|Home | About | Journals | Submit | Contact Us | Français|
Objective To determine the relative effectiveness of second generation ablation techniques in the treatment of heavy menstrual bleeding.
Design Network meta-analysis on the primary outcome measures of amenorrhoea, heavy bleeding, and patients’ dissatisfaction with treatment.
Data sources Nineteen randomised controlled trials (involving 3287 women) were identified through electronic searches of the Cochrane Library, Medline, Embase and PsycINFO databases from inception to April 2011. The reference lists of known relevant articles were searched for further articles. Two reviewers independently selected articles without language restrictions.
Eligibility criteria for selecting studies Randomised controlled trials involving second generation endometrial destruction techniques for women with heavy menstrual bleeding unresponsive to medical treatment.
Results Of the three most commonly used techniques, network meta-analysis showed that bipolar radiofrequency and microwave ablation resulted in higher rates of amenorrhoea than thermal balloon ablation at around 12 months (odds ratio 2.51, 95% confidence interval 1.53 to 4.12, P<0.001; and 1.66, 1.01 to 2.71, P=0.05, respectively), but there was no evidence of a convincing difference between the three techniques in the number of women dissatisfied with treatment or still experiencing heavy bleeding. Compared with bipolar radio frequency and microwave devices, an increased number of women still experienced heavy bleeding after free fluid ablation (2.19, 1.07 to 4.50, P=0.03; and 2.91, 1.23 to 6.88, P=0.02, respectively). Compared with radio frequency ablation, free fluid ablation was associated with reduced rates of amenorrhoea (0.36, 0.19 to 0.67, P=0.004) and increased rates of dissatisfaction (4.79, 1.07 to 21.5, P=0.04). Of the less commonly used devices, endometrial laser intrauterine thermotherapy was associated with increased rates of amenorrhoea compared with all the other devices, while cryoablation led to a reduced rate compared with bipolar radio frequency and microwave.
Conclusions Bipolar radio frequency and microwave ablative devices are more effective than thermal balloon and free fluid ablation in the treatment of heavy menstrual bleeding with second generation endometrial ablation devices.
Heavy menstrual bleeding is a common incapacitating problem in women of reproductive age,1 which can severely affect their quality of life.2 3 When medical treatments fail to provide adequate relief, surgical interventions, including hysterectomy or destruction of the endometrium, can be considered.4 Initially, rollerball ablation, transcervical resection, and laser ablation were the predominant endometrial destruction techniques performed under direct hysteroscopic vision. These led to a rapid decrease in the number of hysterectomies performed.5
Over the past decade, a second generation of non-hysteroscopic techniques, which are safer, technically easier to perform, involve shorter hospital stays, and can be performed under local anaesthesia, have become dominant.6 7 These involve devices that are sited and activated to treat the whole endometrial cavity. Destruction is achieved through various methods, the most commonly used8 being high temperature fluids within a balloon (Thermachoice and Cavaterm)9 10 11 or application of microwave (Microsulis)12 13 or bipolar radiofrequency electrical energy (Novasure).14 Less commonly used ablative techniques include free fluid at high temperature (Hydrothermablator),15 endometrial laser intrauterine thermotherapy (ELITT),16 and cryoablation (HerOption).17 Other than free fluid thermal ablation, these are “blind” techniques. A recent meta-analysis of individual patient data found that second generation techniques were at least as effective, in terms of patients’ dissatisfaction, as first generation techniques and, importantly, gave rise to fewer operative complications, were considerably quicker, and were less likely to need to be performed under general anaesthesia.18 Thus they are clearly preferable to first generation techniques.
When more than two treatments options are available, the use of network meta-analysis, which allows us to combine both direct and indirect estimates for treatment effects, has become increasingly common.19 20 Randomised comparisons within trials are preserved and the main assumption—that there are no systematic differences between the trials that could bias the indirect measurements—is similar to those underlying any standard meta-analysis.
There have been numerous randomised controlled trials comparing the different second generation techniques with each other or with first generation devices, yet with the sheer number of potential comparisons there are gaps in direct evidence for efficacy of some techniques, while other comparisons have low statistical power because of the small numbers randomised. We carried out a network meta-analysis to assess the relative effectiveness of six second generation techniques to provide both users and recipients of these devices with an improved evidence base to inform their decision making.
We carried out an extensive literature search in the Cochrane Library, Medline, Embase, and PsycINFO from inception to April 2011. Our search term combinations consisted of MeSH subheadings, text words and word variations for “heavy menstrual bleeding,” and the various types of first and second generation endometrial destruction techniques. This search strategy was adapted to suit each database and was restricted to “humans” and “females” (see appendix 1). Bibliographies of known relevant primary articles and reviews were searched by hand to identify articles missed by the electronic searches. Were also contacted experts in the specialty to uncover any grey literature. A comprehensive database was constructed with Reference Manager 12.0 to store all identified references. No language restriction was applied.
Studies were selected in a two step process. Firstly, we scrutinised the citations identified by the electronic searches and subsequently obtained full manuscripts of citations that met or were thought to have met the predetermined inclusion criteria. We included studies if they recruited women with heavy, abnormal, excessive, or prolonged uterine bleeding that was unresponsive to initial medical treatment and the study design was a randomised controlled trial comparing endometrial ablation methods.
Two reviewers (LJM and RC) independently inspected all the manuscripts to determine if they met the above criteria. When possible we obtained individual patient data from the primary author using methods described in the related meta-analysis.18 When raw data were not available, two independent reviewers (RC and LJM) used predesigned forms to extract aggregate data from manuscripts. Any disagreements were resolved by consensus or arbitration by a third reviewer (JPD).
Outcome measures used to evaluate the treatments were those most commonly used across studies (table 11):): rate of amenorrhoea (converted from a pictorial bleeding assessment score of zero when these data existed, otherwise as reported), rate of heavy bleeding (likewise converted from a pictorial bleeding assessment score of >10024 when data existed, otherwise as reported), and rate of dissatisfaction with treatment (see related paper for a full description of how dissatisfaction was calculated).18 We used data at 12 months (the most popular time point) as the focus for analysis or results at two years if 12 month data were not available.
For all three outcomes we first assessed direct estimates in individual trials by calculating point estimates and 95% confidence intervals. We then calculated a traditional weighted average meta-analysis yielding a Mantel-Haenszel odds ratio, with heterogeneity investigated with Cochran’s Q25 and I2 statistics.26
We carried out the network meta-analysis by fitting a linear mixed model to the log odds ratio from each trial with the GLIMMIX procedure in SAS, using maximum likelihood to estimate treatment effects. Two models were attempted; the first included a fixed treatment effect and allowed the effect of the study to vary randomly. The second model also included a random effect for the difference in treatment in each trial by including a study by treatment interaction parameter.27 The inclusion of a random study effect allowed the recovery of any differences between trials, which has greater importance when not all of the treatments are included in every trial, as is the case here.28 We also included a random effect for treatment pair (that is, to separate those trials comparing different treatments types) in both models to allow for any inconsistency or “incoherence” in the model.29 This effect (ω, the standard deviation of the random effect) allows for any indirect estimates to be inconsistent with any direct estimates of the same treatments and can be calculated where there any closed “loops” of treatments exist (for instance, A v B, B v C, A v C). Increased incoherence increases standard errors, and hence the size of the confidence interval, of differences between treatments. If large levels of incoherence are present it might not be sensible to combine estimates with this method,30 analogous to having high levels of heterogeneity in a standard meta-analysis. We used Revman v5.0 (Cochrane Collaboration, Denmark) and SAS v9.2 (SAS Institute, Cary, NC) software for analysis.
We identified 719 citations through the electronic literature searches (fig 11)) and excluded 672 after screening titles and abstracts. A further 28 were excluded for being duplicate publications or not involving direct comparisons of endometrial ablative techniques. After detailed evaluation of the citations, 19 primary articles met the selection criteria and were included in the network meta-analysis (see appendix 2 for full details of these 19 studies). These consisted of five head to head comparisons of second generation devices31 32 33 34 35 (744 women) and 14 comparing a second generation device with a first generation (hysteroscopic) device9 12 14 15 16 17 36 37 38 39 40 41 42 43 (2543 women) (fig 22).). Table 1 shows further descriptions of device type and the frequencies used in the analysisanalysis.. As the fixed and random effect network meta-analysis models produced similar results only those from the latter are described here.
Direct comparisons show an increased rate of amenorrhoea with bipolar radio frequency ablation compared with thermal balloon ablation (odds ratio 4.56, 95% confidence interval 2.24 to 9.26; P<0.001) (table 22).). Free fluid thermal ablation was associated with reduced rates of amenorrhoea (0.36, 0.18 to 0.73; P=0.005) and increased rates of heavy bleeding (4.88, 1.32 to 18.11; P=0.02, table 33)) compared with bipolar radio frequency ablation. We found no other convincing differences. When there was more than one study in a comparison, estimates between studies were highly consistent (all I2=0% apart from the comparison of amenorrhoea rates with microwave ablation versus first generation techniques where I2=17%).
Results from the network meta-analysis (tables 2 and 3 and fig 33)) concurred with the results of the direct evidence, with an increased rate of amenorrhea with bipolar radio frequency ablation compared with thermal balloon ablation (2.51, 1.53 to 4.12; P<0.001). Results were also similar for the comparison between free fluid thermal ablation and bipolar radio frequency ablation for amenorrhoea (0.36, 0.19 to 0.67; P=0.004) and heavy bleeding (2.19, 1.07 to 4.50; P=0.03). There was an increase in the rate of amenorrhoea with radio frequency ablation compared with cryoablation in the indirect comparison (0.20, 0.09 to 0.49; P=0.002). Microwave ablation was associated with an increased rate of amenorrhoea compared with thermal balloon ablation (1.66, 1.01 to 2.71; P=0.05) and cryoablation (0.31, 0.13 to 0.74; P=0.01) and some reduction in the rate of heavy bleeding compared with free fluid ablation (2.91, 1.23 to 6.88; P=0.02). The endometrial laser intrauterine thermotherapy (ELITT) was associated with increased rates amenorrhoea compared with all the other devices, with wide confidence intervals reflecting the small number of women evaluated with this device in randomised controlled trials (table 22).). No other compelling differences between devices were noted. There were relatively low but variable rates of heavy bleeding across the studies (such as 0-32% for thermal balloon ablation and 8-18% for bipolar radiofrequency ablation).
Direct comparisons showed some evidence of reduced rates of dissatisfaction with bipolar radio frequency ablation compared with thermal balloon ablation (0.39, 0.16 to 0.91; P=0.03) (table 44),), though this difference was not as convincing in the network meta-analysis (0.56, 0.28 to 1.09; P=0.09). Increased dissatisfaction was seen with free fluid thermal ablation compared with bipolar radio frequency (9.40, 1.14 to 77.18; P=0.04) and a similar result was seen in the network meta-analysis (4.79, 1.07 to 21.48; P=0.04). There were no other obvious differences. Direct estimates between studies were also consistent for dissatisfaction where more than one study in a comparison existed (0%<I2<4%). Overall rates of dissatisfaction were generally low (such as 0-7% for bipolar radiofrequency ablation).
Estimates of incoherence, ω, from the network model of amenorrhoea rate were relatively high at 0.23 compared with other studies that have used this technique to evaluate multiple treatments simultaneously.19 30 44 When we removed the single trial comparing microwave ablation with thermal balloon ablation,33 this estimate was reduced to 0.13. This trial’s direct estimate yielded no significant difference between treatments (1.13, 0.70 to 1.82; P=0.6), which conflicted with the estimate expected by any indirect comparison where the network analysis favoured microwave ablation (1.66, 1.01 to 2.71; P=0.05). A sensitivity analysis of the network analysis with this trial excluded produced a higher estimate in favour of microwave ablation (2.23, 1.21 to 4.09; P=0.02), with some other minor adjustments to the other estimates (free fluid ablation was now associated with a lower amenorrhoea rate also compared with microwave ablation: 0.45, 0.21 to 0.97; P=0.04). Incoherence parameters for heavy bleeding and dissatisfaction, along with the interaction parameter for study by treatment, were estimated to be so close to zero that they were negligible and so effectively did not contribute to any estimates.
Of the three most popular second generation endometrial ablation devices (thermal balloon, bipolar radio frequency, and microwave) examined in this network meta-analysis, bipolar radio frequency and microwave ablation techniques seem to have an advantage over thermal balloon ablation in terms of increased amenorrhea rate in women with heavy menstrual bleeding. Compared with free fluid ablation, both bipolar radio frequency and microwave ablation techniques were associated with a reduction in the number of women still experiencing heavy bleeding. Radio frequency ablation was also associated with an increased amenorrhoea rate and a reduced number of women dissatisfied with treatment compared with free fluid ablation. Endometrial laser intrauterine thermotherapy (ELITT) was also associated with high rates of amenorrhoea compared with hysteroscopic ablation in the small single study that evaluated this technique. Not surprisingly this translated to estimates of amenorrhoea higher than the other devices when we combined results in the network meta-analysis. Conversely, cryoablation was associated with lower rates of amenorrhoea compared with bipolar radio frequency and microwave ablative techniques. Rates of dissatisfaction with treatment or ongoing heavy bleeding were generally low across all studies. Apart from the comparison of microwave ablation versus thermal balloon ablation, estimates from the network meta-analysis were consistent with available direct estimates between second generation devices in terms of direction and significance of treatment effect. The implication of this inconsistency from this one comparison is unclear.
We used the relatively new approach of network meta-analysis to synthesise data, and this provided effect estimates when direct comparisons were not available. We used optimal methods, complying with guidelines on reporting of systematic reviews and meta-analyses.23 We carried out an extensive literature search, with no language restrictions, minimising the risk of missing information. The collection of individual patient data from some studies improved estimates.
The review was not intended to be a comprehensive review of all facets of these devices, and we focused only on outcomes of bleeding after treatment reported by patients and patients’ dissatisfaction with treatment. Many other factors, such as ease of use and safety of device, are taken into account in the decision regarding a particular device. For example, the time to complete treatment is significantly shorter with bipolar radiofrequency ablation, and it results in greater coverage of the endometrial surface than thermal balloon ablation, in an outpatient setting.31
Because of the limited amount of data (several comparisons consisted of only one trial) and the fact that the traditional meta-analysis did not show any evidence of heterogeneity between estimates (I2 was generally 0% and never greater than 17%) we did not attempt to further subgroup results by manufacturer or type (Thermachoice, Cavaterm, and the discontinued Vesta all are types of thermal balloon ablative device). This consistency could have occurred by chance, and there could be some differences because of the design of similar products. This could be reasonably evaluated only should more data become available. We considered that the first generation ablative techniques of transcervical endometrial resection and rollerball electrocoagulation were equivalent so we used them as a generic intermediate comparator.45 46
Estimates of incoherence were relatively high for the analysis of amenorrhoea rate (less when we omitted one of the studies in a sensitivity analysis), though how this impacts on adjusted estimates is debatable, given that confidence intervals of estimates were increased as a result. Like all meta-analysis, we must consider limitations with combining estimates from different studies because of clinical heterogeneity such as differences in populations of patients, even with statistical adjustments in the analysis. For example, the small study that compared endometrial laser intrauterine thermotherapy (ELITT) with first generation techniques seemed to show a large treatment effect. Given that none of the other studies has shown a comparable effect, there could be something particular about this study, such as the lower rate of amenorrhoea for the first generation comparator, that separates it from the rest, but in the absence of raw data, we were unable to investigate this further. Conversely, the absence of a significant difference does not necessarily mean that techniques are comparable but could be a consequence of small underpowered trial populations. This is reflected in wide confidence intervals for some of the estimates. We did not take into account quality of study in these analyses; the only requirement was that the study needed to be randomised in a specified population of patients.
The usefulness of the network meta-analysis in this setting was highlighted during the course of this review when we updated the analysis to include the results of the most recent study of bipolar radio frequency and free fluid endometrial ablation32; the first to compare these two treatments directly. Results from this study for the primary outcome measure of amenorrhoea rate at 12 months suggested improvements with radio frequency, with an odds ratio of 0.36 (0.17 to 0.73). Before inclusion of this later study, our original estimated odds ratio from the network meta-analysis was 0.33, which is similar albeit with a slightly wider confidence interval (0.11 to 0.98). Such an accurate estimate could of course occur by chance, but it does show how useful this analytical method could be where direct head to head assessments of treatments are not available.
Rates of dissatisfaction and heavy bleeding are consistently low for second generation techniques, which, as a group, represent an excellent conservative alternative to hysterectomy. It is clear from the results of our review that bipolar radiofrequency and microwave ablation technique are associated with higher amenorrhea rates than other methods, although it should be reiterated that only hysterectomy can guarantee complete cessation of menstrual bleeding. Endometrial laser intrauterine thermotherapy (ELITT) was associated with highly significant improvements in amenorrhea rates compared with all other methods, albeit from one small study, but this device is no longer marketed (C Dunn, Karl Storz, personal communication).
Data from UK hospital episode statistics show a significant increase in the overall number of inpatient endometrial ablative techniques (fig 44),), though radiofrequency ablation techniques are recently the most rapidly rising techniques used. These data do not include ambulatory procedures, which are also increasing rapidly but not adequately coded in the hospital episode data. More recently, the microwave endometrial ablation system from Microsulis has been withdrawn from the European market after Hologic acquired the intellectual property for the device (Microsulis Medical, press release, 2011). Ambiguous categorisation of first generation techniques in the episode data precludes comparison of relative rates, but consensus is that such techniques have rapidly been superseded.
With respect to costs and cost effectiveness, one study has directly compared thermal balloon and microwave ablative techniques.47 The mean total health service cost per patient was £181 (£70 to £434) more (equivalent to about €219, $287) for the thermal balloon system than for the microwave ablative device (at 2010 costs), though the study found no significant differences between the total non-health costs and health benefits of the two devices. The incremental cost effectiveness ratio showed microwave ablation, on average, dominated thermal balloon ablation at one year after the procedure.
Our previous review indicated that levels of dissatisfaction with second generation techniques, as a class of procedure, were slightly lower than those associated with first generation techniques. Second generation techniques were also quicker, associated with faster recovery times and fewer adverse procedural events, and could be carried out under local anaesthesia.18 Thus they should be offered when no structural or histological abnormality is present and the woman wants to retain her uterus. While chances of pregnancy are reduced after endometrial ablation, they can occur and such pregnancies have a high risk of adverse outcomes, hence continued contraception is strongly advised.48
As to the question of the best second generation technique, bipolar radiofrequency ablation seems to offer benefits over thermal balloon in terms of amenorrhea rates and is procedurally quicker than thermal balloon and microwave ablation.31 33 34 35. Any impact on other outcomes was not significant in either direct comparisons or network meta-analysis. Only two studies have evaluated the use of these three devices under local anaesthesia,31 33 so while outpatient hysteroscopic treatment is feasible, it would be premature to state any advantage for either anaesthetic approach. The decision should be left to the women’s preference and the operator’s experience, though the withdrawal of the microwave ablation device reduces available choices.
Further large scale rigorous randomised trials to compare existing and emerging ablative techniques should use meaningful and standardised measures of satisfaction and menstrual bleeding and be run independently from the manufacturers of the devices. Trials incorporating different anaesthetic approaches will also help to refine the evidence available to women and gynaecologists. Concerns with regard to the safety of the “blind” techniques with respect to extrauterine injury should lead to further evaluation of visual ablative techniques such as the next generation of free fluid ablation. Adequate assessment of relatively rare adverse events, however, would require large cohorts similar to the MISTLETOE study.49
We thank all authors of identified trials who sent data through the previous related review.18
Members of Collaborative Group
The following authors provided us with individual patient data from their trials for this review: J Abbott, University of New South Wales, Sydney, Australia; M Y Bongers, Maxima Medical Centre, Veldhoven, Netherlands; J-L Brun, Hopital Universitaire Pellegrin, Bordeaux, France; T J Clark, Birmingham Women’s Hospital, Birmingham, UK; J Cooper (2004 trial), data supplied by Microsulis Medical, Hampshire, UK; K G Cooper, Aberdeen Royal Infirmary, Aberdeen, UK; S L Corson (2001 trial), data supplied by Boston Scientific Corporation, Marlborough, US; J Hawe, Countess of Chester Hospital, Chester, UK; W R Meyer; data supplied by Ethicon, NJ, US; J P M Penninx, Maxima Medical Centre, Veldhoven, Netherlands; A M Sambrook, Aberdeen Royal Infirmary, Aberdeen, UK; I A van Zon-Rabelink, Medical Spectrum Twente, Enschede, Netherlands.
Contributors: JPD, KSK, and SB conceived and designed the review. JPD, RC, and LJM performed the literature search and data extraction. LJM conducted the network meta-analysis. JPD and LJM drafted the paper for critical review by KSK, KC, BWJM, and SB. All authors gave final approval of the manuscript. JPD is guarantor.
Funding: This study was partly funded by the National Institute for Health Research (NIHR) Health Technology Assessment project (reference 05/45/02).50 Further support came from the University of Birmingham Clinical Trials Unit, which is funded by an infrastructure grant from the NIHR. The views and opinions expressed therein are those of the authors and do not necessarily reflect those of the HTA programme, NIHR, NHS, or the Department of Health.
Competing interests: All authors have completed the ICMJE uniform disclosure form at www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare: no support from any organisation for the submitted work; no financial relationships with any organisations that might have an interest in the submitted work in the previous three years; JPD, LJM, KC, and BWJM were authors of studies included in the analysis.
Ethical approval: Not required.
Data Sharing: No additional data available.
Cite this as: BMJ 2012;344:e2564