the model, which can be entered from any other state based on age-specific mortality rates reported for females in UK life tables. 44

FIGURE 9.

Simplified schematic for the EA arm of the Markov model. The transition from ‘Post EA’ to ‘Well post total hysterectomy’ is included as a one off transition at month 12 to account for the small proportion of patients who had a total hysterectomy instead of EA for their index treatment.

The structure of the model in the LASH arm is similar to that in the EA arm (Figure 10). However, following index treatment women either enter a temporary health state to capture postoperative complications or they transition to ‘convalescence post LASH’. The complication state serves the same purpose as it does in the EA arm. The convalesce state is included to capture the longer time to full recovery observed for LASH than for EA in HEALTH. From the post-LASH health states, a monthly probability of requiring further related gynaecological surgery is also applied. This includes further surgery related to ongoing bleeding or pain following LASH, specifically removal of the cervical stump, laparoscopy to investigate/treat pain and laparoscopic bilateral salpingo-oophorectomy (BSO). Following further surgery post LASH, women enter a post-surgical state dependent on the type of surgery received. For women who undergo cervical stump removal, a probability of severe post-surgical complications is applied and a convalescence period of 3 months is also assumed. Following convalescence, women are assumed to be well for the remainder of the modelled time horizon. For women who require laparoscopy for pain or laparoscopic BSO, no further postoperative complications or convalescence period are modelled, but women attract lower health state utility in the cycle leading up to surgery. After surgery they are also assumed to be well for the remaining time horizon.

FIGURE 10.

Simplified schematic for the LASH arm of the Markov model.

Clinical input parameters

The key clinical input parameters in the model are tabulated in Table 19. The rates of post-surgical complications following EA and LASH were based on HEALTH data, as were the rates of subsequent gynaecological surgery for HMB out to 1 year following index surgery. Figure 11 shows the Kaplan–Meier (KM) plots for further gynaecological surgery by treatment arm in HEALTH (conditioned on receipt of index surgery); the estimated 12-month probabilities were 3% in the EA arm and 0.7% in the LASH arm. These probabilities were transformed into monthly probabilities for application over the first 12 cycles in the model.

TABLE 19 - Clinical input parameters applied in the model

Observed further surgery included hysterectomy following EA and trachelectomy following LASH.

Applied in sensitivity analysis to adjust down the incidence of hysterectomy following EA.

Further surgery beyond 12 months post LASH includes laparoscopic surgery to remove the cervical stump, laparoscopy to investigate/treat pain and laparoscopic BSO.

Variable	Point estimate	Standard error	Distributional form	Source
Probability of readmission for complications post EA	0.033	0.01	Beta	HEALTH
OR for postoperative complications (LASH vs. EA)	1.107	0.420	Lognormal	HEALTH
Probability of hysterectomy post EA (by 12 months)a	0.03	0.0098	Beta	HEALTH
Probability of further surgery for bleeding post LASH (by 12 months)a	0.0069	0.0049	Beta	HEALTH
Probability of hysterectomy post EA (beyond 12 months up to 10 years)			Applied and tested deterministically
Weibull rate parameter (λ)	0.119			Cooper et al.5
Weibull shape parameter (γ)	0.397			Cooper et al.5
Log HR for hysterectomy (second- vs. first-generation techniques)b	–0.274	0.0532	Lognormal	Bansi-Matharu et al.46
Inferred log HR for further surgery post LASH vs. EAc	–0.4886	0.122	Lognormal	Calibrated to Lieng et al., 200817; see Extrapolation of subsequent surgery following laparoscopic supracervical hysterectomy
Probability of severe complications post hysterectomy	0.0102	0.001	Beta	Maresh et al.47; Bhattacharya et al.10
Probability of severe complications post trachelectomy	0.0102	0.001	Beta	Assumption

FIGURE 11.

Kaplan–Meier plot of time from index operation to further related surgery by treatment allocation.

Extrapolation of subsequent hysterectomy following endometrial ablation

Given what is known about the probability of hysterectomy following EA, it is anticipated that the KM curves for time to further surgery will continue to diverge over time. In addition, as the follow-up of HEALTH is currently truncated at 15 months post randomisation and waiting times for the index surgery were 3–4 months on average, only 69% of women in the EA group and 56% in the LASH group were observed to 12 months post surgery at the time of writing. Coupled with the current waiting times for subsequent surgery among those who need it, HEALTH data do not currently give an accurate picture of further surgery rates by treatment allocation group.

Given these limitations, external data were used to inform rates of subsequent surgery beyond 12 months in the model. A focused search of MEDLINE was conducted to identify randomised trials comparing EA and LASH using the Ovid interface (see Appendix 4). This identified one small RCT which included long-term follow-up of patients. Zupi et al. 48 reported that, after a mean follow-up of 14.4 years, 20 out of 71 patients (28.1%) required a reoperation for menstrual bleeding in the EA arm of the trial, compared with none in the LASH arm (out of 82 available for analysis). They did report that 6 out of 82 women (7.3%) required further surgery in the LASH arm for indications other than menstrual bleeding (five requiring a laparoscopy to investigate pain). Although this RCT provides evidence in favour of a lower rate of subsequent related surgery following LASH, it is a small, single-centre study. Given this limitation, a further focused literature search was undertaken to identify cohort studies reporting rates of hysterectomy following EA or rates of cervical stump removal, laparoscopy and BSO following LASH (see Appendix 4). Priority was placed on identifying large population-based cohort studies reporting long-term rates. This yielded three5,46,49 population-based cohort studies reporting rates of hysterectomy following EA and eight17,50–56 smaller observational studies reporting rates of further surgery following LASH (see Appendix 4 for details).

For the probability of hysterectomy following EA, we used data from a large observational study carried out using routine Scottish health service data. Based on linked health episode data, Cooper et al. 5 reported on 14,078 women identified as having received primary EA for HMB between 1989 and 2006 in Scotland. Over a median duration of 6.8 years follow-up, 2779 (19.7%) women were observed to receive a subsequent hysterectomy. Data were extracted from the published KM curve using digitising software (WebPlotDigitizer, V4.1, 2018; URL: https://automeris.io/WebPlotDigitizer/) and regression methods were then used to fit a Weibull distribution to the observed time-to-event data (see Appendix 5 for details).

A limitation of the study by Cooper et al. 5 was an inability to discriminate between EAs carried out using second- and first-generation techniques, and a number of studies have suggested that some second-generation techniques may incur a lower risk of post-ablation hysterectomy. 46,57 However, a very similar rate of post-ablation hysterectomy has been reported following second-generation procedures carried out between 1997 and 2007 in a Finnish registry study (1086/5484, 19.8%). 49 The Weibull distribution fitted to the Scottish data was therefore used in the model to derive time-dependent probabilities of transition to hysterectomy in the EA arm (see Appendix 5 for details regarding the derivation of transition probabilities). Sensitivity analysis was used to explore the impact of adjusting the hazard rate downwards using a HR for radiofrequency ablation (HR 0.76, 95% CI 0.69 to 0.85), reported by Bansi-Matharu et al. 46 and based on a large English population-based cohort study. Further sensitivity analysis was also conducted to explore the impact of using time-dependent transition probabilities for hysterectomy following EA, derived directly from the KM curve reported by Cooper et al. 5 The model projections of post-ablation hysterectomy were compared against the 5-year rates reported by Bansi-Matharu et al. 46 as means of external validation (see Model validation).

Extrapolation of subsequent surgery following laparoscopic supracervical hysterectomy

Several studies of variable size and quality were identified to inform the risk of further surgery following LASH (see Appendix 4). The reported incidence of post-LASH cervical stump removal ranged from 0.9% to 23%. 17,50–53,58 Based on length of follow-up, consistency with the observed 12-month rate in HEALTH, and a similar rate of post-LASH menstrual bleeding (24% vs. 19%), a Norwegian cohort reported by Lieng et al. 17 was considered the most useful for informing longer-term rates in the model (see Appendix 4 for details). Lieng et al. 17 reported that, of 308 women undergoing LASH at a university hospital in Oslo during 2004 and 2005, six (1.9%) subsequently underwent laparoscopic adhesiolysis, seven (2.3%) underwent laparoscopic cervical stump removal, one (0.3%) underwent BSO and eight underwent other procedures in the 12–36 months of follow-up after LASH. The other reported procedures included laparoscopic drainage of postoperative abscess (n = 1), laparoscopy with bowel resection for postoperative peritonitis (n = 1), scar correction (n = 3), umbilical hernia repair (n = 1) and tension-free vaginal tape procedures (n = 2). These procedures were not included in the model as they were considered to be either short-term postoperative complications (included in the model based on HEALTH data) or of uncertain association with the LASH procedure. The model was therefore calibrated to yield the combined cumulative incidence of laparoscopic adhesiolysis, laparoscopic cervical stump removal and BSO (4.5%), reported by Lieng et al. 17 by month 36. This was done by applying a HR to the rate parameter of the Weibull distribution used to model time to hysterectomy following EA. Thus, it is assumed that the hazard for further surgery post LASH is proportional to the hazard of hysterectomy post EA. This approach yields a cumulative incidence of further surgery of ≈11% by 10 years post LASH in the model.

Postoperative complications following hysterectomy and removal of the cervical stump

For those women modelled to go on to receive a hysterectomy following EA, or cervical stump removal following LASH, an associated postoperative complication rate is applied. This probability is taken from the previous HTA study conducted by Bhattacharya et al. ,10 originally sourced from Maresh et al. 47 As no data were available to inform the risk of severe postoperative complications following cervical stump removal, the same probability of complications was applied to this procedure.

Health state utilities

The health state utilities applied in the model were derived primarily from the EQ-5D-3L data from HEALTH (Table 20). For the first 12 months in the model, adjusted utility estimates are applied by treatment arm. Beyond 12 months, extrapolation assumptions are applied.

TABLE 20 - EuroQol-5 Dimensions, three-level version-based health state utilities applied to the post-EA and post-LASH health states

Note, utility values are reported for those remaining in the post-EA and post-LASH states, free from subsequent surgery. All values are adjusted for baseline utility and minimisation factors using lineal regression, with adjustment for clustering within centres.

Variable	EA (mean)	LASH (mean)	Mean difference (95% CI)	Distributional form	Source
Baseline	0.702	0.702	–	Beta	HEALTH
6 weeks post surgery (no complications)	0.829	0.820	–0.009 (–0.05 to 0.032)	Beta + normal for (increment)	HEALTH
6 weeks post surgery (postoperative complications)	0.786	0.739	–0.047 (–0.275 to 0.182)	Beta + normal for (increment)	HEALTH
6 months post surgery (from cycle 3)a	0.815	0.817	0.0019 (–0.041 to 0.044)	Beta + normal for (increment)	HEALTH
12 months post surgery (from cycle 9)a	0.787	0.825	0.038 (–0.013 to 0.090)	Beta + normal for (increment)	HEALTH

Post-endometrial ablation and post-laparoscopic supracervical hysterectomy health state utility (to 12 months)

The baseline utility estimate for women entering the model was taken as the mean baseline EQ-5D-3L index score observed across treatment groups. Following treatment in the first cycle of the model, the cohort is distributed between the relevant post-treatment states. In cycle 2, the cohort is dichotomised by whether or not early postoperative complications occur. Health state utility values were therefore estimated by treatment allocation group and the occurrence of postoperative complications resulting in readmission to hospital. This was done by regressing the 6-week utility data on indicators for treatment allocation, readmission for complications (yes/no) and the interaction between treatment allocation and readmission. The regression also adjusted for baseline utility and age (< 40 years vs. ≥ 40 years) as a minimisation factor used in the randomisation process. Ordinary least squares regression was used for the analysis of the utility data, given the moderately large sample size, but with cluster robust standard errors. 59 The method of recycled predictions was used to recover the adjusted mean value in the base group, and the estimated incremental effects and robust standard errors from the regression models were used to define the distributions on effect differences in the Markov model. The analyses to inform the model utility inputs were based on available complete data for the relevant variables.

Beyond cycle 2 in the model, those remaining in the post-EA or post-LASH health states were assigned the corresponding treatment arm-specific 6-month utility values derived from HEALTH data. Those remaining in the post-EA or post-LASH treatment states were then modelled to receive the relevant treatment arm-specific 15-month utility value from cycle 9 (month 9) post surgery. This assumes that, on average, the 15-month post-randomisation utility estimate from HEALTH corresponds to 12-month postoperative utility, and that the change in health state utility between 6 months and 12 months post surgery is linear. For those modelled to incur a hysterectomy following EA, or subsequent surgery post LASH, further health state utility assumptions were applied as described below under Extrapolations of post-endometrial ablation and post-laparoscopic supracervical hysterectomy health state utility.

Extrapolations of post-endometrial ablation and post-laparoscopic supracervical hysterectomy health state utility

Table 20 shows that health state utility was slightly increased following EA, relative to LASH, in the short term (6 weeks), but, by 15 months post randomisation (≈12 months post surgery), there was an emerging difference that favoured LASH. Although this difference was not significant (p = 0.14), the direction of effect is consistent with superior satisfaction and MMAS scores observed in the LASH group at 15 months (see Tables 10 and 11). A lower utility value among those remaining in the post-EA health state at 12 months is also consistent with the expected higher incidence of further surgery (hysterectomy) beyond 12 months. It may reflect an increased number of women experiencing treatment failure or recurrence by 12 months post EA, who are yet to return for further surgery.

Therefore, the model extrapolated the estimated 12-month (post-surgery) between-group difference in health state utility over an extended time horizon (Figure 12). However, it is anticipated that it will be those women who have a poorer outcome at 12 months following EA who return in the future for a hysterectomy. Thus, the average utility score among those remaining in the post-EA state would be expected to rise over time. Conversely, it is possible that the recurrence of symptoms following EA has yet to peak and that the utility curves will continue to diverge further beyond 12 months before beginning to converge. It is also uncertain to what extent the post-EA and post-LASH utility curves may converge, and how quickly. In our base-case analysis, we adopted a conservative approach (in favour of EA) and assumed that the mean difference in utility between the post-EA and post-LASH health states would diminish over time in proportion to the total expected number of post-EA hysterectomies completed. For example, from a baseline of 12 months post surgery, 50% of the further post-EA hysterectomies will have been completed after a further 3 years, by which time the extrapolated utility difference is approximately half of that observed at 12 months. By 10 years in the model, the utility difference is assumed to have diminished to zero. We explored the impact of applying alternative extrapolation assumptions in sensitivity analysis, including retention of the full 12 months post-surgery utility difference over the entire duration of the model and allowing it to diminish over a shorter time.

FIGURE 12.

Observed and extrapolated health state utility post LASH and post EA.

Health state utilities associated with further surgery

Regarding the heath state utility of those transitioning to hysterectomy following EA, similar assumptions to those used in the HTA by Bhattacharya et al. 10 were applied. Women were assigned the baseline utility value of 0.702 in the cycle preceding hysterectomy. Following that, a utility value appropriate to the post-hysterectomy state was applied. For severe postoperative complications following hysterectomy, a utility value of 0.49 was sourced from Bhattacharya et al. ,10 originally from Clegg et al. ,60 and applied for a single cycle. For the convalescence period following hysterectomy (3 months), a utility value of 0.74 was applied. 10,61 Finally, it was assumed that health state utility in the ‘well post hysterectomy’ health state is equivalent to that in the ‘well post LASH’ health state. Table 21 provides a summary of health state utility inputs applied for the further surgery states.

TABLE 21 - Health state utility values applied for further surgery states

SEM, standard error of the mean.

Variable	Mean	SEM	Distributional form	Source
Symptomatic requiring further surgery	0.702	0.012	Beta	HEALTH
Severe postoperative complication following hysterectomy or removal of cervical stump	0.49	0.049	Beta	Clegg et al.60
Convalescence post hysterectomy or removal of cervical stump	0.74	0.05	Beta	Sculpher61
Well post hysterectomy	0.827		Beta	Assumption (see Health state utilities associated with further surgery)

For women requiring cervical stump removal following LASH, the same utility assumptions pertaining to total hysterectomy were applied. For post-LASH laparoscopy or BSO, we applied baseline utility in the cycle preceding treatment and the cycle of treatment, but then assumed a return to the ‘well post LASH’ state. Thus, overall, the model base case assumes that, after 10 years, women ultimately end up at the same level of health state utility, although the pathway to that outcome varies.

Model validation

To assess the internal validity of the model we compared the 12-month model-based cost and QALY estimates with the 15-month trial-based estimates. These data are provided in Table 23. Although the expected QALYs were lower in both arms in the model, the estimated incremental difference was similarly very small. The lower average QALY estimates are accounted for by the simplifying assumption of omitting the pre-treatment waiting period and running the model for 12 months rather than 15 months. The mean treatment arm costs were also very similar between the model and trial-based analyses, as was the incremental cost. The small differences are attributable to the fact that the model was populated using the available complete data for each input parameter, rather than the multiple imputation data set used in the trial-based analysis.

TABLE 23 - Trial- and model-based estimates of mean costs and effects, at 15 months post randomisation and 12 months post surgery, respectively

Variable	EA	LASH	Difference
Trial-based estimates
Mean cost (£)	1282	2886	1604
Mean QALY	0.974	0.978	0.004
Model-based 12-month estimates
Mean cost (£)	1277	2890	1612
Mean QALY	0.8119	0.8140	0.0021

To assess the validity of projected rates of further surgery beyond 12 months, the modelled cumulative incidence following EA and LASH was plotted over the 10-year time horizon (Figure 13). The model projected that cumulative incidence of hysterectomy following EA was 3.3%, 6.8% and 12.8% after 1, 2 and 5 years, respectively. This compares to estimates from a large population-based English study of 5.6%, 9.6% and 13.5%, respectively, at these corresponding time points. 46 The incidence was lower in the model because the lower 12-month rate from HEALTH was applied in year 1. There were few data against which to validate the projections for further surgery following LASH. As previously discussed under Extrapolation of subsequent surgery following laparoscopic supracervical hysterectomy, the projected estimates are in line with the incidence observed at 12–36 months in a Norwegian cohort,17 and reach ≈11% by 10 years post surgery when combined with the extrapolation assumptions.

FIGURE 13.

Modelled incidence of further surgery following EA and LASH.

Results

Base-case analysis

Table 24 presents the results of the base-case analysis. Over the modelled 10-year time horizon, intention to treat with LASH resulted in an increased cost to the health service of £1362 per woman, for an expected QALY gain of 0.111 per woman, compared with EA. The corresponding ICER was £12,314 per QALY gained for LASH compared with EA. The chance of LASH being cost-effective ranged from 53% to 80% at WTP per QALY thresholds of £13,000 and £30,000, respectively. It can be noted that extending the time horizon of the evaluation from 1 year to 10 years reduced the incremental cost of LASH by £250 (£1362 at 10 years vs. £1612 at 1 year). This is due to the incorporation of expected costs of further surgery. The QALY gain associated with LASH resulted primarily from extrapolation of the estimated difference in 12-month post-EA and post-LASH health state utility. Further temporary reductions in utility associated with further surgery accounted for only a very small proportion of the incremental QALY.

TABLE 24 - Base-case model results

Strategy	Cost (£)	Δ cost (£)	QALYs	Δ QALYs	ICER (£)	Probability cost-effective
						£13,000	£20,000	£30,000
EA	2089		6.938			0.468	0.291	0.201
LASH	3452	1362	7.049	0.111	12,314	0.532	0.709	0.799

Figures 14 and 15 further illustrate the uncertainty surrounding the estimated incremental costs and effects of LASH compared with EA. LASH remained the significantly more costly strategy based on extrapolation to 10 years, with all the estimated points being above zero on the y-axis of Figure 14. Most points (≈93%) also lie to the right of zero on the x-axis, indicating a 93% chance that LASH will generate more QALYs based on the simulation. The cost-effectiveness acceptability curve for LASH (see Figure 15) therefore asymptotes to 93%, as the threshold of WTP per QALY increases towards infinity. This is analogous to finding a p-value (one-sided) of 0.07 for an estimated difference in QALYs favouring LASH.

FIGURE 14.

Incremental cost-effectiveness scatterplot (LASH vs. EA).

FIGURE 15.

Cost-effectiveness acceptability curve (LASH vs. EA).

Discussion

The modelling exercise reported in this chapter indicates that intention to treat with LASH has the higher probability of being cost-effective in the longer term across a range of plausible values of WTP per QALY gained. The key sources of uncertainty in the model relate to the extrapolation of the observed difference in health state utility between the post-EA and post-LASH health states, and the impact that the incidence of further surgery will have on this parameter. The ICER for LASH ranged from as low as £5377, when the observed 15 months post-randomisation (12 months post surgery) utility difference was maintained over the entire time horizon of the model, to £21,992 when it was assumed to diminish to zero by 5 years post surgery. The further collection of patient-reported outcome data and the need for subsequent surgery at 3–5 years would help to reduce this uncertainty.

A strength of the modelling is its basis on randomised data from a large multicentre and pragmatic RCT. HEALTH utilised a detailed approach to costing the initial surgical episode and captured subsequent complications out to 15 months post randomisation (≈12 months post surgery). However, the censoring of patients for longer-term post-surgery health state utility is a limitation based on the trial design. Had a larger percentage of the cohort reached a minimum of 12 months post surgery by 15 months post randomisation, we may have observed a more marked difference in health state utility between the randomisation groups.

A further limitation relating to the duration of HEALTH was the need to rely on external non-randomised data to inform the expected longer-term incidence of further surgery by treatment allocation group. Although good data were available on the incidence of hysterectomy following EA, limited published data were available to inform the incidence of subsequent surgery post LASH. Therefore, conservative assumptions (favouring EA) were applied in the base-case analysis, with further surgery following LASH modelled to reach ≈11% by 10 years. This is substantially higher than the long-term incidence of further gynaecological surgery following total hysterectomy, reported to be 4% at a median duration of 11 years, based on Scottish population data. 5

As indicated above, the 15-month post-randomisation follow-up also necessitated the application of assumptions to extrapolate observed differences in health state utility over time. Again, conservative assumptions were applied, in that the observed 15-month (post-randomisation) difference in health state utility between the post-EA and post-LASH health states was modelled to diminish to zero by 10 years. This may underestimate the QALY gain for LASH compared with EA if the observed difference in utility at 15 months were to diverge further before converging, and/or converge less quickly over time. Nevertheless, the QALY gain for LASH compared with EA remains uncertain as it accrued primarily during the extrapolation phase of the model. A further uncertainty relates to the fact that the difference in health state utility observed at 15 months post randomisation (≈12 months post surgery) was not statistically significant. This uncertainty was propagated through the model and was reflected in the probabilistic model output. Furthermore, the direction of the effect observed for the EQ-5D-3L score at 12 months was consistent with the observed significant differences in the primary clinical outcomes.

Currently, no published studies have assessed the cost-effectiveness of LASH compared with EA, but several studies have assessed the cost-effectiveness of total hysterectomy compared with EA from a UK NHS perspective. These have generally found that hysterectomy is likely to be cost-effective. 9,10,43 These previous studies utilised a similar model structure to the one described in this chapter, but made less conservative assumptions regarding the extrapolation of the estimated difference in post-EA and post-hysterectomy utility. Thus, the above-mentioned studies reported a larger QALY gain for total hysterectomy compared with EA than we report here for LASH compared with EA. Nevertheless, when applying a cost-effectiveness threshold ratio of £20,000 per QALY gained, our results indicate a relatively high probability that LASH offers a cost-effective alternative to EA for women with HMB. It may also prove a more acceptable alternative to total hysterectomy in terms of risks of AEs. If applying the more recently suggested cost-effectiveness threshold of £13,000 per QALY,63 there is greater uncertainty surrounding the cost-effectiveness of LASH. It will therefore be important to assess the longer-term risks of subsequent surgery following LASH and to formalise comparisons with total hysterectomy through indirect treatment comparisons and decision modelling. The first of these issues will be addressed through extended follow-up of the HEALTH cohort.

In conclusion, the cost-effectiveness analysis based on HEALTH participant data indicates that LASH cannot be considered cost-effective by 15 months post randomisation (≈12 months post surgery). However, based on extrapolation over a more relevant time horizon, there is a relatively high probability that LASH offers a cost-effective alternative to EA at thresholds of WTP per QALY gained typically applied in the UK NHS (£20,000–30,000). Longer-term follow-up of HEALTH participants will be beneficial for reducing the current decision uncertainty.

Aim and overview

Heavy menstrual bleeding is a common condition that affects many women of reproductive age and significantly impairs their QoL. For those for whom medical treatment is ineffective, NICE recommends either EA or hysterectomy. 43 In comparison with EA, total hysterectomy is a more definitive procedure but is associated with higher surgical morbidity and slower postoperative recovery. 8 Two small randomised trials have suggested that a less invasive form of hysterectomy, LASH, is superior to EA but with similar morbidity and recovery. 12,13 Our large, pragmatic, randomised trial was designed to determine whether or not LASH was more effective than EA without incurring additional risks or recovery time.

Our primary clinical outcomes were satisfaction and condition-specific QoL (MMAS). Additional secondary clinical outcomes were collected along with economic end points.

Summary of findings

Strengths and limitations of the trial

To our knowledge, this was the first large randomised trial comparing LASH with EA and the first study to undertake a cost-effectiveness analysis. The pragmatic nature of the trial and inclusion of centres across the UK has generated results that are robust, reliable and generalisable.

The main weakness of our trial was a higher loss to follow-up than anticipated. As soon as we became aware of this problem, we implemented a number of strategies, including cash incentives (an unconditional £25 gift voucher issued with the first reminder) and telephone follow-up from the trial office, which enabled us to increase our response rates from 74% to 85% for the primary outcome questionnaire.

As a surgical trial recruiting within NHS hospitals with variable waiting lists, we were unable to enforce a standard time between randomisation and surgery. In view of the NHS waiting list target of 12 weeks for routine operations, we chose to collect primary outcome data at 15 months post randomisation, believing that this would give us a 12-month duration of follow-up. Although this was appropriate for most women, a minority (10%) faced delays in receiving treatment. This occurred as a result of patient unavailability through either illness or preference, theatre staff shortages in some centres and seasonal pressures on elective hospital admission. However, these were balanced across the groups, although those allocated to LASH did have to wait slightly longer for their procedures than those allocated to EA.

A strength of the economic modelling based on randomised data from a large trial was that it enabled a detailed approach to costing of the initial surgical episode and the capture of subsequent events, including complications. However, the censoring of patients for subsequent surgery and longer-term post-surgery health state utility is a limitation, as is the need to rely on external non-randomised data to inform the longer-term risk of further surgery. Although data on the incidence of hysterectomy following EA are relatively robust, this is not currently the case for the chance of subsequent surgery post LASH.

The 15-month time horizon post randomisation also required us to make assumptions about differences in health state utility over time. Again, conservative estimates were applied, in that the observed 15-month (post-randomisation) difference in health state utility between the post-EA and post-LASH health states was modelled to diminish to zero by 10 years, which may underestimate the QALY gain for LASH compared with EA if the observed difference in utility at 15 months were to diverge further before converging, and/or converge less quickly. Future longer-term follow-up of the HEALTH cohort would help to reduce the current uncertainties in the economic model. As is most commonly practised in the UK NHS, EA procedures in HEALTH were carried out in theatre as admitted patient care, the majority (95%) under general anaesthetic. Scope may exist for cost-effectiveness to move in favour of EA if it were to prove widely acceptable to deliver in an ambulatory outpatient setting in the NHS.

To our knowledge, no published studies assess the cost-effectiveness of LASH compared with EA, but several studies9,10,43 have assessed the cost-effectiveness of total hysterectomy compared with EA. These have generally found that hysterectomy is likely to be cost-effective from a UK NHS perspective. 9,10,43

Interpretation in the context of available literature

At 15 months post randomisation (approximately 12 months post surgery), our results show high levels of satisfaction in both groups, which are similar to those reported in an individual participant data meta-analysis of total hysterectomy compared with EA. 8 The key difference between the two is that total hysterectomy, but not LASH, guarantees amenorrhoea. In our trial, both types of surgery resulted in high rates of patient satisfaction, but women had 2.5 times higher odds of being in a more favourable satisfaction category following LASH without any appreciable increase in surgical or postoperative risk.

Both LASH and EA resulted in significant improvements in MMAS scores from baseline that are higher than those reported in a previous trial evaluating medical treatment for HMB. 64 Women randomised to LASH had almost twice the odds of being in a more favourable MMAS score category, and a higher proportion of women (69% vs. 54%) reported the maximum score of 100.

The amenorrhoea rate of 60% achieved in the EA arm of our trial is higher than reported in previous randomised trials, in which rates of around 40% are commonly described. 65 However, those women allocated to LASH reported an amenorrhoea rate of 80%, which is lower than the 90–95% rates quoted by a Cochrane review16 but consistent with results reported by Lieng et al. 17

Women who underwent EA spent half as much time in theatre as those who underwent LASH, required significantly less postoperative analgesia and were discharged after a median of 3 hours post procedure (compared with 22 hours following LASH). Surgical morbidity and postoperative complications were comparable in both groups.

Fourteen women in the EA arm were readmitted for hysterectomy within 15 months after randomisation. Long-term hysterectomy rates of almost 20% following EA are described in population studies. 5,49 Rates for further surgery after LASH in the longer run are less well known as follow-up of only 3 years has been reported and rates are quoted as 7% at this point. 17 An English population-based study showed that 10% of women aged > 45 years have further surgery 5 years following EA, compared with about one-third of women aged < 35 years. 46 We found no such association with age in this trial. There were poorer outcomes following EA in women with fibroids, although more fibroids were diagnosed in the LASH arm because of laparoscopy. There was no association with cavity length, which is at odds with a meta-analysis of trials on second-generation EA, which found that a uterine cavity length of > 8 cm had an adverse impact on patient satisfaction. 8

A Cochrane review of total versus subtotal hysterectomy suggests a cyclical bleeding rate of 5–10% following LASH,16 whereas a higher rate of 23% was reported by Lieng et al. 17 Despite a standardised procedure of cervical canal cautery after removal of the body of the uterus, the observed rate of cyclical bleeding in the LASH group was 19%. This is in some way explained by the fact that, of 330 women allocated to the LASH group, 21 women had no treatment and 11 women had EA. Even so, it seems that our rates of post-LASH cyclical bleeding lie between those quoted by the Cochrane database16 and the longer-term rates quoted by Lieng et al. 17

The potential risk of malignancy is always an important consideration in planning conservative surgical treatments for HMB. In EA, in which the uterus, tubes and ovaries are conserved, an endometrial biopsy is undertaken before surgery to exclude endometrial atypia. As the cervix is retained in both techniques, ongoing cervical screening is required in all women. The risk of cervical stump carcinoma in women with a previously normal Pap smear is no more than 0.3%66 and is not considered to be a justification for total over subtotal hysterectomy in countries with cervical screening programmes,16 especially where immunisation against human papillomavirus (HPV) virus is the norm.

At the time when this trial was launched, concerns were raised by the US Food and Drug Administration about the potential risk of disseminating cells from undiagnosed uterine malignancy associated with the use of morcellators during laparoscopic hysterectomy in women with fibroids. 67 This prompted us to modify our eligibility criteria to exclude women with fibroids measuring > 3 cm, as the risk of malignancy is associated with fibroid size. The results from our trial are reassuring, as specimens from the 308 women who underwent hysterectomy have not shown any evidence of histological atypia or malignancy. The initial estimate of risk of 1 in 350 for unexpected/unknown malignancy within a presumed fibroid by the Food and Drug Administration has been revised, and a recent review puts this risk figure closer to 1 in 2000. 68 Updated estimates of leiomyosarcoma (LMS) rates at hysterectomy have been published, and show that women aged > 50 years with larger presumed fibroids are at particular risk. These are exclusion criteria for participation in HEALTH. 69 Although we have been reassured by the absence of any histological abnormalities in this trial, continued vigilance is required, and morcellation avoided and total hysterectomy performed if there are concerns of possible sarcomatous change within a fibroid based on morphological appearance or rapid growth. Dissemination of morcellated material can be avoided by using containment bags, although these have not been formally assessed. An alternative is to remove unmorcellated specimens in a bag through a culdotomy, but this may be associated with different morbidities and also needs to be assessed. It must be remembered that if a woman with unknown LMS has conservative treatment of her HMB, such as EA, she continues with an unknown and untreated LMS. It is worth noting that the routine removal of fallopian tubes during LASH could halve the subsequent risk of epithelial ovarian cancer,70 without any increase in surgical risk.

Conclusions

Laparoscopic supracervical hysterectomy is superior to EA in terms of clinical effectiveness. EA is quicker, cheaper and associated with an earlier discharge and shorter recovery than LASH. It is less costly in the short term, but its higher failure rate means that LASH is more likely to be considered more cost-effective than EA by 10 years post procedure.

Project Management Group

Siladitya Bhattacharya (co-chairperson), Kevin Cooper (co-chairperson), Suzanne Breeman, Rebecca Bruce, Justin Clark, Jed Hawe, Robert Hawthorn, Rodolfo Hernández, Graeme MacLennan, Kirsty McCormack, Alison McDonald, John Norrie, Mini Paulose, Kevin Phillips, Danielle Pirie, Graham Scotland, Neil Scott and Samantha Wileman.

Independent members of the Trial Steering Committee

Henry Kitchener (chairperson), Patrick Chien, Barbara Farrell and Isobel Montgomery.

Independent members of the Data Monitoring Committee

Jane Norman (chairperson), Peter O’Donovan and Andy Vail.

Recruitment sites

We would like to thank the staff and participants of the following sites:

Aberdeen: Kevin Cooper (PI), Premila Ashok, Stamatios Karavalos, Mini Paulose, Danielle Pirie, Lucky Saraswat, Sherif Saleh, Ruth Valentine and Sarah Wallage; Basildon: Yatin Thakur (PI), Kellie Allen, Preethi Angala, Yee Yin Chan, Mamdouh Ghobrial, Neerja Gupta, J Jadhav, Elisabeth Joseph, Syathe Kalburg, Shaneen Mannan, Claire McCormack, Anne Nicholson, Stacey Pepper, Emily Redman and Donna Southam; Basingstoke: Christian Phillips (PI), Jackie Moody, Clare Rowe Jones, Tim Sayer; Birmingham: T Justin Clark (PI), Yousri Afiffi, Fiona Beale, Andrea Galloway, Janesh Gupta, Virginia Iqbal, Shanteela McCooty, Paul Smith, Helen Stevenson and Jennifer Taylor; Cornwall: Dominic Byrne (PI), Benita Adams, Alyson Andrew, Susan Bates, Samy Bishieri, Jacqueline Dingle, Anna Fouracres, Fiona Hammonds, Rob Holmes, Farah Lone, Jonathan Lord, Tom Smith-Walker, Jessica Summer, Lisa Trembath, Liz Verity and Abigail Weeks; Chester: Jed Hawe (PI), Jonathan Ford, Elizabeth Gallimore, Nabil Haddad, Helen Jefferey, Nichola Kearsley, Sandra Leason, Michael J McCormack, Janneke Van Rij and Simon Wood; Durham: Partha Sengupta (PI), Vicki Atikinson, Remko Beukenholdt, Jonathan Bradley, Jean Dent, Jacqueline Jennings, Seema Sen, Velur Sindhu and Eileen Walton; Edinburgh: Stuart Jack (PI), Hilary Critchley, Jennifer Devlin, Paul Dewart, Catherine Hall, Andrew Horne and Lucy Whitaker; Fife: Chu Lim (PI), Julie Aitken, Laura Beveridge, Keith Boath, Isla Smith and Omar Thanoon; Forth Valley: Oliver Milling-Smith (PI), Jacqueline Gardiner, Shahzya Huda and Kara Sewnauth; Glasgow: Robert Hawthorn (PI), Hassan Ali, Carol Archibald, Steingrimur Bjornsson, Chris Hardwick, Therese McSorley, Lorna McKay and Kirsteen Paterson; Hull: Kevin Phillips (PI), Jane Allen, Frank Biervliet, Helen Bexhell, Stephen Boakye, Helen Brown, Adrian Fauvre, Melony Hayes, Jaishree Hingaroni, Piotr Lesny, Kim Mitchelson, Androniks Mumdzjans and Alexander Oboh; Kent: Mr Chappatte (PI), Elias Kovoor and Tracey Nolan; Kilmarnock: Inna Sokolova (PI), Danielle Gilmour and Margo Henry; Newcastle: Tony Chalhoub (PI), Sue Harbertson, Alison Kimber, Kayleigh Lennox, Victoria Murtha, Mark Roberts and Michelle Russell; Northampton: Clemens von Widekind (PI), Lucy Dudgeon, Kathy Hall, Rachael Hitchcock and Gillian Smith; North Tees: Somon Roy (PI), Patricia Bage, Misra Bano-Mahroo, Dolan Basu, Sarah Croft, Sharon Gowans, Anne Marie Jones, Susan Kelsey, Hany Mostafa, Sue Mullens, Atul Nalawade, Santhosh Puthuraya, Alison Samuels, Rebecca Tate, Ratan Wadia and Maggie Wilkinson; Poole: Tyrone Carpenter (PI), Anne Chalk and Daniel Webster; Plymouth: Robert Freeman (PI March 2017–present), Jonathan Frappell (PI October 2014–March 2017), Sarah Caukwell, Heidi Hollands and Abigail Patrick; Preston: Brice Rodriguez (PI), Sarah Francis, Anne Gardner, Annabel Grossmith, Sean Hughes, Catherine Langley, Lauren Murphy, Claire Noor, Katie Robinson, Katy Sanders, Mark Tattersall, William Wilson-Theaker, Robyn Wittersheim and Katey Yeowart; Sheffield: Ted Baxter (PI), Sheila Duffy; Stockport: Tarique Salman (PI August 2015–September 2018), Andrew Pickersgill (PI June 2014–August 2015), Lindsay Barber, Helen Cochrane and Julie Grindey; Southampton: Dimitrios Miligkos (PI), Agnieszka Burtt, Jane Forbes, Abby Rand, Sameer Umranikar, Fiona Walbridge and Susan Wellstead; South Tees: Pinky Khatri (PI), Hazel Alexander, Sarah Croft, Helen Cuthbert, Neil Hebblethwaite, Aethele Khunda, Padma Manda, Dayang Mohammed, Suzie Peatman, Julie Potts and Sanjay Rao; Sunderland: Nick Matthews (PI February 2016–September 2018), Jonathan Chamberlain (PI September 2014–February 2016), Amna Ahmed, Helen Cameron, Kirsten Herdman, Kim Hinshaw, Princy Mathew, Judith Ormonde, Janet Scollen, Aarti Ullal, Eileen Walton and Menem Yossry; Surrey: Saikat Banerjee (PI), Claire Atkinson, Lauren Brown, Michelle Davis, Rebecca Gilbert, Shaheen Khazali, Beth Peers and Lisa Sharpe; Whipps Cross: Funlayo Odejinmi (PI), Bashir Dawlatly, Anwen Gorry, Prudence Jones, Zandile Maseko, Anupama Shahid and Sotitis Vimplis; Winchester: Keith Louden (PI), Renee Behrens, Jane Martin, Dawn Trodd and Caroline Wrey Brown; Wirral: Thomas Aust (PI), Ash Alam, Khadija Ashraf, Mark Doyle, Michael Ellard, Julie Grindey, Stella Mwenechanya, Libby Shaw, Gillian Steele, Vasileios Minas and Jeremy Weetch; Worcester: Angus Thomson (PI), Victoria Cashmore, John Hughes, Dawn Kelly and Mamta Pathak; Worthing: Natasha Waters (PI), Vivienne Cannons, Marian Flynn-Batham, Sarah Funnell, Fani Gkrozou and Sarah House.

Contributions of authors

Kevin Cooper (Co-Chief Investigator and Professor) contributed to the conception and the design of the trial, the conduct of the trial, the interpretation of the results and the writing/editing of the report.

Suzanne Breeman (Trial Manager, Triallist) was responsible for the day-to-day management of the trial, contributed to the interpretation of the data and the writing/editing of the report.

Neil W Scott (Statistician) contributed to the design of the study, conducted the statistical analyses, contributed to the interpretation of the data and the writing/editing of the report.

Graham Scotland (Health Economist) contributed to the conception and design of the study, the analysis of the health economics data, the drafting of the health economics chapters, plus commented on other chapters of the report.

Rodolfo Hernández (Health Economist) contributed to the analysis of the health economics data, the drafting of the health economics chapters, plus commented on other chapters of the report.

T Justin Clark (Consultant Obstetrician and Gynaecologist) contributed to the conception and design of the study, the recruitment of participants, the interpretation of the data and made significant contributions to the drafting of the report.

Jed Hawe (Consultant Obstetrician and Gynaecologist) contributed to the conception and design of the study, the recruitment of participants, the interpretation of the data and made significant contributions to the drafting of the report.

Robert Hawthorn (Consultant Gynaecologist) contributed to the conception and design of the study, the recruitment of participants, the interpretation of the data and made significant contributions to the drafting of the report.

Kevin Phillips (Consultant Obstetrician and Gynaecologist) contributed to the conception and design of the study, the recruitment of participants, the interpretation of the data and made significant contributions to the drafting of the report.

Samantha Wileman (Triallist) contributed to the design of the trial, the conduct of the trial and made significant contributions to the drafting of the report.

Kirsty McCormack (Triallist) contributed to the conception and design of the study, the conduct of the trial and made significant contributions to the drafting of the report.

John Norrie (Professor, Statistician and Triallist) contributed to the conception and design of the trial, the conduct of the trial, the interpretation of results and made significant contributions to the drafting of the report.

Siladitya Bhattacharya (Co-Chief Investigator and Professor) contributed to the conception and the design of the trial, the conduct of the trial, the interpretation of the results and the writing/editing of the report.

The Health Services Research Unit and the Health Economics Research Unit are core funded by the Chief Scientist Office of the Scottish Government Health and Social Care Directorates.

Publications

Cooper K, McCormack K, Breeman S, Wood J, Scott NW, Clark J, et al. HEALTH: laparoscopic supracervical hysterectomy versus second-generation endometrial ablation for the treatment of heavy menstrual bleeding: study protocol for a randomised controlled trial. Trials 2018;19:63.

Cooper K, Breeman S, Scott NW, Scotland G, Clark J, Hawe J, et al. Laparoscopic supracervical hysterectomy versus endometrial ablation for women with heavy menstrual bleeding (HEALTH): a parallel-group, open-label, randomised controlled trial [published online ahead of print September 12 2019]. Lancet 2019.

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 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, NETSCC, 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, NETSCC, the HTA programme or the Department of Health and Social Care.

Within-trial economic evaluation: further details on the statistical analysis