Trial oversight

Study oversight was provided by a trial steering committee (TSC) and a data monitoring committee (DMC). The TSC provided independent supervision for the trial, providing advice to the chief investigator and sponsor on all aspects of the trial throughout the study. The DMC adopted the DAMOCLES charter³⁴ to define its terms of reference and operation in relation to oversight of the trial.

The trial had a favourable ethical opinion from the London (Bloomsbury) National Research Ethics Service Committee (REC No 14/LO/1602, 25 September 2014) and clinical trial authorisation from the MHRA. Amendments to the protocol, required as a consequence of the two USMs, were based on the MHRA guidance to monitor the safety of existing and new participants.

Patient and public involvement

The idea for the UCON trial was initially reviewed by a clinical studies group, including non-clinical members representing women’s health support groups, although none exist specifically for menstrual problems. A woman with lived experience and a professional understanding of the impact of HMB on women’s working lives, was invited to join the co-applicant team, to provide an independent lay perspective on the treatment options, the outcome measures and the approaches to recruitment. We also had another lay representative on the TSC, who responded to an invite via the Royal College of Obstetricians and Gynaecologists (RCOG) Women’s Voices panel.

At the time of first USM, when we were planning to reopen the trial to recruitment, we once again went via RCOG Women’s Voices. We conducted a small survey to elicit women’s concerns about the use of ulipristal, the addition of the blood tests for safety monitoring and the resumption of recruitment. The respondents were overwhelmingly in favour of the continuation of the trial and supportive of the information to be provided to existing and prospective participants and the safety measures.

Trial design

The UCON trial was a randomised, open-label, parallel-group, multicentre trial of UPA compared with LNG-IUS in women presenting to primary and/or secondary care with HMB. An embedded mechanism of action (MoA) study was also included (see Chapter 4 for the methods and results of studies pertaining to mode of action of UPA).

Recruitment

UCON participants were recruited from gynaecology outpatient departments in ten NHS participating sites across the UK (see Figure 2). Patients with HMB were identified either from general practice (via screening of HMB related codes) or by research nurses in secondary care who screened patient referral letters. Invitation letters (including the participant information sheet) were sent to potentially eligible patients who were then given opportunity to speak to the research nurse about the study by telephone. If the patient expressed an interest, they were invited to a screening visit where they were assessed for eligibility and written informed consent obtained. The gynaecologist providing clinical care discussed treatment options and established potential eligibility based on clinical history and treatment preferences. Potential participants at the Edinburgh site were offered the option to contribute to the MoA study in addition to the main study.

FIGURE 2

Identification and screening of participants for the UCON trial.* Following first USM (February 2018).

At the screening visit, a transvaginal and/or abdominal ultrasound scan was conducted (unless the patient had an adequate ultrasound scan within the 3 months prior to randomisation) and an endometrial biopsy taken (unless an adequate endometrial biopsy had been taken within the previous 6 months). Blood samples [haemoglobin, serum estradiol, with addition of liver function tests (LFTs) from 20 March 2018], were taken, clinical history was elicited, and a menstrual blood loss diary was provided to the participant.

The next appointment was at least one menstrual cycle after the screening visit. The results of the ultrasound scan, endometrial biopsy and later, the LFTs, were reviewed and eligibility for the trial determined. The menstrual blood loss diary was collected and the other patient questionnaires (see Outcomes) were completed. If eligible, the woman had a urinary pregnancy test and ongoing consent was confirmed before randomisation. Where ultrasound scan, endometrial biopsy or LFTs rendered the patient ineligible, appropriate treatment was offered. Reasons why screened women were not randomised were noted.

Eligibility criteria

Women were eligible for the randomised trial if they met all the inclusion criteria and had none of the exclusion criteria, which were determined at the screening and baseline visits by scans, tests and review of medical history.

Randomisation

Once final eligibility was established, women were randomised to the UCON trial (see Figure 3). Randomisation was performed using a secure online randomisation service provided by the Birmingham Clinical Trials Unit (BCTU). Participants were allocated in an equal (1 : 1) ratio to UPA or LNG-IUS using a minimisation procedure via computer-based algorithm (based upon the method described by Taves³⁵) to avoid chance imbalances in important prognostic variables. Strata used in the minimisation were:

FIGURE 3

Patient pathway.* Following first USM (February 2018), LFTs performed once a month during each 12-week course of treatment. If the test was abnormal (liver enzyme levels more than three times the upper limit of normal), treatment stopped and participant (more...)

• age: ≤ 35 years or > 35 years
• body mass index (BMI): ≤ 25 kg/m² or > 25 kg/m²
• presence of any fibroid > 2 cm, as determined by the ultrasound scans
• duration of symptoms: < 1 year or ≥ 1 year
• individual site.

For Edinburgh participants only, a further minimisation variable was included for those agreeing to participate in the MoA study. Participants not randomised at Edinburgh were designated ‘not applicable’ in the minimisation algorithm for the purposes of trial entry. To avoid any possibility of the treatment allocation becoming too predictable a random element was incorporated into the algorithm; participants were allocated to the minimised allocation with probability 0.8 and otherwise to the opposite intervention.

Investigational medicinal product information

UPA and the LNG-IUS are both investigational medicinal products under MHRA definitions.

Blinding

As the treatments are so different in route of administration, the participants, investigators, research nurses and other attending clinicians were not blinded to the treatment allocation.

Adherence monitoring

Participants randomised to UPA were provided with their prescription immediately. In the UPA group, women received a reminder to collect their repeat prescriptions from the hospital pharmacy. Self-reported adherence with treatment was evaluated by participants using the follow-up questionnaires (see Secondary outcome measures). For UPA, participants were given categorical choices that best represented their adherence with study medication: took medication not very often (once per week or less); took medication some days (2 to 3 days per week on average); took medication most days (5 to 6 days per week on average); took medication every day. Women who were considered adherent if medication was taken every day or most days. Pill counting was considered unfeasible due to the duration of treatment.

Those women randomised to LNG-IUS were encouraged to have it fitted promptly by the gynaecologist at the baseline visit. Women were counselled to expect some disturbance to their menstrual cycle but encouraged to persist with this treatment option. Retention of the LNG-IUS was captured by self-report; participants were considered adherent provided they did not report removal of the device.

Withdrawal from treatment

A participant could be told to cease the trial treatment if, in the opinion of the gynaecologist or GP, it was medically necessary to do so. Participants could also voluntarily stop UPA or request to have the LNG-IUS removed at any time; however, women were encouraged to continue follow-up following cessation or change of trial treatment to minimise attrition bias. If a participant did not return for a scheduled visit (or attend a telephone clinic appointment, once face-to-face visits were restricted due to the COVID-19 pandemic), attempts were made to contact her and, where possible, review adherence and safety data. All attempts were made to capture reasons for cessation or change of treatment.

Following the first USM, those participants prescribed UPA were allowed to complete their current course of treatment but not start any subsequent course. The second USM required participants taking UPA to cease treatment immediately and not take any further courses. The trial continued with follow-up of all participants, regardless of adherence (enforced or non-enforced) to 12-months post randomisation.

Withdrawal from the trial

Participants could voluntarily withdraw their consent to study participation at any time. If a participant did not return for a scheduled visit, attempts were made to contact her and where possible, review adherence and safety data. Reasons for withdrawal were captured where possible. If a participant explicitly withdrew consent to have any further data recorded their decision was respected and recorded on the electronic data capture system. All communication surrounding the withdrawal was noted in the patient’s medical notes and no further data collected for that participant.

Outcomes

Secondary outcome measures

Secondary outcomes are listed below and were collected at time points shown in Table 1.

TABLE 1

Schedule of outcome assessments

1. Condition-specific quality of life score as measured by the MMAS at the other assessment points (see Study assessments).
2. Menstrual bleeding, captured by validated pictorial blood loss assessment chart (PBAC)⁴² via menstrual blood loss diary. Summary scores range from 0 (amenorrhoea), with increasing scores indicating worse bleeding (no upper limit). It was used to generate the incidence of amenorrhoea (=0), light (1–10), normal (10–100) and heavy menstrual bleeding (> 100).
3. Cycle regularity (ordinal four-point scale).
4. Duration of period (ordinal three-option scale).
5. Visual analogue scales (VAS; 0 = best outcome, 10 = worst outcome) for pelvic pain during periods, intercourse and at other times.
6. Uterine fibroid symptom and quality of life (UFS-QoL) instrument,⁴³ which contains a health-related quality of life domain and a symptom domain. Scores range from 0 at worst to 100 at best. This instrument was only given to women diagnosed with fibroids.
7. Sexual activity questionnaire (SAQ),⁴⁴ which is a valid, reliable and acceptable measure for describing the sexual functioning of women in terms of pleasure, discomfort and habit. Scores for pleasure range from 0 (lowest level) to 18 (highest level), scores for discomfort range from 0 (greatest) to 6 (none), and scores for habit range from 0 (worst outcome) to 3 (best outcome).
8. Generic quality of life (EQ-5D-5L).⁴⁵ The descriptive system has five dimensions with five levels, that creates the EQ-5D index score (−0.59 = worst outcome, 1.0 = best outcome). The EQ-5D health thermometer is a visual analogue scale for self-rated judgement of current health status (0 = worst imaginable health, 100 = best imaginable health).
9. Satisfaction with treatment outcome measured on a five-point Likert scale.
10. Participant rating of effect of treatment on HMB over 12 months measured on a four-point Likert scale.
11. Whether participant was willing to recommend the treatment to a friend (yes/no).
12. Surgical intervention (hysterectomy, endometrial ablation and other gynaecological surgery).
13. Adherence to trial treatments and reasons for changing treatment, as reported by the participant.
14. Serious adverse events (SAEs) and reactions, further defined in the corresponding section below.
15. Clinical measurements via pelvic ultrasound: uterine volume, evidence of adenomyosis, presence of fibroids, largest fibroid volume, endometrial thickness, endometrial appearance (regular/irregular), evidence of ovarian cysts.
16. Clinical measurement via endometrial biopsy: primary diagnosis (normal/benign/hyperplasia/malignant) and further sub-diagnosis if non-normal including presence or absence of PAEC or other non-physiological changes (UPA group only).
17. LFTs, including ALT and AST, and other tests used in local protocols, from 20 March 2018 (UPA group only).
18. Haemoglobin.
19. Serum estradiol.

Study assessments

Assessment times were at approximately 3, 6 and 12 months post randomisation and at other time points (see Table 1). Additional assessments related to the mechanism of action study are detailed in Chapter 4. Owing to the nature of the UPA treatment, with three courses taken over a 48-week period and the restrictions imposed by the USMs, women in the UPA had a specific assessment schedule for collection of outcomes, as follows:

1. The participant-completed questionnaires (MMAS, UFS-QoL, SAQ, EQ-5D-5L) were to be completed in the final week of each on-treatment cycle.
2. The menstrual blood loss diary while on treatment was to be completed over the final four weeks of each treatment cycle in the UPA group. The UPA group were also asked to complete the diary during the first four weeks off treatment before the start of the next treatment cycle.
3. The post treatment endometrial biopsy was to be completed after four weeks off treatment, which would be at around 48 weeks after UPA was commenced. If UPA treatment finished early, a biopsy was taken four weeks after cessation, or as soon as was feasible if access to gynaecology clinics was affected by the COVID-19 pandemic.
4. If PAEC was observed in the post treatment biopsy specimen, a repeat endometrial biopsy was taken around 13 weeks (15 months) after the completion of treatment, and then again around 26 weeks (18 months) post treatment if PAEC persisted. If UPA was ended prematurely, if PAEC was observed in the post treatment biopsy, repeat biopsies were performed 3 and 6 months thereafter if necessary, or when access to clinics was feasible.
5. With effect from 20 March 2018, blood samples were collected from women for LFTs each month while on UPA treatment and two to four weeks after each UPA course, including after the third course. After the second USM, when all participants were told to stop UPA treatment, blood samples were delayed or not performed due to restrictions on gynaecology services.
6. LFTs were also indicated for women who presented with signs or symptoms suggestive of liver injury (such as nausea, vomiting, malaise, right hypochondrial pain, anorexia, asthenia, jaundice) and UPA treatment was stopped. Such participants were closely monitored and referred for specialist hepatology evaluation as clinically indicated. During the period after the second USM, participants who had had to cease UPA treatment were telephoned to determine if they were exhibiting symptoms suggestive of liver injury.

Assessments were scheduled for an equivalent time, at 3, 6 and 12 months post randomisation in the LNG-IUS group.

The schedule for outcome assessment is summarised in Table 1.

The BCTU collected participant reported outcomes (e.g. MMAS, menstrual blood loss diary) postally at 3 and 6 months and then transcribed them to a secure web-based database. Twelve-month outcomes required clinic assessment, including an ultrasound assessment, blood sample (for haemoglobin and serum estradiol) and for those allocated to UPA, an endometrial biopsy. This final assessment was delayed for those participants whose 12-month clinical assessment fell during the period when non-COVID-19 research or face-to-face clinical assessment was suspended.

Adverse events and serious adverse events

The adverse event profile for LNG-IUS is well defined, as the system has been licenced for over a decade, and hence the collection of expected adverse events is not required. For example, elective admission for LNG-IUS insertion or elective admission for hysterectomy would not need to be classified as a serious adverse event. The focus for safety reporting of UPA was on changes to the endometrium, with repeat endometrial biopsies scheduled for those women where PAEC were observed (see Study assessments).

Reasons for change or cessation of treatment were collected, including decisions driven by perceived side effects of treatment such as weight gain. Postal questionnaires collected information on admission to hospital, gynaecological investigations or treatments (e.g. hysteroscopy or endometrial ablation), relevant diagnoses (e.g. endometrial thickening), or used any new medications. Pregnancy was considered an adverse event if the woman was compliant with either trial treatment, but not if she intentionally stopped treatment.

UPA treatment was stopped if any woman developed liver enzyme levels (ALT or AST) more than three times the upper limit of normal and participants were closely monitored and referred for specialist hepatology evaluation if clinically indicated.

A serious adverse event was defined as any event or reaction that was life-threatening, required emergency hospitalisation, resulted in death or persistent or significant disability or if the woman became pregnant, when a congenital anomaly was diagnosed in her baby. All diagnoses of endometrial cancer, ovarian cancer, cervical cancer, breast cancer or ductal carcinoma were defined as serious adverse events. The local investigator had to assign seriousness, severity, causality and expectedness (if deemed related) to the serious adverse event before reporting. Those categorised by the local investigator as both suspected to be related to the trial drugs and unexpected were subject to expedited reporting.

Impact of urgent safety measures on trial populations

The enforced non-compliance because of the temporary (and subsequently permanent) withdrawal of UPA had substantial implications for the sample size and validity of the data reported by participants. It was therefore necessary to redefine the analysis populations, considering the restrictions that prevented women taking their courses of UPA might influence their responses as well as any other new biases that may be apparent in either group due to, for example, knowledge of the safety concerns around UPA. The general principle was that a revised primary analysis population would be agreed – free from as much confounding and bias as possible – and supplemented with a number of additional planned sensitivity analyses, which may be more exploratory in nature. All changes were documented in a revised statistical analysis plan. Revisions were reviewed and approved by a statistician independent to the trial to ensure that they were necessary and appropriate from a methodological perspective and not because of observing accumulating interim data.

The original planned primary analysis population comprised all participants, regardless of adherence to treatment, in keeping with the principles of intention to treat. Following the urgent safety measures, the primary analysis population (population A; Figure 4) would now comprise participants with questionnaire responses received prior to the first USM (12 February 2018), along with questionnaire responses from participants recruited following the study restart (phase 2; from 18 October 2018) provided that these were returned before the second USM (17 March 2020). There was considered to be no additional risk of bias with these participants as a result of the urgent safety measures. However, there remained some concerns about whether those participants randomised in phase 2 were completely comparable with the earlier population, given that they were informed of the risk of liver damage by UPA and had to return to hospital for monthly LFTs. We planned to investigate any potential impact of this through examination of interaction of treatment effect by recruitment period. The same approach would be used for all assessment times (3, 6 and 12 months), provided that a previously agreed threshold for late returns was not breached. Participant responses would be still included, regardless of adherence to treatment, in keeping with principles of intention to treat, to limit any potential for confounding biases.

FIGURE 4

Revised analysis populations.¹ UPA participants could finish current course of treatment (up to 12 weeks); LNG-IUS participants unaffected² UPA participants were advised to cease treatment immediately; LNG-IUS participants unaffected³ Defined as taking (more...)

The first exploratory sensitivity analysis (population B1; Figure 4) would extend the primary analysis population to include participants who chose to complete their current course of UPA (for up to 12 weeks of treatment) following the first USM. To allow for courses to be completed and questionnaires to be returned, a date 15 weeks after the first USM was considered an appropriate cut-off date for responses to be included. The equivalent group of participants in the LNG-IUS group completing assessments in this time window would also be incorporated. Only adherent participants in both groups were to be included in this population to ensure consistency with this ‘per protocol’ approach; those who had ceased taking UPA (or took it sporadically) for any reason or had LNG-IUS removed were also excluded. A second sensitivity analysis (population B2; Figure 4) would extend the first exploratory sensitivity analysis further by including responses from adherent LNG-IUS participants, who were unaffected by the USM in terms of being instructed to stop treatment but accepting there may be some further confounding biases accrued.

Any responses from participants randomised to UPA but received after enforced cessation of treatment would contribute to an observational cohort (population C; Figure 4), giving some valuable indication of the impact of stopping UPA. All sensitivity analyses would be limited to the condition-specific quality of life score as measured by the MMAS and some the other most important secondary outcomes (see Statistical principles) to reduce the possibility of overinterpretation of data.

Sample size, including impact of urgent safety measures

The trial was designed to be able to detect a clinically useful difference in MMAS score between the two groups at 12 months with high power. The ECLIPSE trial,¹⁰ which evaluated the effectiveness of LNG-IUS against standard treatment for HMB using MMAS as the primary outcome, demonstrated a difference of 13 points between the groups with a SD of 24 points. This size of difference is equivalent to approximately 0.5 SDs, which is often considered a medium-sized effect and likely to be at least minimally important.⁴⁶ To detect a difference of 0.5 SDs with 90% power (p = 0.05) required 86 women in each group (172 in total). To allow for a 20% loss to follow-up or pregnancy, the sample size was inflated to 220 women.

Prior to the first halt of recruitment on 12 February 2018, the trial had recruited 198 participants. At this stage, it was anticipated that recruitment would restart and therefore plans were made for recruiting new participants with a revised sample size. The aim was to recruit to the original target but in such manner that enough participants would be unaffected by enforced non-compliance or knowledge of the USM during the follow-up period (i.e. to gain 172 quality of life MMAS responses at 12 months in population A; Figure 4). This would mean an inflation from 220 participants to a target of 302 participants. This figure was derived from the number of participants who had completed 12-month assessment prior to the first USM (89), taking into account the total number who had been randomised up to this point (198). An additional 104 participants, gaining data on 83, would be required to reach 172 responses. The trial ultimately recruited 236 participants before recruitment was terminated after the second USM.

Statistical principles

A comprehensive statistical analysis plan was drawn up prior to any analysis and provided to the independent data monitoring and trial steering committees for review. The baseline characteristics of the trial population were tabulated for all randomised participants as well as for those participants in the primary analysis population A (those providing the primary outcome at 12 months were used for this purpose). Categorical data were summarised with frequencies and percentages. Normally distributed continuous variables were summarised with means with SDs, otherwise medians with interquartile ranges were presented.

The general analytical approach for all outcomes employed suitable regression models, dependent on the underlying data type and incorporating repeated responses at all assessment times where possible. Estimates were adjusted for the minimisation parameters and baseline response (where available). If repeated responses were made, models included variables for participant and assessment time (categorical) and to allow for varying treatment effect over time, a time by treatment interaction parameter. All estimates of differences between groups (mean differences or odds ratios) were presented with two-sided confidence intervals (CIs). Analysis was conducted for all outcomes for population A, along with the following for populations B1 and B2: quality of life MMAS scores; amenorrhoea and heavy menstrual bleeding; surgical interventions; the clinical measurements from pelvic ultrasound; and haemoglobin and estradiol. Plots of MMAS and PBAC bleeding score responses over time were presented for the exploratory population C (UPA group only).

For the primary outcome, the initial analytic approach incorporated a linear regression model estimating mean differences in quality of life MMAS responses between the two groups. Upon inspection of pooled data as part of data validation processes, a high degree of skew in the responses was thought likely. A reserve method for analysis was specified a priori should the regression residuals indicate skewed data, and this was ultimately the analysis performed. A generalised estimating equation (GEE) model⁴⁷ was used with a cumulative logit link for ordered MMAS scores, categorised as ≤ 50, 51 75, 76–99, = 100. These categories have been used previously in similar trials of HMB with MMAS as the primary outcome.¹⁷ The GEE model took into account correlated longitudinal data; a general unstructured covariance matrix was assumed. Cumulative odds ratios and 95% CIs for the treatment group parameter were produced and the statistical significance (p-value) of the treatment group variable determined by an associated chi-squared test. Questionnaire responses were considered valid provided that they had been completed before the subsequent time point (out to 18 months post randomisation for the 12-month assessment); if responses were late they were not included in the analysis but sensitivity analysis was performed with their inclusion.

MMAS scores at three’ and six months’ follow-up were analysed as part of the aforementioned model. Bleeding scores from the PBAC were converted into the following categories: (1) the proportion with amenorrhoea (= 0) and ‘any bleeding’ (score > 0) as well as (2) non-heavy (score ≤ 100) and heavy (score > 100) bleeding. These outcomes, along with cycle regularity, were analysed in a similar manner to the dichotomised MMAS scores. Duration of period was another ordinal response and was analysed in a similar manner to the MMAS categorised scores. Data from patient-reported outcomes (UFS-QoL, EQ-5D, VAS and SAQ) returning continuous scores were analysed using linear regression models for repeated measures to estimated mean differences between the two groups at each time point.

Continuous outcomes assessed by pelvic ultrasound or blood samples, such as uterine volume, largest fibroid volume, haemoglobin and serum estradiol, were analysed using linear regression models. Satisfaction and participant rating of treatment were analysed using ordinal logistic regression. Binary clinical observations from the pelvic ultrasound and willing to recommend to a friend was analysed using logistic regression. There were too few events to analyse the number of surgical interventions formally, so only summary statistics are presented. The number of serious adverse events was analysed using a chi-squared test. Observations from the endometrial biopsies taken after the end of UPA treatment and LFTs taken during UPA courses were tabulated.

Interim analyses

Interim assessment of effectiveness and safety outcomes were performed on behalf of the DMC (see Acknowledgments) on an annual basis throughout the study. These analyses were performed with the use of the Haybittle–Peto approach,⁴⁸ therefore no adjustment was made in the final p-values to determine significance.