Surgical interventions for uterine prolapse and for vault prolapse: the two VUE RCTs

Article history

The research reported in this issue of the journal was funded by the HTA programme as project number 11/129/183. The contractual start date was in November 2012. The draft report began editorial review in February 2018 and was accepted for publication in August 2018. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The HTA editors and publisher have tried to ensure the accuracy of the authors’ report and would like to thank the reviewers for their constructive comments on the draft document. However, they do not accept liability for damages or losses arising from material published in this report.

Declared competing interests of authors

Robert Freeman reports speaker fees [from Bard Medical (Covington, GA, USA), Astellas Pharma Inc. (Tokyo, Japan) and Pfizer Inc. (New York City, NY, USA)] and grants from the National Institute for Health Research (NIHR) Health Technology Assessment (HTA) programme. John Norrie reports membership of the following NIHR boards: the Cardiopulmonary Resuscitation Decision-making Committee, HTA Commissioning Board, HTA Commissioning Sub-Board (Expression of Interest), HTA Funding Boards Policy Group, HTA General Board and HTA Post-board Funding Teleconference; the NIHR Clinical Trials Unit Standing Advisory Committee; the NIHR HTA and Efficacy and Mechanism Evaluation Editorial Board; and the Pre-exposure Prophylaxis Impact Review Panel.

Copyright statement

© Queen’s Printer and Controller of HMSO 2020. This work was produced by Hemming et al. under the terms of a commissioning contract issued by the Secretary of State for Health and Social Care. This issue may be freely reproduced for the purposes of private research and study and extracts (or indeed, the full report) may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NIHR Journals Library, National Institute for Health Research, Evaluation, Trials and Studies Coordinating Centre, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

2020 Queen’s Printer and Controller of HMSO

Relationship to PROlapse Surgery: Pragmatic Evaluation and randomised Controlled Trials in women with anterior or posterior pelvic organ prolapse

The PROlapse Surgery: Pragmatic Evaluation and randomised Controlled Trials (PROSPECT) in women with anterior or posterior pelvic organ prolapse (POP)1 was a large HTA-funded UK-based RCT of anterior or posterior prolapse surgery with or without the use of mesh (HTA reference number 07/60/18) that was also undertaken by the research team (with long-term follow-up under way). The methods developed in PROSPECT1 were used to inform the design, conduct, analysis and reporting of the VUE trial.

Description of the underlying health problem

Pelvic organ prolapse is the descent of pelvic organs from their normal anatomical position.

Prolapse can occur in three compartments: the anterior (cystocele) vaginal wall, the posterior (rectocele) vaginal wall and the uterus or the apex (vault in those who have had a hysterectomy). Women can have prolapsed in one or more compartments at the same time.

There is little epidemiological research into this condition because it has a variety of presentations and they do not all cause symptoms, particularly in the early stages. 2 Commonly reported symptoms include a feeling of dragging or heaviness in the vagina, an uncomfortable bulge distending the introitus, urinary symptoms (e.g. incontinence and voiding difficulty), bowel symptoms (e.g. incomplete emptying) and sexual dysfunction.

The prevalence of prolapse varies depending on age, race and symptoms, with anterior prolapse being the more prevalent. 3–8

Standard management

Women with prolapses may be managed conservatively with pelvic floor muscle training (PFMT), pessaries or surgery. Management of associated conditions, such as lower urinary tract symptoms (urinary incontinence or overactive bladder syndrome), bowel problems (constipation or faecal incontinence), sexual dysfunction and local vaginal atrophy caused by oestrogen deficiency if post-menopausal, may also be required.

Conservative management for women with prolapses

There is not enough evidence to inform the use of mechanical devices (pessaries or rings), although these are often used as a first-line treatment or can be used for women who are unfit for surgery or wish to avoid surgery. 9–12

Mechanical devices can be very efficacious, but further research is required to identify the best type of device, the potential long-term effects and the use of supplementary treatments, such as oestrogen. 9,10

Conservative physical treatments, such as PFMT, are also often recommended as first-line management. The most recent reviews found evidence supporting the use of PFMT to prevent and reduce prolapse symptoms and severity, as well as improvement in urinary and bowel symptoms. 11,13

Vaginal oestrogen can also be used to reduce atrophic symptoms for post-menopausal women, or for before and after surgery. However, the evidence supporting vaginal oestrogen use is limited and inconclusive. 14,15

Surgical management for women with prolapses

Surgery for POP is common, with recent estimates indicating that women have a lifetime risk of between 6% and 20% of undergoing surgery for prolapse. 3,5,16,17

In England and Wales in 2016–17, 24,784 women were admitted to hospital with a main diagnosis of female genital prolapse, and 29,729 operations were performed (some women had more than one type of prolapse operation). 18 The majority of operations (83%) were in women having an anterior repair (n = 11,224), posterior repair (n = 6855) or both operations (n = 6502). Around 21% of women (n = 5148) had an apical procedure. It was identified, from PROSPECT1 and NHS digital data,18 that around one-third of women with prolapse have a hysterectomy for uterine prolapse.

The demand for surgical prolapse repair may increase given that projections have predicted a rise of around 1.4 million women in the UK aged 50–85 years (most likely to have surgery) between 2017 and 2027. 19

Little is known about the long-term effectiveness and safety of different types of operations for uterine or vault prolapse, although the National Institute for Health and Care Excellence (NICE) has provided some guidance. 11,20,21

Surgical management for women with uterine or vault prolapses

Women may present with an isolated uterine or vault (apical) prolapse or in combination with other compartments. Numerous surgical techniques now exist for apical prolapse, but none has been properly evaluated in terms of adequately powered multicentre RCTs. 22–24

Uterine prolapse

If there is uterine descent (however small the uterus), removal of the uterus (hysterectomy) is standard practice at the time of prolapse repair in most parts of the world. Uterine descent is the most common indication for hysterectomy;25 however, hysterectomy for uterine descent is not an evidence-based practice.

At the onset of the VUE trial, surgery for uterine prolapse was broadly divided into two approaches: uterine removal (hysterectomy) or uterine preservation.

Vault prolapses

After a hysterectomy, vault prolapse occurs in women when the vaginal vault descends into or out of the vagina despite vault support procedures carried out at the time of hysterectomy.

At the onset of the VUE trial, a variety of techniques to suspend or reposition the vault were available. The techniques were broadly divided into vaginal or abdominal approaches and included the following:

• Vaginal approaches:

  • vaginal sacrospinous fixation or sacrospinous colpopexy (vault attachment to the sacrospinous ligament, either bilaterally or on one side only; this is traditionally performed using sutures, but mesh could also be used)

  • transvaginal mesh kits that suspended the vault.

• Abdominal approaches:

  • abdominal sacrocolpopexy (attachment of the vault to the sacrum, with a mesh bridge); this could be an open, laparoscopic or robotic laparoscopic procedure.

Current recommendations from the National Institute for Health Care and Excellence

An interventional procedures review26 and a Cochrane review22 have been conducted on the use of mesh in uterine or vault prolapse. NICE considered the evidence from independent reviews, RCTs and non-randomised studies, and in 2017 developed updated guidance on:

• sacrocolpopexy with hysterectomy using mesh for uterine prolapse repair – see Interventional Procedures Guidance (IPG) 57723

• uterine suspension using mesh (including sacrohysteropexy) to repair uterine prolapse – see IPG58427

• infracoccygeal sacropexy using mesh to repair uterine prolapse – see IPG58228

• infracoccygeal sacropexy using mesh to repair vaginal vault prolapse – see IPG58129

• sacrocolpopexy using mesh to repair vaginal vault prolapse– see IPG583. 24

Of these procedures, only the standard operation of sacrocolpopexy using mesh for vault prolapse (IPG583) and uterine suspension using mesh (including sacrohysteropexy) to repair uterine prolapse (IPG584) are now considered to have enough evidence for safety and efficacy, such that they can be used under standard arrangements. 24 The uncertainty regarding the other procedures now means they require strict clinical governance arrangements. A clinical decision-making tool is currently being developed. 27

The decision to test uterine and vault prolapse surgical procedures

The most recent Cochrane review included 30 RCTs comparing surgical procedures for women with an apical vaginal prolapse. 22 The reviewers did not, however, separate the results for uterine and vault prolapse. The reviewers, and others, concluded that there is still insufficient information about any of the surgical options to guide management of uterine or vault prolapse and identified a need for adequately powered RCTs. 22,26,30 Thus, the evidence base for treating either of these groups of women is clearly inadequate in terms of patient-reported outcomes [e.g. subjective prolapse symptoms, effect on quality of life (QoL)], cost-effectiveness and safety.

Questions addressed by the VUE trial

Study design

The methods developed in PROSPECT1 were used to inform the design, conduct, analysis and reporting of the VUE trial.

The VUE trial comprised two parallel RCTs (i.e. the Uterine trial and the Vault trial) to determine the clinical effectiveness (including safety) and cost-effectiveness of surgical treatment, primarily in terms of improvement in prolapse symptoms in women having either uterine or vault POP surgery.

Participating centres were asked to randomise participants as close to surgery as possible to minimise participant dropout; therefore, there was a delay from date of consent to date of randomisation. For the majority of sites, randomisation occurred around 2 weeks prior to surgery. Participant questionnaires were issued at 6 months after surgery and at 12 months after randomisation. This meant that if a woman did not receive surgery, no 6-month follow-up questionnaires were issued, but 12-month questionnaires were issued to all women. Women who received surgery were also reviewed in clinics 12 months after surgery (Figure 1).

FIGURE 1.

Flow diagram, with anticipated sample sizes. POP-Q, Pelvic Organ Prolapse Quantification System.

Further details of the study design, methodology and management [eligibility, consent, comparison of health technologies, treatment allocation, data collection and processing, sample size calculation, avoidance of bias (and blinding), serious adverse event (SAE) reporting] have been described previously. 31 All trial case report forms (CRFs) and participant completed questionnaires are included in the report methodology (www.journalslibrary.nihr.ac.uk/programmes/hta/11129183/#/; accessed February 2019).

Study outcome measures

The VUE trial used the International Urogynaecological Association (IUGA)/International Continence Society’s POP outcome recommendations and definitions. 32

Three primary outcome measures were identified (www.journalslibrary.nihr.ac.uk/programmes/hta/11129183/#/; accessed February 2019):

1. Women’s symptoms of prolapse [as measured using the patient-reported Pelvic Organ Prolapse Symptom Score (POP-SS)]33 at 12 months after randomisation. This scale was derived from seven questions judged to be most directly related to prolapse symptoms and has been shown to reflect the range and intensity of symptoms experienced by women, as well as being responsive to change over time. 34 Scores were determined for each of the seven items (ranging from 0 for ‘never’ to 4 for ‘all of the time’), with an overall POP-SS out of 28. Women who only partially completed the seven-item response schedule (defined as having completed six out of seven items in the scale) were assumed to have no symptoms when no response had been given to that individual item. Participants with more than one missing item were considered to have a missing overall score. Women were considered to be symptomatic if their overall score was > 0.

2. Quality of life (condition specific, measured as the woman’s rating of the overall effect of prolapse symptoms on everyday life on a 0–10 visual analogue scale in which 10 is the worst).

3. The primary economic outcome measure of cost-effectiveness was the incremental cost per quality-adjusted life-year (QALY), based on the EuroQol-5 Dimensions, three-level version (EQ-5D-3L). 35

Secondary outcomes

Other outcome measures included objective prolapse measurement; urinary, bowel and sexual symptoms [using the International Consultation on Incontinence (ICI) suite of validated questionnaires];36 intra-operative and postoperative complications, including the need for additional surgery (repeat surgery for prolapse recurrence, further prolapse surgery, or incontinence, and surgery required for AEs); cost; and cost-effectiveness (www.journalslibrary.nihr.ac.uk/programmes/hta/11129183/#/; accessed February 2019).

Objective prolapse measurement

Objective prolapse staging was carried out using the Pelvic Organ Prolapse Quantification (POP-Q) system37 and classified from stage 0 to 4 for anterior, posterior and apical at baseline and 12 months after surgery. The leading edge of the most descended compartment was used to define the overall stage.

The POP-Q system measures the maximum descent of each of the three prolapse compartments (anterior, posterior and apical) relative to the hymen; measurements inside the vagina are negative, whereas those outside the vagina are positive. An algorithm was employed to ensure that POP-Q staging was correctly calculated from the component measurements of the POP-Q (i.e. Aa, Ba, C, D, Bp, Ap and total vaginal length) in which common recording errors (e.g. Ba less than Aa) were corrected or queried. If data were discrepant, they were corrected by consultation with the local hospital records to obtain additional data, to achieve as complete a set of prolapse staging as possible, separately in each compartment. If the POP-Q data were missing, the surgeon’s qualitative record of stage was accepted for both overall and individual compartments (i.e. surgeons could specify the individual stage without giving the POP-Q measurements).

Usually, using the classic Bump criteria for the POP-Q system, any measurement from –1 cm (inside the hymen) to + 1 cm outside the hymen counts as stage 2. 37 Stage 2 was further subdivided into prolapse at the hymen or within (–1 cm to 0 cm, stage 2a or less) compared with prolapse at > 0 cm (stage 2b). 38,39 Thus, women were classified as having objective prolapse if the leading edge was at any point outside the hymen (measured at > 0 cm, stage 2b or more).

Urinary, bowel and sexual symptoms

Symptoms related to other aspects of pelvic floor dysfunction were measured using the ICI suite of validated questionnaires. 36

Urinary incontinence was assessed using the International Consultation on Incontinence Questionnaire (ICIQ) – urinary incontinence Short Form questionnaire (ICIQ-UI SF). Other urinary symptoms were recorded by the ICIQ – female lower urinary tract symptoms (ICIQ-FLUTS) instrument. The latter provides subscales for filling, voiding and incontinence symptoms.

The ICIQ bowel symptom questionnaire was not finalised when the VUE trial began. As in PROSPECT,1 draft questions were drafted to produce a short summary of relevant bowel symptoms in line with the ROME criteria40 to define constipation (see Appendix 1, Table 39).

Vaginal and sexual symptoms were assessed using the ICIQ – vaginal symptoms (ICIQ-VS) questionnaire. 36 The ICIQ-VS provides a brief and robust measure to assess the impact of vaginal symptoms and associated sexual matters on QoL and the outcome of treatment. The questionnaire provides subscales for vaginal symptoms, sexual matters and the overall impact of vaginal symptoms on QoL. Women who were sexually inactive were asked whether this related to their vaginal or prolapse symptoms or for another reason (including no partner).

Choice of validated outcome measures

Outcome measures were chosen to reflect the current international standards of reporting to ensure that the findings would be relevant to patients, clinicians and policy-makers. 32 The outcomes were measured at baseline to provide values for later statistical adjustments. The primary measure of prolapse symptoms was the woman’s subjective report using the POP-SS, developed and validated in a variety of populations for both research and clinical practice. 33

Temporary trial suspension

In June 2014, the Scottish Government requested that all Scottish NHS Health Boards considered suspending the use of transvaginal mesh implants in the surgical treatment of stress urinary incontinence and POP to enable an independent review to consider the ongoing debate on complication rates and AEs from the use of transvaginal mesh implants. The VUE trial was temporarily suspended (in Scottish centres only) for a period of 3 weeks during this time to ensure compliance and sponsor, Trial Steering Committee (TSC) and Data Monitoring Committee (DMC) satisfaction with all trial processes across Scottish sites. The debate on the use of transvaginal synthetic mesh continues. 41,42

Blinding

Baseline data were reported by women before randomisation using self-completion questionnaires. Outcome assessors were blinded to randomisation; participants were unblinded if they requested the information. Surgeons were not blinded to the allocation procedure.

Sample size

Important changes to the methods after trial commencement

Statistical analysis

The predefined statistical analyses are included in the report methodology (www.journalslibrary.nihr.ac.uk/programmes/hta/11129183/#/; accessed February 2019). These methods apply to both the Uterine and the Vault trials, but each trial is reported and analysed separately.

A single principal analysis was undertaken 12 months after the last woman was randomised. All main analyses were based on an intention-to-treat principle in which women with observed outcome data were analysed according to their randomised allocation. All outcomes in both trials were described with the appropriate descriptive statistics when relevant: mean and SD for continuous and count outcomes, or medians and interquartile range if required for skewed data, numbers and percentages for dichotomous and categorical outcomes.

The analysis of the primary outcome POP-SS estimated the mean difference (MD; and 95% CI) between intervention and control groups at 12 months after randomisation using a linear mixed model with surgeon fitted as a random effect that adjusted for the minimisation covariates and the baseline score. The full model is available in Appendix 1, Table 40. A similar analysis was used to analyse the primary outcome at 6 months after surgery.

All secondary outcomes were analysed in a similar manner but using the appropriate generalised linear model (GLM) (logistic regression for dichotomous data, such as subjective prolapse failure) or time to event methods (Cox regression for time to further surgery). If possible, all models were adjusted for minimisation variables and baseline values.

Health economic evaluation

A within-trial economic evaluation assessing the costs, QoL and cost-effectiveness of the interventions according to a NHS and patient perspective was undertaken for both the Uterine and the Vault trials. All analyses were completed following the intention-to-treat principle. Trial data were used to populate a Markov decision-analytic model, developed to extrapolate the 12-month data from the Uterine trial over a longer time horizon (see Chapter 9). Methods are applicable to both the Uterine (see Chapter 5) and the Vault (see Chapter 8) trials, that is, within-trial economic evaluations unless otherwise stated.

Health services perspective costs

The NHS resource-use data were collected using CRFs completed by the recruiting officers/research nurses or gynaecologists and patient-completed questionnaires administered at baseline, at 6 months after surgery and at 12 months after randomisation (www.journalslibrary.nihr.ac.uk/programmes/hta/11129183/#/; accessed February 2019). The following categories of resource usage and costs were collected from a NHS perspective:

• intervention delivery costs [based on a component costing approach for the base case and using Healthcare Resource Group (HRG) tariffs as a sensitivity analysis]

• inpatient costs (cost of re-admissions for repeat prolapse procedures, incontinence procedures and treatment of AEs)

• costs of consultations with health-care professionals [outpatient consultations and primary care consultations, such as general practitioners (GPs), nurses, physiotherapists, etc.]

• medications and treatments relating to prolapse symptoms (e.g. oestrogens, pessaries, antibiotics).

Intervention costs

Two approaches to intervention costing were considered. The base-case analysis used a component costing approach. Operative details were recorded at the time of surgery using CRFs completed at each site by the recruiting officer/research nurse or gynaecologist (www.journalslibrary.nihr.ac.uk/programmes/hta/11129183/#/; accessed February 2019). Resource use data collected included patient time in theatre, grade of operating gynaecologist, grade of anaesthetist, other nursing staff, type of anaesthetic used and the use of prophylactic antibiotics. Details of concomitant surgery and catheterisation were also collected. Intervention resources included the type of mesh for the apical procedures and for the concomitant procedures. Detailed information regarding initial hospitalisation length of stay was available and was used to assign appropriate hospitalisation costs to each group, based on per-night costs for a gynaecology ward. The component costing approach focused on the costs that were predicted to differ across the randomised groups. Microcosting information was not collected on items of resource use that were not expected to differ across groups.

Secondary care resource use over follow-up

Inpatient resource use included further prolapse surgery, further urinary incontinence surgery, further hospitalisations for complications and any other hospital resource use deemed relevant to the VUE trial. Data on occurrence of a hospitalisation were primarily collected from participant-administered questionnaires at 6 months after surgery and at 12 months after randomisation, and supplemented by asking women about hospitalisations at their final clinic visit (12 months after surgery). The joint approach ensured the greatest possible level of detail for analysis. All reported hospitalisations were then further investigated in a post-coding exercise to (1) determine that they were related to the VUE trial, (2) obtain further details on the category of procedure (e.g. further prolapse procedures: anterior, posterior, apical, combination) and (3) ensure that procedures were not double counted. The post-coding exercise involved investigation of site notes to verify the participant-reported data. The number of attendances for outpatient visits at gynaecology clinics were collected from the patient-administered questionnaires.

Other consultations over follow-up

Participants reported contacts with primary care health professionals for prolapse-related symptoms in the 6- and 12-month questionnaires (after surgery and after randomisation, respectively). This included visits to the GP, practice nurse, district nurse, physiotherapist (at a GP practice), physiotherapist (at a hospital clinic) and any other related consultations with health professionals. When collating the data on primary care resource usage, the following assumptions were made to ensure the best use of available data:

• when women answered ‘no’ to seeing a health professional, resource use was assumed as zero visits

• when women returned a questionnaire and left all items of resource use blank, data were assumed to be missing

• when women returned a questionnaire and answered one or more categories as ‘yes’, leaving the remaining blank, it was assumed that unticked boxes related to zero resource use.

Medications and treatments

Data on treatments and medications prescribed to women (i.e. oestrogens, bladder medications, catheters, pessaries and antibiotics) for prolapse-related symptoms were collected using participant questionnaires.

Total costs to the NHS

Resource use data were multiplied by the relevant national average unit cost. Health services costs were summed across categories to generate a total cost to the NHS of the intervention and the follow-up care for each trial participant. Unit cost sources were the British National Formulary (BNF)43 and the NHS Business Services Authority’s online drug tariff catalogue for medication resource use, Information Services Division (ISD) Scotland was used for intervention (ward-specific) length of stay,44 NHS Reference Costs 2015–1645 was used to provide information for secondary care procedures and visits and the Personal and Social Services Research Unit (PSSRU) provided unit costs of health and social care for other consultations with health professionals. 46 All costs were reported in 2016–17 GBP. No discounting was required because of the single-year time horizon. Detailed information on unit costs applied to each resource-use item, including any assumptions that were made are provided in Table 1.

Participant perspective costs

This wider cost perspective identified the effect of any shifts in the balance of care between the NHS and the patients and their families. Participant perspective costs included out-of-pocket expenditure on over-the-counter medications/products and private care to treat prolapse-related symptoms. The opportunity cost to women (and companions) of attendance at primary and secondary care consultations was also included, as was the cost of time off work (for those in paid employment).

To avoid additional patient burden through completion of lengthy questionnaires, it was assumed that the unit cost of time and travel to health-care professionals was similar to PROSPECT. 1 PROSPECT data for all procedures (primary, secondary and cohort) were used to calculate unit costs of time and travel for inpatient, outpatient and primary care visits (see Table 1). These unit costs were then multiplied by the number of appropriate visits observed in the VUE trial to obtain a cost of time and travel tailored to the VUE trial participants. For time off work, the appropriate national average price of 1 hour of work is used (see Table 1). For the remaining aspects of personal incurred costs (i.e. private medical costs, over-the-counter medications, other expenses), data are sourced directly from the participant questionnaires and multiplied by the relevant participant-reported expense.

A wider cost perspective was obtained by summing together the total NHS and participant perspective costs and is considered as a sensitivity analysis.

Outcome measurement, quality of life and quality-adjusted life-years

The EQ-5D-3L generic QoL instrument was completed by participants at baseline, at 6 months after surgery and at 12 months after randomisation. 53 The EQ-5D-3L instrument divides health status into five dimensions: mobility, self-care, usual activities, pain/discomfort and anxiety/depression. Each of these dimensions have three levels, so 243 possible health states exist. EQ-5D-3L responses were presented in graphical format illustrating the percentage of respondents with any problems on each domain, split by randomised arms of the trial.

Responses to the EQ-5D-3L questionnaire were translated into utility values using the UK general population tariffs, based on the time trade-off method, generating utility on a scale of worst possible health state (–0.564) to best possible health state (1). 35 Respondents who died during the trial were assigned a utility of 0 at each time point following death.

The QALYs were then derived by multiplying the utility score at each time point by the length of time in that health state. An area under the curve approach was used, assuming a linear extrapolation between utility measurement time points.

Statistical analysis of cost and quality-adjusted life-year data

All components of costs were described with the appropriate descriptive statistics when relevant: mean and SD for continuous and count outcomes, and numbers and percentages for dichotomous and categorical outcomes (e.g. numbers reporting problems on EQ-5D-3L). All analyses were conducted using Stata version 14.1^® (StataCorp LP, College Station, TX, USA) software. Cost data were analysed using GLMs, with adjustment for minimisation covariates specific to each trial and baseline measure when appropriate (e.g. adjustment of QALYs and costs for baseline utility score). 54 All models include a cluster effect for surgeon. GLMs allow correction for the potential for skewed cost data (i.e. a small proportion of participants incurring very high costs) by allowing specification of an appropriate distributional family and link function to best fit the data. Different distributional families offer alternative specifications to reflect the relationship between the mean and variance of the estimates under consideration. 55,56 A modified Park test was used to select the most appropriate distributional family for each analysis. Then, a combination of tests was run on the preferred family to identify the most appropriate link function (Pearson correlation test, Pregibon link test and a modified Hosmer and Lemeshow test). The preferred link function was chosen as the one that performed best on the greatest number of the three tests, confirmed by examination of the respective p-values of the tests. The process follows that outlined by Glick et al. 55 and was implemented in Stata 14.1 using the user written ‘glmdiag’ command. The coefficient on treatment in the respective analysis models represents the incremental treatment effect on costs and/or QALYs. When different link functions are used, marginal effects are estimated on the cost and QALY scale, respectively and CIs calculated using the delta method. 55,56 The base-case cost data area analysed using a gamma family with a log-link and the base-case QALY data were analysed with a Poisson family and a power equal to 0.75 link function.

Incremental cost-effectiveness analysis

Missing data

Missing data can pose significant problems for data analysis, especially surrounding data reported using participant-administered questionnaires. A decision rule on imputation was taken as follows. If > 10% of complete-case costs or complete-case QALYs were missing, or if > 15% of complete cost and QALY pairs were missing, then imputation was considered for the base-case analysis. The analysis meets the prespecified criteria for imputation and so the base-case analysis was based on multiple imputation of missing data. 58 Data were imputed using the iterative chained equations approach within Stata 14.1’s ‘mi impute’ procedure. Missing EQ-5D-3L data were imputed using predictive mean matching (the mean of five nearest values) to account for multiple time points per respondent (i.e. at baseline, at 6 months after surgery and at 12 months after randomisation). Missing cost data were imputed at the category level (operation cost, hospitalisation cost, consultation cost and other treatment cost) using a multivariate regression approach. All imputation models were adjusted for minimisation variables specific to each trial and baseline EuroQol 5 Dimensions (EQ-5D)-derived utility score. Imputations were completed separately for the Uterine and Vault trials. Ten imputations were considered sufficient to generate stable and reliable estimates for analysis. Imputed data were then analysed using the appropriately specified GLMs described.

Assessment of uncertainty

Sensitivity analyses

A range of deterministic sensitivity analyses were conducted to assess the impact of important choices surrounding assumptions and analysis models on the cost-effectiveness findings. All sensitivity analyses were conducted using data sets with multiple imputation of missing cost and QALY data, and each analysis is subjected to the same assessment of sampling uncertainty with the production of scatterplots and CEACs for each sensitivity analysis undertaken. The following sensitivity analyses were explored.

Study recruitment

The trial outline and methodology for recruitment to the VUE trial have been described previously31 (see Chapter 2). Women attending gynaecology outpatient clinics who chose to have surgery for symptomatic uterine POP and women on the waiting list for uterine prolapse surgery were invited to participate in the Uterine trial. Women were asked if they were willing to be randomised to either a uterine preservation or a vaginal hysterectomy for their uterine prolapse. The centres and surgeons who participated in the VUE trial, the numbers recruited, and the rate of recruitment are detailed in Appendix 2 (Table 47 and Figure 19).

Non-recruited women

Of the 1544 women approached to participate in the Uterine trial, 979 did not enter the study because they declined (n = 774, 50.1%), were ineligible (n = 177, 11.5%) or not timely identified or seen (n = 28, 1.8%; see Appendix 2, Tables 48 and 49 for more details). Women’s preference (n = 431, 55.7%) for a particular surgery [most commonly the preference was for a vaginal hysterectomy (n = 324, 41.9%; see Appendix 2, Table 49)], was the most common reason for declining. Ineligible reasons included ‘a specific operation is necessary’ (n = 77, 43.5%), ‘unsuitable due to medical history’ (n = 28, 15.8%) and ‘not suitable for surgery’ (n = 13, 7.3%). Other reasons included not wanting to be randomised (n = 149, 19.3%), deciding against surgery (n = 219, 28.3%) and deciding to try a pessary (n = 34, 4.4%).

Recruited women: baseline characteristics

The baseline characteristics of the 563 women who agreed to participate in the Uterine trial and were truly eligible are described in Table 2 (see also Appendix 2, Table 50 and Figure 3).

Epidemiological characteristics

There was no difference between the randomised groups in respect of age, body mass index (BMI), parity or delivery mode history (with the majority of women having a normal vaginal delivery; see also Appendix 2, Table 50).

Previous treatment for prolapse

Overall, around one-third of women had undergone PFMT with the same proportion of women having used a vaginal pessary (ring or other type) before surgery. Fewer than 1 in 10 women had undergone previous surgery for prolapse (Table 2).

Prolapse symptoms at baseline

There were no differences in prolapse symptoms at baseline. Overall, women in the Uterine trial had been symptomatic for around 4 years, and had been bothered by their symptoms for just over 2 years (Table 3). The mean POP-SS before surgery was 13.6 out of a maximum score of 28 (Table 3) and the score ranged from 0 to 28. Using a POP-SS of > 0 to indicate presence of symptoms, around 94% of women had at least one symptom. The prolapse-related effect on QoL score (‘overall, how much do your prolapse symptoms interfere with your everyday life?’) ranged from 0 to 10 out of 10, with a mean value of 6.7 out of 10 (Table 3).

The most common individual prolapse symptom was ‘a feeling of something coming down from or in your vagina’ reported in 91.7% of women. This was reported ‘most or all of the time’ in 72.0% of women and was the most bothersome symptom in around half (49.7%) of the women.

Around one-third (33.0%) of women found that the prolapse caused hygiene problems, and almost one in five (17.8%) needed to relieve pressure or discomfort from the prolapse using their fingers (Table 3).

Generic quality of life

The mean QoL score and visual scales (EQ-5D-3L) were 0.752 and 75.1 points, respectively.

Preoperative objective prolapse measurements

The leading edge of the most descended compartment relative to the hymen was used for the overall POP-Q stage. The majority (94.1%) of women had an overall objective prolapse and 53.5% had a uterine prolapse beyond the hymenal ring [stages 2b or more (Table 4; see also Appendix 2, Table 51)]. Around half (49.4%) of the women had an overall stage 3 prolapse of any compartment (see Appendix 2, Table 51), with 43.2% of women specifically having a stage 2 uterine prolapse.

Planned concomitant surgery

Planned surgery was based on preoperative findings on clinical examination and women remained in the group to which they were allocated, irrespective of the actual procedure performed (Table 5). In order to take into account minimisation criteria, centres were asked to specify in advance which concomitant surgery was also thought to be necessary.

All women were expected to undergo a uterine prolapse procedure. Most women (85.5%) were expected to undergo a concomitant prolapse repair. Planned procedures were equally distributed between the two randomised groups.

Vaginal and sexual symptoms at baseline

Vaginal and sexual symptoms were measured using the ICI-validated instruments36 (www.journalslibrary.nihr.ac.uk/programmes/hta/11129183/#/; accessed February 2019). These symptoms were common and had effects on QoL. The majority of women were not sexually active (70.0%) and in around one in five of the women this was most often because of their prolapse symptoms. Among the women who were sexually active, or whose reason for no sex life was not ‘due to prolapse symptoms,’ a few (n = 5) had dyspareunia (pain with intercourse) at baseline (Table 6).

Urinary and bowel symptoms at baseline

Urinary symptoms were relatively common and affected QoL. Women were counted as symptomatic if they had the symptom ‘most or all of the time’. Nocturia and urgency were common types of urinary symptom. Overall, around one in five (22.6%) women reported at least some urinary incontinence (using the ICIQ-UI SF)36 and this was slight or moderate in the majority (68.9%) of cases (see Appendix 2, Table 52). The mean QoL because of urinary symptoms was 3.9 out of 10 (Table 7).

Few women reported bowel symptoms, and these did not appear to have a large impact on QoL with the mean 2.7 out of 10 (Table 7). Around one-third (30.2%) of all women reported faecal incontinence at least occasionally [defined as loss of solid or liquid stool, but not including loss of flatus (wind)] and this was severe in < 9% of women. Few women (6.2%) reported constipation.

Discussion

Surgery actually received

Of the 563 women included, 469 received their randomised allocation (76.8% in the uterine suspension group and 89.8% in the vaginal hysterectomy group) (Table 8). Reasons for not having the allocated treatment were the woman’s choice (n = 27), no apical descent [with alternative prolapse surgery (n = 17) and without surgery (n = 5)], alternative apical procedure (n = 7), other clinical factors (n = 10) or reason unknown (n = 28).

The use of mesh for the apical or concomitant procedure was recorded and is detailed in Appendix 3, Table 53.

In relation to the randomised groups, more women in the vaginal hysterectomy group received a concomitant anterior repair (50.2%) than in the uterine preservation group (40.4%). Overall, the numbers receiving a concomitant procedure in other compartments or for incontinence were similar between the groups (see Table 8).

Description of surgical characteristics

Most procedures were performed by a consultant gynaecologist (73.6%) with the patient under general anaesthesia (80.3%), and most patients received prophylactic antibiotics (92.0%) (see Appendix 3, Table 54). The operation was significantly longer in the uterine preservation group (114 minutes) than the vaginal hysterectomy group (103 minutes, MD 9.79 minutes, 95% CI 3.50 to 16.07 minutes). Women in the vaginal hysterectomy group had significantly higher blood loss (166 ml) than those randomised to uterine preservation (125 ml, MD –42 ml, 95% CI –62.68 to –21.73 ml). There was no difference in length of hospital stay (1.9 days, see Appendix 3, Table 55). There was no difference in time from the woman’s randomisation to her surgery between the groups (see Appendix 3, Table 55).

Outcomes

Women’s prolapse symptoms and effect on everyday life

At 6 months after surgery, women reported a reduction in their POP-SS from a mean score of 13.7 out of 28 to 4.9 out of 28 in the uterine preservation group and from 13.5 out of 28 to 4.0 out of 28 in the vaginal hysterectomy group (MD 0.85, 95% CI 0.00 to 1.71). The effect of prolapse on QoL also improved at 6 months after surgery and again there was no significant difference between the groups.

The primary outcome was POP-SS at 12 months after randomisation; there was no statistically significant difference between the groups. The MD in the POP-SS for uterine preservation (4.2, SD 4.9) compared with vaginal hysterectomy (4.2, SD 5.3) adjusted for baseline variables was –0.05 (95% CI –0.91 to 0.81) (Table 9 and Figure 4).

FIGURE 4.

Mean and standard error POP-SSs at baseline and follow-up.

There was also no significant difference on the effect of prolapse on QoL at 12 months after randomisation [uterine preservation 1.7 (SD 2.5) vs. vaginal hysterectomy 1.5 (SD 2.5); MD 0.12, 95% CI –0.26 to 0.49]. The intracluster correlation for the surgeon was 0.04 (95% CI 0.00 to 0.44).

Distribution of individual POP-SSs and prolapse-related effects on QoL are also given in Appendix 3, Figures 20 and 21.

Each individual prolapse symptom also improved at 12 months after randomisation (Table 10) and was already apparent at 6 months after surgery (see Appendix 3, Table 56).

The most bothersome symptoms across both groups at 6 months after surgery and at 12 months after randomisation were different from those at baseline. The most bothersome symptom at baseline, ‘a feeling of something coming down’ reduced overall from 49.7% to 19.5%. Symptoms of incomplete emptying of either bladder or bowel became more common, but ‘actions necessitated by prolapse symptoms’ were improved (see Table 10, see also Appendix 3, Table 56).

Secondary outcomes

Urinary and bowel symptoms

Overall, at 12 months after randomisation, the proportion of women who had any urinary incontinence increased from 22.6% at baseline to 35.8%, although the proportion with severe urinary incontinence decreased from 13.0% to 1.6% (Table 13). Significantly more women reported urge incontinence at 12 months after randomisation in the uterine preservation group (n = 32, 13.9%) than in the vaginal hysterectomy group (n = 12, 5.0%; MD 2.71, 95% CI 1.32 to 5.55). Following surgery, de novo urinary incontinence was similar across both groups and reported in around one-quarter of the women. Further urinary symptoms can be found in Appendix 3, Table 58.

Although bowel urgency increased to 14.3% across both treatment groups at 12 months after randomisation, the presence of severe faecal incontinence was slightly improved. However, significantly more women in the uterine preservation group (n = 15, 6.6%) had severe faecal incontinence than in the vaginal hysterectomy group (n = 7, 3.0%), though these were small numbers (MD 4.64, 95% CI 1.49 to 14.42).

There was no evidence of a difference in any other urinary or bowel symptoms outcomes between the randomised groups.

Satisfaction with treatment at 12 months after randomisation

At 12 months after randomisation, most women reported that they were better than before surgery, with no statistically significant difference between the groups (90.8% vs. 95.3%; OR 0.49, 95% CI 0.22 to 1.05) (Table 14). Similarly, the proportions of women completely or fairly satisfied were not statistically significantly different (85.6% vs 90.6%; OR 0.62, 95% CI 0.34 to 1.13). However, significantly more women (95.0%) in the vaginal hysterectomy group would recommend it to a friend compared with women in the uterine preservation group (88.3%) (OR 0.39, 95% CI 0.18 to 0.83).

Further treatment required for failure or adverse events at 6 and at12 months

When women reported that they had been re-admitted to hospital, the information was verified with the centres and post-coded accordingly to ensure accuracy of data and resolution of discrepancies. A hospital re-admission was counted as a SAE if it was related to the initial prolapse surgery. Further surgery for prolapse (recurrence if in the same compartment), or for continence surgery, was differentiated from re-admission for surgery-related complications, such as bleeding or infection (Table 15).

Overall, 29 women were re-admitted for further surgery in the first 12 months (Table 15). Twenty-three women had further prolapse surgery within 12 months [uterine preservation (n = 15) vs. vaginal hysterectomy (n = 8)]. Ten women in the uterine preservation group had a further apical procedure compared with three women in the vaginal hysterectomy group during that time. More women in the uterine preservation group (n = 8) had a re-admission for an anterior repair compared with those women in the vaginal hysterectomy group (n = 1). Overall, there was no evidence of a difference between the groups regarding further treatment.

One woman did not have her VUE procedure (or any procedure) at the time of admission (clinical decision), but pursued the apical procedure by another health-care provider 12 months later.

Serious and related adverse events in the first 12 months after surgery

Overall, there were 30 SAEs (5.7%) reported in the first 12 months following surgery (more than one SAE could be reported per woman) (Table 16). Hospitalisation was the most common type of SAE. There were no differences between the groups in any SAEs.

Specifically, of the 129 procedures (apical and concomitant) using a mesh implant one mesh exposure/extrusion was identified and required surgical treatment in the first 12 months after surgery.

Non-serious and related adverse events in the first 12 months after surgery

Non-serious AEs or complications were also recorded (Table 17). Overall, there was a total of 18 (3.6%) within the 12 months after surgery. There were no differences between the groups in any non-serious AEs.

Subgroup analysis

There was no evidence of a difference between treatment groups in any of the planned subgroup analyses (age < 60 and ≥ 60 years, previous anterior or posterior surgery or previous incontinence surgery) (see Appendix 3, Table 60).

Sensitivity analyses

A multiple imputation model was used to impute missing values in the primary outcome, assuming a missing-at-random mechanism. No evidence of a difference between the two groups after the imputation (uterine preservation vs. vaginal hysterectomy; MD –0.02, 95% CI –0.83 to 0.78; p = 0.96) was found.

Discussion

Conclusion

Vaginal hysterectomy and uterine preservation surgeries are equally effective in relieving uterine prolapse in relation to symptoms and objective descent, although vaginal hysterectomy was more likely to be recommended to a friend. The presence of urinary incontinence may worsen after either procedure, but severe incontinence appears to remain static; however, an increased risk of re-operation for an anterior pelvic floor repair in those women having preservation procedures was identified.

NHS perspective resource use and costs

Patient participant perspective resource use and costs

Women with prolapse symptoms have a high rate of contact with health-care professionals, including a large number of consultations in both primary and secondary care. The opportunity costs of time and travel, purchase of over-the-counter medications and other related expenses were similar across the groups. A small number of women reported private care, and the cost was higher for women in the uterine preservation group.

The number of women reporting that they had to take time off from work, as a proportion of the trial population, was small. In total, in the Uterine trial 130 out of 572 (23%) randomised women reported the need to take time off work for prolapse symptoms [64/279 (23%) and 66/283 (23%) of those randomised to the uterine preservation and vaginal hysterectomy groups, respectively]. Among those women who required sick leave from work, the average number of days of leave required was 35 days (SD 65 days) and 22 days (SD 40 days) in the uterine preservation and vaginal hysterectomy groups, respectively, but differences were not statistically significant. See Appendix 4, Table 65 for full details of all costs incurred by the women.

Among the women with complete data for participant-incurred costs (46% in the uterine preservation group vs. 54% in the vaginal hysterectomy group), the average costs were £1266 (SD £4127) and £460 (SD £1505) in the uterine preservation and vaginal hysterectomy groups, respectively (MD £609, 95% CI £47 to £1172). The large SDs in both groups indicate that some women experienced very high personal costs.

When summing participant and NHS perspective costs together to inform a wider-perspective analysis of prolapse costs, the uterine preservation group remains significantly more costly (MD £911, 95% CI £359 to £1463).

Generic quality-of-life outcomes

The proportion of women reporting any health problems on each EQ-5D-3L domain (i.e. a score of 2 or 3 points) at baseline, at 6 months after surgery and at 12 months after randomisation is presented in Figure 5. Data are presented as reported in the respective questionnaires. Missing data are not reported in the figure. At baseline, < 70% of the women in both groups reported problems related to pain and discomfort and between one-quarter and one-third of women had at least some problems with usual activities or anxiety and depression. The reporting of any pain or discomfort fell to about 30% for all women by 12 months after randomisation, with fewer than one in five women reporting problems with usual activities or anxiety/depression. The fewest number of problems were experienced in self-care, with a small proportion of women reporting any problems by 12 months after randomisation. There are no obvious differences between the groups evident from the graphical presentations. The proportion of women experiencing problems decreased in both groups, indicating that both procedures had a positive impact on generic QoL. This finding echoes the findings from the condition-specific QoL data (see Chapter 4).

FIGURE 5.

Percentage of respondents with any problems on each EQ-5D-3L domain at (a) baseline, (b) 6 months after surgery and (c) 12 months after randomisation.

Full details of the utility scores resulting from EQ-5D-3L responses at each time point and calculated QALYs are reported in Table 18. In terms of covariates included within the analysis model, baseline utility was the only significant predictor of overall QALYs, re-enforcing the importance of adjusting for this baseline measure. 54

Complete data were available across domains and time points, enabling QALY calculation for 198 out of 279 (71%) and 209 out of 283 (74%) women in the uterine preservation and vaginal hysterectomy groups, respectively. There were no differences in QALYs between the uterine preservation (0.845 years) and vaginal hysterectomy (0.866 years) groups (MD 0.006 years, 95% CI –0.016 to 0.028 years). Multiple imputation of missing EQ-5D data underpinning the QALY calculation indicates a similar lack of difference across the groups.

Cost-effectiveness results

Complete-case cost and complete-case QALY data are presented (Table 18), making best use of all available data. For the base-case analysis, multiple imputed missing data are used because only 372 out of 562 (66%) participants had fully complete profiles of costs and QALYs to enable calculation of the ICER. A range of deterministic sensitivity analyses are also reported. All analyses of cost-effectiveness also indicate the probability that uterine preservation is a cost-effective use of NHS resources at five alternative threshold values (£0, £10,000, £20,000, £30,000 and £50,000) of WTP for a QALY gained.

Base-case analysis

The base-case analysis using the imputed data set shows significant additional costs for the uterine preservation group (MD £235; 95% CI £6 to £464), for no significant difference in QALYs (MD –0.004 years, 95% CI –0.026 to 0.019 years). As uterine preservation is associated with additional costs for a negative point estimate of the impact on QALYs, uterine preservation is therefore dominated by vaginal hysterectomy. The scatterplot (Figure 6) illustrates this finding visually. There is little uncertainty regarding the additional cost of uterine preservation, with almost all of the increments from the bootstrap lying above the horizontal line. However, regarding QALYs, the scatterplot indicates substantial uncertainty, with no clear evidence of one strategy being preferable.

FIGURE 6.

Scatterplot of the cost-effectiveness plane (imputed data set).

The CEAC (Figure 7) shows that, based on the NMB, there is a high probability that vaginal hysterectomy is the most cost-effective treatment option, with little evidence to support the cost-effectiveness of uterine preservation. This is driven by the additional costs of delivering the uterine preservation intervention, and the higher re-operation rate, for no QALY benefit. The probability of uterine preservation being considered to be cost-effective is 15% if society is willing to pay up to £30,000 for 1 QALY gained. A full set of scatterplots and CEACs for each analysis undertaken can be found in Appendix 4, Figures 22–9.

FIGURE 7.

Cost-effectiveness acceptability curve (imputed data set).

Deterministic sensitivity analyses

The study has produced a range of additional deterministic sensitivity analyses to address areas of uncertainty in assumptions, data collection and analysis modelling approaches. All deterministic analyses are based on the imputed data set unless otherwise stated (e.g. complete-case analysis). See Chapter 2 for the rationale behind each of the analyses. A number of observations have been made on the results.

First, the findings are not sensitive to the type of model used in the imputed data set with very similar results for the base case and a sensitivity analysis using seemingly unrelated regression, in which the probability of cost-effectiveness at a threshold of £30,000 per QALY gained reduces from 15% (base case) to 14%. Additionally, it is reassuring that the results are not sensitive to missing data assumptions, with similar findings for both the imputed and the complete-case analyses. The probability of uterine preservation being cost-effective at a threshold of £30,000 per QALY gained remains < 20% for both the imputed and the complete-case scenarios.

Second, the estimates of incremental costs are highly sensitive to the approach taken to intervention costing. The base-case component costing approach estimated additional costs because of the use of mesh and the slightly longer (around 10 minutes) operation time. However, when mapping Office of Population, Censuses and Surveys (OPCS) codes onto HRGs, using national reference cost spell-based tariffs, uterine preservation attracts a lower cost compared with vaginal hysterectomy. The use of a HRG approach to costing the intervention means that total costs to the NHS over follow-up are lower in the uterine preservation group, MD –£395; 95% CI –£619 to –£171. Considering that cost savings may be achieved for marginal (non-significant) reductions in QALYs, the base-case ICER is £98,750. The commonly accepted maximum price society would be willing to pay to get one additional QALY of benefit is between £20,000 and £30,000. The threshold could also be interpreted as the maximum amount of money that would need to be freed up in the NHS to compensate for a QALY loss. Using a HRG-based costing approach indicates that society would save £98,750 for every QALY lost, indicating a cost-effective use of resources. The probability that uterine preservation is cost-effective increases from 15% in the base case to 85% under this scenario, illustrating that the results are highly sensitive to the choice of intervention costing approach taken.

Third, because of differences in time points for questionnaire triggers (i.e. at 6 months after surgery and at12 months after randomisation) and the potential for comparatively long periods of time to elapse between randomisation and surgery, it is possible that respondents received their 12-month questionnaire before their 6-month questionnaire. The sensitivity analyses show that estimates of cost-effectiveness are not sensitive to any of the following assumptions in sensitivity analysis:

1. an analysis dropping the 6-month EQ-5D data, if the 6-month questionnaire is returned on or after the 12-month date

2. an analysis dropping any 6-month EQ-5D data within 3 months of the 12-month data

3. an analysis using the exact date of questionnaire report as the measure of time used in the calculation of QALYs.

Finally, the incorporation of a wider-perspective costing analysis, including both NHS and participant-incurred and reported costs, does not alter the estimates of cost-effectiveness. The probability that uterine preservation is cost-effective remains very low (5%) at a threshold of £30,000 per QALY.

Discussion

The findings of the economic evaluation indicate that uterine preservation is unlikely to be cost-effective compared with vaginal hysterectomy over a 12-month time horizon. Although the findings are unfavourable for uterine preservation, there are a number of factors that need to be considered when arriving at a decision on cost-effectiveness. First, the cost-effectiveness results are sensitive to the costing approach taken for the intervention. Arguments can be made for the use of either a component-costing or HRG approach. Component costing is chosen as the base case because it is grounded in the economic concepts of resource scarcity and opportunity cost. Therefore, component costing more accurately reflects the resources displaced for each surgery, and explicitly accounts for differences in operation times, length of stay and materials costs (e.g. mesh implants) between the groups observed in the trial. However, the HRG costing approach is more reflective of the way in which hospitals are paid for carrying out procedures, and it could be argued that HRGs reflect the true cost outlay to the NHS under current payment mechanisms. The disadvantage of the HRG approach is that it groups many similar surgeries together. It does not directly consider the opportunity cost of consultant time, etc. for these specific procedures. That is because HRGs are not based on the intricate level of detail that can be obtained from the trial data. Although the choice of intervention costing has a substantial impact on overall costs and cost-effectiveness for the trial-based analysis, the choice may be less influential in determining lifetime cost-effectiveness from the decision model (see Chapter 2).

Second, the validity of any policy recommendations based solely on the within-trial economic evaluation could be correctly criticised because the time horizon is very short. Indeed, it is highly unlikely that a 12-month time horizon is sufficient to capture the important costs and QoL impact of different surgical treatments. The economic modelling (see Chapter 9) aims to address this concern by extrapolating the trial results over a woman’s lifetime. Additionally, it is planned to follow-up women for a longer period to get a better understanding of the costs and the outcomes associated with treatment. These data will be used to validate and update the trial and economic modelling results in the future.

Study recruitment

The trial outline and methodology for recruitment have been described previously31 (see Chapter 2). Women who attended gynaecology outpatient clinics with symptomatic vault POP and then chose to have surgery, and women on the waiting list for vault prolapse surgery, were invited to participate in the Vault trial of the VUE trial. Women were asked if they were willing to be randomised to either a vaginal or abdominal vault procedure. The centres and surgeons who participated in the VUE trial, the numbers they recruited and the rate of recruitment are detailed in Appendix 5 (Table 66 and Figure 30).

Non-recruited women

Of the 544 women approached regarding trial participation, 335 women did not enter the study because they declined (n = 211), were ineligible (n = 115) or not timely identified or seen (n = 9) (see Appendix 5, Tables 67 and 68). Reasons for declining to participate included decided against surgery (n = 58, 27.5%), did not want mesh (n = 8, 3.8%) and preference for the route of surgery (n = 85, 40.3%), with more women preferring an abdominal (n = 53, 25.1%) rather than a vaginal procedure (n = 32, 15.2%). Clinical reasons for non-participation included ‘specific operation is necessary’ (n = 73, 63.5%) and ‘unsuitable due to medical history’ (n = 21, 18.3%).

Recruited women: baseline characteristics

The baseline characteristics of the 208 women included in the Vault trial are described in Table 19 (see also Appendix 5, Table 69).

Epidemiological characteristics

There was no difference between the randomised groups in respect of age, BMI, parity or delivery mode history [with the majority of women having a normal vaginal delivery (see Appendix 5, Table 69)].

Previous conservative treatment

Around one-quarter to one-third (29.8%) of women had PFMT supervised by a physiotherapist for prolapse symptoms and around one in five (20.2%) women had previously used a vaginal pessary (ring or other type) before surgery (Table 19).

Overall, around 1 in 10 women (13.0%) had previous supervised PFMT or used drug treatment (9.7%) for urinary incontinence.

Previous surgery

All women in the Vault trial had a previous hysterectomy. Slightly more women had had a vaginal hysterectomy (54.8%) than an abdominal hysterectomy (46.2%) (Table 19). Of those women with a previous vaginal hysterectomy, in just over one-third (36.6%) of the women this was for prolapse symptoms. A previous abdominal hysterectomy was rarely undertaken for prolapse reasons and was most likely due to ‘other reasons’ (31.8%), for example menorrhagia or fibroids. The reason for a previous hysterectomy (vaginal or abdominal) was unknown in around 12% of women (Table 19).

In four women, their VUE vault repair was a repeat vault procedure, having already had a previous vault repair (three of which were for prolapse and one for ‘other reasons’).

Just over one-quarter (27.4%) of all women had previously had an anterior repair and four women (3.8%) had two previous anterior repairs. Fewer women (approximately 15%) had a previous posterior repair, and one woman had two previous posterior repairs. Previous vaginal repair was noted in an additional 10 women, but the compartment of previous surgery was unknown (see Table 19).

Overall, approximately 7% of women had undergone previous continence surgery.

Prolapse symptoms at baseline

Women in the Vault trial had been symptomatic for approximately 3 years, and were bothered by their symptoms for approximately 2 years. The overall mean POP-SS was 15 out of 28 (and ranged from 1 to 28). Using a POP-SS of > 0 to indicate presence of symptoms, 100% of women who completed these questions had at least one symptom (this section was missing for seven women). The effect on QoL score (‘overall, how much do your prolapse symptoms interfere with your everyday life?’) was high and ranged from 0 to 10 out of 10, with a mean of 7 out of 10 (Table 20).

The most common individual prolapse symptom was ‘a feeling of something coming down from or in your vagina’, reported in 94.7% of women. This was reported ‘most or all of the time’ in over three-quarters (78.9%) and was the most bothersome symptom for almost half (47.2%) of the women.

Over one-third (36.6%) of women found the prolapse caused hygiene problems, and 15.9% needed to relieve pressure or discomfort from the prolapse using their fingers (Table 20).

Generic quality of life

The mean score for generic QoL and visual scales (EQ-5D) were 0.735 and 74.1 points, respectively.

Preoperative objective measurements

Over one-third (38.9%) of women had a stage 0 or 1 vault prolapse. Using a more clinically meaningful definition of leading edge of prolapse beyond the hymen (POP-Q stage 2b or more), 91.4% of women had an overall objective prolapse (see Appendix 5, Table 70), and 39.4% had an objective vault prolapse [stages 2b or more (Table 21)]. The stage of vault prolapse was missing in 3.9% of women.

More women had an objective (stage 2b or more) anterior (80.3%) than a posterior (52.4%) prolapse (Table 21).

Planned concomitant surgery

Planned surgery was based on preoperative findings on clinical examination (Table 22). The VUE trial was designed so that women would remain in the group to which they were allocated, irrespective of the actual procedure performed. In order to randomise women appropriately, taking account of minimisation criteria, gynaecologists were asked to specify in advance which concomitant surgery would also be necessary.

All women were expected to undergo a vault prolapse procedure from the preoperative clinical planning. A concomitant anterior prolapse repair was planned more frequently (32.7%) than either posterior repair (13.5%) or both anterior and posterior repair (21.7%). Surgery for incontinence was rarely planned (2.9%).

Planned procedures were equally distributed between the two randomised groups.

Vaginal and sexual symptoms at baseline

Vaginal and sexual symptoms were measured using the ICI-validated instruments36 (Table 23) (www.journalslibrary.nihr.ac.uk/programmes/hta/11129183/#/; accessed February 2019). These were common and had important effects on QoL. The majority of women were not sexually active (72.6%) and for about one-quarter of women this was most often because of their prolapse symptoms. No women reported dyspareunia at baseline.

Urinary and bowel symptoms at baseline

Urinary symptoms also had some impact on QoL (the overall mean QoL due to urinary symptoms was 4/10) (Table 24). Around one in six (16.8%) of all women reported at least some urinary incontinence, as measured using the ICIQ-UI SF;36 however, this was slight or moderate in the majority (79.3%) of cases (Appendix 5, Table 71). Urgency was the most common type of urinary symptom, reported in around 28.9% of all women. Women were counted as symptomatic if they had the symptom ‘most or all of the time’.

Over one-third (38.0%) of all women had faecal incontinence at least occasionally [defined as loss of solid or liquid stool, but not including loss of flatus (wind)], and this was severe in around 12.0% of women. Few women (8.2%) reported constipation.

Discussion

Description of surgical characteristics

Most procedures were performed by a consultant gynaecologist (87.5%), were performed under general anaesthesia (90.4%) and most patients received prophylactic antibiotics (91.8%) (Appendix 6, Table 73).

The duration of the operation was significantly longer in the abdominal group (146 mean duration, SD 54 minutes) than in the vaginal group (82 mean duration, SD 34 minutes; MD 60.48, 95% CI 49.80 to 71.16 minutes). There was no difference in blood loss (98 ml in the abdominal group vs. 108 ml in the vaginal group) or length of hospital stay (2.1 days in the abdominal group vs. 1.8 days in the vaginal group) between the groups (Appendix 6, Table 74).

There was no difference in time from the woman’s randomisation to her surgery between the groups (Appendix 6, Table 74).

Outcomes

Women’s prolapse symptoms and effect on everyday life

At 6 months after surgery, women reported a reduction in their POP-SS (maximum score 28) from an overall mean score of 15 out of 28 at baseline to 5.3 out of 28 in the abdominal group and 6.1 out of 28 in the vaginal group (Table 26). The between-group difference was not statistically significant (MD –1.17; 95% CI –2.64 to 0.31). The effect of prolapse on QoL also improved at 6 months after surgery and there was no significant difference between the groups.

The primary outcome was POP-SS at 12 months after randomisation; there was no statistically significant difference between the groups. The MD in the POP-SS for abdominal (5.6, SD 5.4) compared with vaginal (5.9, SD 5.4) adjusted for baseline variables was –0.61 (95% CI –2.08 to 0.86) (see Figure 10, and Appendix 6, Figure 31). The intracluster correlation for surgeon was 0.04 (95% CI 0.01 to 0.12).

FIGURE 10.

The mean and standard error POP-SSs at baseline and follow-up (6 months after surgery and 12 months after randomisation).

There was also no significant difference on the effect of prolapse on QoL at 12 months after randomisation [abdominal 2.3 (SD 3.0) vs. vaginal 2.6 (SD 2.8); MD –0.25, 95% CI –1.10 to 0.59].

The distribution of individual POP-SSs and prolapse-related effects on QoL are also given in Appendix 6, Figures 31 and 32.

Each individual prolapse symptom also improved at 12 months after randomisation (Table 27) and was already apparent at 6 months after surgery (see Appendix 6, Table 75).

The most bothersome symptoms across both treatment groups at both 6 months after surgery and 12 months after randomisation were different from those at baseline. The most bothersome symptom at baseline, ‘a feeling of something coming down’, reduced overall from 47.2% to 10% in the abdominal versus 34% in the vaginal group. Symptoms of incomplete emptying of either bladder or bowel became more common, but ‘actions necessitated by prolapse symptoms’ were low (see Table 27, see also Appendix 6, Table 75).

Secondary outcomes

Subgroup analysis

There were no differences in any of the planned subgroup analyses [aged < 60 years and ≥ 60 years, previous anterior or posterior surgery, or previous incontinence surgery (see Appendix 6, Table 79)].

Sensitivity analyses

A multiple imputation model was used to impute missing values in the primary outcome, POP-SS, assuming a missing-at-random mechanism. No evidence of a difference between the two treatment groups after the imputation was found (abdominal vs. vaginal: MD –0.32, 95% CI –1.71 to 1.06; p = 0.65).

Discussion

Primary and secondary outcomes

Around 85.5% of women responded to the 12-month questionnaire, with slightly more women attending for the 12-month clinic assessment.

The primary outcome was POP-SS at 12 months after randomisation, and there was no difference between an abdominal and a vaginal procedure. The MD in the POP-SS for the abdominal group was 5.6 (SD 5.4) compared with a MD of 5.9 (SD 5.4) in the vaginal group. Adjusted for baseline variables, the MD was –0.61 (95% CI –2.08 to 0.86) and suggests that both procedures are equally clinically effective in improving vault prolapse symptoms at 12 months.

At 12 months after randomisation, prolapse-related effects on QoL scores had improved and were apparent at 6 months after operation. The most bothersome symptom at baseline, ‘a feeling of something coming down’, reduced from 47.2% overall at baseline to 10% in the abdominal group and 34% in the vaginal group. Incomplete emptying of the bladder or bowel were the most bothersome at both 6 and 12 months for both randomised groups of women, but did not affect ‘actions necessitated by prolapse symptoms’, which were reduced from baseline.

Overall, objective prolapse scores (i.e. POP-Q) improved for both randomised groups, particularly in terms of prolapse protrusion beyond the hymen (stage 2b or more: 32.6% after an abdominal vs. 46.9% after a vaginal procedure; this was not statistically significant). Objective vault prolapse (stage 2b or more) fell to 6.0% in the abdominal group and 9.0% in the vaginal group 12 months after surgery (adjusted for baseline values). Again, this was not statistically significant.

More women were sexually active after surgery. There were no statistically significant differences in vaginal and sexual symptoms between either of the randomised groups of women at 12 months.

Overall, at 12 months after randomisation, the proportion of women who had any urinary incontinence increased from around one in six women to one in three in the abdominal and one in four in the vaginal group; severe urinary incontinence remained the same as baseline. In addition, de novo urinary incontinence following surgery was found in one in four women in the abdominal group and one in five in the vaginal group. Few women had continence surgery to alleviate their symptoms within the first 12 months (n = 4). There was no evidence of a difference in any urinary symptom outcomes between the randomised groups of women.

There were no other significant differences in bowel symptoms between the randomised groups of women.

There were no differences between the randomised groups of women in terms of satisfaction with treatment at 12 months after randomisation. Most women reported that they were better than before surgery (92.8% vs. 86.2%). The majority of women were completely or fairly satisfied (87.8% vs. 82.6%) and would recommend the surgery to a friend (87.2%).

Overall, 15 women were re-admitted for further surgery in the first 12 months after surgery (see Table 32). Eleven women had further prolapse surgery within 12 months [abdominal preservation (n = 5) vs. vaginal (n = 6)], and this was not statistically significant. One woman in the abdominal group had a further apical procedure compared with four women in the vaginal group in that time (again, this was not statistically significant). There was no difference in re-admission rates for anterior repair between both groups. Three women in the abdominal group had a re-admission for a posterior repair.

Overall, 12 SAEs (5.8%) were reported in the first 12 months following surgery and there were no differences between the groups in any serious adverse effects. Specifically, of all the 106 procedures using a mesh implant one mesh exposure/extrusion from a concomitant continence procedure was identified requiring surgical treatment in the first 12 months after surgery relating to a risk ratio (RR) of around 1% for mesh exposure.

Conclusion

Both abdominal and vaginal surgeries are equally clinically effective in relieving symptoms of vault prolapse in the short term. There is no statistically significant difference for the apical measurements. The objective descent measured by POP-Q is significantly different in relation to the anterior wall prolapse 12 months after surgery (though the numbers were small). This, however, did not affect re-admissions for anterior wall prolapse in the first 12 months. The presence of urinary incontinence may worsen after either procedure, but severe incontinence appears to remain static.

Overview

This chapter presents the results of the within-trial cost–utility analysis for women in the Vault trial. A single table of all results is presented (Table 35), with separate sections for summary data on costs, QALYs and cost-effectiveness (including deterministic sensitivity analysis results). Further details of the cost-effectiveness data are provided in Appendix 7, Tables 80–84.

NHS perspective resource use and costs

Patient participant perspective resource use and costs

Women with prolapse symptoms have a high rate of contact with health professionals, including a large number of consultations in both primary and secondary care. The costs to women of time spent attending appointments is substantial, with costs per women of £252 in the abdominal group compared with £201 in the vaginal group.

The number of women reporting that they had to take time off from work, as a proportion of the trial population, was small. In total, 38 women, 19 per group, representing 18% of trial participants reported needing to take sick leave. The low proportion requiring sick leave is likely to be because of the proportion in paid employment (c. 25%) and the average age of the trial participants having passed the standard retirement age. Among those women needing to take sick leave, the average number of days off work was 24, with no differences across the groups.

The costings include a few women who incurred a small amount of cost for over-the-counter medications and other related expenses. No women in the Vault trial report any use of private health-care services for prolapse-related symptoms. Full details of all costs incurred by women, collected through the study are reported in Appendix 7, Table 84.

The total average costs to women over the trial follow-up period is substantial, but there were no differences across the groups (MD –£53, 95% CI –£400 to £294). High SDs contribute to wide CIs and indicate that there is substantial variability in the experience of personal costs, with some women experiencing substantially greater financial burden than others.

When summing participant and NHS perspective costs together to inform a wider-perspective analysis of prolapse costs, the abdominal group remains significantly more costly with a MD of £825 (95% CI £208 to £1443).

Generic quality-of-life outcomes

The proportion of women reporting any health problems on each EQ-5D-3L domain (i.e. a score of two or three) at baseline, at 6 months and at 12 months is presented in Figure 11. Missing data are not reported on the graphs. At baseline, a substantial proportion of women in both groups report problems related to pain and discomfort. This would be expected given the need for surgery. Those reporting any pain or discomfort reduces to about half of all women in both groups at 6 months, with similar proportions reporting pain/discomfort at 12 months. The proportion of women reporting problems with self-care was low.

FIGURE 11.

Percentage of respondents reporting any problems on each EQ-5D domain (baseline, 6 months after surgery and 12 months after randomisation).

Full details of the utility scores resulting from these EQ-5D-3L responses at each time point and calculated QALYs can be found in Table 35. In terms of covariates included within the analysis model, baseline utility was the only significant predictor of overall QALYs, re-enforcing the importance of adjusting for this baseline measure.

Complete data were available across EQ-5D domains and time points (baseline, and 6 months and 12 months), enabling QALY calculation for 67 out of 104 (64%) and 78 out of 104 (75%) women randomised to the abdominal and vaginal vault groups, respectively. Based on the complete-case data, there are no statistically significant differences in QALYs between treatment groups, with the abdominal and vaginal vault groups obtaining an average of 0.824 and 0.825 QALYs per person over 12 months, respectively (MD –0.009, 95% CI –0.049 to +0.032). Multiple imputation of missing data indicates a similar lack of difference across the groups (MD 0.002, 95% CI –0.036 to 0.039).

Cost-effectiveness results

Complete-case costs and complete-case QALY data are presented, making best use of all available data (see Table 35). The base-case analysis uses multiple imputation of missing data. Imputation of missing data was chosen because (1) only 129 out of 208 (62%) participants had fully complete profiles of costs and QALYs to enable calculation of the ICER, and (2) missing data for QALYs were imbalanced across the randomised groups. A range of deterministic sensitivity analyses are also reported. All analyses of cost-effectiveness also indicate the probability that abdominal vault repair is a cost-effective use of NHS resources at five alternative threshold values of WTP for 1 QALY gained (i.e. £0, £20,000, £30,000, £50,000 and £100,000).

Base-case analysis

The multiple imputed data show significant additional costs for the abdominal group (MD £570, 95% CI £459 to £682) for no significant QALY gain (MD 0.004, 95% CI –0.030 to 0.040). The base-case cost to the NHS of obtaining one additional QALY is therefore £285,000 per QALY.

The scatterplots (Figure 12) of all bootstrapped replications further illustrates the sampling uncertainty accrued from the bootstrapped replications of the analysis models. There is little uncertainty regarding the additional cost of the abdominal procedure when using the base-case component costing approach with almost all of the increments from the bootstrap lying above the horizontal line. However, regarding QALYs, the scatterplot indicates substantial uncertainty, with no clear evidence of one strategy being preferable.

FIGURE 12.

Vault trial: scatterplot of the cost-effectiveness plane (imputed data set).

The CEAC in Figure 13 shows that at 12 months after randomisation, based on the NMB, there is a high probability that vaginal repair is the most cost-effective treatment option, with little evidence to support the cost-effectiveness of abdominal repair. The probability of abdominal repair being considered to be cost-effective is < 17%, up to a threshold value of WTP for 1 QALY gained of £30,000, increasing to a 29% probability of cost-effectiveness if the threshold was £50,000 per QALY.

FIGURE 13.

Vault trial: CEAC (imputed data set).

Deterministic sensitivity analyses

Deterministic sensitivity analyses are conducted to address areas of uncertainty in assumptions, data collection and analysis-modelling approaches. All deterministic analyses are based on the imputed data set unless otherwise stated (e.g. complete-case analysis). The rationale behind each of the analyses is described in Chapter 2 and a number of observations were made on the results. Results of all deterministic sensitivity analyses are reported in Table 35 and a full set of CEACs and scatterplots for each analysis can be found in Appendix 7, Figures 33–40.

First, based on an analysis of complete cost and QALY pair data, at the 12-month follow-up, abdominal vault repair remains significantly more costly and generates non-significantly fewer QALYs and is thus dominated by vaginal vault repair. Although the additional costs of abdominal vault repair are marginally greater for the complete-case analysis, the additional costs are surrounded by more uncertainty, with wider CIs. Probabilities of cost-effectiveness are therefore slightly higher in the complete-case group, which may initially appear counterintuitive. However, this is because of greater uncertainty around the effect size on costs in the complete-case group, with a higher proportion of simulations falling to the right of the threshold line (see Appendix 7). The important point remains that complete-case data show no evidence of changing the overall findings of the analysis.

Second, it was shown that the findings are robust to the analysis models incorporated. The use of seemingly unrelated regression as an alternative approach to account for the correlation in costs and QALYs (abdominal, –0.202; vaginal, –0.066; trial, –0.146) does not change the overall findings.

Third, the estimates of incremental costs in the Vault trial are sensitive to the approach taken to intervention costing. Using HRG reference costs for intervention procedures increases the MD (abdominal vs. vaginal) to £1212 (95% CI £916 to £1508) and increases the ICER to > £300,000 per QALY. This analysis further reduces the probability that abdominal vault repair might be considered cost-effective compared with vaginal vault repair. Arguments can be made for the use of either approach. Component costing is chosen as the base case because it is grounded in the economic concepts of resource scarcity and opportunity cost. It therefore more accurately reflects the resources displaced for each surgery, and explicitly accounts for differences in operation times, length of stay and materials costs (e.g. mesh and devices) between the groups observed in the trial. Importantly, the approach to intervention costing does not change the overall cost-effectiveness conclusions from the Vault trial.

Fourth, because of differences in time points for questionnaire triggers (i.e. at 6 months after surgery and at 12 months after randomisation), and the potential for comparatively long periods of time to elapse between randomisation and surgery, it is possible that respondents get their 12-month questionnaire before their 6-month questionnaire. The sensitivity analyses show that estimates of cost-effectiveness are not sensitive to any of the following assumptions in sensitivity analysis:

1. an analysis dropping the 6-month EQ-5D data if the 6-month questionnaire is returned on or after the 12-month date

2. an analysis dropping any 6-month EQ-5D data within 3 months of the 12-month data

3. an analysis using the exact date of questionnaire report as the measure of time used in the calculation of QALYs.

However, a note of caution on the last analysis is that some women will have QALYs greater than one over a single year, as some questionnaires were returned well after 12 months (i.e. maximum = 1.80 years).

Fifth, the incorporation of a wider-perspective costing analysis, including both NHS and participant-incurred and -reported costs, does not alter the estimates of cost-effectiveness or the probability of cost-effectiveness at alternative threshold values of WTP for 1 QALY gained.

Discussion

The findings from the Vault trial indicate that there is a low probability that abdominal vault repair is a cost-effective use of resources when compared with vaginal vault repair. However, it should be noted that the findings of the trial-based economic evaluation only partially address the question of cost-effectiveness. To fully understand the cost-effectiveness of an intervention, it is important to capture the longer-term costs and effects of a surgical intervention. For example, it may not be until many years into the future when the true impact of failure rates and complications may be realised, and any cross-group differences observed. Therefore, it is planned to have long-term follow-up of women randomised in the Vault trial to better understand the long-term costs and outcomes of surgery. The Vault trial provides a rich source of data to populate the long-term economic model of the most cost-effective treatment pathway combining the findings of the Uterine and Vault trials into one single model of the treatment pathway. The analysis will be updated as more mature data become available.

Introduction and objective

Chapters 5 and 8 report the results of the within-trial economic evaluations for both the Uterine trial and the Vault trial over a 12-month time horizon. The data provide useful indications of short-term cost-effectiveness. Apical prolapse, however, is a chronic condition and the clinical effects of alternative surgical treatments may persist into the future or emerge over the longer term, well beyond the 12-month follow-up period of the VUE trial. Recent NICE guidance acknowledges that there is no conclusive evidence about the long-term clinical effectiveness or cost-effectiveness of different surgical treatments;20,21 hence, little is known about the most efficient treatment pathway. The VUE trial seeks to address this gap in the evidence base. Longer-term follow-up of women randomised to the VUE trial is ongoing. However, clinical decision-makers need best estimates of the long-term cost-effectiveness data to inform treatment decisions now. The objective of this chapter is to use a decision-analytic model to extrapolate the Uterine trial results over a lifetime horizon to determine the cost-effectiveness of uterine preservation versus vaginal hysterectomy for treating apical prolapse from a NHS perspective.

Markov structure and model description

A probabilistic Markov cohort model was developed using TreeAge Pro™ 2016 software (TreeAge Pro 2016, R1.0.; TreeAge Software, Inc., Williamstown, MA, USA) to represent the treatment pathway for women with apical prolapse. The model is structured to mimic progression through the pathway of care. A cohort of 1000 women, with an average age of 63 years (trial baseline population), suffering from symptomatic prolapse requiring surgical intervention enter the model at a treatment decision point (uterine preservation vs. vaginal hysterectomy). The women then follow a treatment pathway, based on risk of requiring further surgery. A maximum of four (vaginal hysterectomy followed by a maximum of three repeat procedures) or five (uterine preservation followed by a maximum of four repeat procedures, including vaginal hysterectomy) surgeries are modelled depending on the starting arm. Figure 14 describes the model structure for the uterine preservation arm.

FIGURE 14.

Markov model structure for treatment pathway.

The sequence of surgeries and transitions allowed in the model were developed in discussion with clinicians and trial collaborators. The model allows the cohort to have further surgery for prolapse in the following sequence: (1) uterine preservation, (2) vaginal hysterectomy, (3) vaginal vault, (4) abdominal vault and (5) repeat abdominal vault or other surgical procedure. The treatment pathway described in Figure 14 is identical for both model arms, the only difference being the number of potential surgical procedures.

After each surgical procedure, the cohort can enter into one of four mutually exclusive states:

1. Well after surgery – this reflects the proportion of the cohort in any given cycle who do not experience failures or complications requiring surgery, or die naturally. They may, however, still experience some prolapse-related symptoms or other complications (that do not require hospitalisation). Minor treatment costs may be incurred, such as physiotherapy, oestrogens, etc. Once entering this state, the cohort remains there over successive model cycles until they experience failure, complication or die.

2. Complications – the cohort may enter the complications state at any point in the model following surgery, reflecting on the fact that women may experience complication risks over both the short and the longer term following prolapse surgery. Complications have been classified in accordance with the trial reporting of SAEs. The complications state reflects major complications that require hospitalisation. The duration of the complications state is 6-month cycles, reflecting the average time required to recover from complications. The model allows the cohort to have multiple complications at any time following a surgical procedure. As the cohort progresses through successive surgeries for failures, each complication state relates to the most recent surgical procedure. For example, it is assumed that ‘complications following vaginal vault’ relate to the vaginal vault procedure, and not the index decision of uterine preservation versus vaginal hysterectomy.

3. Further surgery for failures – failure requiring a repeat surgical intervention triggers progression of the cohort along the sequence of surgeries described in Figure 14. A failure requiring surgical repair is considered to be any repeat prolapse surgery, whether it occurs in the same or a different compartment.

4. Long-term containment – the proportion of the cohort having progressed through all surgical options described and still experiencing symptoms indicative of treatment failure are assumed to be managed conservatively, and so do not receive further surgical intervention. Conservative management includes the use of pessaries, physiotherapy and regular outpatient consultation.

5. Long-term well – this state reflects the proportion of the cohort who have progressed through the full sequence of prolapse surgeries and are no longer experiencing symptoms indicative of a failure of the final procedure. In the absence of any data to confirm the proportion progressing from the final surgery state to either long-term containment or long-term well, clinical expert opinion is relied on (Dr Christine Hemming, University of Aberdeen, 2017, personal communication), assuming that approximately 75% of the cohort would have no further problematic symptoms over a 10-year period, with 25% continuing to be managed conservatively.

6. Dead – the modelled cohort may die at any point in the model, following female-specific general population all-cause mortality rates. 62 No additional surgical mortality rate is included because no women died during the surgical procedures in the VUE trial.

The cohort of 1000 women progress through the model health states in monthly Markov cycles following the sequence of surgical treatments and health-state transitions described above. A monthly cycle length has been chosen to reflect the time increments for which data regarding time of treatment failure and complications requiring surgery were available. Tunnel states were used to ensure that the proportion of the cohort in each surgery and complication state remain there for 6 monthly cycles to allow for sufficient time to recover from treatment, and experience a longer time period of lower utility (QoL). The model time horizon is 30 years’ duration (equivalent to 360 monthly cycles) in the base case, this is deemed sufficient to capture the important life-time costs and consequences of an average 63-year-old woman entering the model. Costs and QALYs occurring beyond the first year of the model are discounted at a rate of 3.5% per annum in line with current NICE guidelines. 63 All costs are reported in 2015–16 GBP and the analysis is conducted from the perspective of the UK NHS.

Model parameterisation: health-state transition probabilities

Survival analysis is used to estimate transition probabilities to further surgery and complication states using time-to-failure events observed in both the Uterine and the Vault trials. For example, failure and complications following uterine preservation are derived from the corresponding arm of the Uterine trial data. Similarly, transition probabilities following vaginal and abdominal vault procedures are derived from the corresponding arms of the Vault trial. As noted in the Introduction and objective, long-term follow-up is ongoing. Two-year follow-up data on complications and failures were available for 208 out of 562 (37%) participants randomised to the Uterine trial and 140 out of 208 (67%) participants randomised to the Vault trial. The available 2-year data supplement the existing 1-year follow-up data to generate a richer data set. The methods used to extrapolate data from both trials are similar.

Complications

The process for incorporating complications into the model was identical to that described for failures requiring further surgery. For the purposes of the transition probabilities, complications are defined here as any complication that required admission to hospital. Transitions to complications states were based on time since the most recent surgical procedure. For the Uterine trial, the adjusted hazard ratio for uterine preservation versus vaginal hysterectomy was 1.126 (95% CI 0.478 to 2.651); the unadjusted hazard ratio was 1.077 (95% CI 0.453 to 2.552). For the Vault trial, the hazard ratios for abdominal versus vaginal were – 0.689 for adjusted (95% CI 0.154 to 3.078) and – 0.885 for unadjusted (95% CI 0.342 to 2.294). Overall, there were no significant differences between any of the treatments in terms of complications requiring hospitalisation. CIs were wide, in part due to the relatively low number of complications in both trials.

Figure 16 presents the Kaplan–Meier curves for complications using all available 2-year follow-up data. As can be seen from the Kaplan–Meier curves and the estimated hazard ratios, there is no evidence to suggest any differences in time-to-complication events across the groups based on currently available data.

FIGURE 16.

Kaplan–Meier curves for time to complication requiring hospitalisation.

As with the analysis of time to failure, the time-to-complication event is based on Weibull regression models. A similar process for selecting the models was used. Details of the full set of models considered are available in Appendix 8, Tables 85–8. Uncertainty surrounding transition probabilities was again incorporated into the model by fitting multinormal distributions to the correlations between the relevant coefficients, obtained from the Cholesky decomposition matrix of the regression model.

Mortality parameters

When women move through the model, the chance that they might die is based on the annual rates of age-specific all-cause mortality for women (Office for National Statistics interim life tables). 62 As there were no intraoperative deaths reported during the trial, surgical mortality was not included in the model.

Model parameterisation: resource use and costs

Intervention delivery costs are based on the component costing approach used for the trial-based cost-effectiveness analysis. Sensitivity analysis considers the use of HRGs in the model given the sensitivity of findings in the trial analyses to the choice of intervention costing approach. Costs for the remaining health states (failures, complications and well) are obtained from the trial data. Definitions of failures and complications are as described in Complications for obtaining time-dependent state-transition probabilities. Women who were not categorised as having complications or failures were categorised as being in the ‘well’ health state. This does not mean that these women were asymptomatic, and so the costs of the ‘well’ state implicitly assumes a proportion of the cohort have costs of consultations with health professionals and treatments based on the data reported in the trials. To avoid double costing, all regression analyses were based on total costs over follow-up, less the initial intervention costs. GLM regression models with robust standard errors (SEs) were then used to determine the impact of health-state membership on costs. Health-state dummy variables were interacted with the randomised treatment group to obtain treatment-specific health-state costs. The marginal effects of the relevant coefficients and SEs (estimated using the delta method) were used to populate the model. Sensitivity analysis considers the impact on results of assuming that the health-state costs are not treatment specific, that is, calculating health-state costs based on all trial women, regardless of randomised treatment group. Costs obtained from the trial data are adjusted to reflect the monthly time cycle of the model. Exceptions to this include full costs applied for complication and failure states at the point of entry to capture the full resource impact of high-cost surgical/hospitalisation procedures.

All cost parameters are defined in the model as statistical distributions and are assumed to follow a gamma distribution. Table 36 reports the mean, SE (SD of the sampling distribution) and the alpha and beta parameters used to define the gamma distributions.

Model parameterisation: utilities

The process of obtaining health-state utilities mirrors that described for the cost parameters. Health-state membership was determined (failure, complication and well). The utility measure obtained closest to the time-to-complication/time-to-failure event in the trial was used to reflect the utility of that state specific to each woman in the trial. GLM regression models were used to obtain the effect of state membership on utility and interactions between state. Treatment dummies were used to determine treatment-specific health-state utilities for the base-case analysis. Sensitivity analysis explores the impact of using health-state utilities that are not treatment specific. All utilities are adjusted for age and gender (female)-specific general population norms in accordance with best practice guidance66 to allow the average utility of health states to fall as the cohort ages. The utilities accrued in each health state are adjusted to reflect the monthly time cycle of the model. All utility data are included in the model as statistical distributions. Uncertainty in utility parameters is characterised and incorporated by sampling from beta distributions for the utility of each modelled health state. Alpha and beta parameters are calculated using the method of the moments approach and the parameters of the distribution are presented in Table 37.

Assessment of cost-effectiveness

The model analysis was conducted using second-order Monte Carlo (probabilistic or second-order) simulation, with 1000 iterations. Each iteration resampled a value from each input parameter based on the defined distributions. The model was therefore fully probabilistic, with 1000 expected values of costs and QALYs generated for each treatment strategy from the sampling distributions. These cost and QALY values were combined into a single measure of efficiency and reported as incremental costs per QALY gained, commonly referred to as the ICER. Interventions reporting an ICER of < £20,000–30,000 per QALY gained are generally considered cost-effective. Interventions that generate cost savings and higher QALYs than a comparator are the dominant treatment and offer an even stronger case for cost-effectiveness. Conversely, interventions that cost more and generate fewer QALYs than a comparator are said to be dominated and are not considered a cost-effective use of resources.

The simulation approach allows fully probabilistic ICERs to be obtained for both the base case and all sensitivity analyses conducted. CEACs were generated by assessing the probabilistic NMB for each treatment strategy. The CEACs allow for calculation of the probability that each strategy is cost-effective at alternative threshold values of WTP for a QALY gained. Similarly, scatterplots of incremental costs and incremental QALYs were obtained for uterine preservation versus vaginal hysterectomy and report the results on the cost-effectiveness plane. Both CEACs and the scatterplots illustrate the uncertainty surrounding the optimal treatment strategy caused by the combined statistical variability in the model’s parameter estimates.

Deterministic sensitivity analysis

Cost-effectiveness acceptability curves illustrate sampling uncertainty, but do not address underlying uncertainty driven by choice of data, structural assumptions, methodological choices or heterogeneity across subgroups of the modelled cohorts. A range of deterministic sensitivity analyses were explored to address the main areas of uncertainty in the model and test the impact of important assumptions on the cost-effectiveness conclusions. Model parameters for all deterministic sensitivity analyses are sampled from the statistical distributions, ensuring that all presented ICERs are fully probabilistic. Sensitivity analyses were conducted to:

1. Explore uncertainty surrounding the choice of model utilities and costs. The base-case analysis uses treatment-specific utilities and costs for each health state in the model. These data are advantageous in that they preserve the impact of treatment on the costs and QoL in any given health state. However, for rare events, they are derived from trial data with small samples and mean values subject to rare but extremely severe health events. Sensitivity analysis explores the impact of using utilities and costs that are defined for each modelled health state, but not split according to randomised treatment in the VUE trial. This increases the sample for estimating costs and utilities, but removes any assumption about treatment-specific effects on these parameters for the model. The base-case analysis (treatment-specific utilities) assumes that baseline utility for the cohort may be different across groups. This could potentially limit generalisability to the wider population. Therefore, a further sensitivity analysis explores keeping treatment-specific utilities for health states, but ensuring that the baseline utility for the cohort is the same in both uterine preservation and vaginal hysterectomy (i.e. 0.760).

2. Determine the impact of intervention costing approach on cost-effectiveness. The trial-based analyses showed that estimates of cost-effectiveness for uterine preservation versus vaginal hysterectomy were highly sensitive to the costing approach used for the interventions. A similar sensitivity analysis is conducted for the model-based analysis to determine whether or not the implications of intervention costing approach remain an important determinant of cost-effectiveness in the longer term.

3. Determine the impact of heterogeneity in the cohort’s characteristics on cost-effectiveness. The mean age of the women in the Uterine trial was 63 years. Sensitivity analysis explores the implications for cost-effectiveness of varying the age of women at the point of entry to the model between 53 and 73 years.

4. Determine the impact of methodological uncertainty on the results. Long-term data on the time to failure and complications are lacking and so estimates have been developed for the model based on survival models, extrapolating short-term findings over the longer term. Although the models represent a good fit to the observed trial data and are clinically plausible, the models may nonetheless fail to accurately capture the true (unknown) long-term complication and failure rates of surgery. The impact of any modelling inaccuracies on cost-effectiveness is likely to be magnified over longer-term time horizons. Therefore, sensitivity analysis shortens the time horizon to 5, 10 and 20 years.

5. Determine the impact of methodological uncertainty surrounding the appropriate discount rate to apply to future costs and QALYs. Sensitivity analyses vary the discount rate between 0% and 6% in line with NICE best practice guidelines. 63

Cost-effectiveness results

Sensitivity analysis results

The results from all sensitivity analyses undertaken are reported in Table 38 alongside the base-case results. A full set of CEACs and incremental scatterplots for each analysis undertaken can be found in Appendix 8, Figures 45–68. In general, the base-case findings are robust to the sensitivity analyses around age adjustment of health-state utilities, discount rates, starting age and starting utility of the modelled cohort. As illustrated by sensitivity analyses 7–9, the uncertainty in cost-effectiveness increases over longer time horizons, when projections of the trade-offs between complications and failures become less clear. For all the time horizons considered, the model shows vaginal hysterectomy to be the most cost-effective strategy.

The cost projections from the model remain somewhat sensitive to the choice of intervention costing approach. These results echo the findings of the trial-based economic evaluation, indicating that the methodology used to assign initial treatment costs is important when projecting total costs. However, the use of HRG costing does not change the conclusions about the most cost-effective treatment option unless society were to place zero value on 1 QALY. Under the HRG costing approach, uterine preservation becomes marginally less costly, but at the expense of QALY losses. Therefore, if society would require compensation of at least £20,000–30,000 for 1 QALY lost, it is unlikely that this analysis would change the cost-effectiveness conclusions.

The model is sensitive to the choice of costs and utilities used to populate the health states, in particular the ‘well’ health state. Treatment-specific costs and utilities both tend to favour vaginal hysterectomy, whereas using averages from the Uterine trial, as a whole, favours uterine preservation. The finding is driven primarily by the treatment-specific utilities and costs for the proportion of the cohort who are in the ‘well’ state. The regression analyses indicate slightly lower costs in the ‘well’ state following vaginal hysterectomy and higher utility compared with vaginal preservation. Turning off treatment-specific costs and utilities removes this difference between the groups. As failure and complication rates are low and transient in the model (state duration for failure and complications is 6 months), the majority of the cohort spend most of the model in some form of ‘well’ state, hence the sensitivity of the results to this assumption.

Interpretation of the model results

The model describes the important long-term trade-offs between the choice of uterine preservation versus vaginal hysterectomy. The model is populated with the best available data on failures and complications following surgery. However, the maximum follow-up to date of 2 years means data are immature, and it is unlikely that all the important trade-offs between failures and complications have been observed across the groups. It is therefore crucial that the model analyses are updated once more mature data on time-to-failure and complication events become available. From a decision-making point of view, the ICER is therefore sensitive to the true long-term failure and complication rates, and in particular cross-group differences. In addition, the model describes a set of scenarios and assumptions that favours uterine preservation and vaginal hysterectomy, respectively.

Within the modelling analysis and structure of the care pathway, advantages of uterine preservation include:

• Keeping the cohort out of a final health state in which no other options are available apart from conservative management and long-term containment for longer. This means that the cohort avoid long-term costs of containment and the QoL implications of having no other potentially curative surgical options available.

• The use of HRG intervention costing. The implications of HRG compared with component costing have been discussed in Chapters 4 and 7.

Issues favouring the cost-effectiveness of vaginal hysterectomy include:

• Lower intervention costs, based on a shorter surgical procedure, using the component-based costing approach described in Chapter 2.

• Potentially lower short-term failure rates. Although differences in time to event are not statistically significant, point estimates of the hazard ratios favouring hysterectomy are propagated through the model. Fewer failures mean a lower likelihood of costly surgical procedures and the QoL benefits of experiencing a failure later than with uterine preservation.

Trial recruitment and treatment choice

Around 50% of parous women report symptoms of POP8,72 and have a lifetime risk of between 6% and 20% of undergoing surgery for prolapse,3,5,16,17 this is dependent on symptoms, QoL and functional impact.

Women within the VUE trial had requested surgical treatment for different reasons; the most bothersome symptom was ‘a feeling of something coming down’ (see Chapters 4 and 7). This symptom has been found to have a sensitivity of 84% and a specificity of 94% for prolapse at, or beyond, the hymenal ring. 73 Other symptoms prompted women to consider surgery, such as faecal incontinence (more common in the Vault trial), dragging in the abdomen or back or incomplete bladder emptying, although there is no clinical evidence that confirms that these symptoms have any direct relationship attributable to prolapse74 and surgery may potentially increase the risk.

A considerable number of eligible women chose not to participate in the Uterine trial during the initial planned recruitment time, which affected recruitment rate, and a 15-month recruitment extension was required. The main reason given to recruitment staff for non-participation was of the women’s wish to have their uterus removed (see Chapter 3). This finding suggests the complex nature of the decision-making process and it would be interesting to further explore reasons through a qualitative study (see Appendix 10) given that previous literature suggests that 31–60% would choose uterine preservation. 25,75,76 The recruitment rate for the Vault trial was not affected (see Chapter 6).

During the recruitment phase of the VUE trial, debate on the use of mesh implants in POP repair surgery intensified. The potential insertion of mesh, though not necessarily transvaginally, was an important decision-making point for many women. However, some women were unaware of the publicity and their knowledge depended on their own social circle. This also highlights, and reinforces, the need for high-quality standardised patient information provision on the options available for women before surgery, to allow informed decision-making and person-centred care. 41,42,77 In both the Uterine and the Vault trials, two women had a mesh exposure from 235 mesh procedures at 12 months after surgery, and both underwent surgical mesh removal. This provides a mesh exposure rate per procedure of 0.9% for the first 12 months. One of the two mesh exposures was due to a transvaginal concomitant procedure. Mesh exposure rates have been considered and reported as much greater for transvaginal mesh41,42,77 and further monitoring of the VUE trial participants is taking place to identify if the natural timeline when using mesh for apical surgery is different given the different approach. The exposure rates may also be lower than national statistical data, as the surgeons involved in the VUE trial are recognised as being highly skilled; therefore, the study rates may not accurately reflect actual rates.

Clinical findings before operation

The POP-Q scoring system for objectively quantifying prolapse is considered highly reliable for research purposes, though whether or not apical prolapse should use the same standards as anterior and posterior prolapse has been debated. It has been suggested that the scoring system does not allow sufficient detail for apical descent, such as in women with isolated cervical elongation or a large anterior wall prolapse that obscures the vault location. 78–81 At present it remains unchanged,82 as there is no unanimous view and no alternative to replace the present system. In the VUE trial, a number of women had minimal apical descent but were included in the trial (see Chapters 3 and 6), suggesting that there may be some women for whom scoring by POP-Q has been less accurate. 82 Of note, women in the Vault trial were found to have higher POP-SSs (i.e. symptoms were more bothersome), but their POP-Q scores were not necessarily as severe as those women in the Uterine trial (see Chapters 3 and 6). This may suggest a reduced tolerance to prolapse or functional symptoms, or perhaps this cohort actually suffers more from prolapse as a result of the absent uterus.

In addition to apical descent in the VUE trial population, anterior compartment descent was also present and not only led to more planned anterior repairs preoperatively, but may have also contributed to increases in baseline POP-SSs (see Chapters 3 and 6). This anatomical coexistence has been recognised previously in studies and may possibly be caused by pelvic floor dynamics, irrespective of uterine presence. However, it is possible that anterior prolapse may have been over diagnosed, particularly in those with vault prolapse, where the apical and anterior compartment are less well defined,83,84 and highlights the complexity and outpatient limitations of specific compartment preoperative surgical assessment and planning.

Some POP-Q anterior compartment and apical scoring (C+/–D; dependent on uterine presence) may have related to observer error, though the majority of clinicians had participated in the previous RCT, PROSPECT,1 and were considered experienced in the scoring process.

Diurnal variation of the prolapse presentation and bowel/bladder distension may have also affected vaginal compartments and the overall POP-Q scores.

The findings suggest that objective clinical assessment of prolapse by POP-Q is not necessarily the highest priority for surgical decision-making given that women who had low POP-Q scores did not consistently have reduced POP-SSs, and this has been reported elsewhere. 85

Overall, this preoperative clinical process and subsequent randomisation of women in the VUE trial aligns with the pragmatism of clinical day-to-day practice, enabling better insight into surgical practice for this cohort of women.

Surgical procedures and their differences

The majority of women did receive their randomly allocated surgery, though a number did not (see Chapters 4 and 7). It is possible that factors, such as lifestyle change, weight loss, oestrogen treatment and physiotherapy, could have benefited some women, causing them to change their mind about the need for surgery. It is also possible that the presence of a uterus may improve results from these conservative approaches. 86

Some women also received different surgery to that proposed preoperatively after examination under anaesthetic, suggesting that a preoperative POP-Q assessment may be useful and cost-effective if symptoms or significant lifestyle changes have occurred since surgical listing. This pre-emptive action may be useful for women whose initial examination was by a less experienced clinician. It is also possible that assessment under anaesthetic has allowed a more thorough evaluation of the vaginal compartments, but may also overestimate the degree of prolapse as the pelvic floor is paralysed. This reaffirms the need for comprehensive preoperative counselling and consent (both verbal and written) in relation to the potential change in surgical plans under anaesthesia. 39

Although both the Uterine and the Vault trials were evaluating different approaches to apical repair, associated factors, such as the use of different types of anaesthesia, are also likely to have influenced final outcome measures or surgical decisions within theatre. Procedures undertaken by a vaginal approach were more likely to involve a spinal or regional anaesthetic (see Chapters 4 and 7). Abdominal procedures were more likely to require a general anaesthetic, particularly so in those receiving laparoscopic procedures when respiratory and operating position requirements are important. Overall, abdominal procedures (open/laparoscopic) for both trials were significantly longer than vaginal procedures (see Chapters 4 and 7), and given that prolapse surgery affects the more elderly population, this approach may have exposed women to an increased anaesthetic risk (particularly in those with associated morbidity). This was not identified within the VUE trial, but the relatively healthy women in the trial and self-selected group of participating surgeons may not be representative to provide reassurance for more widespread use.

Recently, robotic laparoscopic surgery has been considered for abdominal pelvic floor procedures, although small RCTs to date have indicated that the operation is even longer and operating costs are higher. 87 These early findings are unlikely to favour its general clinical use in the current NHS and further evaluation is required.

As agreed in the VUE trial protocol,31 synthetic mesh was utilised in some of the transvaginal procedures, whether for concomitant surgery or the apical compartment. Transvaginal mesh use changed during the lifetime of the VUE trial, with a reduction in use for prolapse surgery and a consistent number (though small) for incontinence. The ongoing debate on the use of mesh implants and disseminated publications from national and professional organisations during the time of the trial was likely to be a causal factor affecting women’s and surgeons’ choice of surgical approach. 88 This trend is also in line with a recent UK national prolapse survey, and at present the use of transvaginal mesh has been paused by NHS England and NHS Northern Ireland. 23,24,27–29,77,89–91

Mesh implants were required in all abdominal procedures (see Appendices 3 and 5), though information on specific anatomical site mesh application was not requested/retrieved on the intraoperative CRF. As a result, it is not possible to identify those surgeons who indirectly or specifically repaired the anterior compartment either by an additional anterior leaf of mesh or by their individual technique during abdominal surgery. This does, however, emphasise the pragmatic approach of the variability in surgical practice in relation to mesh position for individual patients.

More concomitant anterior and posterior repairs were undertaken in the vaginal groups for both trials (see Chapters 4 and 7). This may be related to the local surgical field given the good visibility of other compartmental descent compared with abdominal procedures, and its outcome is discussed in The VUE trial of the future.

Overall, intraoperative complications and subsequent SAE rates in the VUE trial, irrespective of surgical approach, were lower than previously published rates for pelvic floor surgery. 92 This finding may be because of the surgical experience in the self-selected cohort of participating gynaecologists in the trial, as described in Trial recruitment and treatment choice.

Women’s symptoms after operation

Prolapse symptoms outcomes at 6 months after surgery, at 12 months after randomisation, with a clinic appointment for POP-Q assessment 12 months after surgery were collected (see Chapters 4 and 7).

The use of this method of data collection may be considered by some clinical investigators as controversial, but has robust statistical support and is commonly used in drug trials (Clinical Trials of an Investigational Medicinal Product). By data retrieval at 6 months after surgery, the POP-SS is very specific and aligned to time allocation for surgery. However, in routine clinical practice, women do undergo a waiting time before surgery, hence the outcome data at 12 months after randomisation is equivalent to a patient’s normal clinical pathway. This post-randomisation analysis at 12 months offers the expectations in improvement 12 months after treatment allocation occurred. The 12 months after surgery clinical assessment enables the surgeon to assess objective changes after the procedure. This time differential (c. 60 days) will become irrelevant for outcome measures over the longer term, which will be critical in informing health-care commissioners, health-care professionals and consumers to provide estimates of treatment effects currently not available in the literature.

Financial considerations for all

Women who were in paid employment and required time off work for surgery were found to require longer convalescence in the uterine preservation group in the Uterine trial (see Chapter 5). This may affect future generations in relation to surgery options given the increasing pensionable age and employment legislation, particularly in relation to unpaid leave, and will undoubtedly become integral in decision-making for the women themselves.

At 12 months, the randomised surgical approaches for each trial were found to be clinically effective in the relief of symptoms and objective surgical outcome. However, in both trials the short-term cost-effective impact and the health economic findings suggest that vaginal surgical approaches for apical prolapse may be more beneficial (see Chapters 5 and 8). The differential was based on additional re-operation rates for uterine preservation cohorts in the Uterine trial as well as longer operating times. The differential was more evident in the Vault trial when direct comparison of abdominal and vaginal approaches were made. This finding may suggest that current HRG classification for health-care costs in terms of pelvic floor surgery may require more frequent revision and amendment to accurately reflect costs and the rate of change in surgical techniques offered by clinicians. Regular long-term follow-up of women in both trials, particularly in relation to health economics, will enable more accurate costings for the present apical surgical techniques. This will identify if the predictions presented in this publication hold true for the long-term future.

The VUE trial of the future

Currently, the women participating in the VUE trial are being followed up for 6 years after randomisation, though this time frame does not provide a thorough evaluation of long-term outcomes given the expected increase in lifespan. A follow-up of up to 12 years is hoped to be achieved to enable longer-term evaluation of outcomes. This would include women who did not receive any form of surgery for prolapse.

Women participating in the VUE trial were generally in their seventh decade, with a slightly older cohort in the Vault trial (see Chapters 3 and 6). Longer-term follow-up may be more difficult because of age-related diseases and/or death. Alternative methods for data collection and outcome measures require recognition of abilities and cognitive function (i.e. age specific) over time.

Although there was no difference in POP-Q scores in the Uterine trial between the two groups, in the Vault trial there was a statistical difference in relation to the anterior wall prolapse after a vaginal vault repair (although the numbers were small). Anatomical correction does not guarantee functional improvement in neighbouring systems, with the risk of deterioration, such as de novo urinary incontinence or faecal incontinence, as was evident in both trials (see Chapters 4 and 7). Women participating in the VUE trial did experience bowel and bladder symptoms that required hygiene measures after their operation despite improved POP-Q scores (see Chapters 4 and 7). The cause of bowel and bladder dysfunction is a complex issue, but the resultant effects on QoL and finance for an elderly woman are substantial. This emphasises the need for assessment of potential short-, medium- and long-term changes made by different surgical techniques given the expected overall age-related hand–eye co-ordination, mobility changes and potential life expectancy of women in the future.

Preservation of the uterus is not without risk. Long-term information about additional investigations and treatments required for those women who had uterine preservation will be particularly important and may require analysis taking into account actual treatment received for accurate risk information. This would also enable the additional personal, financial and health impact to be scrutinised rather than prolapse symptomatology in isolation as an outcome measure. This is likely to be considered an important measure for health-care providers, clinicians and women alike.

The majority of women within the VUE trial were postmenopausal and results are generalisable for this cohort of women. It is somewhat difficult to extrapolate these outcomes for women under the age of 45 years who may also present with an apical prolapse that remains unresponsive to conservative treatments. These women are more likely to be fertile and require cervical screening, both of which may be unacceptable if uterine removal is an option. Uterine preservation prolapse surgery is possible and used presently for those women pursuing fertility preservation and future pregnancy (although excluded from the VUE trial). Although comprehensive adequately powered long-term data will not become available from the VUE trial for this age group, subanalysis of those women randomised to uterine preservation who were under 45 years of age may assist in information on its benefits or otherwise for apical prolapse.

Limitations

Although few participating women were lost to follow-up overall in both the Uterine and the Vault trials, the proportion of women who were successfully included was < 50% of those initially identified. Although not all of these women would have been eligible for the VUE trial, it may have allowed for an increased generalisability.

The surgeons who participated in the VUE trial chose the uterine preservation procedure they were most familiar with, which, although allowed pragmatic practice, did allow for greater heterogeneity in the uterine preservation arm of the Uterine trial. However, some highly skilled laparoscopic surgeons familiar with the VUE trial surgical procedures did not take part and may have influenced the uterine preservation cohort outcomes in relation to health economic and clinical outcomes. There were several reasons for this: surgical bias to particular procedures, site difficulties with timing of ethics approval, and research and development processes.

A small number of participating women did not undergo surgery and this cohort was not offered a 12-month post-surgical examination for POP-Q evaluation. The women were, however, included in the POP-SS outcomes at 12 months. The retrieval of POP-Q scores from these women who had no surgery may have impacted the POP-Q outcomes and also provided some information on the natural progression or regression of prolapse in women with apical prolapse.

The questionnaire response rate and that of clinic attendance have also identified a greater clinical attendance than questionnaire return. This suggests that women may place greater importance on objective assessment of prolapse by a health-care professional than their symptoms through questionnaire completion. However, it is possible women do not understand the importance of their feedback through serial structured questionnaire completion over time and this requires further evaluation.

Health-care costs were greater for abdominal procedures; however, the cost may still be artificially lower than the actual costings given that surgical equipment costs within each trial were assumed to be equivalent, and this may not be the case if disposable instruments were used. This information was not collected in the VUE trial other than additional costings for mesh implantation and should be considered for any further trials in the future. In addition, there was a greater exposure to health-care professionals after an operation if women had undergone uterine preservation; this may be artificially high at present given the limited experience of primary care with these procedures and the requirement of more specialist advice. If this hypothesis is true, then a trend to less health-care input will occur. Further investigation into primary health-care behaviours in relation to new interventions may be useful to inform a health-care economic analysis.

The blinding of participants was particularly difficult given the different surgical approaches and the required counselling before the operation. Blinding participants to their surgical procedure by performing sham incisions to conceal the route of surgery was not considered necessary or ethical. This would have caused falsely increased operating times, pain relief and recovery times in those having vaginal operations and unlikely to have had an impact on women’s 12-month primary outcomes. However, it was more useful to blind the POP-Q assessor at the 12-month examination, although this was sometimes difficult if the assessor had also been the surgeon or if the patient divulged the procedure before examination.

Future generations

In conclusion, this publication reports primarily on a postmenopausal population of women with apical prolapse who experienced different obstetric care from today (primarily in terms of fewer caesarean sections and forceps deliveries). The change in obstetric practice, in addition to societal changes tending towards smaller families, has a potential to reduce the pelvic floor insult associated with childbirth. 93 Linkage to increases in caesarean section rate and reduced instrumental vaginal delivery has demonstrated a reduction in lifetime risk for prolapse surgery. 67,71 In addition, the impact from reduced hysterectomy rates through changes in gynaecological practice may also improve pelvic floor health. 94–97 Over time it is therefore possible that prolapse requiring surgical treatment may be significantly reduced or delayed for several years in life. This may limit the opportunity for further large RCTs for both uterine and vault prolapse in the future because of a potential reduction in the number of women requiring treatment. However, this hypothesis does not take into account the increasing incidence of obesity worldwide, which may also affect prolapse98 and its related symptoms. 99 Further trials for assessment of uterine preservation for prolapse may also not be possible over time given the increased risk of gynaecological cancers related to obesity.

With the increasing information on the impact of mesh-related procedures worldwide, the use of synthetic mesh implants must also be scrutinised in the long term. The use of mesh and its positioning is particularly important in relation to all structures within the pelvis and abdomen. Subsequent abdominal or pelvic surgery may become more difficult if adherent mesh from a prolapse repair is nearby to the operating field, causing increased open procedures and increased lengths of hospital stay. This must also be considered and reviewed in the long term given the increasing longevity of human life.

Although evidence to date has not been favourable for biological graft support in pelvic floor surgery, some trials have shown promise. 100–103 It is therefore possible that the present synthetic mesh may also be further adapted or totally replaced by more advanced biological graft material for prolapse surgery to allow clinically effective and potentially safer treatments. This research and development will undoubtedly take many years and high resources and may await studies, such as the VUE trial, to identify their role, if any, before manufacture.

Conclusion

There was no evidence that an abdominal procedure was more clinically effective for either a uterine or vault prolapse in terms of all measured outcomes in the short term, but the abdominal procedures were slightly more expensive to perform. To identify if predictions described in this publication hold true (see Chapter 9 for long-term predictions over the next 30 years in relation to any benefit difference), long-term follow-up of the VUE trial participants will provide a unique opportunity to collect long-term information on the most clinically effective, safe and cost-effective treatment for apical prolapse.

Members of the Project Management Group

Dwayne Boyers, Suzanne Breeman, Kevin Cooper, Lynda Constable, Janice Cruden, Andrew Elders, Mark Forrest, Robert Freeman, Cathryn Glazener, Beatriz Goulao, Suzanne Hagen, Christine Hemming, Mary Kilonzo, Graeme MacLennan, Alison McDonald, Isobel Montgomery, John Norrie and Anthony Smith.

Independent members of the (1) Trial Steering Group and (2) Data Monitoring Committee

1. Tony Falconer (Chairperson), Pamela Warner, Trish Emerson (Patient Representative) and Ian Ramsay.

2. Jim Thornton (Chairperson), Angela Crook, Lucia Dolan, Gill Gyte (Patient Representative) and James Neilson.

Members of the VUE trial group responsible for recruitment in the clinical centres

Contributions of authors

Christine Hemming [Consultant Gynaecologist and Chief Investigator (April 2016 to present), NHS Grampian and University of Aberdeen] contributed to the conception and the design of the trial, recruitment of participants, the conduct of the trial, the interpretation of the results and the writing/editing of the report.

Lynda Constable [Trial Manager, Co-chief Investigator (April 2016 to present)] was responsible for the day-to-day management of the trial, the interpretation of the results and the writing/editing of the report.

Beatriz Goulao (Statistician) conducted the statistical analyses, contributed to drafting the report and reviewed the final report.

Mary Kilonzo (Health Economist) contributed to the conception and the design of the trial, conducted the analysis of the health economics data and the writing of the health economics chapters, and commented on other chapters of the report.

Dwayne Boyers (Health Economist) conducted the analysis of the health economics data and the writing of the health economics chapters and commented on other chapters of the report.

Andrew Elders (Statistician) contributed to the conception and the design of the trial, oversaw the statistical analyses and provided commentary on the draft chapters of the report.

Kevin Cooper (Consultant Obstetrician and Gynaecologist, NHS Grampian) contributed to the conception and design of the trial, the recruitment of participants and the interpretation of the data, and provided commentary on the draft chapters of the report.

Anthony Smith (Consultant Gynaecologist) contributed to the conception and design of the study, the recruitment of participants and the interpretation of the data, and provided commentary on the draft chapters of the report.

Robert Freeman (Consultant Gynaecologist) contributed to the conception and design of the study, the recruitment of participants and the interpretation of the data, and provided commentary on the draft chapters of the report.

Suzanne Breeman (Trial Manager, Triallist) contributed to the conception, design and conduct of the trial, and provided commentary on the draft chapters of the report.

Alison McDonald (Senior Trial Manager, Triallist) contributed to the conception, design and conduct of the trial, and provided commentary on the draft chapters of the report.

Suzanne Hagen (Professor of Statistics and Deputy Director of the Nursing, Midwifery and Allied Health Professionals Research Unit) contributed to the conception and the design of the trial, and provided commentary on the draft chapters of the report.

Isobel Montgomery (Patient Representative) contributed to the conception and the design of the trial, participated in trial meetings from the perspective of a patient who has undergone prolapse surgery and reviewed the report.

John Norrie [Professor of Statistics and Director of CHaRT (to 2016)] contributed to the conception, design and conduct of the trial, and provided commentary on the draft chapters of the report.

Cathryn Glazener [Chief Investigator (to April 2016) and Emeritus Professor] contributed to the conception, design and conduct of the trial, the interpretation of the results and the writing/editing of the report.

Data-sharing statement

All data requests should be submitted to the corresponding author for consideration. Access to anonymised data may be granted following review.

Publication

Glazener C, Constable L, Hemming C, Breeman S, Elders A, Cooper K, et al. Two parallel, pragmatic, UK multicentre, randomised controlled trials comparing surgical options for upper compartment (vault or uterine) pelvic organ prolapse (the VUE Study): study protocol for a randomised controlled trial. Trials 2016;17:441.

Patient data

This work uses data provided by patients and collected by the NHS as part of their care and support. Using patient data is vital to improve health and care for everyone. There is huge potential to make better use of information from people’s patient records, to understand more about disease, develop new treatments, monitor safety, and plan NHS services. Patient data should be kept safe and secure, to protect everyone’s privacy, and it’s important that there are safeguards to make sure that it is stored and used responsibly. Everyone should be able to find out about how patient data are used. #datasaveslives You can find out more about the background to this citation here: https://understandingpatientdata.org.uk/data-citation.

Disclaimers

This report presents independent research funded by the National Institute for Health Research (NIHR). The views and opinions expressed by authors in this publication are those of the authors and do not necessarily reflect those of the NHS, the NIHR, NETSCC, the HTA programme or the Department of Health and Social Care. If there are verbatim quotations included in this publication the views and opinions expressed by the interviewees are those of the interviewees and do not necessarily reflect those of the authors, those of the NHS, the NIHR, NETSCC, the HTA programme or the Department of Health and Social Care.

Final model (Uterine and Vault trials)

The Uterine and Vault trials had the same primary outcome and were analysed using the same statistical model. The notation followed by the model, used in both trials, is presented in Table 40.

The final model is the following:

The surgeon-level random effects b_i is assumed to be normally distributed with mean 0 and variance σA2, i.e. b_i ∼ N(0, σA2). The error term was assumed to be:

Unit costs for trial-based economic evaluation