Hysterectomy, endometrial ablation and Mirena® for heavy menstrual bleeding: a systematic review of clinical effectiveness and costeffectiveness analysis

Incidence of surgical operations for heavy menstrual bleeding

Of 51,858 hysterectomies in the public sector in England in 1999–2000, it is likely that half were for HMB. 29 Between 1999–2000 and 2004–5, 6500 fewer hysterectomies were performed. 30 In contrast there were 826 hysteroscopic EAs in England in 1989, rising to 7173 in 1992–3, before falling to 3847 in 1997–8. In 2004–5, 9701 EAs were performed, of which over half (5457) used second-generation (non-hysteroscopic directed) methods. 30 With just 7179 hysterectomies performed for HMB over this period, the predominant operation for HMB was now ablation. The use of LNG IUS has increased concurrently, although the widespread use of this device for contraception as well as for the control of HMB across a number of clinical settings (primary care, sexual reproductive health as well as gynaecology clinics in secondary care) makes it difficult to gather accurate data on uptake rates.

Hysterectomy

Hysterectomy is defined as the surgical removal of the uterus. It offers a definitive treatment for menorrhagia and guarantees amenorrhoea, but is particularly invasive and carries risk of significant morbidity. 31 Hysterectomy can be performed through a number of routes. In abdominal hysterectomy the uterus is approached through the anterior abdominal wall, while vaginal hysterectomy involves surgical removal of the uterus through the vagina. In laparoscopic hysterectomy surgery is accomplished without the need for a laparotomy. Laparoscopic hysterectomy includes three subtypes: (1) laparoscopically assisted vaginal hysterectomy; (2) laparoscopic hysterectomy; and (3) total laparoscopic hysterectomy. In addition, laparoscopically assisted supracervical hysterectomy involves removal of the body of the uterus while the cervix is retained.

Hysterectomy can also be categorised on the basis of the extent of the operation and organs removed. Removal of the uterus and cervix constitutes total hysterectomy, while excision of the body of the uterus while conserving the cervix is defined as subtotal hysterectomy. Removal of the uterus alone is conventionally known as simple hysterectomy, while additional removal of the fallopian tubes and ovaries or ovaries alone is referred to as salpingo-oophorectomy or oophorectomy, respectively. Oophorectomy is usually performed in the presence of ovarian pathology but can also be carried out prophylactically to avoid the risk of cancer. Removal of the ovaries in cases of HMB is incidental. 15 Of 37,000 hysterectomies performed in the UK in 1994–5, two-thirds were abdominal (4% of these were subtotal hysterectomies) and 57% were accompanied by removal of tubes and ovaries. 5

Hysterectomy is generally performed as an inpatient procedure. The need for general anaesthesia, prolonged hospital stay and delayed recovery makes it a potentially expensive treatment. 32 Overall, 1 in 30 women suffers a major adverse event after hysterectomy, and the mortality rate is 0.4–1.1 per 1000 operations. Around 3% of women suffer from perioperative problems such as haemorrhage, trauma to other pelvic organs and anaesthetic problems. Immediate postoperative complications include sepsis, bleeding and venous thromboembolism. Adverse events following hysterectomy are summarised in Table 2. Although delayed complications including urinary incontinence, fatigue, pelvic pain, hot flushes and sexual problems have been reported,5,33–35 satisfaction rates following hysterectomy are very high. 31

Endometrial ablation

Endometrial ablative techniques, which aim to destroy functionally active endometrium along with some underlying myometrium,36,37 offer a conservative surgical alternative to hysterectomy. The first-generation ablative techniques including endometrial laser ablation (ELA),38,39 TCRE40 and rollerball endometrial ablation (RBEA) were all endoscopic procedures. Although none guarantees amenorrhoea, their effectiveness (in comparison with hysterectomy) has been demonstrated in a number of RCTs. 41–46

National audits47,48 revealed that, although first-generation ablative techniques were less morbid than hysterectomy, they were associated with a number of complications, including uterine perforation, cervical laceration, false passage creation, haemorrhage, sepsis and bowel injury. In addition, fluid overload associated with the use of 1.5% urological glycine (non-ionic) irrigation fluid in TCRE and RBEA resulted in serious and occasionally fatal consequences due to hyponatraemia. 49,50 Mortality from these techniques has been estimated at 0.26 per 1000. 47,48

Second-generation ablative techniques represent simpler, quicker and potentially more efficient means of treating menorrhagia, which require less skill on the part of the operator. Examples of second-generation ablative techniques are fluid-filled thermal balloon EA (TBEA), radiofrequency (thermoregulated) balloon endometrial ablation, hydrothermal EA, 3D bipolar radiofrequency EA, microwave endometrial ablation (MEA), diode laser hyperthermy, cryoablation and photodynamic therapy. The most common techniques in the UK are TBEA (ThermaChoice^®, Ethicon, Livingston, UK and Cavaterm™, Pnn Medical SA, Morges, Switzerland)51–53 and MEA,54,55 while the NovaSure^® (Hologic Inc., Bedford, MA, USA) device56 is becoming more widely used.

Second-generation endometrial ablation techniques

Since the 1990s, several new methods of EA have been developed. These are often referred to as second-generation techniques. They do not require direct visualisation of the uterine cavity and employ a variety of means to destroy the endometrium – circulation of heated saline within the uterine cavity, use of a diode laser [endometrial laser intrauterine thermo-therapy (ELITT)], punctual vaporising methods, photodynamic methods, radiofrequency, microwaves, a balloon catheter filled with heated fluid and cryotherapy. The treatments are much less dependent on the skill of the surgeon than first-generation techniques, and much more dependent on the reliability of the machines used to ensure safety and efficacy. Complications associated with second-generation techniques include equipment failure, uterine infection, perforation, visceral burn, bleeding and cyclical pain. A limited number of randomised trials indicate that these procedures appear to be as effective as first-generation ablative techniques. 61 In addition, some have the added benefit of being performed under local anaesthetic.

Microwave endometrial ablation

Microwave EA uses microwave energy (at a frequency of 9.2 GHz) to destroy the endometrium with a tissue penetration depth of  6 mm. An 8-mm applicator inserted through the cervix delivers the microwaves using a dielectrically loaded waveguide. 62 Power is controlled by the surgeon using a footswitch and the temperature inside the uterus is monitored by thermocouples on the surface of the waveguide. Prior to microwave ablation treatment, oral and vaginal thinning agents may be given. Immediately prior to MEA, hysteroscopy is performed to exclude false passages, wall damage and perforation.

Following measurement of uterine cavity length, the cervix is dilated to Hegar 8 or 9 under general or local anaesthetic and the uterine cavity length is measured again. Next, the microwave probe is inserted until the tip reaches the fundus. Graduated centimetre markings on the applicator shaft confirm the length and if these three measurements of uterine length are the same the device is activated. 63 When, after a few seconds, the temperature reaches 80 °C, the probe is moved laterally so that the tip is placed in one of the uterine cornua. The temperature briefly falls and rises again and when 80 °C is reached again the probe is moved to the other cornual region and the procedure is repeated.

Maintaining a temperature of 70–90 °C, the probe is withdrawn with side-to-side movements. The temperature measured by the thermocouple is actually the heat transmitted back from the tissue through the plastic sheath to the applicator shaft. Tissue temperature is higher than these measured levels during active treatment. As a marker on the probe appears at the external os, the applicator is switched off to avoid treating the endocervix. The procedure takes 2–3 minutes. 62 Postoperative analgesia is provided as required.

Thermal balloon endometrial ablation

Thermal balloon EA aims to destroy the endometrium by means of heated liquid within a balloon inserted into the uterine cavity, which should be of normal size and regular shape. Available devices such as ThermaChoice and Cavaterm have an electronic controller, a single-use latex or silicone balloon catheter with a heating element, thermocouples and an umbilical cable. As the balloon must be in direct contact with the uterine wall, the device is unsuitable for women with large or irregular uterine cavities.

In the case of the ThermaChoice device, following dilatation of the cervix to about 5 mm, the balloon is introduced within the uterus and filled with sterile fluid (5% dextrose in water) which causes it to expand to fit the cavity. Once intrauterine pressure is stabilised to 160–180 mmHg, the temperature of the fluid is raised to 87°C and maintained for 8 minutes. Pressure, temperature and time are continuously monitored and the device is switched off if safety parameters are breached. The heat produced by the device causes destruction of the endometrium. Postoperatively, oral analgesia is prescribed and the treated endometrium sloughs off over the following week to 10 days.

The Cavaterm device acts in a similar fashion. The cervix is dilated to about 6 mm and a silicone balloon is inserted and filled with sterile 5% glucose solution to a pressure of 230–240 mmHg. The liquid is heated at a target temperature of 78°C for 10 minutes.

Endometrial thinning agents are not recommended although curettage immediately prior to the procedure can be used. NSAIDs are given to reduce perioperative cramping.

Impedance-controlled bipolar radiofrequency (NovaSure)

The NovaSure system consists of a single-use bipolar ablation device which is inserted into the uterine cavity transcervically after dilatation to 8 mm. This is connected to a generator which functions at 500 kHz and has a power cut-off limit set at a tissue impedance of 50 Ω. The cavity and cervical length are measured and the difference in centimetres is determined; this setting is selected on the shaft of the device. The device is inserted and the trigger is deployed which delivers the bipolar triangular array into the cavity. With gentle tapping and slight rotation in both directions the array fully deploys with the tips sited in each cornua. The distance between the cornuae is displayed and then entered into the generator, and this determines the energy required. A cavity integrity test is then automatically performed, which must be passed before the energy is delivered. Active treatment times are under 2 minutes, during which time suction pulls the walls onto the device. After completion the array is retracted into the device sheath and withdrawn. While the device is versatile, it cannot effectively treat larger cavities (> 11 cm) or distorted cavities. Pre-treatment endometrial thinning is not required and the procedure can be performed under local anaesthetic. A number of randomised trials has been undertaken, one comparing with RBEA56 and the others with thermal balloon devices. 64,65 One of the trials has published follow-up to 5 years. 66 A randomised trial comparing NovaSure with microwave ablation has been completed and awaits publication. It has approval from NICE. 15 Results have been consistent through the trials, with amenorrhoea rates varying between 42% and 56%, high satisfaction rates of over 90% and low hysterectomy rates. Active treatment times vary between 90 and 120 seconds.

Adverse effects associated with second-generation EA devices include the following:16 uterine infection, perforation, visceral burn, bleeding, haematometra, laceration, intra-abdominal injury and cyclical pain.

Use of local anaesthetic

Use of local anaesthetic (LA) is a potential advantage of second-generation EA techniques, although this may not be suitable for all women. In a partially randomised trial of general anaesthetic (GA) and LA,67 the procedure was considered acceptable under GA in both preferred (100%) and randomised (97%) groups. However, under LA, 97% of those who chose this method and 85% of those allocated to LA found the procedure acceptable.

Selecting an appropriate treatment for heavy menstrual bleeding

The introduction of new EA techniques over the last two decades has been accompanied by a series of randomised clinical trials aimed at evaluating their clinical effectiveness and cost-effectiveness. Initially, first-generation EA techniques such as TCRE and laser EA were compared with hysterectomy. 31 Subsequent trials, which compared alternative first-generation techniques such as TCRE, laser EA and RBEA, established TCRE as the gold standard for this group of treatments. As less invasive and more user-friendly second-generation techniques such as MEA became available, these were compared with earlier methods of ablation like TCRE and RBEA. Although not all techniques have been subjected to head-to-head comparisons in the context of randomised trials, an overview of the literature demonstrates that MEA (second generation) has been shown to be comparable with TCRE (first generation) – which in turn has been shown to be an effective alternative to hysterectomy (gold standard). However, questions about the long-term clinical effectiveness and cost implications of alternative forms of surgical treatment remain unanswered. Published data report no more than 5 years of follow-up. 46,68 Inevitably, some women treated by EA will eventually require repeat ablation or hysterectomy. Following hysterectomy, a proportion of women will also develop further complications such as postsurgical adhesions and pelvic floor dysfunction, which may lead to further surgery. The necessity for a head-to-head comparison between the two most common second-generation methods – MEA and TBEA – has been identified. 15 Given the widespread use of ablative techniques as first-line surgical treatment for menorrhagia at the present time, it is uncertain whether it is either necessary or feasible to compare second-generation techniques directly with hysterectomy. At the same time, the need to obtain comparative information on long-term outcomes is clearly accepted, as is the need to identify the best technique for individual women.

From a clinical perspective, the most relevant research questions at the present time are:

1. How do the currently used ablative techniques compare with hysterectomy in the medium to long term?

2. Which among the commonly used second-generation ablation techniques is the most effective and cost-effective?

3. Are there subgroups of women who are most likely to benefit from either hysterectomy or specific types of ablation?

In this project we have performed a series of studies in order to address these questions by analysis of data from national data sets and randomised trials. Long-term outcomes following EA and hysterectomy in a national cohort have been explored by means of record linkage, while the effectiveness of Mirena, hysterectomy and EA have been determined by individual patient data (IPD) meta-analysis of existing trials. The output has been used (along with other data from the literature) to create a model for the utilisation and costs of the different treatments.

Literature search and study selection

The Cochrane Library, MEDLINE (1966–2010), EMBASE (1980 to May 2010) and CINAHL databases (1982 to May 2010) were searched using relevant terms and word variants for population [e.g. menorrhagia, hypermenorrhea, (excessive) menstrual blood loss, HMB, dysfunctional uterine bleeding] and interventions (e.g. hysterectomy, vaginal hysterectomy, total abdominal hysterectomy, subtotal abdominal hysterectomy, laparoscopic hysterectomy, LNG IUS, Mirena coil and all types and variants of first- and second-generation ablative techniques). Variant names (e.g. EA, resection) and different makes for EA (e.g. Microsulis Medical Ltd, Denmead, UK; Cryogen – now American Medical Systems, Minnetonka, MN, USA) were also searched (see Appendix 1). We also hand-searched the bibliographies of all relevant primary articles and reviews to identify any articles missed by the electronic searches. Experts were contacted to identify further studies. To identify any ongoing RCTs, the Meta-Register of Controlled Trials and the International Standard Randomised Control Trial Number register were searched. No language restriction was applied.

Studies were selected in a two-step process. Firstly, we scrutinised the citations identified by the electronic searches and obtained full manuscripts of all the citations that met, or were thought likely to meet, the pre-determined inclusion criteria based on patient entry criteria (women with HMB or abnormal/excessive/prolonged uterine bleeding that was unresponsive to other medical treatment) and study design, the latter limited to RCTs only. We then considered four categories of intervention with the intention of comparing them against each other: hysterectomy (performed abdominally, vaginally or laparoscopically); ‘first-generation’ EA techniques (using operative hysteroscopy, including endometrial laser ablation, TCRE and RBEA); ‘second-generation’ EA techniques [those that use a ‘blind’ device to simultaneously treat the whole cavity, including thermal balloon (Cavaterm, ThermaChoice and Vesta), microwave (Microsulis), laser (ELITT), bipolar radio frequency (NovaSure), cryoablation and hydrothermal ablation]; and LNG IUS (Mirena). Studies making a comparison within these categories could not contribute to the meta-analysis, but these data were also requested to allow further exploration of possible predictors of the primary outcome measure.

Data collection and study quality assessment

Repeated attempts were made to contact corresponding authors via post, email or telephone to access data. Where initial attempts failed, we attempted personal contact via our links through the British and European Societies for Gynaecological Endoscopy. Authors were asked to supply anonymised data for each of the pre-specified outcome measures (both published and unpublished to reduce the chance of selective reporting bias) and were invited to become part of the collaborative group with joint ownership of the final publication. Where the investigators declined to take part in the study or could not be contacted, published data were extracted from manuscripts using pre-designed proformas by two independent reviewers (RC and LJM). Any disagreements were resolved by consensus or arbitration by a third reviewer (JPD). Received data were merged into a master database, specifically constructed for the review. The data were cleaned and results were cross-checked against published reports of the trials. Where discrepancies existed authors were contacted for clarification.

Authors of the protocol reviewed all relevant outcome measures to be used in the meta-analysis from articles identified in the literature search. Level of satisfaction with treatment was the most frequently measured outcome across all identified studies, with 21 out of 30 (70%) using this measure, and was used as the primary outcome measure. Dissatisfaction rates are presented to simplify interpretation of statistical output. Responses of ‘very satisfied’ or ‘satisfied’ were taken as a positive response, likewise ‘very dissatisfied’ or ‘dissatisfied’ as a negative response. Where a ‘not sure’ or ‘uncertain’ response was given these were conservatively taken to be a negative rating of treatment, although sensitivity analysis was undertaken to test the robustness of this assumption. For a small number of studies,78–81 surrogate outcomes for satisfaction were used (major problem resolved/improvement of health state/menstrual symptoms successfully treated/degree of recommendation). This assumption was also tested by sensitivity analysis without these studies (indicated in the results section where important) (see Appendix 2). A more disease-specific QoL tool19 would have been the ideal choice for primary measure, but these data were not available from the studies identified. We have shown from the data in this review, though, that a strong relationship between dissatisfaction and patient QoL is apparent (see Results).

Other outcome measures were bleeding scores (ranging from a minimum of zero to no upper limit),12 amenorrhoea rate (converted from a bleeding score of zero where data existed, otherwise as reported), heavy bleeding rate (converted from bleeding scores of > 10012 where data existed, otherwise as reported), EQ-5D utility score,82 SF-36 scores,20 duration of surgery/hospital stay, general anaesthesia rates, postoperative pain score (standardised from visual analogue and ordinal scale scores onto a 0–10 scale), time to return to work/normal activities/sexual activity, dysmenorrhoea/dyspareunia rate and proportion undergoing subsequent ablation/hysterectomy or discontinuing use of Mirena. Pre-defined subgroups were age at randomisation (≤ 40 vs > 40 years), parity (nulliparous vs parous), uterine cavity length (≤ 8 vs > 8 cm), presence or absence of fibroids/polyps and, where available, severity of bleeding at baseline (bleeding score ≤ 350 or > 350).

All selected trials were assessed for their methodological quality, using received data sets where available in addition to the reported information. Quality was scrutinised by checking the adequacy of randomisation, group comparability at baseline (examining baseline characteristics for any substantive differences), blinding (where appropriate), use of intention-to-treat (ITT) analysis, completeness of follow-up, compliance, reliability using a priori sample size estimation and generalisability using description of the sample recruited. Adequacy of randomisation was assessed with subquestions examining information on sequence generation, the process of allocation and allocation concealment.

Statistical analysis

To minimise the possibility of bias IPD and aggregate data (AD) were combined in a two-stage approach. 83 IPD were reduced to AD to allow studies with AD only to be combined with those where IPD were obtained. Unless specifically stated in the text of the results section, all estimates shown are from all available data (both IPD and AD). Point estimates and 95% confidence intervals (CIs) were calculated for individual studies at each time point. For the primary outcome measure, differences in effect estimates between trials and the pre-defined subgroups of patients are displayed using odds ratio (OR) plots; results from other outcome measures are summarised in tables in the appendices for ease of reference. Heterogeneity was investigated using Cochran’s Q,84 I² statistics24 and Higgins et al. 85 Subgroup analyses to explore the causes of heterogeneity were undertaken if the p-values of these tests were < 0.1. Differences between studies contributing IPD and those with AD only were examined in the same fashion to check that the latter results were consistent with those we received IPD for. Further details are given in the Results section if any inconsistency exists. Likewise, further details are given on any obvious publication bias if noted from the assessment of funnel plots. Only a limited amount of data were available for studies comparing Mirena with ED, so Mirena was compared with first- and second-generation studies combined as well as separately. Assumption-free ‘fixed-effect’ methods were used to combine dichotomous outcome measures and estimate pooled ORs using the method of Peto et al. ,86 and, for continuous variables, weighted mean differences (WMD) were calculated87 at each time point. Data at less than 12 months were combined and are described as results at 6 months. Results from the limited number of studies with follow-up longer than 2 years are not referred to in the text but are given in the appendices.

The primary outcome measure of dissatisfaction was investigated comprehensively using received data. Results at 12 months, where the majority of studies had collected data, were used as the focus for analysis. Where responses were not available at this time point, data were substituted in the first instance from 2 years and, failing that, from 6 months. If it was not possible to make a direct comparison between treatments (e.g. hysterectomy vs second-generation EA), indirect estimates were made88 using a logistic regression model89 allowing for trial and treatment. 90 Estimates using dissatisfaction at any time were also examined, along with an analysis allowing for the correlation of the repeated measurements using generalised estimating equations (IPD only). 91

Access to IPD also allowed the inclusion of patient-level covariates to examine possible predictors of dissatisfaction. First, covariates were considered individually, while allowing for differences between trial estimates. If considered statistically important (p < 0.1), covariate parameters were included together in a multivariate analysis to examine adjusted estimates. In addition to the analysis of the primary outcome measure described above, as a sensitivity analysis IPD were also used to explore the effect observed in compliance rates for comparisons between first- and second-generation EA (unfortunately there were insufficient data to extend this analysis to Mirena comparisons). For example, for those women ‘satisfied’ with treatment but subsequently undergoing a hysterectomy, positive responses were substituted with negative ones. The relationship between dissatisfaction and responses from the SF-36 QoL questionnaire was examined at the patient level using a regression model, allowing for trial. Given the number of analyses performed, any interpretation of p-values greater than the conservative threshold of 0.01 has been cautious owing to the likelihood of increased type 1 error rates. revman v5.0 (Cochrane Collaboration, Denmark) and sas v9.2 (SAS Institute, Cary, NC, USA) software were used for analysis.

Trials and patients

A total of 556 potentially relevant citations were identified by electronic searches. After detailed evaluation, 30 trials were eligible for inclusion in the review (Figure 1). Of these trials, seven compared hysterectomy with ED techniques. Six of these studies involved first-generation techniques. 41–45,78 The seventh study used a combination of first and second generation in equal proportions92 and was included here as a first-generation comparison, with a sensitivity analysis performed without the trial. One study compared hysterectomy93 with Mirena. Fourteen studies compared first-generation EA techniques with those of second generation53,54,56,79,94–103 and eight studies compared Mirena with EA, three of which were first generation80,104,105 and five second generation. 81,106–109 Characteristics of these studies are shown in Appendix 2. Data from a further five studies,64,65,110–112 which involved comparisons within first- and second-generation EA, were also received.

FIGURE 1.

Study selection process for systematic review and IPD meta-analysis of randomised trials comparing hysterectomy, EA techniques and Mirena for HMB (details of selected trials in Appendix 2).

Trials comparing hysterectomy with EA and those comparing first- and second-generation ED involved women of a similar age, with average ages of 40.6 years [standard deviation (SD) 5.1 years] and 41.0 years (SD 4.9 years), respectively. Women involved in trials comparing Mirena with ED were slightly older, with an average age of 43.6 years (SD 3.5 years). Eligibility criteria for women with uterine pathology varied between trials; inclusion of women with fibroids was generally limited by size or number. Where included, they amounted to a maximum of 30% of the women in each individual study.

A high proportion of data was received from trials involving hysterectomy (seven of eight studies; 1278 of 1363 women), with less for trials of EA techniques (7 of 14 studies; 1359 of 2448 women) and those involving Mirena (three of eight studies; 177 of 494 women) (see Appendix 2). Overall, we received some IPD from 65% (2814/4305) of women involved in the trials, although only eight studies were able to provide all requested variables. 41,42,53,94,95,99,102,109 The remaining studies had some missing information, with limited details on patient follow-up covering subsequent operations (e.g. hysterectomy following Mirena). See the section on statistical analysis for details on how data from studies providing IPD were utilised.

Hysterectomy versus first-generation endometrial ablation

More women were dissatisfied at 12 months following first-generation EA than hysterectomy [12.6% vs 5.3%; (57/454 vs 23/432); OR 2.46; 95% CI 1.54 to 3.93; p = 0.0002] (Figure 3), with no significant heterogeneity between study estimates (p = 0.9; I² = 0%). This estimate of effect size was consistent with, although slightly less than, the estimate from the repeated measures analysis (IPD only) over all time points (OR 3.75; 95% CI 2.18 to 6.46; p < 0.0001) and an analysis using dissatisfaction at any time point (OR 3.37; 95% CI 2.14 to 5.31; p < 0.0001). There was no evidence of any differences between subgroups (see the Data collection and study quality assessment section), including between studies providing IPD or AD (test for heterogeneity: p = 0.9).

FIGURE 3.

Dissatisfaction at 12 months – hysterectomy versus first-generation EA.

First- versus second-generation endometrial ablation techniques

Similar dissatisfaction rates were seen with first- and second-generation EA [12.2% vs 10.6% (123/1006 vs 110/1034); OR 1.20; 95% CI 0.88 to 1.62; p = 0.2; test for heterogeneity: p = 0.7)] (Figure 4). Comparison estimates were obtained from the repeated measures analysis of IPD (OR 1.21; 95% CI 0.84 to 1.74; p = 0.3), the analysis using dissatisfaction at any time (OR 1.22; 95% CI 0.91 to 1.62; p = 0.2), and also an analysis adjusting for patients who went on to receive a hysterectomy (OR 1.25; 95% CI 0.93 to 1.67; p = 0.1). Results were consistent over all subgroups, including those studies providing IPD or AD only (test for heterogeneity: p = 0.8).

FIGURE 4.

Dissatisfaction at 12 months – first- versus second-generation EA.

Mirena versus endometrial ablation techniques

Rates of dissatisfaction with Mirena and second-generation EA were similar [18.1% vs 22.5% (17/94 vs 23/102); OR 0.76; 95% CI 0.38 to 1.53; p = 0.4] (Figure 5), although the latter rate was twice as high as that seen for second-generation EA when it was compared with first-generation EA (see Figure 4). The slightly older age of women in these studies (see Trials and patients) could be a possible explanation for this increase, although given the small number of women studied in these trials this difference in rate could easily have arisen by chance. The combined estimate of this and the one study that compared Mirena with first-generation EA80 (test for differences between subgroups: p = 0.2) also showed no evidence of a difference (OR 0.94; 95% CI 0.50 to 1.77; p = 0.9). Heterogeneity of estimates overall was of borderline statistical significance (p = 0.09; I² = 54%). Overall rates of dissatisfaction were 17.2% (22/128) for Mirena and 18.2% (25/137) for both first- and second-generation EA. Lack of IPD prohibited any further investigation of subgroups or repeated measures. Sensitivity analysis performed without two studies where surrogates for dissatisfaction were used significantly reduced the data available for analysis but did not change the findings.

FIGURE 5.

Dissatisfaction at 12 months – first- and second-generation EA versus Mirena.

Indirect comparisons of hysterectomy with second-generation endometrial ablation techniques and Mirena

Indirect estimates (Figure 6) suggest that hysterectomy is also preferable to second-generation EA [5.3% vs 10.6% (23/432 vs 110/1034); OR 2.32; 95% CI 1.27 to 4.24; p = 0.006) in terms of patient dissatisfaction. This is confirmed by the repeated measures analysis over all three time points, which per force only include IPD (OR 3.06; 95% CI 1.59 to 5.90; p = 0.0008). The evidence to suggest hysterectomy is preferable to Mirena was weaker [5.3% vs 17.2% (23/432 vs 22/128); OR 2.22; 95% CI 0.94 to 5.29; p = 0.07], but given the lack of precision from Mirena comparisons this was not a surprising result and should be cautiously interpreted.

FIGURE 6.

Dissatisfaction at 12 months summary. Odds ratios (95% CIs) are shown. Estimates > 1 indicate increased dissatisfaction for the second treatment listed. Indirect estimates are represented by a dotted line.

Predictors of dissatisfaction

For second-generation EA, IPD showed that uterine cavity length was the strongest predictor of dissatisfaction (p = 0.02), with shorter uterine cavity length (≤ 8 cm vs > 8 cm) associated with reduced rates (OR 0.59; 95% CI 0.38 to 0.93; p = 0.02) (Table 5). Absence of fibroids/polyps also showed a trend towards reduced dissatisfaction (p = 0.07), although no further adjusted estimates including both parameters were attempted as only three studies had data on fibroids/polyps. There were no convincing associations with any of the variables for hysterectomy or first-generation EA.

Hysterectomy versus endometrial ablation and Mirena

These comparisons focused on recovery times and QoL, as estimates of postoperative menstrual blood loss are redundant after hysterectomy (see Appendix 4). EA offered quicker surgery (WMD 32 minutes; 95% CI 30 to 34 minutes; p < 0.0001), shorter hospital stay (WMD 3.0 days; 95% CI 2.9 to 3.1 days; p < 0.00001), faster recovery periods (time to return to normal activities: WMD 5.2 days; 95% CI 4.7 to 5.7 days; p < 0.00001) and less pain postoperatively (WMD 2.5 points; 95% CI 2.2 to 2.9 points; p < 0.0001), although estimates of differences for some of these parameters should be used with caution given the high variability between studies (see Appendix 4). One study92 suggested no obvious difference in EQ-5D utility scores, while another78 suggested differences in favour of hysterectomy in the general health (WMD 9.6 points; 95% CI 5.7 to 13.5 points; p < 0.0001), social function (WMD 24 points; 95% CI 21 to 27 points; p < 0.0001) and vitality (WMD 13 points; 95% CI 9.3 to 16 points; p < 0.0001) domains of the SF-36 questionnaire. Perioperative adverse events associated with hysterectomy were relatively few (0.5%–2.0% each), but UTIs were more common with hysterectomy (43/530; 8.1%) than with EA (9/585; 1.5%). Of the women who were initially treated with EA, 15% had undergone a hysterectomy within 2 years.

No differences in EQ-5D scores were seen at 6 or 12 months in the single study comparing hysterectomy with Mirena (see Appendix 5), while the only statistically significant effect observed in the SF-36 questionnaire was in the pain domain, which favoured hysterectomy (WMD 9.6 points; 95% CI 2.7 to 16.6 points; p = 0.007). All results were consistent over subgroups.

First- versus second-generation endometrial ablation techniques

The proportion of women with amenorrhoea or still experiencing heavy bleeding was similar in both groups at all time points apart from at 2 years, where there was a borderline significant difference in favour of second-generation techniques (amenorrhoea: OR 0.64, 95% CI 0.41 to 0.99, p = 0.04; HMB: OR 1.85, 95% CI 1.04 to 3.32, p = 0.04) (see Appendix 6). High heterogeneity for the estimate of amenorrhoea rate at 12 months (OR 1.12, 95% CI 0.93 to 1.35, p = 0.3) appeared to be due to two outlying studies (Duleba98 and Perino100), the results of which could not be verified as IPD were not available. However, analysis without these studies gave very similar results (OR 1.10, 95% CI 0.90 to 1.35, p = 0.4) with lowered heterogeneity (p = 0.1; I² = 36%). Change from baseline analysis of bleeding scores showed no evidence of a difference at any of the time points. Inconsistency of estimates for this outcome at 12 months was due to a single study;103 sensitivity analysis without this study also showed no change to the overall result (WMD –0.3; 95% CI –27.5 to 27.0; p = 1.0; heterogeneity: p = 0.4; I² = 10%). Two studies54,99 using the SF-36 questionnaire and one small study94 using the EQ-5D questionnaire showed no consistent difference between first- and second-generation techniques, in terms of change from baseline results.

Second-generation EA was quicker (WMD 15 minutes; 95% CI 14 to 15 minutes; p < 0.0001) and less likely to need general anaesthesia than first generation (OR 0.16; 95% CI 0.12 to 0.20; p < 0.0001), although highly significant heterogeneity makes estimates difficult to interpret. Less frequent use of general anaesthesia with second-generation EA translated to a slightly quicker time to return to work (WMD 1.36 days; 95% CI 0.69 to 2.03 days; p < 0.0001) and time to return to normal activities (WMD 0.48 days; 95% CI 0.20 to 0.75 days; p = 0.0008), although the overall estimate for the latter was somewhat inconsistent (heterogeneity: p = 0.04; I² = 59%). Postoperative pain was similar following either method of EA, although estimates from different studies varied widely (heterogeneity: p < 0.0001; I² = 89%) without any obvious explanation. Adverse events were relatively low in both groups (each < 2%), but perioperative complications such as uterine perforation (OR 0.20; 95% CI 0.07 to 0.57; p = 0.003), excessive bleeding (OR 0.14; 95% CI 0.07 to 0.55; p = 0.005), fluid overload (OR 0.12; 95% CI 0.04 to 0.36; p = 0.0001) and cervical laceration (OR 0.12; 95% CI 0.05 to 0.33; p < 0.0001) were lower with second-generation EA. The number of women requiring a subsequent hysterectomy was lower for second-generation EA, but these differences were not large enough to be statistically significant within the first 2 years (12 months: OR 0.77, 95% CI 0.47 to 1.24, p = 0.3; 2 years OR 0.68, 95% CI 0.41 to 1.13, p = 0.1). Overall rates were 3.3% (74/2265) and 7.6% (71/939) at these time points. Any differences amongst subgroups were confined to single time points only. Results from studies providing IPD were consistent with those with AD only.

Mirena versus endometrial ablation techniques

Fewer women experienced HMB after Mirena at 6 months (OR 0.23; 95% CI 0.09 to 0.57; p = 0.001) and at 2 years (OR 0.08; 95% CI 0.01 to 0.50; p = 0.007), although total numbers here were small compared with the estimate at 12 months, where there was no evidence of any difference (OR 0.74; 95% CI 0.34 to 1.61; p = 0.5) (see Appendices 7 and 8). Amenorrhoea rates were similar at all time points, although the overall estimate at 12 months, along with the estimate for HMB, was somewhat inconsistent (heterogeneity: p = 0.05 and 0.06, respectively; I² = 59% and 55%, respectively). Changes in bleeding scores favoured EA at 12 months only (WMD 38 points; 95% CI 15 to 60 points; p = 0.0009), a result consistent over all studies (heterogeneity: p = 0.5; I² = 0%). Other outcome measures could not separate the two treatments. Two studies107,109 provided SF-36 changes from baseline scores, and no differences were found in any of the domains. The number of women subsequently undergoing a hysterectomy was slightly higher for Mirena, although total numbers in this comparison were very small; rates at 12 months were 2.3% (2/86) for EA and 6.7% (6/89) for Mirena. A high proportion of women originally prescribed Mirena discontinued use of this treatment – 15.7% (30/191) at 12 months, rising to 27.6% (29/105) by 2 years. Reported adverse events were low with Mirena; around only 3% reported an expelled/migrated coil within the first month. These results were from studies of first- and second-generation studies combined where first-generation data existed, and were consistent over both types of EA.

Principal findings

In this review, access to IPD enabled a more rigorous analysis than is possible from published data from trials comparing second-line treatments for HMB. The primary outcome measure of dissatisfaction was shown to be strongly related to increased QoL. Based on direct and indirect comparisons using all available data, the review found that both first- and second-generation EA techniques were associated with greater dissatisfaction than hysterectomy, although rates were low for all treatments and absolute differences were small. Recovery times and length of hospital stay were longer for hysterectomy. Dissatisfaction levels with second-generation techniques were slightly lower than those associated with first-generation techniques. In addition, second-generation methods of EA were quicker, had faster recovery times, were associated with fewer adverse procedural events and could be offered under local anaesthetic. Fewer women subsequently underwent hysterectomy after second-generation EA than with first-generation EA, but this difference was not statistically significant. Shorter uterine cavity length was associated with lower levels of dissatisfaction for second-generation EA. Comparisons of EA with Mirena suggest comparable efficacy, although studies involving the latter treatment were generally small and consequently imprecise. Substantial discontinuation of Mirena use was noted and makes interpretation of findings for this treatment difficult.

Research questions

1. What is the risk of further gynaecological surgery following EA compared with that following hysterectomy in women with HMB?

2. What is the risk of further gynaecological surgery following different types of hysterectomy in women with HMB?

3. What is the risk of gynaecological cancer following EA compared with that following hysterectomy in women with HMB?

4. What is the association of age with risk of further gynaecological surgery following EA compared with that following hysterectomy in women with HMB?

Hysterectomy codes used

Operation codes used

Cancer codes used

Principal findings

The sociodemographic profile of women who underwent either first- or second-generation EA was different to that of those who received hysterectomy for HMB. Hysterectomy was more likely to lead to surgery for pelvic floor repair and for stress urinary incontinence. Vaginal hysterectomy was associated with a higher chance of further surgery and surgery for pelvic floor prolapse compared with hysterectomy carried out through the abdominal route. The incidence of cancers was generally low in both groups (< 1.6%), with endometrial cancer following EA having an incidence of 0.02%.

Strengths and limitations

To our knowledge this is the first large population-based study using national data on long-term outcomes in women who had received alternative surgical treatments for HMB. Use of ICD codes allowed us to define both the cause of HMB as well as the nature of surgery, but, as the diagnosis of dysfunctional uterine bleeding was performed by a process of exclusion, it is possible that the hysterectomy cohort could have included a few women with other causes of HMB.

This was a retrospective cohort study based on routinely collected national data. Like any observational study, it is not free from the usual problems of bias and confounding. Additionally, the analysis was compromised by the limited availability of key socioeconomic as well as clinical variables. Although the numbers of women in the hysterectomy and ablation cohorts were large, a major drawback was our inability to discriminate among the individual types of first- and second-generation EA or adjust for the experience of the operator as has been done in previous national audits. 47 We were also unable to analyse the long-term outcomes following laparoscopic hysterectomy as numbers were small; these were therefore grouped with abdominal hysterectomy.

Interpretation of findings

Our findings suggest that women from higher socioeconomic groups are more likely to have EA than hysterectomy. This inverse correlation between hysterectomy rates and social class has been noted by some authors121 but not by others. 122 None of these studies had a control group of women with HMB who went on to have EA.

Our results confirm previous work46 that suggests that women treated initially with EA for HMB are more likely to need subsequent surgery for the same condition than those treated by hysterectomy, which is a more definitive operation. Although they were excluded from the current analysis, we noted that around one in five women had a subsequent hysterectomy following initial EA, while repeat ablations and exploration of the uterus accounted for a substantial minority of cases. As has been noted previously,46,112,123 the survival curve indicates that most repeat surgery for persistent HMB occurred within the first 2 years of initial surgery.

Women were more likely to undergo a TVT procedure for stress urinary incontinence after hysterectomy than after EA – corroborating the results of some previous studies that suggested a link between hysterectomy and urinary incontinence. 34,124 The biological justification of this has been debated in the past and could be due to compromise to the pelvic floor caused by surgical damage to muscular, connective or neurological tissue. The lower risk of genital fistula repair after ablation is explained by the higher probability of pelvic organ damage during hysterectomy.

The risk of gynaecological cancer following EA has been identified as a key clinical and research question in the past. Our results are in agreement with Krogh et al. ,125 who found no significant increase in the incidence of endometrial cancer after ablation in a Danish population. We found no significant difference in the risk of ovarian cancer between women undergoing EA and those who had a hysterectomy. In contrast, Loft et al. 126 reported that the risk of ovarian cancer was lower in women who had hysterectomy (with conservation of at least one ovary) than in those who had not [relative risk (RR) 0.78; 95% CI 0.60 to 0.96]. Owing to an average age of 41 years at initial surgery and the relatively short follow-up (median duration 6.2 and 11.6 years for EA and hysterectomy groups, respectively), many of the women in our study have yet to reach the peak age of incidence for many of the cancers in question and hence the incidence of a common malignancy like breast cancer is low in both groups.

We are able to explore the impact of different types of hysterectomy versus ablation and show that the overall risk of further surgery was significantly higher in the ablation group irrespective of type of hysterectomy. Compared with the ablation group, the odds of TVT were significantly higher in women after a total hysterectomy, but not after subtotal hysterectomy, thus fuelling the ongoing debate on the potential association between total (as opposed to subtotal) hysterectomy and urinary stress incontinence. 123,127–129 Subtotal hysterectomy has other potential disadvantages, as shown in a small series of women: laparoscopic subtotal hysterectomy led to a 22.9% (16/70) chance of potentially difficult further surgery to remove the cervical stump. 130 The chances of undergoing surgery for pelvic floor prolapse and urinary stress incontinence were significantly higher in women who had undergone vaginal hysterectomy than in the ablation group. It is impossible to rule out the possibility that the decision to opt for a vaginal route for a hysterectomy may have been informed by prior knowledge of a degree of descent or pelvic laxity. Comparison of different types of hysterectomy revealed that the vaginal route was more likely to be associated with future surgery for prolapse and incontinence. This is supported by data from Blandon et al. ,131 who reported that, in comparison with women without prolapse, women who had a hysterectomy for prolapse were at increased risk for subsequent pelvic floor repair. In the absence of follow-up data from randomised trials, the existing literature on this subject is conflicting as described by Thakar and Sultan. 123

There was no significant association between age and risk of further gynaecological operations.

Clinical implications

The results of this study provide clinicians and women with useful data on expected outcomes after EA which can be used to counsel women regarding options of treating HMB. The lower perioperative complications of the less invasive ablation procedures need to be balanced against the 8% chance of repeat surgery for the same symptoms, although the chance of long-term pelvic floor problems may be less. It is also useful to confirm data from other follow-up studies on smaller cohorts that indicate that most repeat procedures occur within 2 years of the initial operation. Our data are broadly reassuring in terms of identifying the risk of endometrial cancer after ablation.

Research implications

This study underlines the limitations of the available literature in this field, which include retrospective studies that have been prone to selection and reporting bias as well as lack of data on key confounders. In the absence of adequately powered large randomised trials with high rates of follow-up, well planned prospective cohort studies with pre-determined end points are needed. It is also important to consider the need for large national audits of EA especially now that new second-generation ablation technologies are being adopted.

Cost-effectiveness model

A state transition (Markov) model was developed by the authors using Microsoft excel. The structure was informed by clinical input. The model examines the progress of four hypothetical cohorts of women with HMB who are treated by one of four alternative strategies: Mirena coil; first-generation EA techniques; second-generation EA techniques; or hysterectomy. Given the reliance on secondary data and the nature of available data, the model-based economic evaluation takes the form of a cost–utility analysis and was carried out from the perspective of the NHS in a secondary care setting. An incremental approach was used for the reporting of the results.

Structure of the economic model

In the model, a cohort of 10,000 eligible women was compared for each strategy (Mirena coil; first-generation EA; second-generation EA; and hysterectomy). The starting age of women in the model is 42 years, and the model runs for a total of 10 years and assumes that all women will become menopausal at the age of 52 years, (the average age of menopause in the UK). These assumptions are those used by Garside et al. 16 Each model cycle is 1 month long and represents a typical menstrual cycle. The death rate from causes other than procedures for HMB was based on values for women in the Government Actuary’s Department life tables of England and Wales for the years 1998–2008. 133

The model is based on the clinical pathways presented in Figure 16. The pathway for women undergoing any EA technique (first or second generation) is shown in Figure 17. The pathway for women undergoing hysterectomy is shown in Figure 18. Health states are shown in boxes and arrows show the transitions that can occur. The health states and pathways are the same for both types of EA technique.

FIGURE 16.

Summarised version of clinical pathway in the model.

FIGURE 17.

Clinical pathway for EA (first- or second-generation techniques).

FIGURE 18.

Clinical pathway for hysterectomy.

Definition of health states for endometrial ablation pathways

• Menorrhagia – all women in the cohort have pre-operative HMB.

• EA techniques – women undergo EA by first- or second-generation techniques.

• Complications – following EA, some women will experience severe postoperative complications. Perioperative complications are included in the EA state.

• Well – following EA, complications or treatment failure, women are satisfied with treatment.

• Symptomatic – following EA, complications or well, HMB may recur (treatment failure) at any time. Women may be retreated (repeat ablation), become ‘well’ or have a hysterectomy after initial or repeat ablation.

• Repeat ablation (RB) – if HMB recurs postoperatively, women may choose to have a second EA which occurs 6 months after the initial ablation. Only one repeat EA is permitted and it is always a first-generation technique (RB).

• Hysterectomy – if women become symptomatic after the first ablation, they may choose to have a hysterectomy. Hysterectomy is also an option after a failed repeat EA. This operation occurs 6 months after the decision.

• Death – it is possible to die from natural causes. At hysterectomy and EA, women may also die as a direct result of the surgical procedure.

Definition of health states for hysterectomy pathways

• Menorrhagia – all women in the cohort have pre-operative HMB.

• Hysterectomy – all women undergo hysterectomy.

• Complications – following hysterectomy, some women will experience severe postoperative complications. The effects of these may last for 1 month. Operative complications are included in the hysterectomy state.

• Convalescence – following hysterectomy both with and without complications, a period of convalescence is experienced. This may last up to 3 months.

• Well – following convalescence women are satisfied with treatment.

Definition of health states in Mirena pathways

• Menorrhagia – all women in the cohort have pre-operative HMB.

• Mirena – all women have Mirena inserted.

• Well – following Mirena, women are satisfied with treatment.

• Symptomatic – following Mirena or well, HMB may recur (treatment failure) at any time. Women may be retreated, have a second-generation EA or remain symptomatic.

Mirena

The lifespan of Mirena is assumed to be 5 years. 15 If it is successful, treatment is repeated at 5 years. Treatment failure in Mirena is assumed to be more evident in the first year. 27,93 We also assumed a 2% insertion failure rate where procedure is repeated within a month. 27 If Mirena is unsuccessful, women move to the second-generation pathway. No adverse events associated with Mirena were available from the literature.

First- and second-generation endometrial ablation techniques and hysterectomy

Large national audits of hysterectomy and first-generation EA techniques were used as sources for perioperative and severe postoperative adverse events5,47 and are presented in Table 13. Minor postoperative complications were not modelled. For second-generation EA techniques, complication rates of MEA were used. 134

Some of the data (specifically proportions in various health states at points in time after initial treatment) related to model outputs rather than model inputs. We applied a method that may be called ‘probabilistic calibration’ whereby model inputs for the relevant parameters were sampled uniformly across the plausible range, and cost and QALY outcomes were weighted according to the likelihood function comparing the model proportions in the various health states with the data.

Repeat ablation (rollerball)

If EA of any type fails, repeat ablation or hysterectomy is offered. In the model it is assumed that 60% of those with recurrent menorrhagia (symptomatic state) will have a repeat EA and 40% will have a hysterectomy. 28 Only one repeat ablation is offered, which is RB (first-generation EA). If the treatment fails a second time, only hysterectomy is available.

Complication rates in the repeat ablation are assumed to be double those incurred for the initial ablation. 135

Repeat procedure

The transition probability for requiring a repeat procedure (ablation or hysterectomy) is time-dependent and is reduced by a constant factor each month. This reflects a decreasing hazard, which is obvious from the IPD data.

Mirena

The cost of insertion of Mirena was estimated at £130.27. This procedure is assumed to be performed in a menstrual clinic as an outpatient procedure. The total cost includes those for the device, the initial consultation (10 minutes with a nurse and 30 minutes with a specialist registrar) and a sterile pack for use during Mirena insertion. Cost of discontinuation (£28.34) includes the cost of the consultation and the consumables (sterile pack) used for removal of the device.

First-generation techniques

The cost of first-generation EA was estimated at £1238. The source was a study which compared the costs of treating women with menorrhagia by hysterectomy or hysteroscopic surgery. 32 Costs in this study included pre-surgery treatment for EA, technical equipment (which varied for each method), hospital costs, gynaecological outpatient costs and retreatment. We excluded retreatment from our estimate because this is a separate procedure included in the model.

Second-generation techniques

The cost of second-generation techniques was estimated at £1101. The source for the costs of MEA and TBEA was the HTA report by Garside et al. 16 Statistical weights for the weighted cost mean were obtained from a study reporting NHS hospital episode statistics of EA from 1989–90 to 2005. 30

Hysterectomy

The cost of hysterectomy was estimated at £2162 and the source was a study comparing the costs of treating women with menorrhagia by hysterectomy or hysteroscopic surgery. 32 For women who had a hysterectomy after a failed repeat ablation, we included an additional cost of a consultation with a GP for referral from primary to secondary care at £46. 136

Repeat ablation (rollerball)

The cost of repeat ablation was the same as the cost of the first-generation techniques described above. We also included an additional cost of a GP consultation for referral from primary to secondary care, at £46. 136

Adverse events

The source for the costs of adverse events was NHS reference costs (2009). 139 We used the same cost for both perioperative and severe postoperative complications for each of the procedures included in the model. The varying severity of complications of the two different types of EA and hysterectomy is reflected in costs as well. The cost of complications was £2161 for the first-generation techniques, £1198 for the second-generation techniques and £3008 for hysterectomy.

The MISTLETOE study was the source for perioperative and severe postoperative adverse events of the first-generation EA techniques. 47

For second-generation EA techniques, complication rates of MEA were used as a proxy. 134 We chose not to use the complications rates from MISTLETOE for the second-generation EA techniques as well for the following reasons:

• The MISTLETOE study included radiofrequency and cryoablation only, and did not include MEA or TBEA, which are the commonest second-generation procedures used at the present time.

• Radiofrequency is no longer performed as it proved to be unsafe.

• The number of cryoablations in MISTLETOE is very low (n = 36) and there were no perioperative complications within that group.

• Severe postoperative complications in the MISTLETOE study are more frequent in the second-generation group than in the first-generation group, which is counterintuitive, as complications in second-generation techniques should be less frequent.

In using the complication rates of MEA from Parkin134 (n = 1400), we acknowledge that this study might underestimate the true incidence of complications. All data reported in this study were from a specialist centre and reported complication rates might not be representative of other second-generation techniques. There were no adverse events associated with Mirena.

Analysis 1

For Analysis 1, for the first-generation EA, second-generation EA and Mirena strategies, we assumed that repeat procedures (ablation or hysterectomy) are allowed at any age, but with a decreasing hazard.

The deterministic results are presented in Table 16.

The Mirena strategy

The results show that Mirena is both less costly and produces more QALYs than first-generation EA and thus dominates it. However, although the Mirena strategy costs only half as much as the second-generation EA strategy, the second-generation EA is more effective at producing QALYs. Thus, the ICER for second-generation EA compared with Mirena is £50,000 per QALY. This means that for every woman who is treated with second-generation EA instead of Mirena there is an additional cost of £50,000 to achieve an additional QALY.

Consider for example a threshold ICER of £5000 per QALY; according to our model (not shown), the expected net monetary benefit per woman for hysterectomy is £34,386 at this valuation. This is higher than the expected net benefit for any of the other three options and so hysterectomy is the preferred option given parameter uncertainty. However, replications of the model accounting for approximately 26% of the probability favour different options. So the model probability that hysterectomy is the preferred option is 74%. This is the probability shown on the cost-effectiveness acceptability frontier (CEAF) (Figure 19). The limitation of the CEAF as a decision aid is that it does not tell us, in the remaining 26% of cases, whether changing to another option would make a large or a small difference to the expected net benefit.

FIGURE 19.

Cost-effectiveness acceptability frontier for Analysis 1.

Figure 20 shows the difference between the expected net benefit of the optimal strategy allowing for perfect information and the expected net benefit of the optimal strategy given current information for any given threshold ICER. If the preferred option could be chosen after completely resolving parameter uncertainty, then the estimated net benefit per woman would be £34,812 (from the model, not shown). The additional £426 is known as the ‘per woman’ EVPI. Similar remarks apply at other threshold ICERs. At thresholds below 1600 per QALY (i.e. the kink in Figure 19), the preferred option is Mirena, and so the CEAF shows the probability that Mirena is the preferred option. As the threshold approaches zero the decision is effectively to choose the least costly option – this is Mirena in all replications of the model. So the probability shown on the CEAF goes to one, and the EVPI becomes zero. As the threshold ICER becomes very large, the decision is effectively the option with the highest QALY output. There is some residual uncertainty in the model and so the probability on the CEAF does not actually approach one. Had the EVPI been measured in health units, the height of the EVPI curve would tend to a fixed non-zero limit. However, because it is measured in monetary terms, the curve has an increasing slope.

FIGURE 20.

Probability that the preferred option is cost-effective for any given threshold ICER.

In summary, for every replication in this model, Mirena was the least costly option. For very low-threshold ICERs, Mirena is preferred with certainty (so EVPI tends to zero and CEAF tends to one). In some replications of the model (accounting for 18% of the probability), hysterectomy was not the most effective option (in terms of total number of QALYs). Therefore, the CEAF never goes above 82% and the EVPI does not go to zero.

At an ICER of £1600, the preferred option changes from coil to hysterectomy, so there is a discontinuity in the CEAF and a discontinuity in the gradient of the EVPI curve.

Analysis 2

For Analysis 2, we assumed that if symptoms do not recur within 2 years of the initial ablation, then they are unlikely to do so later, and therefore no repeat procedure takes place thereafter. Thus, we have to limit the time as to when a repeat procedure (ablation or hysterectomy) may occur to 2 years. Similarly, as regards to repeat ablations, if women are not symptomatic within 4.5 years of the initial ablation (2 + 2 years + 6 months from decision to repeat procedure), it is assumed they will never become symptomatic (Table 17).

First-generation EA is dominated by all the other strategies.

The hysterectomy strategy

Hysterectomy is the strategy which produces the most QALYS. As this strategy is both cheaper and produces more QALYs than the first-generation EA strategy, it is considered to dominate the latter. Hysterectomy is more expensive but produces more QALYs than the Mirena strategy and the ICER representing the value of the additional benefit of this strategy compared with Mirena is £1440 per additional QALY. Despite being more costly, the hysterectomy strategy produces more QALYs than the second-generation EA strategy and the cost per unit of benefit for this comparison is £970 per additional QALY.

The detailed explanation of the interpretation of Figures 21 and 22 is similar to the explanation given for the comparable figures for Analysis 1.

FIGURE 21.

Cost-effectiveness acceptability frontier for Analysis 2.

FIGURE 22.

Per-woman expected value of perfect information.

In summary, Figure 21 presents the probability that the preferred option is cost-effective for any given threshold ICER.

In this model, hysterectomy dominates the graph. For all but a few replications, and the few account for a negligible probability, Mirena was the least costly option and for very low-threshold ICERs Mirena is preferred with certainty (so EVPI tends to zero and CEAF tends to one); however, as the threshold ICER increases hysterectomy becomes the preferred option.

However, it is clear that in some replications of the model (accounting for 20% of the probability), hysterectomy was not the most effective option (in terms of total number of QALYs). Therefore, the CEAF never goes above 80% and the EVPI does not go to zero, which would suggest that there is value in finding out whether if the uncertainty was removed hysterectomy would remain the most cost-effective option.

At an ICER of £1440, the preferred option changes from the Mirena coil to hysterectomy, so there is a discontinuity both in the CEAF as well as in the gradient of the EVPI curve presented in Figure 21.

Figure 22 presents the difference between the expected net benefit of the optimal strategy allowing for perfect information and the expected net benefit of the optimal strategy given current information for any given threshold ICER.

Subgroup analysis

Utilities are changed from the base case which used mean utility values to median utility values

In the base case analysis the mean value for the health state of ‘well’ post ablation was lower (0.73) than the mean value for the health state of first-generation ablation (0.76). Furthermore, the health state of ‘well’ post Mirena coil is not available in the report by Sculpher132 and so a value was taken from a different study by Hurskainen et al. 27 As a result the health state for ‘well’ post Mirena coil is lower than the health state of ‘well’ post repeat ablation. Thus, although mean values are most appropriate, the use of mean values for utilities that are available in alternative published sources in our base case presents some slightly counterintuitive results.

Other studies have tackled this issue by using median values from different studies rather than the mean, which leads to a more intuitive set of values being used. But there is no other sensible justification for using median rather than mean values.

We carried out two sensitivity analyses to explore the impact of using the median reported values as opposed to the means.

Assumption that all (100%) women undergo hysterectomy after a failed repeat ablation

In the sensitivity analysis we change this assumption and assume that 10% of women remain symptomatic and only 90% follow the hysterectomy strategy after a failed repeat ablation

The deterministic results for this sensitivity analysis are presented in Table 21.

In this analysis, although the total cost and QALYs for each strategy are different to those in the base case, the overall result is unchanged from the base case.

Two sensitivity analyses deemed worthy of investigation before the main analysis was undertaken were not carried out for the following reasons:

1. Subgroup analysis on women below 40 years of age.  On scrutinising the data it was clear that women below the age of 40 years had a 100% satisfaction rate with the Mirena coil strategy and therefore it was very clear that any subgroup analysis on the basis of age would be dominated by the strategy of coil. The base case uses ‘all ages of women’ – and so to specifically analyse the subgroup of women older than 40 years would prove fruitless because that result was captured by the main analysis.

2. Sensitivity analysis – varying the cost of anaesthetic.  In the base case analyses, the cost of second-generation techniques (MEA and TBEA) was based on the estimates used by Garside et al. 16 The cost estimate from Garside et al. 16 already takes into account local versus general anaesthesia. It was thus considered that any recalculations/re-estimations to explore the minor difference in costs of LA versus GA observed in Seymour et al. 142 would add more uncertainty.

Statement of principal findings

The results of base case Analysis 2 show that the strategy of hysterectomy is most cost-effective. Hysterectomy dominates first-generation EA and, although more expensive, produces more QALYs than the other strategies. The ICER for hysterectomy compared with Mirena is £1440 per additional QALY. Compared with the second-generation EA strategy, the ICER for hysterectomy is £970 per additional QALY. These results suggest that hysterectomy would be considered the most cost-effective strategy in the light of the acceptable thresholds used by NICE, which tends to accept new technologies if the ICER is within £20,000 per additional QALY.

The results of this study were robust to all the sensitivity and subgroup analysis that were carried out with the exception of the sensitivity analysis carried out on the QALY data. The results of the main analysis reported in this study are based on an analysis that used the reported values of QALYs that are available in the published literature, specifically the ‘mean’ reported QALY values. When we carried out the same analysis using the median QALY values, the results changed and second-generation EA became the most cost-effective strategy, dominating both the first-generation ablation strategy and the hysterectomy strategy. In the sensitivity analysis the second-generation EA strategy was more expensive than the Mirena strategy but also produced more QALYs. The ICERs that resulted suggested that second-generation EA would be considered the most cost-effective strategy.

Strengths and limitations

The major strength of the economic component of this study is that it is based on a state-of-the-art Markov model which was informed by data from an IPD meta-analysis of randomised trials. A multidisciplinary team including economists, expert clinicians and statisticians provided input into the model structure primarily based on the evidence in the literature. All assumptions used in the model were based on the available evidence as far as possible. Assumptions were agreed with the team before the analysis was carried out and without knowledge of how these assumptions would affect the results.

In terms of limitations, not all aspects of outcome have been included because of the limited time scale in our model and the lack of long-term data. For strategies such as hysterectomy, for instance, there is a finality associated with the procedure which may have an effect on women and, in the long term, have implications on QoL that have not been captured in our model. These include the possibility of long-term complications such as urinary incontinence for which surgery is required. Furthermore, the utilities used reflect the satisfaction of the outcomes only. It is clear that once women have had a hysterectomy their satisfaction is high since in contrast to the other interventions they experience no bleeding at all. But the utility measure does not capture the anxiety prior to hysterectomy associated with major surgery and GA. It is conceivable that such anxiety may lead to decisions that avoid the strategy and to try other options for as long as possible.

In addition, the fact that the complexity of the model contributed to a long running time has some effect on the extent of sensitivity analyses that were undertaken. Both the main analysis and sensitivity/subgroup analysis had a long model running time and required laborious recalibration, which meant that additional sensitivity and subgroup analyses were not undertaken without serious consideration.

Strengths and weaknesses and assumptions in relation to other studies

With regard to the utility values that have been used in the current study the following points are worth noting.

First, the paper by Sculpher132 has been extensively referenced in the literature. Other studies (including Garside et al. 16) have used the median values from that paper and not the means. However, for economic evaluation it is the mean values that are most appropriate to use. 145,146

There are minor inconsistencies in the results presented by Sculpher132 when looking at the QALY values for ‘well post TCRE’ and ‘convalescence post TCRE’ compared with the ‘well post hysterectomy’ and ‘convalescence post hysterectomy’ states: the mean QALY value for the ‘well post TCRE’ state is 0.73 while the mean QALY value for the ‘convalescence post TCRE’ state is 0.76. One would expect that ‘well post TCRE’ state would have a higher value than the ‘convalescence post TCRE’ state, which is the case for the median QALY values. This is also true for the mean and median QALY values of the ‘well post hysterectomy’ and ‘convalescence post hysterectomy’ states. Despite these apparent inconsistencies, we used the mean values as they were presented in the literature and considered it possible that this might be explained by the greater initial relief that a woman might experience after a hysterectomy than after a TCRE, although no explanation was put forward in the original literature.

Comparison with other studies

Our results are consistent with those reported by Garside et al. ,16 who compared MEA and TBEA with TCRE, RB and hysterectomy. A state-transition (Markov) model was used and assumed a hypothetical cohort of 1000 patients for a period of 10 years. Our model used the same assumptions as Garside et al. 16 regarding the age of women entering the model, who were assumed to be 42 years; the model ran for 10 years and women became menopausal at age 52 years.

Garside et al. concluded that hysterectomy is cost-effective compared with MEA and TBEA. They found that TBEA dominated all other ablation techniques.

In addition, when compared with hysterectomy, MEA and TBEA were found to be less costly but provided fewer QALYs. Garside et al. ’s ICER for hysterectomy compared with TBEA was £2410 per QALY, and for hysterectomy compared with MEA their ICER was £2108 per QALY. The authors found that, when comparing MEA and TBEA with TCRE, RB and hysterectomy, the model was highly sensitive to the utility values associated with being well following ablation. Garside et al. 16 recommended that results are interpreted with caution owing to the sensitivity of the model to the utility values used.

The results of our study do not concur with the result of the trial by Hurskainen et al. 27 from Finland, which compared Mirena with hysterectomy. Mirena was found to be cost-effective at 5 years when compared with hysterectomy. There was no statistically significant difference in QoL scores at 5 years, as measured by the EQ-5D instrument, between the two treatment groups. Mean direct costs in the Mirena arm remained significantly lower ($1892) than in the hysterectomy arm ($2787) despite 40% of women in the Mirena arm going on to have a hysterectomy.

Economic modelling undertaken to inform NICE guidelines on HMB (NICE CG44, 2007)15 shows that Mirena is cost-effective when compared with both hormonal and non-hormonal treatment. It generates more QALYs at a lower cost than any other medical or surgical treatment strategy considered. This analysis also considered surgery as a comparator treatment. The surgical strategy produced fewer QALYs at a higher cost than Mirena. The NICE model assumed that, within the 5-year lifespan of Mirena, some women who experienced failure would move straight to hysterectomy (based on data from Hurskainen et al. 27). In contrast, in our study the assumption (based on advice from clinical colleagues) was that all women who experienced failure with the Mirena strategy would follow the second-generation EA pathway in the first instance.

However, while hysterectomy clearly comes out on top, the available long-term follow-up data on Mirena use are so inconsistent that we have to be cautious in our interpretation.

Meaning of the study

The results of this study suggest that hysterectomy is more cost-effective than either ablation or Mirena. These results are based on ‘mean’ reported values of utilities in published studies and are highly sensitive to the utility values that are used. When ‘median’ values for utility estimates are used, the strategy of second-generation ablation becomes the most cost-effective strategy. To the current authors there is no clear justification for using median values but they have been used in similar prominent studies without such justification. We assume that this is because some of the ‘mean’ values reported in the literature appear inconsistent. Replacing ‘mean’ values that appear inconsistent with ‘median’ values without clear justification will bias any results. The clear sensitivity of the results to the utility values serves to highlight the importance of using appropriate and robust data for utilities.

The study shows that first-generation ablation techniques are less cost-effective than second-generation techniques whatever utility values are used. Based on available data, Mirena does not come through as a cost-effective strategy compared with second-generation ablation or hysterectomy.

Hysterectomy

1. #1 EXPLODE “hysterectomy”/all sub-headings

2. #2 “vaginal hysterectomy”/all sub-headings

3. #3 “total abdominal hysterectomy”

4. #4 “subtotal abdominal hysterectomy”

5. #5 “laparoscopic hysterectomy”

6. #6 #1 OR #2 OR #3 OR #4 OR #5

Ablation

1. #1 EXPLODE “hysteroscopy”/all sub-headings

2. #2 (“transcervical resection”) NEAR “endometrium”

3. #3 “TCRE”

4. #4 “endometrial ablation”

5. #5 “laser ablation”

6. #6 “electrosurgery”

7. #7 “rollerball”

8. #8 “thermal balloon”

9. #9 “hypertherm$”

10. #10 “thermotherapy”

11. #11 “photodynamic therapy”

12. #12 “phototherapy”

13. #13 “cryoablation”

14. #14 “microwave ablation”

15. #15 “radiofrequency”

16. #16 “saline irrigation”

17. #17 “laser interstitial”

18. #18 “Thermachoice”

19. #19 “Cavaterm”

20. #20 “ELITT”

21. #21 “Vesta”

22. #22 “Novasure”

23. #23 “Microsulis”

24. #24 “Cryogen”

Mirena

1. #1 EXPLODE “contraceptive”/all sub-headings

2. #2 “Mirena^® coil”/all sub-headings

3. #3 “levonorgestrel”

4. #4 “intra uterine device”

5. #5 #1 OR #2 OR #3 OR #4

Duration of surgery (minutes)

Duration of hospital stay (days)

Surgery pain score

Time to return to work (days)

Time to return to normal activities (days)

Time to return to sexual activity (days)

Proportion with dyspareunia

SF-36 scores (absolute values)

SF-36 scores (change from baseline)

EQ-5D scores (absolute values)

EQ-5D scores (change from baseline)

Proportion requiring repeat endometrial ablation

Proportion requiring hysterectomy after endometrial ablation

Number of patients with adverse events – periprocedure

Number of patients with adverse events – postoperatively (within 1 month)

SF-36 scores (absolute values)

SF-36 scores (change from baseline)

EQ-5D scores (absolute values)

EQ-5D scores (change from baseline)

Proportion discontinuing Mirena

Proportion requiring hysterectomy after Mirena

Number of patients with adverse events – periprocedure

Number of patients with adverse events – postoperatively (within 1 month)

Proportion with amenorrhoea

Proportion with menorrhagia

Bleeding/pictorial blood loss assessment chart scores (change from baseline)

Proportion with dysmenorrhoea

Duration of surgery (minutes)

Use of general anaesthesia

Surgery pain score

Time to return to work (days)

Time to return to normal activities (days)

Proportion with dyspareunia

SF-36 scores (absolute values)

SF-36 scores (change from baseline)

EQ-5D scores (absolute values)

EQ-5D scores (change from baseline)

Proportion requiring repeat endometrial ablation

Proportion requiring hysterectomy after endometrial ablation

Number of patients with adverse events – periprocedure

Number of patients with adverse events – postoperatively (within 1 month)

Proportion with amenorrhoea < 12 months

Proportion with amenorrhoea – 12 months

Proportion with amenorrhoea – 2 years

Proportion with amenorrhoea – 3 years

Proportion with heavy bleeding – < 12 months

Proportion with heavy bleeding – 12 months

Proportion with heavy bleeding – 2 years

Proportion with heavy bleeding – 3 years

Bleeding/pictorial blood loss assessment chart scores (change from baseline) – < 12 months

Bleeding/pictorial blood loss assessment chart scores (change from baseline) – 12 months

Bleeding/pictorial blood loss assessment chart scores (change from baseline) – 2 years

Bleeding/pictorial blood loss assessment chart scores (change from baseline) – 3 years

Proportion with dysmenorrhoea – < 12 months

Proportion with dysmenorrhoea – 12 months

Proportion with dysmenorrhoea – 2 years

SF-36 general health (absolute values) – 12 months

SF-36 physical function (absolute values) – 12 months

SF-36 mental health (absolute values) – 12 months

SF-36 vitality (absolute values) – 12 months

SF-36 physical role limitation (absolute values) – 12 months

SF-36 emotional role limitation (absolute values) – 12 months

SF-36 social function (absolute values) – 12 months

SF-36 bodily pain (absolute values) – 12 months

SF-36 general health (absolute values) – 2 years

SF-36 physical function (absolute values) – 2 years

SF-36 mental health (absolute values) – 2 years

SF-36 vitality (absolute values) – 2 years

SF-36 physical role limitation (absolute values) – 2 years

SF-36 emotional role limitation (absolute values) – 2 years

SF-36 social function (absolute values) – 2 years

SF-36 bodily pain (absolute values) – 2 years