- Top of page
Objective To compare the effectiveness of two second generation endometrial ablation techniques (microwave and thermal balloon endometrial ablation) with first generation techniques of endometrial ablation to treat heavy menstrual bleeding in women.
Search strategy We searched the Cochrane Library (issue 3, 2002), the National Research Register, MEDLINE (1966 to August 2002), Embase (1980 to August 2002) and Web of Science Proceedings (all years). We also searched reference lists and contacted experts and manufacturers in the field.
Selection criteria Randomised controlled trials and controlled trials of microwave endometrial ablation and thermal balloon endometrial ablation versus transcervical resection and rollerball ablation, alone or in combination, to treat heavy menstrual bleeding were included.
Data collection and analysis Two reviewers independently selected studies for inclusion and extracted data. As there was considerable clinical and methodological heterogeneity among the studies, meta-analysis was not undertaken and results are presented descriptively.
Results Two randomised controlled trials of microwave endometrial ablation and eight trials (six randomised controlled trials) of thermal balloon endometrial ablation were included in the review. No significant differences were found between first and second generation techniques in terms of amenorrhoea, bleeding patterns, pre-menstrual symptoms, patient satisfaction or quality of life. Microwave endometrial ablation and thermal balloon endometrial ablation had significantly shorter operating and theatre times than first generation techniques. Adverse effects were few with all techniques, but there were fewer peri-operative adverse effects with second generation techniques.
Conclusion Microwave endometrial ablation and thermal balloon endometrial ablation are alternatives to first generation techniques for treating heavy menstrual bleeding. No head-to-head trials of microwave endometrial ablation and thermal balloon endometrial ablation have been undertaken and there is not yet enough evidence of differences in clinical effectiveness between these two techniques.
- Top of page
Electronic databases searched were the Cochrane library (2002), National Research Register (2002), MEDLINE (1966–2002/08), PubMED (previous 180 days from August 2002), Embase (1980–2002/08) and Web of Science Proceedings (from 1980). The full search strategy is available from the authors. In addition, the bibliographies of articles were searched for further relevant papers and recognised experts in the field and relevant manufacturers of ablation equipment were approached for further information.
Systematic reviews, randomised and non-randomised controlled trials of microwave or thermal balloon ablation versus transcervical resection or rollerball ablation for women with heavy menstrual bleeding were included.
Studies were excluded if they were animal models, preclinical/biological and uncontrolled studies were excluded as were narrative reviews, editorials and opinion pieces. Only studies available in English were included.
Abstracts were initially examined for inclusion independently by two researchers (RG and KS) and disagreements resolved by discussion. Full text papers were then obtained and decisions about inclusion made independently by two researchers (RG and KW) and disagreement resolved by a third (KS).
Data were extracted by one researcher (RG) and checked by a second (KW). Actual numbers were extracted where possible and, where necessary, results were recalculated on an intention-to-treat basis. Key outcomes were satisfaction with treatment and quality of life after surgery, time to return to normal activities, menstrual bleeding postsurgery, levels of further surgery, adverse events (intra-operative—fluid overload, perforation, visceral burn, haemorrhage; post-operative—pelvic pain, readmission, haemorrhage), operation details and resource use.
Systematic reviews were assessed for quality using the QUOROM checklist.3 This is omitted from this article for reasons of space but is fully reported elsewhere.4 The quality of randomised controlled trials was assessed in relation to aspects of internal validity (appropriateness of randomisation, blind assessment of outcomes, number of women randomised, excluded and lost to follow up, use of intention-to-treat analysis, use of a power calculation, timing, duration and location of study) and external validity (patient characteristics, inclusion and exclusion criteria, length of follow up) and results are presented descriptively. The generalisability of studies (Table 3) was rated as ‘high’ if there was a full description of the patient characteristics and inclusion and exclusion criteria, ‘medium’ if there was some description of these and ‘low’ if no details of patient group or inclusion and exclusion criteria were given. Main outcome measures and other outcomes in the included studies were extracted and tabulated together with methods of assessing these outcomes (Table 2).
Table 3. Methodological characteristics of included controlled trials. Details of randomisation and allocation concealment not applicable (N/A) where trials are non-randomised controlled trials.
|Author/Date||Randomised?||Adequate randomisation||Adequate allocation concealment||Comparability of groups||Same intervention to all pts?||% loss to follow up||Sample size calc.||ITT||Generalisability||Main outcome measured independently|
|Cooper et al. 19997||Y||Y||Y||Y||Y||9||Y||Ya||H||Y|
|Bongers et al. 20008||N||N/A||N/A||Y||N||U 38?||Y||Yb||M||Y|
|Meyer et al. 19989||Y||Y||U||Y||Y||46||Y||N||H||Y|
|Gervaise et al. 199910||N||N/A||N/A||Nc||Y||0||N||N/A||M||U|
|Soysal et al. 200114||Y||Y||Y||Y||Y||0||N||Y||H||Y|
|Pellicano et al. 200213||Y||Y||U||Y||Y||29||N||N||H||Y|
Table 2. Outcomes and their method of measurement.
|Author/Date||Main and other outcomes||Method of measurement||Details|
|Cooper et al.,19997 MEA||Patient satisfaction||Patient questionnaire||Totally or generally satisfied|
|Treatment acceptability||Patient questionnaire||Treatment acceptable, cure or acceptable symptomatic improvement, recommend treatment to others|
|Menstrual symptoms||Patient questionnaire||5-point scale for pain and bleeding added achieve composite bleeding and pain score|
|Operative details||Theatre questionnaire||Covered operating times, complications, need for post-op. analgesia, length of stay|
|Surgical re-intervention||Hospital data||Hysterectomy, further ablation|
|AEs||Hospital data||Intra-operative difficulties (procedure abandoned, equipment failure, blunt perforation, haemorrhage), post operative readmission pelvic pain|
|Microsulis 2002, MEA||Menstrual bleeding||Monthly PBAC diary||Success taken as ‘Score ≤75’|
|Operative details||Clinical notes||Type of anaesthesia, anaesthesia time, treatment time|
|Menstrual bleeding||PBAC||Amenorrhoea = PBAC score of zero.|
|Menstrual symptoms||FU Physician enquiry||Presence of dysmenorrhoea|
|Patient satisfaction||FU Physician enquiry||Very satisfied/satisfied or dissatisfied|
|Treatment acceptability||FU Physician enquiry||Acceptance of operation positive or negative|
|Surgical re-intervention||Clinical notes/FU||Hysterectomy, further ablation|
|Bongers et al., 200018 Thermachoice||Surgical re-intervention||FU Physician enquiry||Hysterectomy, further ablation|
|Menstrual bleeding||FU Physician enquiry||Duration of menstruation|
|Menstrual symptoms||FU Physician enquiry||Dysmenorrhoea|
|Patient satisfaction||FU Physician enquiry||4-point scale: Perfectly satisfactory, satisfactory, no treatment effect, complaint worsened|
|Meyer et al, 199819 Thermachoice||Patient satisfaction||FU Physician enquiry||3-point scale—very satisfied, satisfied, not satisfied|
|Menstrual bleeding||PBAC||Mean reduction in PBAC score, score ≤100,|
|Menstrual symptoms||FU Physician enquiry||Change in dysmenorrhoea, change in PMS|
|QoL||FU Physician enquiry||Ability to work outside the home, impact on QoL (severe to none),|
|Gervaise et al, 199910 Thermachoice||Amenorrhoea||Telephone interview||Absence of menses|
|Menstrual bleeding||Telephone interview||Characteristics of menstrual flow, number of pads used|
|Menstrual symptoms||Telephone interview||Degree of dysmenorrhoea|
|Operating time||Clinical notes||Mean operating time, ≤30 minutes|
|AEs||Unclear||Intra-operative complications, endometriosis, pregnancy|
|Surgical re-intervention||Telephone interview||Hysterectomy, further ablation|
|Soysal et al., 200114 Thermachoice||Blood loss||PBAC, Hb levels||Score, ≤75|
|Patient satisfaction||Physician enquiry||3-point scale but only ‘not very satisfied’ reported|
|Operative details||Clinical record||Operating time from insertion of main surgical tool to its removal|
|AEs||Clinical record||Intra-operative complications, post-operative pain score on 10-point linear scale.|
|Surgical re-intervention||Clinical record||Hysterectomy, further ablation|
|Zon-Rabelink, 200115 Thermachoice||Menstrual bleeding||PBAC||Score <185|
|Patient satisfaction||Not stated||Quality of life score (unspecified)|
|Operative details||Not stated||Operation time, post-op pain medication|
|AEs||Clinical FU||Intra-operative complications (perforation, laceration, electrolyte disbalance)|
|Brun, 2002 Cavaterm||Menstrual bleeding||Patient diary card||Adaptation of the PBAC—success bleeding score <80,|
|Satisfaction||Physician enquiry||4-point scale (excellent, good, moderate, bad)|
|QoL||Physician enquiry||Time to return to normal life after operation|
|AEs||Not stated||Intra-operative complications (cervical channel burns)|
|Pellicano et al., 200213 Cavaterm||Patient satisfaction||Patient questionnaire||4-point scale (excellent, good, moderate, no improvement)|
|Menstrual bleeding||Physician enquiry||Length of bleeding|
|Menstrual symptoms||Patient questionnaire||Pain on a 5-point scale (1 = no pain to 5 = intolerable pain)|
|Operative details||Clinical records||Operation time, blood loss, discharge time|
|QoL||Physician enquiry||Time to resumption of normal activities|
|AEs||Clinical record||Intra-operative (fluid overload, cervical tear, uterine perforation, conversion of operation to hysterectomy) post-operative (pain, fever, UTI, haemorrhage, blood transfusion, incontinence)|
|Surgical re-intervention||Clinical record||Reoperation rate|
|Romer, 19985 Cavaterm||Satisfaction||Physician enquiry||Satisfied with treatment|
|Menstrual bleeding||Physician enquiry||Bleeding patterns|
|Surgical re-intervention||Clinical FU||Further surgery|
Study results were tabulated and described using forest plots where appropriate to illustrate outcomes although study results were not statistically pooled as there were important differences in the population groups and outcome measures such as different definitions of heavy menstrual bleeding and primary outcomes, such as bleeding patterns measured in different ways (Tables 1,2). All reported outcomes are based on intention-to-treat analysis and where necessary these were recalculated by the authors.
Table 1. Characteristics of studies included in the review.
- Top of page
We identified 215 abstracts of which 67 full text articles were acquired. Thirteen trial reports relating to 10 studies were included in the review (see Fig. 1). These included one study which had been published in German but which was provided to us in English translation by the device manufacturer5; an unpublished report provided by a thermal balloon equipment manufacturer [Brun JL, et al. French randomised comparative multicentre study of Cavaterm (TM) versus rescetion in patients with menorrhagia: interim results. 2002. Unpublished work.] and details of a report submitted to the USA Federal Drug Administration (FDA) supplied by the makers of microwave endometrial ablation technology [Microsulis Americas. Microwave endometrial ablation FDA trial. 2002. Unpublished work].
Patient characteristics and surgical details for included trials are shown in Table 1 (median population 143, range 20–322). Overall, 1561 women were included. The main differences in study populations were definitions of heavy menstrual bleeding and the inclusion or exclusion of women with fibroids.
Table 2 shows the reported outcomes and methods of measurement.Table 3 describes methodological characteristics of the studies included. Methods of defining heavy menstrual bleeding and successful treatment varied. Some used the Pictorial Blood Loss Assessment Chart (PBAC). The PBAC is a measure of menorrhagia based on the number and staining of pads or tampons and clots passed.6 Different methods of measuring patient satisfaction are also reported.
Only the studies by Brun et al. [Unpublished work, 2002], Cooper et al.,7 Bongers et al.8 and Meyer et al.9 reported a sample size calculation. The study by Brun et al. (thermal balloon endometrial ablation vs transcervical resection) failed to recruit the specified sample size.
The studies by Cooper et al.7 (microwave endometrial ablation vs transcervical resection/rollerball ablation) and Meyer et al.9 (thermal balloon endometrial ablation vs rollerball ablation) were based on expected differences between groups that were not seen in practice. The possibility of type II errors therefore remains, although the differences in effect are very small and unlikely to be clinically significant.
There were significant differences at baseline between the two groups in the non-randomised study by Gervaise et al.,10 with the transcervical resection group having lower parity (1.9 vs 2.4) and more postmenopausal women (27%vs 7%), although the number of pads used per cycle was similar.
The study by Bongers et al.8 (thermal balloon endometrial ablation vs transcervical resection) was not randomised and compared women undergoing thermal balloon endometrial ablation with a historical control group. However, the authors report that consecutive women were recruited prospectively and baseline characteristics of the two groups were comparable.
It is not possible to blind surgeons to procedure type. While it would be possible to blind those who assessed outcomes or carried out statistical analyses, and possibly the women undergoing the procedure, no study reported this.
Meyer et al.9 (thermal balloon endometrial ablation vs rollerball ablation) randomised 275 patients but 15 electively withdrew before the procedure. Reports of the study population are inconsistent between the three main reports of this study. An intention-to-treat analysis was not performed. Only the three-year follow up paper reports on how many of the original 275 women were allocated to treatment and control.11 Forty-six percent of the recruited participants were lost to follow up by 60 months. Furthermore, in the five-year paper,12 patients who underwent repeat surgery are excluded from calculations of bleeding and pain outcomes.
The failure rate for ablation techniques, as measured through repeat ablation or hysterectomy, is time dependent with endometrial regeneration and menopause determining risk. Shorter study follow up among younger women may under-estimate the costs and disadvantages of endometrial ablation. Length of follow was limited for most trials (Table 1).
Outcomes reported in Table 4 is given (a) as reported in the trials, and (b) with intention-to-treat figures shown in parentheses.
Table 4. Satisfaction with treatment and its acceptability % (% intention to treat).
| ||Length of follow up (months)||Intervention||Very/Perfectly satisfied||Satisfied||Not very satisfied/no effect||Not satisfied/complaints worse||Pt satisfaction||Totally or generally satisfied|
|Cooper et al. 19997||12||MEA||–||–||–||–||–||–||–||77 (69)|
| ||TCRE/RB||–||–||–||–||–||–||–||75 (69)|
| ||TCRE/RB||–||–||–||–||–||–||–||67 (64)|
|Microsulis 2002||12||MEA||–||98 (98)a||–||2 (2)||–||–||–||–|
| ||RB||–||99 (99)a||–||1 (1)||–||–||–||–|
|Bongers et al. 20008||3||TBEA||66 (66)||20 (20)||10 (10)||4 (4)||–||–||–||–|
| ||TCRE||80 (80)||11 (11)||8 (8)||1 (1)||–||–||–||–|
|6||TBEA||63 (51)||10 (8)||16 (13)||11 (9)||–||–||–||–|
| ||TCRE||57 (52)||7 (7)||35 (32)||1 (1)||–||–||–||–|
|12||TBEA||63 (52)||13 (10)||10 (8)||14 (12)||–||–||–||–|
| ||TCRE||57 (52)||2 (1)||37 (28)||9 (7)||–||–||–||–|
|24||TBEA||60 (36)||4 (3)||11 (6)||25 (16)||–||–||–||–|
| ||TCRE||43 (27)||6 (4)||17 (11)||34 (21)||–||–||–||–|
|Meyer et al. 19989||12||TBEA||86 (78)||10 (9)||–||4 (4)||–||–||–||–|
| ||RB||87 (72)||12 (10)||–||1 (1)||–||–||–||–|
|24||TBEA||86 (77)||10 (9)||–||4 (3)||–||–||–||–|
| ||RB||87 (66)||11 (9)||–||2 (1)||–||–||–||–|
|36||TBEA||88 (72)||9 (6)||–||3 (2)||–||–||–||–|
| ||RB||92 (67)||6 (4)||–||2 (1)||–||–||–||–|
| ||RB||–||100 (44)||–||–||–||–||–||–|
|Soysal et al. 200114||12||TBEA||–||–||33 (31)||–||–||–||–||–|
| ||RB||–||–||39 (39)||–||–||–||–||–|
|Zon-Rabelink 200115||24||TBEA||–||80 (80)||–||–||–||–||–||–|
| ||RB||–||75 (73)||–||–||–||–||–||–|
|Pellicano et al. 200213||3||TBEA||–||–||–||–||27 (59)||13 (28)||0||–|
| ||TCRE/RB||–||–||–||–||21 (42)||12 (24)||9 (18)||–|
|12||TBEA||–||–||–||–||20 (43)||10 (22)||5 (11)||–|
| ||TCRE/RB||–||–||–||–||12 (24)||12 (24)||10 (20)||–|
|24||TBEA||–||–||–||–||16 (35)||12 (26)||5 (11)||–|
| ||TCRE/RB||–||–||–||–||2 (4)||18 (36)||3 (6)||–|
|Romer 19985||12||TBEA||–||100 (100)||–||–||–||–||–||–|
| ||RB||–||100 (100)||–||–||–||–||–||–|
| ||Length of follow up (months)||Intervention||Cure or acceptable improvement||Treatment acceptable||Menstrual loss acceptable||Would recommend treatment?|
|Cooper et al. 19997||12||MEA||78 (70)||94 (84)||–||91 (81)|
| ||TCRE/RB||76 (70)||90 (84)||–||89 (82)|
|24||MEA||–||–||96 (89)||90 (84)|
| ||TCRE/RB||–||–||88 (84)||90 (87)|
|Microsulis 2002||12||MEA||–||99 (99)b||–||–|
| ||RB||–||100 (100)b||–||–|
A wide range of outcomes were reported in the trials.Table 2 details the a priori primary and secondary outcome measures defined in the trials. PMS and dysmenorrhoea outcomes are not reported here. While important symptoms, they are not indications for surgery for either first or second generation endometrial ablation techniques. Not all studies reported these measures but in those that did, there were no significant differences between the groups for these measures when an intention-to-treat analysis was conducted.
Figure 2 illustrates the findings for amenorrhoea at 12 months for first generation versus second generation endometrial ablation techniques and Fig. 3 at 24 months. At 24 months, only the non-randomised Gervaise study10 (thermal balloon endometrial ablation vs transcervical resection) indicates a more favourable outcome for transcervical resection. The populations in the two arms of this study were different at baseline.
Figure 2. Forest plot of amenorrhoea at 12 months—first generation versus second generation endometrial ablation methods (random effects model, results not pooled).
Download figure to PowerPoint
Figure 3. Forest plot of amenorrhoea at 24 months—first generation versus second generation endometrial ablation methods (random effects model, results not pooled).
Download figure to PowerPoint
Differences in bleeding patterns between the treatment and control groups were not statistically significant for any of the various measures used in different trials.
Only the studies of Brun et al. [Unpublished work, 2002] and Gervaise et al.10 did not report patient satisfaction. The others use slightly different measures of satisfaction which are detailed in Table 4 with the outcomes recorded as they are reported in the papers. Only Pellicano et al.13 (thermal balloon endometrial ablation vs transcervical resection/rollerball ablation) reported a significant difference between groups (P < 0.001 at all follow up times.)
Only the microwave endometrial ablation study by Cooper et al.7 used a quality of life instrument previously validated for heavy menstrual bleeding, the SF-36 (Fig. 4). Prior to treatment, mean scores were lower across six of the eight items than a general population of the same age, and the SF-36 pain score was significantly lower in the microwave endometrial ablation group than the transcervical resection group. Following treatment, six of the eight items improved significantly in the microwave endometrial ablation group as did seven items in the transcervical resection group. Analysis of covariance showed that the only difference between the groups was in physical role function for which there was greater improvement with microwave endometrial ablation.
Cooper et al.7 and Meyer et al.9 report on the effect on work outside the home. In both studies, women were significantly less likely to be unable to work or to have two or more days absence from work due to heavy menstrual bleeding after ablation treatment. Differences between groups treated with first and second generation techniques were not significant.
It is not clear how ‘operation time’ is defined where reported.7,10,13,14 The study by Microsulis reports ‘anaesthetic time’ and ‘treatment time’. Differences between procedure times, however described, were significant in all studies in favour of second generation techniques. Cooper et al.7 (microwave endometrial ablation vs transcervical resection/rollerball ablation) also report on the mean post-operative stay and the percentage of women who were fully recovered in four weeks. Results from the two arms were similar.
Not all trials report all adverse effects so it is not always possible to determine whether adverse effects were absent, or not recorded. For example, not all trials report on haemorrhage and pain, although these may be common adverse effects. Those trials which specified adverse effects a priori are shown in Table 2. In order to report the most conservative success rate, percentages based on the number of women available for follow up are given.
Only the trial by Cooper et al.7 (microwave endometrial ablation vs transcervical resection/rollerball ablation) reported equipment failure, which occurred in 9% of microwave endometrial ablation operations and in 2% of transcervical resection operations (P= 0.02). The procedure was abandoned in 4% of both microwave endometrial ablation and transcervical resection procedures. It is reported that the equipment failures for microwave endometrial ablation all occurred early in the study with a prototype microwave generator. Adverse effects were reported in 1% of microwave endometrial ablation procedures (one blunt uterine perforation) by Cooper et al.7 In the control arm, 5% of transcervical resection and rollerball procedures had intra-operative adverse effects with one reported uterine perforation and five cases of haemorrhage. The trial by Microsulis did not report adverse effects.
Only one trial of thermal balloon endometrial ablation [Brun et al., 2002. Unpublished work] reported any intra-operative adverse effects in the thermal balloon endometrial ablation arm, with one patient (3%) reporting cervical burn. In the control arms, 3% of patients had an intra-operative adverse effect in Meyer et al.9 (two cases of fluid overload, one cervical and one uterine laceration), 10% in the Soysal trial14 (two patients with fluid overload and one with cervical laceration), 19% in the Pellicano trial13 (five patients with fluid overload, two with uterine perforation, both of which had an emergency conversion to hysterectomy, and one with cervical laceration) and 27% in the Bongers trial8 (20 patients with fluid overload, defined as >2000 mL intravasation, and one with uterine perforation).
Zon-Rabelink15 (thermal balloon endometrial ablation vs transcervical resection) only lists the types of adverse effects (cervical laceration, uterine perforation and electrolyte imbalance in the rollerball ablation arm). Significantly more post-operative pain relief was required by women who had undergone thermal balloon endometrial ablation than those who had undergone rollerball (P= 0.01).
No significant difference was reported in post-operative adverse events by any of the trials. Rates were generally low—endometriosis (0–4%), urinary tract infection (0–2%), haematometra (0–4%), fever (2–5%), haemorrhage (0–10%), pain (0–2%) and symptomatic hydrosalpinx (0–1%).
In all reports of repeat surgery, hysterectomy was much more common than ablation and only one woman was reported as having a repeat second generation treatment.7 In the microwave endometrial ablation trials, between <1% and 8% of women had undergone repeat surgical treatment for heavy menstrual bleeding at 12 months after microwave endometrial ablation and 1% to 8% after transcervical resection/rollerball ablation (median 4.5%). In the thermal balloon endometrial ablation trials, between 0% and 10% (median 4%) of women treated with thermal balloon endometrial ablation had repeat surgery within 12 months, rising to between 3% and 12% by 24 months (median 7%). The transcervical resection/rollerball ablation arms report 0–21% (median 8%) and 8–36% (median 9%), respectively, for the same follow up. One trial9 reported follow up to 60 months; 11% of women undergoing thermal balloon endometrial ablation and 7% of women undergoing rollerball ablation had had repeat surgery. However, there was considerable loss to follow up by this stage and results for the available population were 25% for thermal balloon endometrial ablation and 16% for rollerball ablation.
One small randomised controlled trial13 (thermal balloon endometrial ablation vs transcervical resection/rollerball ablation) found a significant difference in repeat surgery rate at 24 months with more patients undergoing transcervical resection/rollerball ablation than thermal balloon endometrial ablation having a second surgical treatment (P < 0.01)
- Top of page
Our review of microwave endometrial ablation and thermal balloon endometrial ablation compared with first generation endometrial ablation techniques identified few significant differences in clinical outcomes. In common with the previous review,16 we found second generation techniques quicker to perform and possibly safer, eliminating the risk of fluid overload and being less dependent on surgical skill. With longer follow up, it does not appear that equipment failure for second generation techniques continues to be problematic.
The previous systematic review16 concluded that, overall, second generation techniques had similar success rates and were significantly quicker to perform (WMD 11 minutes, 95% CI −18.6 to −2.6) than first generation techniques and were significantly more likely to be performed under local anaesthetic (OR 7.6, 95% CI 1.1 to 52.7). However, equipment failure was more likely in second generation techniques (OR 4.1, 95% CI 1.1 to 15.0).
The previous review included any type of first generation endometrial ablation compared with any type of second generation endometrial ablation technique. A total of eight studies of second generation techniques were included. However, at that time, only one study of microwave endometrial ablation and two of thermal balloon endometrial ablation were available. Our review identified more studies and longer follow up for included studies. In addition, we included controlled, non-randomised studies to broaden the evidence base. These were included for two reasons. Firstly, the difficulty of blinding practitioners and patients to the type of ablation limits the rigour of randomised controlled trials. Secondly, the nature of heavy menstrual bleeding means that sufferers come from a relatively restricted population in terms of age and sex. Given the rapidly evolving evidence base and existence of non-randomised controlled studies, we carried out a further systematic review of the evidence relating to the effectiveness of microwave and thermal balloon ablation compared with transcervical resection and rollerball (alone or in combination).
We did not search all possible databases, such as PsychLIT and CINAHL, and we restricted our search to papers available in English. This may be considered a weakness of the study. However, we are confident that the strategy for these new technologies, which included contacting experts and industry for papers, means that we have not missed any studies. Recent empirical evidence has shown that restriction to the English language does not automatically introduce bias and that such reviews will often produce results that are similar to those produced by reviews without any language restriction.17
The quality of the included randomised trials was variable. None reported blinding among those in whom it could have been accomplished. While randomisation methods seem sound in almost all, few gave adequate details about the method of treatment allocation. There were further difficulties with important outcomes such as bleeding patterns and satisfaction being recorded in different ways. For completeness, we included two trials that did not attempt to randomise patients. However, there are differences in the baseline populations of one of these trials and we cannot rule out the possibility that there are other systematic differences between the arms. Caution should be used in interpreting their results.
Although superficially, the women included in the trials were treated for the same condition—heavy menstrual bleeding, different criteria to define heavy menstrual bleeding and bleeding after treatment, both subjective and objective, are used. Only one trial7 used the pragmatic approach of allowing the condition to be patient defined. This is usual in clinical practice in the UK. For those using the Pictorial Blood Loss Assessment Chart, both heavy menstrual bleeding and success are defined at different levels. This makes it difficult to compare results across studies, and precludes meta-analysis of bleeding outcomes. There is a lack of consensus about what constitutes ‘excessive’, ‘normal’ or ‘reduced’ bleeding. This may lead to amenorrhoea being given particular weight as an outcome because it is more easily defined. However, this may not be the most important outcome for all women, some of whom will be satisfied with a reduction in bleeding.18 Most studies did not state how researchers measured different types of post-operative bleeding (for example, how they quantified ‘hypomenorrhoea’, ‘eumenorrhoea’ etc.) or the value that women themselves placed on different bleeding patterns.
As heavy menstrual bleeding is rarely life threatening, the essential purpose of any treatment is to improve quality of life. However, only one study in this review7 used a validated quality of life measurement scale. The SF-36 has been shown to be a ‘feasible’ means of looking at Quality of Life (QoL) in this group,19 responding to changes over time. However, the same authors have subsequently suggested that SF-36 has weaknesses in this context.20 SF-36 is a generic measure while heavy menstrual bleeding is a chronic, cyclical condition. In a study of women with heavy menstrual bleeding who had completed the SF-36, some found questions difficult to answer, or inappropriate for heavy menstrual bleeding.20 No study assessed in this review used a condition specific measure of QoL.
Satisfaction is a widely used primary outcome in studies of heavy menstrual bleeding. However, it is difficult to draw conclusions from these data as details of measurement were unclear. Reports of satisfaction are known to be related to patient characteristics, notably age and health status.21 Satisfaction may be related to the experience of the care setting and experience, as well as to the outcomes of treatment. Patient satisfaction measures come in a wide range of formats and are prone to biases related to design and delivery. Single-item satisfaction measures, such as those used in these endometrial ablation trials, may be less valid than well-constructed multi-item scales.21 Details of methods are frequently not reported. It is therefore difficult to know whether ‘satisfaction’ measured in one study is similar to that measured in another. Lack of detailed description also means that it is not possible to assess the extent to which potential biases are addressed.
Adverse effects are important outcomes. Not all included trials collected data on this, and where they do, the same adverse effects are not collected across all trials such as data on the common adverse effects of bleeding and pain. Further, while fluid overload can have serious, even fatal, sequelae, clinically meaningful outcomes such as cardiac failure and pulmonary oedema are not reported.
The included studies show very few significant differences in outcome. The study of Pellicano et al.13 is exceptional in this respect but was of poor methodological quality with no sample size calculation reported and relatively small numbers recruited. Details of allocation concealment were not given. Analysis was by available data and there was loss to follow up (29%).