A randomised controlled trial of the clinical effectiveness and cost-effectiveness of the levonorgestrel-releasing intrauterine system in primary care against standard treatment for menorrhagia: the ECLIPSE trial

Definition and diagnosis

Heavy menstrual bleeding (HMB), also called menorrhagia, is a common problem that can significantly impact on women’s lives and burden individuals and health-care systems. The National Institute for Health and Care Excellence (NICE) states that ‘for clinical purposes, HMB should be defined as excessive menstrual blood loss which interferes with the woman’s physical, emotional, social and material quality of life (QoL), and which can occur alone or in combination with other symptoms’. 1 In many previous studies, attempts were made to define HMB using an objective measure of blood loss, by weighing sanitary products or using the alkaline haematin method, with a loss of more than 80 ml of blood per cycle being considered excessive. Menstruation changes over a woman’s reproductive lifespan, with cycle length decreasing and the duration and heaviness of the period increasing with age. Heaviness is thus subjective, defined relative to what a woman considers normal and is influenced by sociocultural and psychological well-being factors. QoL and actual menstrual blood are not closely linked,2,3 so any interventions should aim to reduce the burden of the symptoms for the woman.

Heavy menstrual bleeding is generally considered to be heavy cyclical menstrual blood loss over a minimum of three consecutive cycles without any intermenstrual or post-coital bleeding. Where no organic pathology is present, the term dysfunctional uterine bleeding is used, whereas abnormal uterine bleeding includes that caused by uterine pathology, for example fibroids or genetic factors (e.g. von Willebrand disease). 4–6 The majority of women with HMB have no pathology that can be attributed as the cause of their symptoms, while approximately 30% will have uterine fibroids and 10% uterine polyps, with malignant causes being rare in the pre-menopausal population. 7

With a subjective definition, diagnosis of HMB is made through careful history taking, with any suggestion of pathology such as intermenstrual or post-coital bleeding prompting further investigation, including physical examination, ultrasound and biopsy if endometrial hyperplasia or cancer is suspected. Objective measures are inconvenient and impractical, and indirect measures of menstrual blood loss, such as pictorial diaries, are not consistently accurate and these too fail to reflect women’s experience of what is burdensome for them. 8 A full blood count is recommended by the NICE guidelines,1 but serum ferritin and oestradiol measurement are not considered worthwhile. Further testing for coagulation or thyroid disorders are dependent on symptoms and personal history.

Epidemiology

In various cohorts studies, reported prevalence has ranged from 4% to 51%, although it is acknowledged that differences in definition, measurement (objective vs. subjective), clinical and cultural setting will undoubtedly influence reporting. 1 A UK survey of women aged 18–54 years indicated an annual community incidence HMB of 25% and prevalence of 52%, with a slight increase with increasing age. 9

Heavy menstrual bleeding accounts for 19% of gynaecologist office consultations in the USA10 and 20% in the UK,11 and each year around 6% of women in the USA consult their general practitioner (GP). 12 Another survey found that while 22% of menstruating women over 35 years considered their periods heavy and interfering with their lives, only 7% had consulted their GP in the preceding 6 months. 13 Despite the many factors influencing women’s decisions not to seek help,14 the number and cost of consultations and treatments impose substantial demands on the NHS. In 2004–5, there were 7179 hysterectomies and 9701 endometrial ablations performed; compared with 1989–90 this represents a 69% decrease and an 11-fold increase in the number of procedures for HMB, respectively. 15 Rates of surgical procedures for HMB are 17.8 per 10,000 women aged 25–44 years in the USA,16 and 14.3 per 10,000 women in the UK. 17

Tranexamic acid

Tranexamic acid is a plasminogen activator inhibitor that exerts an antifibrinolytic effect on the endometrium, reducing menstrual blood loss by 50%. 19 It is taken in high doses, of up to 1 g four times per day, during the first 5 days of the menstrual cycle. Initial concerns regarding a potential increase in adverse thrombogenic events has been refuted by long-term studies showing no excess risk. 20

Mefenamic acid

Mefenamic acid is a non-steroidal anti-inflammatory drug (NSAID), which can reduce the formation of prostaglandins, implicated in the abnormal clotting in the endometrium by inhibiting the enzyme cyclo-oxygenase. It is also taken on days 1–5 days of the menstrual cycle, at a dose of 500 mg three times per day, but it is contraindicated in women with bleeding disorders. In terms of reducing menstrual blood loss, mefenamic acid substantially reduces objectively measured loss, but not to the same extent as tranexamic acid. 21

Combined oral contraceptive pill

Combined oral contraceptives of any formulation reduce the thickness of the endometrium and inhibit ovulation, with a resultant reduction in menstrual blood loss of 43%. 22 There is no evidence of any relative difference with other medical treatments. COCs are not suitable for everyone. They are contraindicated for those with a personal history or risk factors for thromboembolism, a body mass index (BMI) of ≥ 35kg/m² or migraine with aura, and are advised to be used with caution in women over 35 years who smoke or who are overweight or obese. 23

Norethisterone

Norethisterone is a synthetic progesterone which can reduce menstrual blood loss in ovulatory women if 10–15 mg is taken daily from day 5 to day 26 of the cycle, but is ineffective if taken only during the late luteal phase, from day 19 to day 26 of the menstrual cycle. Compared with tranexamic acid or the LNG-IUS, norethisterone is not as effective at reducing objectively measured menstrual blood loss. 21,24

Levonorgestrel-releasing intrauterine system

The LNG-IUS is a highly effective long-acting reversal contraceptive. Since 2009 in the USA, and earlier in Europe, the LNG-IUS has been available for HMB. 23 The LNG-IUS is a T-shaped plastic frame containing 52 mg of levonorgestrel, which is released at a rate of approximately 20 µg per day and is licensed for 5 years’ use, after which it should be removed and replaced if desired. Fertility is quickly restored after removal. The LNG-IUS is contraindicated in women with a congenital uterine anomaly or who have fibroids that distort the uterine cavity, in women with a uterine or pelvic infection, and in those with a history of gynaecological cancers. The adverse events of interest fall into two categories: those related to an intrauterine device, such as dysmenorrhoea, irregular bleeding, ectopic pregnancy and expulsion of the device; and those related to progestogens, such as bloating, weight gain and breast tenderness. The main disadvantage of the LNG-IUS is the disruption to the menstrual cycle, particularly in the first 6 months, with the number of bleeding and spotting days increasing to a median of 18 days and the cycle becoming irregular in approximately 20% of women, before reducing in duration and severity. 25 The proportion of women who are amenorrhoeic a year following insertion has been reported at around 12%. 26,27

Clinical effectiveness of the levonorgestrel-releasing intrauterine system compared with medical treatments

Prior to the initiation of the ECLIPSE trial, the most comprehensive summary of the data for the LNG-IUS came from a systematic review which identified 10 studies of the LNG-IUS for women with heavy menstrual blood loss (≥ 80 ml per cycle). 28 Of these, only five studies were randomised controlled trials (RCTs),20,24,29–31 and only four reported menstrual blood loss reduction in those with the LNG-IUS in place for 3–12 months, in whom the reduction in blood loss ranged from 79% to 96%.

In 2007, the NICE guidance1 considered all pharmaceutical treatments for HMB. In comparing the LNG-IUS with non-surgical options, NICE referred to the Cochrane review,27 which cited three small randomised trials comparing the LNG-IUS with norethisterone,24,32 mefenamic acid32,33 or danazol32 (Danol^®, Sanofi-aventis), although the trial reported by Cameron et al. 32 used a now discontinued formulation of the LNG-IUS. From this review, the odds ratio for amenorrhoea (> 3 months) was 8.67 [95% confidence interval (CI) 1.52 to 49.35] in favour of the LNG-IUS. The odds ratio for the proportion of women unwilling to continue with treatment was 0.27 (95% CI 0.10 to 0.67) and for the proportion of women satisfied with treatment the odds ratio was 2.13 (95% CI 0.62 to 7.33), both in favour of the LNG-IUS.

For COCs, only one small RCT of four treatments existed at the time of the NICE guidelines,34 suggesting that COCs reduce menstrual blood loss to a greater extent than the NSAID naproxen, but less than mefenamic acid or the gonadotropin-releasing hormone antagonist danazol. However, the COC used was a first-generation formulation and no data exist for the lower oestrogen doses now available.

There was sufficient evidence in three systematic reviews for NICE to conclude that a tranexamic acid causes a clinically significant reduction in menstrual blood loss for women with HMB, ranging from 29% to 58% in studies lasting up to 1 year, although long-term data are missing. 11,21,27,35 NSAIDs (mefenamic acid or naproxen) were also found to produce clinically important reductions in menstrual blood loss, ranging from 20% to 49%, but were inferior to danazol and tranexamic acid, although NSAIDs have fewer adverse effects than danazol. 36,37

The NICE guidelines development group stated that, in its interpretation of the evidence for pharmaceutical treatments for HMB, a high value was placed on reduction of menstrual blood loss and minimising adverse effects. The guidelines development group based its assessment first on the clinical effectiveness of treatments and, second, on the cost-effectiveness of treatments. The results of the systematic review showed that the LNG-IUS, mefenamic acid, tranexamic acid and COCs could be considered equivalent in terms of clinical effectiveness. However, the guidelines went on to state that, if both hormonal and non-hormonal treatments are acceptable, then the order in which treatment should be considered should be the LNG-IUS, then tranexamic acid, a NSAID or COC, then norethisterone on days 5–26 of the cycle.

An updated review identified seven further randomised trials,38–44 involving only 718 women in total, which compared the LNG-IUS with non-hormonal and hormonal treatments. In an industry-sponsored review,45 reduction in menstrual blood loss was achieved in only 230 women, from five studies,33,38,43,44,46 in which the LNG-IUS was administered. Reported median blood loss had decreased by 85% by 3 months, and had further decreased, by a total of 94%, by 12 months, remaining constant thereafter to 5 years.

The emphasis of all the previous trials has been on the reduction in menstrual blood loss. Very few trials collected data on QoL or impact of the symptoms. A small study of the LNG-IUS compared with norethisterone found no difference in the proportion of women in whom menstrual blood loss interfered with their QoL after treatment. 24

The trial by Lahteenmaki et al. 31 did report on QoL, measured using visual analogue scales (VASs) rather than any validated scale. The study selected women awaiting a hysterectomy and offered them the option of randomisation between continuing their existing treatment and the LNG-IUS. The lack of blinding and the control group remaining on an unsatisfactory treatment is likely to have influenced the women’s attitude to treatment and subjective assessment of QoL, therefore, is inherently biased against existing therapy. 31

Clinical effectiveness of the levonorgestrel-releasing intrauterine system compared with surgery

Prior to the start of the ECLIPSE trial, there were six RCTs comparing the effectiveness of the LNG-IUS with surgery, two using first-generation ablation methods,29,30 three comparing the LNG-IUS with second-generation thermal balloon ablation47–49 and one comparing the LNG-IUS with hysterectomy. 50 No difference in the rates of amenorrhoea or satisfaction rates were seen compared with endometrial ablation, although menstrual blood loss was reduced to a greater extent with ablation. 29,48

In a subsequent Cochrane review of medical versus surgical treatments for HMB,51 an additional six trials, four49,52–54 comparing the LNG-IUS with thermal balloon ablation, one55 comparing the LNG-IUS with hysterectomy and two38,56 comparing the LNG-IUS with first-generation ablation, were included. There was no statistically significant difference in satisfaction rates up to 2 years or in most domains of the Short Form questionnaire-36 items (SF-36) measure of QoL. As before, endometrial ablation was significantly more effective than the LNG-IUS in controlling bleeding at 1 year (relative risk 1.19, 95% CI 1.07 to 1.32; p = 0.001). Hysterectomy stopped all bleeding and provided greater satisfaction, but caused serious complications for some women. This review considered the long-term data to be weak.

This review was extended by obtaining individual patient data in order to be able to standardise outcome data, such as satisfaction or, the corollary, dissatisfaction. 57 Rates of dissatisfaction with the LNG-IUS and non-hysteroscopic endometrial destruction were similar (18% vs. 23%; odds ratio 0.8, 95% CI 0.4 to 1.5; p = 0.4). Lack of data from individual patients prohibited any further investigation of subgroups or first-generation ablative methods. There was weak evidence to suggest hysterectomy is preferable to the LNG-IUS (dissatisfaction 5% vs. 17%; odds ratio 2.2, 95% CI 0.9 to 5.3; p = 0.07).

As might be anticipated, women having surgical treatment were significantly less likely to require further surgery within 1 year than women who had a LNG-IUS (relative risk 0.13, 95% CI 0.05 to 0.33; p = 0.00002). None of the hysterectomy group required additional surgery, while about 3% of women who had endometrial ablation required extra surgery during the first year, compared with about 15% of women who were allocated the LNG-IUS. One study38 reported that 14% (4/29) of the surgical group underwent repeat ablation within 3 years (with two having subsequent hysterectomy), while 30% (9/30) of women allocated to the LNG-IUS group discontinued its use and ‘were offered’ endometrial resection. In a Finnish trial58 of 117 women with a LNG-IUS, 60 women no longer had it in situ 5 years after randomisation; 50 of these women had had a hysterectomy (42% of those randomised to the LNG-IUS) and one had had endometrial ablation.

The NICE guidelines1 highlighted the high level of subsequent surgery associated with the LNG-IUS and recommend that endometrial ablation may be offered as an initial treatment for HMB after full discussion with the woman of the risks and benefits and of other treatment options, while hysterectomy should not be used as a first-line treatment solely for HMB.

Cost-effectiveness of the levonorgestrel-releasing intrauterine system compared with other treatments

The first trial to incorporate a cost-effectiveness analysis was the Finnish trial cited above. 50,58 There was no statistically significant difference in QoL scores between the two treatment groups at 5 years, as measured by the European Quality of Life-5 Dimensions (EQ-5D) instrument. Mean direct costs remained significantly lower in the LNG-IUS arm (US$1892) than in the hysterectomy arm (US$2787), despite 42% of women in the LNG-IUS arm going on to have a hysterectomy. This trial, however, compared the LNG-IUS with hysterectomy in women referred to hospital. No economic analysis relevant to the use of the LNG-IUS in a UK primary care setting is available, nor had the relative cost-effectiveness of the LNG-IUS relative to medical treatment been assessed at the start of the ECLIPSE trial.

The health economic assessment performed in the process of developing the NICE guidelines populated a decision model with data predominantly from the Finnish study, and made significant assumptions regarding the effectiveness and discontinuation rates of COCs, tranexamic acid and mefenamic acid. The results showed that the LNG-IUS is the more effective treatment option when long-term use of a treatment is required, as the LNG-IUS generated more quality-adjusted life-years (QALYs), at a lower cost, than any other pharmaceutical treatment strategy. The base-case result was £840 per QALY gained. 1 The uncertainty in this result was not explored at all and so the result could be highly misleading.

Primary objective

To assess the clinical effectiveness of the LNG-IUS compared with standard medical treatment for women seeking treatment for HMB in primary care in the short term (2 years following randomisation).

Secondary objective

• To assess the clinical effectiveness of initial treatment with the LNG-IUS compared with standard medical treatment for women seeking treatment for HMB in primary care in the medium term (5 years following randomisation).

• To determine the cost-effectiveness of the LNG-IUS compared with standard medical treatment in the short and medium term.

• To explore the perspectives of trial participants or women with HMB who declined to be randomised in a longitudinal qualitative study.

• To measure the reliability and validity properties of the Menorrhagia Multi-Attribute Scale (MMAS).

Outcome measures used in the ECLIPSE trial

Method

Population

Women between 25 and 50 years of age who presented to their GP with HMB involving at least three consecutive menstrual cycles were eligible to participate. Women were excluded if they intended to become pregnant over the next 5 years, were taking hormone replacement therapy or tamoxifen (Soltamox^TM, Rosemount Pharmaceuticals), had intermenstrual bleeding (between expected periods) or post-coital bleeding or findings suggestive of fibroids (abdominally palpable uterus equivalent in size to that at 10–12 weeks’ gestation) or other disorders, or had contraindications to or a preference for either the LNG-IUS or usual medical treatments. Women with heavy, irregular bleeding were ineligible unless the results of endometrial biopsy were reported to be normal; no further investigations were mandated by the protocol. All patients provided written informed consent.

Randomisation

Patients were assigned to a study group by telephone or a web-based central randomisation service at the University of Birmingham Clinical Trials Unit. A computerised, minimised randomisation procedure was used to achieve balance between the groups with respect to age (< 35 years or ≥ 35 years), BMI (≤ 25 kg/m² or > 25 kg/m²), duration of symptoms (< 1 year or ≥ 1 year), need for contraception (yes or no) and heavy HMB alone or HMB accompanied by menstrual pain. GP practice was not included as a minimisation variable to avoid any chance of the allocation becoming too predictable.

Study interventions and compliance

Eligible women who provided written informed consent were randomly assigned to either the LNG-IUS or usual medical treatment. Usual treatment options included mefenamic acid, tranexamic acid, norethisterone, a combined oestrogen–progestogen or progesterone-only oral contraceptive pill (any formulation), or medroxyprogesterone acetate injection and were chosen by the physician and patient on the basis of contraceptive needs or the desire to avoid hormonal treatment. 1,18 The particular medical treatment to be used was specified before randomisation. Subsequently, treatments could be changed (from one medical treatment to another, from the LNG-IUS to medical treatment, or from medical treatment to the LNG-IUS) or could be discontinued because of a perceived lack of benefit, side effects, a change in the need for contraception, referral for endometrial ablation or hysterectomy, or other reasons, according to usual practice. 1,18 Treatment changes reported by patients were confirmed with the GP.

Outcome measures and follow-up

The primary outcome measure was the condition-specific MMAS,61,71 which is designed to measure the effect of HMB on six domains of daily life (practical difficulties, social life, psychological health, physical health, work and daily routine, and family life and relationships). Summary scores, which range from 0 (severely affected) to 100 (not affected), were assessed at 6 months, 1 year and 2 years after randomisation. Details on how summary scores are calculated can be found elsewhere. 61 The MMAS has a high degree of reliability and internal consistency,71 has good content and construct validity,61,71 is responsive65,72 and is acceptable to respondents. 61,65,71,72 Secondary outcome measures included general health-related QoL and sexual activity. To assess QoL, we used three instruments: the SF-36 version 2 [with scores ranging from 0 (severely affected) to 100 (not affected)], the EQ-5D [with scores ranging from −0.59 (health state worse than death) to 100 (perfect health state)] and the EQ-5D VAS [with scores ranging from 0 (worst health state imaginable) to 100 (most perfect health state imaginable)]. The validated SAQ measures pleasure [with scores ranging from 0 (lowest level) to 18 (highest level)], discomfort [with scores ranging from 0 (greatest) to 6 (none)] and frequency (assessed relative to perceived usual activity as an ordinal response). 73 Scores were obtained before randomisation and by mail at 6 months, 1 year and 2 years after randomisation. Data were collected from participating clinicians regarding all serious adverse events, defined as adverse events that resulted in death, disability or hospitalisation. Patients were also asked to report any hospitalisations and adverse events leading to discontinuation of the study drug.

Study oversight

The study sponsor was the University of Birmingham. Study oversight was provided by an independent steering committee and an independent data and safety monitoring committee, whose three reviews of interim data provided no reason to modify the trial protocol on the basis of pragmatic stopping criteria,74 other than to support revision of the recruitment target, which is described in Statistical considerations. The study was conducted in accordance with the protocol, which is available at www.nets.nihr.ac.uk/projects/hta/020602. Approval of the study was obtained from the South-West England Multicentre Research Ethics Committee, and clinical trial authorisation was received from the Medicines and Healthcare products Regulatory Authority. All medications and devices were prescribed by providers through the NHS. The manufacturers of the LNG-IUS and other therapeutic agents used in the study were not involved in any aspect of the trial.

Statistical analysis

The study was originally designed to have 95% power at p < 0.01 to detect small to moderate (0.3 SD) differences in mean MMAS score between groups at any one time point. 75 Following interim review of recruitment rates, in consultation with the independent data monitoring committee, the power and type 1 error parameters were revised to more conventional levels of 90% and p = 0.05, to allow timely completion of the study. This required an enrolment of 470 patients; we increased the sample size to 570 to allow for up to 20% loss to follow-up.

Analyses was by intention to treat with continuous measures compared using multilevel repeated-measures models,76 including parameters allowing for participant, treatment, time and baseline score. Responses including all three time points were considered in the primary analysis. Differences between slopes were assessed by including treatment-by-time interaction parameters in the models. If these were not significant (p > 0.05), then they were dropped from models and constant treatment differences over time were assumed. In a similar fashion, treatment-by-subgroup interaction parameters were included in the above model to test for the importance of the prespecified subgroup parameters (same as the minimisation variables, see Randomisation). Treatment effect estimates within subgroups are presented where interaction parameters were statistically important. Differences between groups at each assessment time point were also examined by analysis of covariance (adjusting for baseline score). Changes from baseline score within groups were examined using paired t-tests.

A number of sensitivity analyses were performed on the primary outcome to test the robustness of the results. 77,78 Analyses were performed excluding data received subsequent to a crossover to the alternative treatment group and also once study treatment was ceased altogether. Further sensitivity analyses included assigning the best possible score (100) to those women who indicated that they were no longer experiencing bleeding and did not complete the MMAS, and a carried forward (pre-intervention) score for those undergoing hysterectomy or endometrial ablation. Further analyses included the exclusion of MMAS data from late returned booklets (defined as more than 3 months past the due date) and an analysis assuming all missing responses were the worst possible score (0). Missing MMAS responses were also simulated using a multiple imputation approach with analysis performed on the resulting 20 simulated data sets. 78

Kaplan–Meier plots were constructed for time to first treatment change with women censored at date of last follow-up or, if appropriate, date of death, withdrawal or loss to follow-up. A Cox proportional hazards model was used to construct hazard ratios. Standard statistical methods were used to test the statistical significance of other responses (chi-squared tests for dichotomous data, Cochrane–Armitage test for trend and Wilcoxon signed-rank test for ordinal data). Effect sizes are presented with 95% CIs and two-sided p-values for all. SAS version 9.2 (SAS Institute Inc., Cary, NC, USA) was used for analyses.

Patients and follow-up

Between February 2005 and July 2009, a total of 571 women with HMB from 63 UK centres (median centre recruitment was three, interquartile range one to seven) were randomly assigned to either the LNG-IUS (285 women) or usual medical treatment (286 women). Baseline characteristics were similar between the two treatment groups (Table 2).

For 215 (75%) of the women assigned to usual medical treatment, the initial prescription, decided prior to randomisation, was for mefenamic acid, tranexamic acid or a combination of the two drugs (Table 3); 55 (19%) of the women in the usual treatment group required contraception. Study questionnaire booklets were returned by 478 (84%) of the patients at the 2-year time point (see Figure 1);38 of these women, 45 (9%) could not complete the MMAS appropriately because their menstrual bleeding had ceased, but they completed other parts of the booklet and this information was used to inform the sensitivity analysis.

Of the 285 women randomly assigned to the LNG-IUS, 24 (8%) did not have the intrauterine system (IUS) inserted: 10 chose usual medical treatment, six chose no treatment and, in eight, insertion of the system was unsuccessful and, therefore, usual medical treatment was subsequently instituted (Figure 1).

FIGURE 1.

Enrolment, randomisation and follow-up to 2 years of the study patients. See Appendices 1 and 2 for reasons for discontinuation of treatment.

Women in the LNG-IUS group were almost twice as likely as those in the usual treatment group to still be receiving their assigned treatment at 2 years (64% vs. 38%; p < 0.001) (Figure 2). The most common reasons cited for discontinuation of the LNG-IUS were lack of effectiveness (37%) and irregular or prolonged bleeding (28%). Of the 163 women who discontinued usual medical treatment, 80 (49%) switched to the LNG-IUS. The most common reason for discontinuation of usual medical therapy was lack of effectiveness (53%). Reasons for discontinuing therapy are summarised in Appendices 1 and 2.

FIGURE 2.

Time to first treatment change during 2 years’ follow-up. Data are for women who crossed over from the assigned study treatment to the other study treatment and for those who discontinued treatment.

Primary outcome: Menorrhagia Multi-Attribute Scale

In both groups, total scores on the MMAS were significantly improved, compared with baseline scores, at 6 months, and this improvement was maintained at 1 and 2 years (Figure 3 and Table 4) However, improvements in scores were significantly greater among women assigned to the LNG-IUS than among those assigned to usual treatment over the course of the 2 years (mean difference in scores 13.4 points, 95% CI 9.9 to 16.9 points; p < 0.001). This difference was apparent by 6 months and maintained thereafter.

FIGURE 3.

Primary outcome in the two treatment groups up to 2 years’ follow-up. The primary outcome was the score on the MMAS (scores range from 0 to 100, with lower scores indicating greater severity). Mean MMAS scores are shown for the two groups at 6, 12, and 24 months. Error bars indicate 95% CIs. Overall, the average difference in scores between the women treated with the LNG-IUS and those treated with the usual medical therapy was 13.4 points (95% CI 9.9 to 16.9 points; p < 0.001).

The effect in favour of the LNG-IUS was apparent in all six individual domains of the MMAS at every time point through examination of the frequencies of responses (Table 5).

In a sensitivity analysis that excluded women who crossed over from the assigned treatment to the other study treatments, improvement with the LNG-IUS, as compared with usual medical treatment, increased (mean difference in scores over the course of 2 years, 17.8 points, 95% CI 14.1 to 21.5 points; p < 0.001). Other sensitivity analyses yielded results that were not materially different from the results of the primary analysis (p < 0.001 for all comparisons) (see Appendix 3).

In subgroup analyses, there was a significant interaction between treatment and BMI (p = 0.004). The benefit of the LNG-IUS was greater in women with a BMI > 25 kg/m² (16.7 MMAS points, 95% CI 12.6 to 20.9 MMAS points; p < 0.001) than in those with a BMI of ≥ 25 kg/m² (5.4 MMAS points, 95% CI −1.0 to 11.8 MMAS points; p = 0.10). This finding appeared to be attributable to the superior outcome with usual medical treatment in leaner women (see Appendix 4). Improvements with the LNG-IUS were similar in both subgroups. None of the other tests for subgroup interaction was significant (p > 0.10).

Generic quality of life and sexual activity

The SF-36 domains were generally significantly improved from baseline in both groups at all time points, although the scores for women in the LNG-IUS group were better than for those in the usual treatment group in seven of the eight domains in the analyses over all time points (Table 6); mental health was the only domain for which there were no significant between-group differences. The improvements appeared to be greatest at 6 months, but had lessened by the 2-year follow-up assessment (see Appendix 5). No significant differences were seen between treatments with respect to the EQ-5D instrument; scores were significantly improved from baseline in both groups at 2 years, but not at earlier assessments (see Appendix 6). The treatments did not differ significantly with respect to the scores for the pleasure, discomfort and frequency domains of the SAQ (see Appendix 7).

Adverse events

There were no serious adverse reactions attributable to study treatments. There was no significant difference between the two groups in the frequency of unrelated serious adverse events (58 in the usual treatment group and 49 in the LNG-IUS group; p = 0.59). There was one death in the LNG-IUS group; the cause of death was recorded by the coroner as inconclusive and the LNG-IUS was not in situ. Unrelated serious adverse events are summarised in Appendix 8.

Surgical interventions

The frequency of surgical interventions for HMB within 2 years did not differ significantly between the two groups. Hysterectomy was performed in 6% of the women in each group; endometrial ablations were performed in 4% of women in the LNG-IUS group and in 6% of those in the usual treatment group (p = 0.44).

Trial methods

For details on the population, randomisation and outcome measures the reader should refer to Chapter 3.

Statistical considerations

The sample size assumptions for MMAS scores were the same as previously detailed in Chapter 3, Statistical consideration. Four hundred and seventy patients were required to have 90% power (p = 0.05) to detect 0.3 SD in MMAS score. We ultimately received 424 responses at 5 years’ follow-up. A post-hoc calculation suggests this many responses would provide 87% power (p = 0.05) to detect this same size of difference. For progression to surgical intervention (hysterectomy or endometrial ablation), using an assumed rate of 35% in the standard arm (a figure that was set out in the protocol), 424 women (the number followed up to 5 years; see Figure 4 for details) would provide 80% power (p = 0.05) to detect an absolute reduction of 12%, that is, 35% down to 23%.

The statistical methods used to analyse the data were similar to those set out in Chapter 3; however, given the long length of time between this and the previous assessment point at 2 years, we felt a repeated-measures analysis incorporating all accumulated responses to no longer be clinically meaningful and therefore this has not been included.

Patients and follow-up

See Chapter 2 for details on the demographics and prescribed treatments of those randomised and Figure 4 for losses to follow-up. Study questionnaire booklets were returned by 424 (74%) of the participants. One hundred and fifteen of the women (27%) declined to complete the MMAS, but indicated on the form that they were no longer bleeding; their score was assumed to be the maximum achievable (100, ‘no problems’). These women completed other parts of the questionnaire booklet and a sensitivity analysis of MMAS responses without any assumption was also performed.

FIGURE 4.

Enrolment, randomisation and follow-up up to 5 years of the study patients. a, See Figure 5 for time to first treatment change survival estimates; b, see Appendices 9 and 10 for reason reported for discontinuing treatment.

The 5-year allocated treatment retention rate was estimated to be 47% (95% CI 40% to 52%) in the LNG-IUS group and 15% (95% CI 11% to 20%) in the usual treatment group (Figure 5). Reported reasons for discontinuation of treatment are summarised in Appendices 9 and 10. Lack of efficacy was cited 41% (94/228) of the time in the usual treatment group and 24% (36/148) of the time in the LNG-IUS group. Of the 228 recorded instances of treatment change in women allocated usual medical treatment, 97 (43%) were to the LNG-IUS. In the LNG-IUS group, 39% (57/148) of treatment switches were to usual treatment. There was no significant difference between the groups in the total number of serious adverse events up to 5 years (see Appendix 11).

FIGURE 5.

Time to first treatment change over 5 years’ follow-up.

Menorrhagia Multi-Attribute Scale

After 5 years, scores on the MMAS were significantly improved compared with baseline in both groups (Table 7). This improvement was greater, on average, among women assigned to the LNG-IUS, but the difference was not statistically significant (3.9 points, 95% CI –0.6 to 8.3 points; p = 0.09). The same analysis without any assumption about MMAS scores where a blank form was returned and the woman indicated she was no longer bleeding returned a similar result (5.2 points difference in favour of the LNG-IUS, 95% CI –0.4 to 10.8 points; p = 0.07). Per-protocol analysis (excluding women who switched or stopped taking treatment) – limited to 28 responses in the usual treatment group and 88 in the LNG-IUS group – returned a 15.8-point difference in favour of the LNG-IUS (95% CI 7.4 to 24.2 points; p = 0.0003).

Surgical interventions

There were 53 surgical events (endometrial ablation or hysterectomy) in the LNG-IUS group, compared with 56 in the usual treatment group, that were included in the surgery-free survival analysis (109 events in total). This difference was not statistically significant (hazard ratio 0.90, 95% CI 0.62 to 1.31; p = 0.6) (Figure 6). Analysis excluding participants who crossed over from one group to another returned a similar result (HR 0.96, 95% CI 0.60 to 1.52; p = 0.9). Five-year surgery-free survival rates were estimated to be 80% (95% CI 74% to 84%) in the LNG-IUS group compared with 77% (95% CI 71% to 82%) in the usual treatment group. In total, there were 24 ablations in the LNG-IUS group compared with 31 in the usual treatment group as well as 30 hysterectomies in both groups (115 operations in total; the extra six events coming from patients who had both types of surgery).

FIGURE 6.

Surgery-free (hysterectomy/endometrial ablation) survival analysis over 5 years’ follow-up.

Generic quality of life and sexual activity

Responses to the EQ-5D and SF-36 instruments were generally significantly improved from baseline in both groups (Table 8); the only statistically significant difference between groups was seen in the general health perception domain of the SF-36 and favoured the LNG-IUS (4.7 points, 95% CI 0.6 to 8.8 points; p = 0.02). The treatment groups did not differ significantly with respect to any of the domains of the SAQ.

Sampling

A purposeful sample was selected from women approached for inclusion in the ECLIPSE trial of the effectiveness of medical treatments for HMB in primary care. 93 They had consulted their GP about HMB (lasting three or more consecutive cycles). Women participating in the trial had either agreed to be randomised to the LNG-IUS or to other standard treatment (including oral tranexamic acid, mefenamic acid, norethisterone, contraceptive pill or medroxyprogesterone acetate injection) or had declined to enter the trial because they had a strong preference for a particular medical treatment, which they had then commenced. We selected women from eight general practices across five counties of the UK of varying demographics (age, social and educational background, ethnicity) and parity, who experienced different medical treatments for HMB of differing treatment duration.

Recruitment and consent

Practitioners identified potential participants at the time that interest in participation in the ECLIPSE trial was discussed, and trial participation accepted or declined. Permission was sought to pass contact details to the research team, who sent study information and a request for written consent, translated where appropriate, to participants.

Data generation and analysis

Semistructured face-to-face interviews were conducted by a female researcher (G Prileszky), with a midwifery background, at respondents’ convenience and in their own homes. The support of an interpreter was offered where appropriate (four cases), but all interviewees opted to be interviewed in English. Initial interviews were conducted within 1–3 months of commencing medical treatments. A second interview was conducted, if participants were willing, between 12 and 14 months following commencement of treatment to explore any changes in experience or treatment, or in views occurring over time.

Interviews were digitally audio-taped and transcribed verbatim, and contemporaneous field notes kept of the interview context. Verbatim quotations are annotated with participant number and suffixed with ‘A’ for the initial interview and ‘B’ for the follow-up interview. They followed broad topic areas including menstrual history and experience, social factors influencing experience, relationships and family life, and emotional well-being as well as treatment preference and experience of treatments. The topic prompts were modified and refined following early interviews, with participants encouraged to discuss their perceptions and experiences freely. A sample of participants was also asked to participate in further respondent validation.

Data were analysed using constant comparison94 by the field researcher, with two researchers of different disciplinary backgrounds (academic midwifery and clinical primary care) contributing to development of the analysis. Coding was aided by application of NVivo software (QSR International, Warrington, UK) in identifying emerging themes from the data. Data generation and analysis were iterative, each informing the other, with further theoretical sampling and data collection to extend and challenge earlier data and interpretation, both within and between first and second interview data sets. This tested the integrity and credibility of the analysis until no new themes emerged, which suggested saturation.

Validation

Interim findings were also fed back and reviewed with a sample of six interviewees who were willing to be involved a third time. This was by audio-taped telephone interview following prior distribution of written summaries. Respondents were asked to consider and comment on the findings, enabling the research team to confirm and further refine data interpretation and analysis where appropriate.

Concealment of menstruation

Framing women’s experiences of treatments, women felt a strong societal pressure to remain silent about menstruation, which remained taboo and was amplified by experiencing HMB.

I think in society it’s quite a taboo subject. I work in an all-female environment with one man who is obviously married, who has got children but if we were actually talking about periods we would stop if he came into the room.

11B

It makes me feel not so much like a normal woman really when I have a heavy period because you feel like you’re the only one that’s having a heavy period. Obviously there are hundreds of women out there that are having them but it’s not common knowledge that a lot of women are having heavy periods like I did, I kept it quiet.

15A

The taboo existed within families and among friends. It was strongest in the workplace and could persist within consultation with a health professional.

A lot of women don’t like to go [to the GP] because it’s this taboo subject, so it has to start somewhere other than a GP. It’s got to start somewhere that you know that it is acceptable to have a problem, there isn’t a right and a wrong.

27B

Women could maintain the societal norm by denying their problems, lying to employers and withholding information from their doctors in order to stay silent about HMB. Not speaking about HMB (and menstruation in general) had a negative impact on most women’s QoL.

I had gone to the doctors a few times and somebody had asked me and I wished somebody had asked me earlier and perhaps if you talk about it then somebody perhaps at work might actually well I am doing this, this and this and then you are more confident then to seek help.

26B

Women also expressed a strong sense that women should ‘put up’ with HMB and not complain about it by maintaining a stoic silence. The decision to seek treatment and therefore break the silence was seen as a weakness or failure and was often associated with reduced self-esteem.

And you just feel a bit, not worthless because worthless is the wrong word – but you think some people must think you are one of these women. That you have just got a heavy period – get on and sort of deal with it – but you can’t physically get on unless you have actually been through a phase of having heavy periods – you can’t. And I am the same because I didn’t understand when people said they had really heavy periods because I never had them before.

26B

Expectations of treatments

Many women interviewed had high expectations of a prompt positive outcome from their treatments, both those prescribed standard treatments and those prescribed the LNG-IUS. Women recounted only positive information tended to be shared by friends about treatment for HMB, using upbeat language. The reported benefits of a particular treatment in the lay press or online underline this, in particular, expectations for the effectiveness for the LNG-IUS. Some women expected a very significant impact, not only on a range of menstrual issues, but also on other areas of life, such as life at home, relationship issues and career. This highlighted how HMB could dominate their lives.

I would never have to worry again, it might be the miracle cure that I am after.

18A

Most women had other concerns in relation to HMB. These included pain (cramping period pain), fatigue, anaemia, irritability, low mood, joint pain, lack of energy and headaches. These were highly individual and often combined. Many women did not think of their menstrual bleeding in terms of blood volume, but were far more concerned about rate of blood flow, the presence of blood clots and controlling this flow in order to maintain concealment of their menstrual state. In addition, related pain, cramps and backache were the most significant issues they wanted to address. Many of these women felt this pain to be severe and debilitating and were completely unable to continue with their normal daily activities during menstruation, resulting in taking time off work.

It’s a strange sensation because it’s coming away from you and you have no control over it . . . I have blood clots sometimes the size of a fist and it’s awful.

15A

But it wasn’t the actual bleeding part, it was more the pain … doubled over with the pain.

5A

Then the last year they have got worse with pain and having to have time off work for a couple of days … It always felt like my insides were dropping out the pain on the first 2–3 days.

5A

Women described complex behaviours and coping mechanisms to enable them to both manage their menstrual bleeding and to conceal their condition not only from work colleagues and friends, but also from family members. Preparatory coping behaviours included carrying extra sanitary protection at all times, frequent changes and checking of blood loss, and having a spare change of clothes (sometimes two).

It was every handbag had some sanitary wear in it just in case . . . I always knew it was best to stock up at that time ready for next time.

29B

Avoidance behaviours included rescheduling activities at home and at work, cancelling meetings both social and professional, taking time off sick from work and remaining at home during menstruation. Other mechanisms included having someone else take care of children, or take their place or shift at work. These coping mechanisms for HMB allowed the woman to rest and ensured that their menstrual difficulties remained concealed. If these failed, women reported feeling embarrassment, guilt and shame, particularly if other people needed to help them. They often became distressed and upset, affecting self-esteem and mood.

It was on our staff party [date] it was and it was mid-afternoon and . . . I just thought I need to go to the ladies but . . . when I got to the ladies toilets . . . I had wet my, soiled my trousers, my underwear was saturated through and the pain and the contractions were really bad and it was just pouring out of me . . . I was too embarrassed to come out . . .

39B

Heavy menstrual bleeding could have a negative and wide-ranging impact on women’s QoL, which was most often related to their ability to function effectively in their daily roles, to be able to cope well with their responsibilities at home, at work and within their relationship with their partner during menstruation. The negative impact of HMB on their roles and responsibilities was influential in women’s decisions to seek treatment.

I would say it’s not trivial. You know to them it may be something that you know if they are a gynaecologist and they are dealing with it may be the least of someone’s problems but to that woman it is the butt of her problems you know and it has an impact on her whole being her whole holistic being revolves around that one thing.

27B

Beyond peers and family, few women had sought information about treatment from other sources such as the Internet and all had a treatment preference prior to their medical consultation. The most requested treatment was the LNG-IUS. Women who expressed a preference for other standard treatments did not want the LNG-IUS because of its effect on fertility and also the invasive nature of the coil. Women who rejected the LNG-IUS because of its contraceptive effect were all in their forties and most strongly wanted to conceive a child.

Initial experiences of the levonorgestrel-releasing intrauterine system

Many women who were treated with the LNG-IUS found the coil fitting to be more distressing than anticipated and some felt unprepared psychologically for the procedure. While many doctors did not recommend the LNG-IUS for nulliparous women, multiparous women still had problems. They recounted feeling pain and discomfort in excess of other medical procedures within their previous experience and used emotive language.

I suppose I was a bit naïve, I suppose a picture might have helped. I was thinking of like a smear test but they actually have to open you up a bit more and obviously that’s not very pleasant.

29B

I had it done [LNG-IUS fitting] and it really hurt. I think you would really have to make that clear to somebody. Even if they have not even had a baby or something. It hurts for a short time I suppose. But you wouldn’t like, if you were doing that to somebody’s neck it would be classed as torture or something you know it really is nasty. You know a bit sort of medieval.

29B

Some women had been told to take oral analgesia before the procedure, but many did not. Most women did not specifically mention or recall local anaesthesia in discussing their experience of the LNG-IUS insertion but those that did were dissatisfied with it.

I don’t think she put any [expletive] numbing lotion there to be fair, there is no way on earth that she put anything to numb me because there was no numbness.

Women felt that health professionals taking consent for the procedure had downplayed the possible intensity of the pain and also the complexity of the procedure itself. Women had a strong sense that while this was a simple procedure for the health professional, it was not simple for them experiencing it.

She said it’s going to be feeling like this, this, and this, you are going to feel a little bit of pain and it should be over and done within like 2 minutes, OK, right, not a problem. Oh how very wrong she was. The little [nationality] nurse I was holding her hand and I had to let go of her hand because I would have broke her fingers because in the end it was literally gripping on to the sides of the bed, tears coming down my eyes, excruciating pain . . . I felt sick, I felt giddy I just wanted to be sick and then she done it and the pain was just excruciating, she said you are going to have to wait like half an hour in here. I couldn’t walk when I got off.

23A

Women then found the LNG-IUS troublesome within the first 3 months of treatment because irregular bleeding interfered with women’s long-established behavioural coping mechanisms.

I was quite anxious because I kept thinking that at some point they could come back or they could start or I could spot and that did actually happen. Initially I was having spotting or a mild period pain and things like that but they just got less and less.

Some considered removal of the LNG-IUS during this time but did not seek to do this partly because they were advised or anticipated it would improve over time, or they thought removal would be difficult, or they might lack an alternative treatment. Some women did not ask for the LNG-IUS to be removed because they did not want to go through the process of revisiting their GP and having a vaginal examination within such a short time.

The other day it [intermittent bleeding following treatment with LNG-IUS] drove me so mad somebody actually offered me pliers and I nearly took them [pause] you know.

27B

Tell you the truth I did actually think about going back to the doctor because I thought . . . perhaps there was something wrong, perhaps the coil perhaps wasn’t working properly or perhaps something had happened and it was sort of making it go a bit weird but then I never sort of got round to making an appointment and then the next month would come and it would be a bit better and then by the third or fourth month it seemed to have settled down anyway.

Most women attending for their routine check, around 6 weeks after IUS fitting, were asked by health professionals about whether or not their menstrual blood loss had reduced. However, women found that explaining blood loss in terms of volume was challenging and, moreover, felt that questioning changes in menstrual blood flow did not reflect a more desired holistic change for them.

I think sometimes [doctors] are dealing with the problem as opposed to the woman and that is the mistake; that is the mistake.

27B

Women expected a period of transition following the initial fitting of the LNG-IUS, perhaps reflecting information given by most health professionals about the initial side effects at the time the LNG-IUS was fitted. However, women were surprised at the amount of disruption to their cycle or amount of irregular bleeding as this was greater than either their peers or health professionals had indicated, and indeed some had thought their menstrual bleeding would cease completely within 1 or 2 months

It took a while, it didn’t do much at the beginning, it took a while for it to do anything. I didn’t think it was going to do anything at first but yeah, it took a while and I noticed it slowing down. That was the big thing for me. It didn’t matter if it didn’t stop or anything, just slowing it down. It has done it in the same year, it has done it bit by bit, it slowed down and slowed down and now for the last 2 or 3 months I have not had anything, not a thing.

15B

Women who had a positive experience of treatment were more inclined to talk about it, while women who had had a negative experience of treatment were more likely to remain silent. Women did not anticipate major changes in their menstrual cycle, such as irregular bleeding and spotting, or that this might last much longer than expected before they saw an improvement in their condition.

Experiences of the levonorgestrel-releasing intrauterine system after 12 months of treatment

Most women persevered with the LNG-IUS and at 12 months were satisfied with this treatment, experiencing a positive and often major impact on their QoL. The treatment took effect gradually over the first 3–6 months.

It started off really, really excellent then we tended to have a bit of a dip where it was a bit hit and miss some months but now it seems to have settled back down again so I don’t know whether that was just a fluke or perhaps whether I was a bit more stressed at work at the time and you know busy in my life that made an impact on that but now on the whole is the best decision I could have made.

26B

Most women had, as expected, experienced amenorrhoea (cessation of periods) with this treatment after 12 months, but even those still having menstrual periods felt significantly better.

I was concerned because obviously they advised that if this [LNG-IUS] didn’t control it, it would be a hysterectomy, which was obviously not what I would like. I was only 40, 39 or 40, at the time and so I had that fitted now for a while and everything is now wonderful so I have had no more problems any more.

42B

It is such a relief, it really is, after the pain that I used to have before and the tiredness and the flooding and the whole, it was almost like your whole life revolved around periods and being comfortable or uncomfortable and you know it was, it was awful and I was feeling constantly tired and thinking this is just, you know, that just awful and now that has all gone and it is just such a relief, it really is.

16B

It just sort of gave us I suppose my bit of life back, my bit of independence back that I could do, you know carry on without thinking you know have I leaked.

26B

Most women with a LNG-IUS were happy to continue with this treatment long term. Several women who had considered hysterectomy prior to commencing their treatment with the LNG-IUS now no longer considered surgery as necessary or appropriate use of health-service resource.

What I have now is not worth going through an operation for . . . If it hadn’t worked and I was still having heavy periods then I would think about it but no, not with the coil, no it’s not worth having that done, taking up space in the hospital just to have that.

5B

Experiences of standard treatments

Women who were prescribed standard treatments (most commonly oral tranexamic acid, but also norethisterone, medroxyprogesterone acetate injection and the oral contraceptive pill) had widely varying perceptions and experiences. They were regularly less satisfied with their treatment, with most having ceased prescribed medication at 12 months.

Women were generally dissatisfied with tranexamic acid, but did acknowledge some reduction in menstrual blood flow. Some were concerned about the physical size of tablets and the number that they had to take each day (two tablets, three times per day is commonly recommended). Most did not perceive significant improvements in the short term and so stopped taking their medication after 2 months.

If I take them it does keep the bleed, you know, under control. If I leave it a little bit too long before I take the next lot I am just like flooding again so it’s, I am not really one to take tablets though, it’s very hard.

32B

It’s like remembering to take X, Y, Z and the amount of tablets all the time, that can be a problem . . . remembering to take tablets and making sure there’s the right spaces in the day between these tablet taking that can be a problem in itself ‘cos you might not be where your tablets are or you might be, you know, hung on at work doing extra hours and that and it’s all that and it’s all that added stress that you don’t really need.

7B

I was taking, I had got good at remembering to take them, take them just before and take the maximum dose and I thought they are not making much difference at all, I still can’t get out of bed without hopping to the bathroom cross-legged.

29B

The experience of the few women prescribed oral and injectable contraception for HMB was also generally not satisfactory.

I hardly have any periods, I think I have 2 days of bleeding, don’t get any pain but 2 weeks before my period is due I get terrible depression where I am always crying. I just feel dead depressed and I just let small things become big things and it’s just when I am coming up to my period.

43B (oral contraceptive)

So I had the injection in [month] I haven’t stopped bleeding since, now it’s not working and I am getting chronic migraines all the time. And I have ballooned in weight in the last 3 months so suddenly it is not working, not suiting me. I am changing my treatment as of next week.

18B (Depo-Provera)

Oral contraception was continued for other reasons such as familiarity, reversible contraceptive effect and lack of interference with other medical treatments.

I have been taking the iron tablets, which help with the dizzy spells and things like that. I did try the other medication [tranexamic acid] but I tried it and I felt, no, I didn’t feel anything changes much and therefore I thought is there any point taking another medicine if it doesn’t make any difference? I still have the pain and I still felt the same and the blood flow seems to be no better so I haven’t taken it again.

28B

Women perceived standard treatments to be a short-term measure. Many who discontinued treatment elected not to try any other treatment, preferring to manage their condition themselves by changing coping behaviours and strategies. This was particularly true of the women who had a long history of HMB and those who anticipated the onset of menopause within the next few years.

Some women were concerned about the physical amount of tablets that they would have to take over the course of their remaining years of menstruation. Most disliked the thought of taking medication, with concerns about the long-term effects of the medications prescribed. For some, this was related to potential unknown side effects.

They [tranexamic acid] did work but I didn’t like the quantity of tablets that I was taking.

18A

I will probably go back to the doctor and get some more tablets . . . unless there was something else that they could try instead rather than stay on them long term. But in the short term, yes . . . I know they research them but they never know until years later the effects. It might not have an effect now but in years to come and I’d rather not take the risk of causing anything long term. If I had a choice I would rather not take the tablets.

30B

Women who were initially treated with standard treatments commonly changed their treatment over time, with six women in this sample eventually opting not to use any treatment at all and manage without: two resuming standard treatment after intervals of using no treatment and four women switching to the LNG-IUS. One woman had a hysterectomy.

Strengths and limitations

The work provides qualitative insights in to how women experience current medical treatments for HMB, with the advantage of a longitudinal design, over a timeframe of around 1 year. Methods were used to enhance rigour, including purposeful sampling, involving researchers from differing disciplines in analysis and respondent validation. The sample is described in some detail, which may aid assessment of relevance of findings beyond the study and trial context. However, it is recognised that this was inevitably a selected sample of women willing to articulate their experiences. Although participants were engaged in the context of clinical trial recruitment, we actively included those unwilling to be randomised and those with preferences for a particular treatment for HMB.

Comparison with existing work

Previous work has found that health professionals may be preoccupied with volume of blood loss, within a biomedical model, rather than addressing the impact on women’s health. 95,96 Although highlighted in more recent research,89 assessing reduction in blood volume for treatment efficacy,97 or what reduction in menstrual blood loss is meaningful to women,98 remains a focus of research. Higher volume of daily menstrual blood loss has been correlated with limitations in functioning and social activities. 99 Women in this study were as concerned with menstrual pain and psychological distress affecting their health and QoL as with HMB itself. This finding is consistent with other work. 95,96 Going beyond reduction in volume of blood loss alone, measures of QoL are now becoming more widely used to determine treatment efficacy. 69,93

More than 1 year following initial treatment with the LNG-IUS most women were satisfied with their treatment and had experienced an improvement in their perceived QoL. This finding concurs with the quantitative results of the ECLIPSE trial in Chapters 3 and 4. 93 Narrative accounts were overwhelmingly positive and a number of women used emotive language to describe the improvement in their health and well-being. Most women had experienced amenorrhoea, but even women who still experienced regular menstrual bleeding felt better. The change in perceptions of health and well-being was always associated with being able to fulfil roles and responsibilities. Some women described an improvement in physical well-being, but the emphasis for all was on the psychological benefits of being able to maintain the societal norms of concealment of menstruation. The later acceptability of the LNG-IUS reported here concurs with Lete et al. ,100 who found that 87.9% of pre-menopausal women in their study would use a second LNG-IUS device.

Women who were prescribed tranexamic acid initially took the medication as prescribed and experienced some benefit, but most did not feel better quickly enough to continue with their treatment. However, others have found long-term use of tranexamic acid to be well tolerated over a 15-month timeframe, although emergent adverse effects such as headaches, menstrual and back pain were commonly reported. 86

Implications for practice

These qualitative findings suggest that discussion with women about medical treatments for HMB should not be confined to orthodox clinical considerations, such as volume of heavy bleeding or a woman’s wish to avoid having an IUS or contraception. Rather, health professionals could also helpfully consider the wider impacts of HMB on women’s lives and respond to these appropriately in sharing decision-making with women about choosing and using treatments. Women’s high expectations of a prompt effect from treatments may be understood in terms of their prolonged toleration of the impacts of HMB and its concealment, which had typically preceded seeking medical help. Potentially high expectations of treatment (and the potential influence of peers) should be appreciated and explored appropriately when discussing treatment.

In assessing treatments with women, practitioners should consider that issues such as effect on rate of blood flow, pain and well-being, as well as wider functioning socially or at work, may be as or more important to individuals than amount of menstrual blood loss. Women’s accounts also underline other points for helping women make more informed decisions. These include avoiding ‘underplaying’ initial problems that may be unanticipated by those considering the LNG-IUS, by ensuring that women are appropriately prepared (physically and psychologically) for fitting the LNG-IUS in practice, and in particular for troublesome changes in pattern of bleeding that may commonly occur and can be prolonged.

Unpredictable irregular bleeding with the LNG-IUS was more problematic than the volume of blood loss for women in this study owing to its impact on their established behavioural coping mechanisms for HMB. Irregular bleeding is recognised as a common reason for discontinuing the LNG-IUS. 48,81 Nevertheless, the current qualitative findings, in line with the strong pragmatic clinical trial evidence reported in Chapters 3 and 4, underline that women can be advised that, if they persevere with leaving their IUS in place, they are likely to experience significant improvement in menstrual symptoms and related QoL.

Women’s perceptions of other treatment options, as a more short-term measure they could control themselves (stopping and starting as they wished), may provide helpful information to offer women considering what may suit their individual circumstances. Information provided may highlight that improvements in the impact of HMB may be expected, but may be variable and may not be as substantial as for the LNG-IUS, in line with the trial evidence. Women may also be given advice regarding practical considerations, such as the inconvenience of taking several large tablets in the case of the most commonly prescribed oral medication, tranexamic acid.

The findings also highlight an important wider context for discussing HMB treatments in relation to effect on fertility and desire to conceive among women in their forties presenting to services with change in menstrual experience, often at a time of other life transitions.

Model structure

The model structure was informed by clinical input and the pathways followed by women in the trial. Several iterations of the structure were made as the model was continuously modified to ensure that it most closely reflected the trial data. Although the structure was not informed by literature, the health states for the model were developed using trial data and by adapting those used previously. 1,103 The inclusion of menopause as a health state was considered but the 2-year time horizon of the analysis and the age of the patients, between 25 and 50 years, meant very few women, if any, would be reaching menopause within 2 years. At the 5-year time point the inclusion of menopause was considered but deemed inappropriate as this specific information was not captured (see Additional model assumptions for the analysis based on the outcome at 5 years). For similar reasons, death was not included as a health state: the interventions do not cause death and the age of the women meant that the likelihood of death from natural causes was minimal. As models are a simplification of reality, it is necessary to include only transitions between health states that commonly occur for the majority of patients, and this was not the case for death and menopause.

Figure 7 presents the clinical pathways and the progress of the two cohorts of women in the ECLIPSE trial who were randomised to either the LNG-IUS or usual medical treatment. The pathway for each treatment can be represented in the same structure.

FIGURE 7.

Clinical pathway for the LNG-IUS and usual medical treatment.

Definition of health states for levonorgestrel-releasing intrauterine system pathway

• HMB: every woman in the ECLIPSE trial cohort suffers from HMB.

• LNG-IUS: all women have the LNG-IUS inserted.

• Well with the LNG-IUS: following the insertion of the LNG-IUS women are satisfied with treatment and remain in this health state.

• Symptomatic with the LNG-IUS: following the insertion of the LNG-IUS, adverse effects may arise and women may feel unsettled with their current treatment. From this state women may remain in the state or change to an alternative treatment, (i.e. usual treatment, surgery or no treatment).

• Surgery: women may choose to have surgery, which includes endometrial ablation or hysterectomy, determined from the data. Women can move to surgery from ‘symptomatic with the LNG-IUS’, ‘change to usual medical treatment’ or ‘no treatment’.

• Post surgery: following ‘surgery’, women will then remain in the ‘post-surgery’ state.

• Change to usual medical treatment: if women have chosen ‘no treatment’ or are ‘symptomatic with the LNG-IUS’ they may choose to begin usual medical treatment. Women may immediately become ‘well with usual medical treatment’ or may experience adverse effects and feel unsettled with usual medical treatment and then eventually become ‘well with usual medical treatment’. Alternatively, women may choose to have ‘surgery’ or ‘no treatment’. If women are in this state at 3 months, the cost of a GP review will be incurred. Women can change to usual medical treatment from the LNG-IUS in the first cycle. (see Model assumptions for further details).

• Well with usual medical treatment: following ‘change to usual medical treatment’ women are satisfied with this treatment and remain in this health state.

• No treatment: following the LNG-IUS, ‘symptomatic with the LNG-IUS’ and ‘change to usual medical treatment’, women may choose to discontinue treatment altogether. Following this, women may decide to ‘change to usual medical treatment’ or have ‘surgery’.

Definition of health states for usual medical treatment pathway

• HMB: every woman in the ECLIPSE trial cohort suffers from HMB.

• Usual medical treatment: all women have usual medical treatment.

• Well with usual medical treatment: following ‘usual medical treatment’ women are satisfied and remain in this health state.

• Symptomatic: following usual medical treatment, adverse effects may arise and women may feel unsettled. From this state, women may have an alternative treatment (i.e. ‘change to the LNG-IUS’, ’surgery’ or ‘no treatment’).

• Surgery: women may choose to have surgery, which includes endometrial ablation or hysterectomy, determined from the data. Women can move to surgery from ‘symptomatic with usual medical treatment’, ‘change to the LNG-IUS’ or ‘no treatment’.

• Post surgery: following ‘surgery’, women will then remain in the ‘post-surgery’ state.

• Change to the LNG-IUS: if women have chosen ‘no treatment’ or are ‘symptomatic with usual medical treatment’ they may choose to begin treatment with the LNG-IUS. Women may immediately become ‘well with the LNG-IUS’ or may experience adverse effects and feel unsettled with the LNG-IUS and then eventually become ‘well with the LNG-IUS’. Alternatively, women may choose to have ‘surgery’ or ‘no treatment’. Women can change to the LNG-IUS from usual medical treatment in the first cycle. (see Model assumptions for further details).

• Well with the LNG-IUS: following ‘change to the LNG-IUS’ women are satisfied with treatment and remain in that health state.

• No treatment: following ‘usual medical treatment’, ‘symptomatic with usual medical treatment’ and ‘change to the LNG-IUS’ women may choose to discontinue treatment altogether. Following this, women may decide to ‘change to the LNG-IUS’ or have ‘surgery’.

Time cycle

A monthly time cycle was used in the analysis as this represents the clinically meaningful changes observed in treatment and resource use. The ECLIPSE trial data showed that, in many cases, treatments had been altered and women had visited their GP within a month of treatment administration. A monthly time cycle was therefore required to reflect the data and incorporate the associated costs and treatment changes into the CUA. A half-cycle correction is not applicable in this model because the monthly time cycles used in the model are genuine discrete time events and this transition did not occur in the middle of a cycle. On this basis, a correction would not have made a substantial difference to the costs and outcomes. 104

The following assumptions were made and developed based on clinical expertise from practising primary care and gynaecology clinicians, in addition to standard treatment protocol. The first set of analyses to be carried out was based on the outcome at 2 years post intervention. Some additional assumptions were required to extend the analysis to include a 5-year time horizon and these are reported separately. For both sets of analyses, the assumptions were agreed prior to conducting the analysis. Unless otherwise stated, the assumptions used for the analysis at 2 years were maintained for the 5-year analysis.

Model assumptions

Calculating transition probabilities

To use the trial data in the model, it was necessary to identify the distribution of women in the states at any time and the transition probabilities between health states.

For every patient, the time in months from randomisation to every event (i.e. ‘stop treatment’) was calculated. For each arm of the trial, the movements of all patients between states for every time cycle were identified to determine the transitions made in each cycle. Then from each state (e.g. the LNG-IUS) the total number of transitions from this state were summed to calculate the transition probabilities to each health state, the number of people moving to each state (e.g. well) was divided by the total number of movements from the LNG-IUS. This process was repeated for every other transition from the state and then repeated for all remaining states.

In cases where women made moves that were not possible in the model, for example returning to allocated treatment after changing to another state, these women were censored at the last point a move was possible. This change only occurred in four cases and is advised against clinically. Therefore, this transition was not considered to be a typical pathway and does not represent the population. As the point in the cycle when the women moved back to ‘usual medical treatment’ is unknown, the women were censored at one cycle prior to the return to randomised treatment.

To be consistent and prevent biasing any data, the two patients who changed to the other treatment arm within the first month of treatment allocation were moved directly to this state without spending any time in symptomatic, as reflected in the data. The treatment was never taken and so considering these patients as symptomatic would not be accurate.

Where the woman ‘changes to the other treatment arm’ to ‘well with the other treatment arm’, one of the model assumptions set out above should again be highlighted. That is, women must spend at least one cycle in ‘change to other treatment’ to then enter the ‘well with other treatment’ state as it will take this long for the treatment effects to be seen.

As women cannot stay in surgery, it is constructed as a transition state and therefore all of the women in ‘surgery’ move to ‘post surgery’ after one cycle. ‘Post surgery’ and ‘well’ are both absorbing states. Therefore, as women cannot leave the state once they have entered, the probability of remaining in these states is fixed to one.

The calculation of the utility values assigned to each health state will be described in Quality-of-life measures following a description of the outcome measures collected in the trial.

Quality-of-life measures

Outcome measures were collected using both the EQ-5D and the SF-36 at baseline prior to randomisation, then by post at 6 months, and 1 and 2 years post randomisation. The booklet questionnaire given to women in the trial contained the generic EQ-5D 3-level questionnaire, which measures the impact of treatment on broader aspects of health-related QoL. 105 The SF-36 was converted into SF-6D using the algorithm. 106 Utility values for each state in the model were calculated by averaging the EQ-5D or SF-6D values for each woman in the given state at any given time.

Utility values for the individual states were calculated by averaging the EQ-5D and SF-6D values obtained by each woman in the given state at any given time. For example, if a woman is randomised to the LNG-IUS and then does not change treatment she is considered to be in the ‘well’ state for the remainder of the analysis, as outlined in the model assumptions previously. Therefore, all of the woman’s utility values collected at 6 months and 1, 2 and 5 years will be assigned to the ‘well’ health state in the model. Similarly, if a woman is initially ‘symptomatic with the LNG-IUS’ and then moves to ‘no treatment’ at 2 or 5 years, the utility values for the intermediate months will be assigned to ‘symptomatic’ and the utilities for 2 and 5 years assigned to the ‘no treatment’ health state. This method was used to derive the utility values because the utility for the state is important, not the values associated with the individual woman’s journey, as decision models are a reflection of the typical population.

Costs and resource use

Costs were collected from a UK NHS perspective. Data on health-care resource use, including GP or gynaecologist consultations, were collected from women alongside other outcome measures. Similar to the utility values, data on costs and resource use from the trial were collated and analysed according to the randomised treatment arm. The general health-care costs for both groups included health-care staff costs and the cost of the interventions. LNG-IUS fitting is estimated to take 20 minutes (informed by clinical experts within trial team), requires both a GP and nurse to be present and also requires disposable consumables. Treatment review by the GP was assumed to last 10 minutes (informed by clinical experts within trial team). Staff costs were calculated using nationally recognised reference costs. 107 The costs of standard medical treatment and the LNG-IUS were estimated from the British National Formulary. 23 Cost data on surgical interventions were taken from a previously published study and inflated to 2011 prices. 103,107 As recommended by NICE, a discount rate of 3.5% was applied to both costs and utilities as the model time horizon is beyond 1 year. 108 All costs are reported in 2011 prices in UK sterling (£) using the UK hospital and community health services index. 96,107 Tables 10–12 present the data used in the analysis.

In cases where women were prescribed a combination of usual medical treatments, a weighted average of the cost was taken. Similarly, repeat prescription costs were calculated based on the average weighted cost of repeat prescriptions in the ‘change to usual medical treatment’ state of the LNG-IUS arm. As the most commonly prescribed usual medical treatments involved GP review for effectiveness at 3 months, it was assumed that GP review occurs at 3 months.

Where one-off costs were incurred, such as the cost of LNG-IUS removal, these costs were assigned to the transition arrow.

Distributions for the probabilistic sensitivity analysis

A Dirichlet distribution was used for the transition probabilities because this distribution allows for a possible movement to more than two states. Where a transition between states is possible, but did not occur in the trial data, a value of 1 is added to all such transitions to enable the Dirichlet distribution to be assigned and to compensate for not observing the transition in the data. 112 In Table 12 it can be seen that some distributions are fixed at the given value. This is because these transitions are not dependent on the data and logically it would not be appropriate for the transition to be made probabilistic. Utility parameters are assigned a beta distribution.

Costs are not assigned distributions because they are primarily unit costs, which are variable, but not uncertain. Where costs are uncertain, the PSA on utilities and transitions will be sufficient to account for this uncertainty. Furthermore, as distributions for these costs and transitions would be particularly complex, the small number of uncertain costs and very small cost values are unlikely to affect the results in a manner sufficient to warrant using a separate distribution.

Part 1a: European Quality of Life-5 Dimensions at 2 years

Table 13 presents a summary of the base-case and the deterministic sensitivity analysis results using the EQ-5D.

The base-case results show that the LNG-IUS costs £100 more than usual medical treatment, as the LNG-IUS costs £430 and usual medical treatment costs £330. However, the LNG-IUS also generated 0.067 more QALYs than usual medical treatment, as the LNG-IUS generated 1.580 QALYs and usual medical treatment 1.513 QALYs. The ICER for the base-case analysis shows that the LNG-IUS generates £1600 per additional QALY when compared with usual medical treatment.

In all three of the deterministic sensitivity analyses, the findings supported the base-case results. However, the ICER in each analysis did differ. Sensitivity analyses 1 and 2 had a slightly less favourable effect, increasing the ICER to £2030 and £1640 per QALY gained respectively, while sensitivity analysis 3 resulted in a more favourable effect on the ICER, with a reduction to £1560 per additional QALY.

The results of the PSA in Figure 8 illustrate the distribution of the incremental costs and effects from 1000 Monte Carlo simulations. It is shown that the majority of the incremental costs and effects lie in the upper right-hand quadrant of the cost-effectiveness plane. This indicates that the LNG-IUS is both more costly and more effective than usual medical treatment. The uncertainty is then summarised in relation to the changes in the decision-makers’ threshold for considering an intervention cost-effective in Figure 9. This figure depicts the CEAC, which shows that, from £2000 per QALY, the LNG-IUS has a greater probability of being the more cost-effective intervention. This probability increases to over 90% at approximately £4000 per QALY and this figure is maintained at the subsequent thresholds.

FIGURE 8.

Results of the PSA (EQ-5D) at 2 years. The sloping line indicates the £20,000/QALY threshold.

FIGURE 9.

Cost-effectiveness acceptability curves for usual medical treatment and the LNG-IUS using the EQ-5D at 2 years. Note that the CEACs have stabilised well before the £20,000 per QALY value. Extending these curves could mean loss of clarity in the early part.

Part 1b: European Quality of Life-5 Dimensions at 5 years

The base-case results at 5 years for the EQ-5D (summarised in Table 14) show that the LNG-IUS costs £10 more than usual medical treatment (the LNG-IUS costs £517; usual treatment costs £507). The LNG-IUS also generates 0.090 more QALYs than usual treatment (the LNG-IUS generated 3.698 QALYs and usual medical treatment 3.608 QALYs). Therefore, the additional costs of the LNG-IUS compared with usual medical treatment was £114 per additional QALY gained.

The results of the PSA at 5 years are presented in Figures 10 and 11 in the CEAC. The CEAC shows that the LNG-IUS has a greater probability of being the more cost-effective intervention from approximately £500 per QALY onwards.

FIGURE 10.

Results of the PSA (EQ-5D) at 5 years. The sloping line indicates the £20,000/QALY threshold.

FIGURE 11.

Cost-effectiveness acceptability curves for usual medical treatment and the LNG-IUS using the EQ-5D at 5 years. Note that the CEACs have stabilised well before the £20,000 per QALY value. Extending these curves could mean loss of clarity in the early part.

Part 2a: Short Form questionnaire-6 Dimensions at 2 years

Table 15 presents the base-case and deterministic sensitivity analysis results for the SF-6D.

The base-case results show that the LNG-IUS costs £100 more than usual medical treatment as the LNG-IUS costs £430 and usual medical treatment costs £330. However, when the SF-6D is used, usual medical treatment is shown to generate 0.002 more QALYs than the LNG-IUS, as usual medical treatment generated 1.200 QALYs and the LNG-IUS 1.198 QALYs. Therefore, it is shown that usual medical treatment dominates the LNG-IUS.

In two of the three deterministic sensitivity analyses, the findings supported the base-case results. In sensitivity analyses 1 and 2, usual medical treatment dominated the LNG-IUS, whereas in sensitivity analysis 3 usual medical treatment does not dominate the LNG-IUS. The LNG-IUS was shown to be more effective than usual medical treatment and more expensive, generating an ICER of approximately £110,000.

The results of the PSA in Figure 12 are presented on the cost-effectiveness plane and show the distribution of the incremental costs and effects from the 1000 Monte Carlo simulations. In this case, when the SF-6D is used to generate the effectiveness outcome, it can be seen that the incremental costs and effects are spread between the upper right-hand and upper left-hand quadrant of the cost-effectiveness plane, indicating that the LNG-IUS is either more costly and more effective or more costly and less effective. This is then summarised in the CEAC in Figure 13 to show that, for any threshold willingness to pay per QALY, usual medical treatment has the greater probability of being the more cost-effective intervention. This probability is 100% at £0 per QALY and decreases to 90% at approximately £20,000 per QALY.

FIGURE 12.

Results of the PSA (SF-6D) at 2 years. The sloping line indicates the £20,000/QALY threshold.

FIGURE 13.

Cost-effectiveness acceptability curves for usual medical treatment and the LNG-IUS using the SF-6D at 2 years.

Part 2b: Short Form questionnaire-6 Dimensions at 5 years

Table 16 presents the results based on using the SF-6D at 5 years. When the EQ-5D values used to generate QALYs were replaced with the SF-6D values, usual medical treatment generates 0.002 more QALYs than the LNG-IUS (usual treatment generated 2.797 QALYs and the LNG-IUS generated 2.795). The costs were unchanged from the previous analysis. Therefore, usual medical treatment is shown to be less costly, but generates more QALYs than the LNG-IUS and, thus, usual medical treatment is said to dominate the LNG-IUS.

The results of the PSA presented in Figure 14 and the CEAC (Figure 15) show that, for any threshold willingness-to-pay per QALY, usual medical treatment has the greater probability of being the more cost-effective intervention.

FIGURE 14.

Results of the PSA (SF-6D) at 5 years. The sloping line indicates the £20,000/QALY threshold.

FIGURE 15.

Cost-effectiveness acceptability curves for usual medical treatment and the LNG-IUS using the SF-6D at 5 years.

Main findings at 2 years

In the primary care setting, treating HMB using the LNG-IUS costs more, but is also more effective than usual medical treatment. The relative cost-effectiveness of the LNG-IUS compared with usual medical treatment is £1600 per QALY. This means every additional QALY costs an extra £1600. The deterministic sensitivity analyses showed the uncertainty in this ICER to be in the range of £1560 to £2030 for an additional QALY gained. As the NICE guidelines recommend new interventions into practice if the ICER is below £20,000 per QALY, the LNG-IUS would be considered cost-effective and recommended as the primary choice for women who require treatment, have no preference against contraception or intrauterine device insertion and no contraindication to LNG-IUS insertion.

However, the importance of selecting the most appropriate QoL instrument is highlighted when the measure used to assess QoL is changed in sensitivity analysis 4 from the EQ-5D to the SF-6D. In sensitivity analysis 4, where utility values from the SF-6D are used rather than the EQ-5D, the cost-effectiveness results differ. In contrast to the findings using the EQ-5D, usual medical treatment is the more cost-effective intervention. Usual medical treatment was found to dominate the LNG-IUS in the base-case and the two deterministic sensitivity analyses. In the third deterministic analysis, the ICER for the LNG-IUS was over £100,000 per QALY, which is much higher than the £20,000/QALY threshold currently set by NICE; therefore, usual medical treatment would be recommended for implementation in clinical practice.

The difference in the cost-effectiveness results derived by using the alternative measures to value QoL has a considerable impact on the cost-effectiveness decision. The different measures resulted in a change, not just in the strength of cost-effectiveness of the same treatment, but also in what was the most cost-effective treatment. Therefore, the recommendation to decision-makers would differ depending on the QoL instrument used.

Main findings at 5 years

The overall findings based on the 5-year time horizon support the findings of the 2-year analysis for both the EQ-5D and the SF-6D. Using the EQ-5D, the LNG-IUS is more costly and more effective than usual medical treatment. The relative cost-effectiveness of the LNG-IUS compared with usual medical treatment is now £114 per QALY. The mean difference in QALYs generated using the EQ-5D between the 2-year and 5-year analysis has increased by 0.023. The costs have reduced from a difference of £100 between the two treatments at 2 years to a difference of £17 at 5 years. The ICER reduced from £1600 per QALY in the 2-year analysis to £114 per QALY in the 5-year analysis. These findings show that at 5 years the LNG-IUS becomes more cost-effective using the EQ-5D, relative to the 2-year analysis. Using the NICE £20,000 per QALY threshold, the LNG-IUS is cost-effective and should be recommended as first-line treatment for HMB.

However, in contrast to the findings, when the EQ-5D is used to estimate the QALY, the results based on the SF-6D show that usual medical treatment is again less costly and more effective. At 5 years, using the SF-6D, usual medical treatment again dominates. While the mean difference in QALYs generated between the treatments is the same for the SF-6D in the 2-year and 5-year analysis (–0.002), the reduction in cost difference between the treatments from 2 years and 5 years (£100 difference between treatments at 2 years and £17 difference at 5 years) has meant that the extent of the dominance has ‘increased’. These findings show that at 5 years, using the SF-6D, usual medical treatment would be recommended as first-line treatment for HMB.

The reasons for the costs converging at the 5-year analysis could relate to the LNG-IUS having an initial one-off ‘upfront’ cost only and usual treatment a small, but continuous cost over time. Alternatively, owing to the non-curative nature of these treatments and the impact on QoL, it could be caused by the numbers of women who changed treatments and treatment arms in both arms of the trial, in addition to the number of women who had surgery, a cure, in both arms.

Strengths and limitations

The strength of this CUA is that it is based on data from the largest multicentre randomised trial undertaken for HMB. As the treatment is aimed at managing the condition, the changes in both QoL and costs throughout the woman’s treatment journey are critical to the analyses and these are most appropriately captured by using the trial data to populate a model. The model structure was developed based on the women’s pathway, data from the trial and supported by the advice of expert clinicians. All assumptions were agreed by the team in the model development stage prior to analysis. A further strength is that all data on resource use and outcomes were collected prospectively alongside the trial and the economic evaluation was an integral component of the trial design from the outset.

Some limitations exist as a result of some of the assumptions required for the model; for example, it was assumed that women are ‘well’ if they do not change treatment. Nonetheless, this assumption is mitigated by the fact that, if women enter the ‘well’ state, but are not well, this will be reflected in the overall utility value for ‘well’. Furthermore, utility data were collected at 6 months and at 1, 2 and 5 years and were averaged, so there were insufficient data available to test the time dependency of health states. A further potential limitation is that baseline differences in QoL data at the outset of the trial were not adjusted to be the same. If we did adjust fully for the difference in baseline we would find the LNG-IUS to be less effective than usual medical treatment, but, in so doing, the assumption would be imposed that the difference would be maintained over the time horizon. This would not be an appropriate assumption since the probability of regression to the mean over time would be ignored. It is therefore acknowledged that the base-case results might be slightly overoptimistic, but it is not clear how much of an adjustment, if any, would be appropriate. The initial EQ-5D score in the LNG-IUS group was 0.042 higher at the outset of the trial than in usual medical treatment and this difference is significant (p < 0.05). These data are based on individuals who have been randomised, so it is assumed that this difference occurred by chance and it does not follow that this initial difference between groups would be sustained over the 2-year or 5-year time horizon in the absence of treatment. Adjustment for baseline, therefore, risks imposing a difference at every point in time over the time period of analysis.

Comparison with other studies

To our knowledge, this is the first study to conduct a CUA using prospectively collected primary data from a trial to directly compare the LNG-IUS with usual medical treatment for women with HMB. Other economic evaluations of the LNG-IUS have been carried out; however, none has drawn a comparison of the LNG-IUS directly against usual treatment using primary data. Another primary study has shown the LNG-IUS to be cost-effective, but the comparator was hysterectomy and the study is considered to have methodological flaws. 58,103 Other studies have compared alternative treatments using model-based analyses and secondary data from reviews,1,113,114 although the studies typically compared the LNG-IUS, various oral treatments and various surgical techniques against one another. Two of these showed the LNG-IUS to be the most cost-effective intervention. 1,113 The remaining study suggested that hysterectomy was the optimal intervention, but the authors acknowledged that insufficient published data on the effectiveness of the LNG-IUS were available at the time. 114 Although the LNG-IUS is shown to be cost-effective against surgical interventions in these studies, it does not provide evidence for the LNG-IUS being cost-effective against other pharmaceutical treatments.

Implications and further research

The results, based on the decision-maker-recommended EQ-5D, provide clear evidence in support of the NICE guidelines that recommend that the LNG-IUS be considered the primary treatment for HMB. The main objective of this study was to provide evidence to decision-makers on the cost-effectiveness of two treatments for HMB; therefore, the primary results are based on the EQ-5D. However, as SF-6D data were also collected, the use of this measure was deemed worthy of exploration in the sensitivity analysis.

When the SF-6D is used to generate QALYs, the main base-case results are reversed and usual medical treatment is shown to be the more cost-effective intervention, and this is the case in analyses at both the 2-year and 5-year end points, hence the recommendation to decision-makers differs depending on the outcome measure used. The conflicting findings suggest that these measures may be capturing different aspects of QoL, which clearly has an impact on the results. Reasons why these instruments produce different results have been further explored by other authors,115,116 but that research is methodological and so far inconclusive, and at the current time does not help to guide which instrument is more appropriate within the context of HMB, particularly when the results are so sensitive to that choice. It may well be that neither instrument is appropriate, with their focus exclusively on health-related QoL and, as evidenced in a recent review of measures for HMB,69 consideration needs to be given to alternative measures, such as willingness to pay, which provides a broader assessment of well-being.

Additional work is ongoing by the authors of this report to explore the impact on the results of the use of WTP in a cost–benefit analysis. This is in contrast to the more commonly used measures, the EQ-5D and SF-6D, which may be limited by the recall periods used in these measures and their focus on health-related QoL.

Evidence for clinical effectiveness

Between February 2005 and July 2009, a total of 571 women with HMB from 63 UK centres were randomly assigned to either the LNG-IUS (285 women) or usual medical treatment (286 women). In the usual medical treatment group, 75% were initially treated with mefenamic acid or tranexamic acid or a combination of these two drugs.

At baseline, the women were substantially affected by HMB, as assessed with the condition-specific MMAS and general (SF-36) health-related scales. There were significant improvements over a period of 2 years in both the LNG-IUS and usual medical treatment groups, which were maintained to 5 years. However, improvements in average MMAS scores were greater within the LNG-IUS group than in the usual medical treatment group over the first 2 years, by an average of 13.4 points in an intention-to-treat analysis. Although MMAS scores and the majority of the SF-36 QoL domains at 5 years still favoured the LNG-IUS, these were no longer statistically significant, and contrast with the relatively large differences we noted at 2 years of follow-up.

The greater improvement with the LNG-IUS over the first 2 years was seen despite 36% of women having had the system removed, generally owing to the lack of effectiveness or to irregular prolonged bleeding, which are well-recognised reasons for discontinuing the LNG-IUS. 61,66 There were higher rates of discontinuation in those allocated usual medical treatments (64% at 2 years’ follow-up). The diminution of a difference between trial treatment groups may be unsurprising given the high proportions of women who, by 5 years, had changed to a treatment that worked for them or had ceased bleeding, with only 15% still taking usual medical therapy and 47% having a LNG-IUS in place.

We expected fewer surgical interventions in the LNG-IUS group, but the rates were similarly low in the both groups, with approximate 10% at 2 years and 20% at 5 years. Compared with previous trials,38,49,51,56,81 this low rate may reflect the eligibility criteria for the trial, since women who had fibroids or other disorders were excluded. However, this finding also suggests that women who present in primary care with an uncomplicated presentation of HMB can be both successfully treated at the initial point of contact, with persisting improvement in impact on QoL and avoid referral to secondary care.

Evidence for cost-effectiveness

Alongside the ECLIPSE trial, an economic evaluation provided robust evidence on the cost-effectiveness of the LNG-IUS compared with usual medical treatment. The general health-related instruments, the EQ-5D and SF-6D, represent the extra-welfarist approach to measure cost-effectiveness. A model-based CUA, based on an outcome of cost per QALY, was carried out.

Using the EQ-5D, at 2 years the relative cost-effectiveness of the LNG-IUS compared with usual medical treatment was £1600 per QALY, which by 5 years was reduced to £114 per QALY. This increase in cost-effectiveness is caused both by the mean difference in QALYs generated increasing by 0.023 and by the costs reducing from a difference of £100 between the two treatments at 2 years, to a difference of £17 at 5 years. However, using the SF-6D to produce utility values, usual medical treatment was assessed as less costly and more effective at both 2 and 5 years.

It can be seen that the difference in the cost-effectiveness results derived by using the alternative measures to value QoL will have a considerable impact on cost-effectiveness decisions. The different measures did not just change the strength of cost-effectiveness of the same treatment, but the most cost-effective treatment itself changed. Therefore, any recommendation to decision-makers will differ depending on the QoL instrument used.

Qualitative insights

There was a need to better understand women’s experiences of medical treatments for HMB from a qualitative perspective, alongside quantitative measures. The qualitative longitudinal design of our study over a time frame of around 1 year involved a purposeful sample of women willing to articulate their experiences. This included women participating in the trial and, in addition, those not willing to be randomised because they had a specific preference for a particular treatment for HMB.

The results show that women’s experiences and expectations of medical treatments for HMB vary considerably and change over time. In addition to a range of practical considerations for treatments, they emphasise that practitioners should consider that rate of menstrual blood flow, pain and well-being, alongside wider functioning socially or at work, may be as or more important to women as menstrual blood loss experienced.

Women had high expectations of a prompt effect from treatments. Unpredictable irregular bleeding with the LNG-IUS was more problematic than the volume of blood loss for women in this study, owing to its impact on their established behavioural coping mechanisms for HMB.

Measurement properties of Menorrhagia Multi-Attribute Scale

We investigated further the psychometric properties of the MMAS. We demonstrated that, in addition to having high face validity, the MMAS has good convergent and discriminant validity and test–retest reliability. This supported our decision to use the MMAS to measure the impact of HMB and its treatment on women in the ECLIPSE trial.

Trial Steering Committee

Jim Thornton (chairperson), Irwin Nazareth (independent primary care practitioner), Jayne Fountain (independent statistician), Bill MacKenzie (independent gynaecologist), Klim McPherson (independent epidemiologist) and Elaine Nicholls (patient representative).

Independent Data Monitoring and Ethics Committee

Mary-Ann Lumsden (chairperson), Nick Freemantle and Amanda Farrin (independent statisticians).

Past and present members of ECLIPSE project management team

Laura Gennard and Lisa Leighton (trial management); Hemi Soneja and Sheethal Madari (clinical research fellows); Pam Whatmough and Gail Prilezsky (research associates); Susan Snoxall, Lucy Ingram, Jackie Ingram, Catherine Warlow, Oonagh Pickering and Susan Sargent (research nurses); Laura Gross and Robert Hills (statistics); Diane Fraser (Professor of Midwifery, cosupervisor involved in qualitative study); Yemisi Takwoingi and Nicholas Hilken (database programmers); Richard Lilford, Khalid Khan and Robert Shaw (clinical advisors); and Carol Cummins (systematic reviews).

The NHS Derby City Primary Care Trust administered the support funding for the ECLIPSE trial.

The authors would like to express particular gratitude to the participating women and our NHS colleagues who supported recruitment for the trial. A complete list of the ECLIPSE trial collaborative group is available in Appendix 14.