Endometrial scratch to increase live birth rates in women undergoing first-time in vitro fertilisation: RCT and systematic review

Burden of fertility treatments

In vitro fertilisation (IVF) is an assisted reproductive technology widely used in women unable to reproduce naturally. IVF (in which the sperm and egg are mixed together in a Petri dish) can be undertaken with or without intracytoplasmic sperm injection (ICSI), in which the sperm is directly placed inside the egg; from hereon we refer to IVF, with or without ICSI, as ‘IVF’. In 2018, about 54,000 patients in the UK underwent IVF. 1 Success rates are modest, but have been increasing over recent years, with an overall live birth rate (LBR) of 23% in the UK in 2018. 1 Undergoing IVF takes an emotional and financial toll on couples,2 causing absence from work,3 feelings of distress, lack of control and stigmatisation,4 with repeated failure sometimes invoking subclinical emotional problems. 5 When treatment fails, there is often a desire to try one of a bewildering selection of expensive IVF ‘add-ons’ in order to improve the chances of success,6 despite a lack of high-quality evidence to support their benefits or safety. 2,7,8 One such add-on is the use of local endometrial trauma, also known as endometrial scratch (ES), which can cost up to £325 per procedure and involves ‘scratching’ the lining of the womb (the endometrium) with a pipelle or similar device. 8

Evidence for endometrial scratch

The use of ES to improve implantation rates in women undergoing assisted conception was first described in 2003. 9 Initially, ES was reported to positively affect the outcome of IVF in women with repeated implantation failure. 10 This was later also demonstrated in unselected populations (i.e. women who had undergone any number of previous IVF cycles),11 including in a 2015 Cochrane review,12 which reported a meta-analysis of randomised controlled trials (RCTs), finding that ES improves the chances of a live birth when undertaken in this unselected population [relative risk (RR) 1.37, 95% confidence interval (CI) 1.05 to 1.79]. However, a more recent systematic review has included subsequent RCTs and highlighted that the evidence for ES improving LBR in this population is unclear (RR of 0.95, 95% CI 0.64 to 1.41). 13 Subsequently, a large high-quality RCT by Lensen et al. 14 in an unselected population found that ES did not result in a higher LBR than ‘usual’ IVF. Despite this, the pros and cons of undertaking ES in an unselected population have been discussed since this publication, with proponents of the procedure arguing that it has low costs and a low level of harms, which, coupled with the possible benefits of taking a tissue sample at the same time as ES, warrant its use in selected populations. 11

The chances of achieving a live birth decrease with successive cycles of IVF; therefore, the aforementioned evidence from unselected heterogeneous populations, in many cases including women who had undergone a previous IVF cycle, are not necessarily generalisable to the first-cycle population. 15 One systematic review by Vitagliano et al. 16 specifically focused on patients undergoing their first IVF cycle and included seven RCTs; it found that there was no evidence that ES followed by IVF, compared with IVF alone, increased the success of treatment, with a RR of live birth (or ongoing pregnancy if LBR was not reported) of 0.99 (95% CI 0.57 to 1.73; p = 0.97). Secondary outcomes (miscarriage, multiple pregnancy and ectopic pregnancy) were also not significantly altered by undertaking ES. Notably, the small sample sizes of the included studies resulted in uncertainty around the effects of ES in women undergoing their first IVF cycle, and the included trials were at either a high or an unclear risk of bias. Consequently, the authors concluded that a robust and definitive trial is required to assess the effect of ES on the chances of success of the first IVF cycle. 16

Definitive studies of IVF ‘add-ons’ are challenging, as large numbers of patients are required to detect changes in LBRs with sufficient power, therefore making them expensive. 2 Consequently, studies in this first-cycle population have been small and have included heterogeneous populations. 16 To date, four RCTs have specifically focused on assessing ES in women undergoing their first cycle of IVF,17–20 all of which were included in the review undertaken by Vitagliano et al. 16 The two largest trials included 418 and 300 participants, and found significant increases in IVF success in women who received ES;19,20 however, according to the Vitagliano et al. 16 review, in one study20 the risk of bias was deemed to be high and in the other participants were not followed up until delivery. 19 Previous studies have also differed in the timing of the ES procedure, with Karimzade et al. 17 undertaking ES on the day of oocyte retrieval, and the majority of other studies undertaking ES in the menstrual cycle prior to the cycle in which embryo transfer was carried out. Notably, Karimzade et al. 17 identified that undertaking ES on the day of oocyte retrieval (compared with in the menstrual cycle prior to the cycle in which embryo transfer was carried out) resulted in a reduction in clinical pregnancy rates [12.3% vs. 32.9%, respectively; odds ratio (OR) 0.25, 95% CI 0.12 to 0.66; p < 0.004].

Three RCTs published since the Vitgaliano et al. 16 review included heterogeneous unselected populations but were not adequately powered to detect clinically important differences in women undergoing their first cycle. 14,21,22 In all three trials, outcomes for women undergoing their first cycle were not presented. However, for the unselected population, all three trials identified no significant differences in pregnancy rates or LBRs between women who received ES and those who received ‘usual’ IVF. 14,21,22 Of note, one of these trials stopped early following an unplanned futility analysis, as the clinical pregnancy rates between the ES and control (sham procedure followed by ‘usual’ IVF) groups were similar. 21

The Human Fertilisation and Embryology Authority (HFEA) has set up a ‘traffic-light’ system to inform patients of the evidence supporting a selection of ‘add-ons’ that fertility units frequently offer and charge extra for – according to HFEA, the evidence to support the benefits of ES is currently unclear. 2

Costs to the NHS and patients

Fertility treatment places a significant cost burden on the NHS and patients, with the cost to the NHS estimated to be around £68M per year (based on 2013 costs). 23 The number of first IVF cycles funded by the NHS has been decreasing over recent years; 58% of first cycles were funded by the NHS in 2017, reducing to 52% in 2018. 1 Given that the average patient undergoes three cycles of IVF,1 and a single IVF cycle can cost £5000, the cost of IVF treatment is a significant burden to patients. 8 In addition, loss of productivity due to the emotional and physical effects of IVF treatment has been estimated at £540 (€596, converted in 2020) per woman per IVF cycle in one study. 3 The cost of IVF is increased further if the couple decides to undergo the ES procedure, despite a lack of evidence to suggest that it is effective or safe. ES is estimated to cost £325 per cycle, a cost that does not take into account the potential pain of receiving the procedure and the potential health-care contacts if the procedure is in fact unsafe. Therefore, any increase in LBRs in the first cycle of IVF will prevent patients from receiving subsequent IVF cycles, thereby reducing this burden.

The endometrial scratch intervention

Endometrial scratch is a routinely performed outpatient procedure in which a pipelle or similar device is inserted into the cavity of the uterus; negative pressure is applied by withdrawing the plunger and the sampler is rotated and withdrawn three or four times so that tissue appears in the transparent tube. The rotation and withdrawal of the plunger ‘scratches’ the endometrium. Risks were identified in a previous study when the procedure was undertaken on the day of oocyte retrieval. 17 The procedure is not known to be associated with any particular risks when undertaken in the menstrual cycle preceding that of IVF therapy, apart from pain during and shortly after the procedure. 12 As with any intrauterine procedure, there is a potential for intrauterine infection. However, women attending for fertility treatment are usually screened for serious vaginal infections, such as chlamydia, to minimise the risk of any spread of infection.

The exact mechanism by which ES may improve implantation is not yet known; however, it is known that implantation of the embryo is a complex process involving the release of several inflammatory mediators including uterine natural killer cells, leukaemia inhibitory factor and interleukin 15. 24 ES may lead to the release of inflammatory cells and mediators such as macrophages and dendritic cells, tumour necrosis factor alpha, interleukin 15, growth-regulated oncogene-α and macrophage inflammatory protein 1B. 25 ES has also been shown to cause the modulation of several endometrial genes that may be involved in membrane stability during the process of implantation, such as bladder transmembranal protein and adipose differentiation-related protein and mucin 1. 26

Internal pilot summary

The main trial was designed with an internal pilot phase with stop–go criteria applied after 6 months of recruitment to assess the feasibility of recruitment and delivery of the ES procedure as per the protocol. The trial met the recruitment and intervention delivery outcomes (see Appendix 2), and therefore the trial continued as planned without any modifications to the trial protocol. Participants from the internal pilot phase were included in the final analysis. Figure 1 is a flow chart of the study design.

FIGURE 1.

Flow chart of the study design. ±, with or without; AEs, adverse events; SAEs, serious adverse events.

Trial registration and ethics

The trial was registered on the ISRCTN registry (ISRCTN23800982) on 31 May 2016. The trial protocol was reviewed and ethics approval was granted by the National Research Ethics Service Berkshire South Central (reference 16/SC/0151). 34 Over the course of the trial, eight amendments were made to the protocol (consisting of seven substantial and one non-substantial amendment), all of which were also approved by the National Research Ethics Service Berkshire South Central (as detailed in Appendix 3).

Participant inclusion and exclusion criteria

Patients who met the following criteria were eligible for the trial. Parts of the text have been reproduced with permission from Pye et al. 35 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: https://creativecommons.org/licenses/by/4.0/. The text below includes minor additions and formatting changes to the original text.

Settings and locations where participants were recruited

During the internal pilot trial, recruitment took place at six IVF fertility units across England. In the main trial, recruitment sites were expanded to include an additional 10 IVF fertility units across England and Scotland. Two of these 16 fertility units were privately run outside the NHS. Appendix 4 provides a list of sites and their involvement during and after the internal pilot phase.

Identification, screening and first approach

Participants, who were attending the IVF clinic prior to starting their first IVF cycle were identified by site staff [i.e. research nurses (RNs) and clinical trials assistants], who screened clinic lists to identify potential participants for the trial. Those who were identified and met the ‘screening criteria’ (aged 18–37 years inclusive, first IVF cycle, ovulatory menstrual cycle) were added to the screening log. A full ‘eligibility’ check was then undertaken, following consultation with the participants within the clinic, to confirm the remaining eligibility criteria.

Potential participants were approached about the trial by post by including an invitation to participate along with their initial clinic appointment letters, during routine clinic appointments before the start of their IVF or at patient information sessions, or they self-identified themselves after becoming aware of the trial from the trial website, recruitment video or posters that were placed within IVF units. Potential participants were approached at any time prior to the start of their IVF cycle, which was defined as the start of stimulation. Eligible participants were provided with written and verbal explanations about the trial; the participant information sheet (PIS) was provided to the individual either by post or in person, depending on how the participant was first approached. Three different versions of the PIS were used owing to variation in trial procedures across the trial sites: one version for the Sheffield and Southampton sites (because of the tissue substudy at these sites; see Participant information sheet – Sheffield and Southampton sites), one for Dundee (where a postal consent process was used; see Participant information sheet – Dundee) and one for all other sites (see Participant information sheet – all sites, except Dundee, Sheffield and Southampton).

When participants had been approached and information about the trial provided, they were given time to consider their decision. Women were either asked about the trial at a subsequent routine visit to the fertility unit prior to the start of their IVF cycle, or contacted by telephone, e-mail or text message at a later date to ascertain their interest.

Informed consent

Informed consent was taken during a routine appointment prior to the start of the IVF cycle. The need to possibly delay the IVF cycle was discussed with participants, if randomisation was being undertaken close to the start of IVF, such that, if randomised to the ES arm, the IVF cycle would need to be delayed in order for ES to be delivered within the mid-luteal phase of the menstrual cycle prior to IVF. A decision to delay was made and agreed by both the participant and her fertility team prior to randomisation.

A written consent form (see Consent form – all sites except Sheffield and Southampton) was provided to the participant by a member of site staff [either the Principal Investigator (PI) or the RN], and, following the participant’s agreement to participate, was signed by both the person taking consent and the participant. Participants recruited at the Sheffield and Southampton sites had the opportunity to consent to participate in a tissue substudy, in which tissue was collected from the endometrium and stored for future analysis (the results and detailed methods of which are not contained in this report); therefore, a different version of the consent form was used at these two sites (see Consent form – Sheffield and Southampton). Consent for this substudy was sought at the same time as consent for the trial. Participants were able to participate in the main trial without consenting to participate in the substudy, but not vice versa. Participants recruited at the Dundee site had the option to use a postal consent process, to lessen the burden of having to travel long distances to the fertility unit to consent and be randomised to the trial. In such cases, consent was obtained by post using an adapted consent form [see Postal consent form (Dundee only)], followed by consent in person post randomisation at the next routine appointment.

Endometrial scratch arm

Participants randomised to the ES arm received ES in the mid-luteal phase (defined as 5–7 days before the expected next period, or 7–9 days after a positive ovulation test) of the menstrual cycle prior to which IVF was being undertaken.

Endometrial scratch is a minor procedure performed in daily clinical practice in an outpatient setting. The procedure, which was already routinely undertaken at many of the fertility units for participants undergoing their second or subsequent IVF cycle, and, in some centres, was already offered to patients undergoing their first cycle of IVF, was undertaken as per local guidelines by a suitably trained doctor or nurse. When undertaking the procedure, a standard operating procedure (SOP) was followed (see Endometrial Scratch Standard Operating Procedure).

The participant was required to use a barrier method of contraception (if necessary) in the menstrual cycle in which the ES was performed. Before ES was undertaken, participants were asked if they had complied with this. If there was a risk of pregnancy (i.e. the participant reported unprotected sexual intercourse prior to ES), the ES was not undertaken and was either rescheduled (and the IVF cycle delayed, as appropriate) or IVF commenced as planned without performing the ES procedure.

Participants were free to withdraw from receiving the ES at any point prior to or during the procedure. Those who did not receive the procedure remained in the trial and were followed up as described in Procedures and assessments.

At two centres (Sheffield and Southampton), endometrial biopsies were taken during the ES procedure (if the participant consented to do so). In order to collect the exact time point of menstruation at which the procedure was undertaken (and the tissue collected), ovulation kits were provided to participants who were randomised to the intervention arm at these centres. Once ovulation was detected [luteinising hormone (LH) surge detected], the participant contacted the fertility unit in order to schedule the ES, and the procedure was undertaken 7–9 days later. The tissue was harvested from the pipelle sampler that was inserted into the uterus to undertake the ES. Samples were collected and stored in formalin-containing specimen pots in the fertility units; the specimen pots will be stored for up to 10 years in liquid nitrogen. The tissue samples were anonymised and coded prior to storage and the liquid nitrogen Dewars were kept locked.

At all other fertility units, participants randomised to receive ES contacted the fertility clinic when their menstrual cycle preceding IVF began; the procedure was then scheduled 5–7 days prior to their next period. Women who failed to contact the site were telephoned by the RN.

To perform ES, a speculum was first inserted into the vagina and the cervix was exposed and cleaned. A pipelle or similar endometrial sampler (as per local procedures) was then inserted into the cavity of the uterus; negative pressure was applied by withdrawal of the plunger. The sampler was then rotated and withdrawn several times so that tissue appeared in the transparent tube. The sampler and speculum were then removed. If no tissue was seen in the transparent sampler, this was an indication that the sampler was not fully inside the uterine cavity and, therefore, the procedure was repeated. In total, the ES procedure took approximately 10–20 minutes. See Appendix 5 for a description of the intervention using the Template for Intervention Description and Replication (TIDieR) checklist. 36

Following ES, ‘usual’ IVF treatment was received as per local protocols and no further interventions were performed beyond usual care.

Treatment-as-usual arm

Participants randomised to the treatment-as-usual (TAU) arm received ‘usual’ IVF treatment in accordance with the local protocols at their fertility unit.

Blinding and masking

The nature of the ES procedure makes it impracticable to blind participating women, clinicians, research staff and outcome assessors, and therefore this was an open-label RCT. However, the trial uses objective pregnancy outcomes that are unlikely to be affected by the placebo effect or assessment bias; for example, achieving a live birth is a definitive primary end point. Therefore, it was unnecessary to use a sham ES procedure in the TAU group. The trial statisticians and health economist were blinded to the treatment allocation throughout the trial.

Primary and secondary outcomes

Safety outcomes

Data on adverse events (AEs) were collected up to 6 weeks post partum in participants (see Table 1). In the fetus/baby, data on severe congenital abnormalities (detected both antenatally and postnatally) and neonatal deaths were also collected up to 6 weeks post partum (see Table 1). Definitions of the different events recorded can be found below.

Baseline data collection and randomisation

Following consent, participants underwent baseline assessments prior to randomisation, which included height/weight, ethnicity, medical history and a fertility assessment, including ovarian reserve assessment.

Randomisation was then undertaken as described in Randomisation and concealment. If randomised to do so, the participant received ES (see Description of trial arms, Endometrial scratch arm), followed by ‘usual’ IVF. If randomised to TAU, ‘usual’ IVF was delivered (see Description of trial arms, Treatment-as-usual arm).

In vitro fertilisation cycle

Following randomisation, and ES (if randomised to this arm), information regarding the received treatment cycle was collected, including information about down-regulation and stimulation phases, egg collection and embryo transfer. During embryo transfer, details of whether the clinician detected fluid on the endometrium or blood on the tip of the catheter were collected. In Newcastle, data regarding fluid on the tip of the catheter could not be collected because ultrasound-guided embryo transfer was not undertaken at this site.

Owing to a perceived lack of equipoise in this patient group, fertility unit staff asked participants randomised to either arm if an ES had been received ‘outside’ the trial.

Embryo grading

The quality of each transferred embryo was recorded. Three stages of embryo development are commonly recognised: morula (day 3 of development), cleavage (day 1, 2 or 3) and blastocyst (day 5 or 6). Embryo grading was undertaken immediately prior to embryo transfer; embryos were subjectively assessed by visual microscopic inspection as per the fertility unit’s protocols and recorded using established grading schemes for cleavage-stage embryos and blastocysts. Sites used either the National External Quality Assessment Service (NEQAS)38 or Gardner blastocyst grading systems,39 in accordance with their normal clinical practice (see Appendix 8 for a list of embryo grading systems used at each trial site). The NEQAS system was updated in April 2017 and, therefore, both systems were used during the trial. 40

For each grading system, we mapped embryo grading to a common quality rating scale created by a principal embryologist, to enable the comparison of embryo quality between treatment groups and sites. A separate quality scale was created for cleavage and blastocyst embryos. Both conversion algorithms can be found in Appendix 9.

For embryo selection at the blastocyst stage, three elements of the blastocyst were graded: the expansion score, trophectoderm and inner cell mass. The conversion algorithm for blastocysts does not take into account expansion score, as blastocysts undergo dynamic expansion, the status of which can change depending on what time on day 5 of development the grading was performed. As fertility units graded blastocysts at different time points and we did not record whether the grades provided were at selection or transfer, or a mixture of the two, it was cleaner to leave out the ‘expansion’ score. Cleavage-stage embryos, which are not rated using the Gardner blastocyst grading system, were rated using a different scale in both old and new NEQAS grading. Morula-stage embryos are not commonly graded.

Intervention delivery and post-intervention pain and tolerability

For those randomised to the ES, the details of the ES procedure were collected when the participant attended the unit to undergo the procedure. Pain was subjectively assessed within 30 minutes post ES, and then at 1 day and 7 days post ES, as described in Outcome measures, Secondary outcomes. This was collected in person directly after the procedure, at which time the participant took part in the pain assessment and was asked if the procedure was tolerable (yes/no) (see Directly post ES pain and tolerability CRF). At the 1- and 7-day post-ES time points, the Likert rating scale was recorded via short message service (SMS) (see One day and seven day pain rating CRF). If the participant failed to respond to the message, a reminder text message was sent. If there was still no response, a member of the fertility unit telephoned or e-mailed the participant to request this information. Part-way through the trial, the wording of the SMS was reviewed and changed because of concerns that participants were misreading the question posed.

Pregnancy test and scan

Pregnancy tests and scans were undertaken as per local procedures and details of these were collected by the fertility units. AE information for the period prior to the pregnancy test was also collected by the fertility unit. In the event that the participant’s IVF cycle had ceased prior to the pregnancy test (i.e. no eggs collected or no embryos generated), the participant was contacted at a similar time point (i.e. 2 weeks post pregnancy test) to collect information on AEs.

Post-treatment cycle follow-up

Participants were followed up for at least 10.5 months post randomisation, or 6 weeks post partum, unless they did not become pregnant following their first IVF cycle (in which case, participants were followed up until the termination of their IVF cycle, e.g. IVF cycle not started, no eggs fertilised or pregnancy not achieved) or their pregnancy ended before full term (e.g. miscarriage or termination). Participants who were randomised to the trial but whose IVF treatment was then delayed were followed up for as long as the study timelines allowed in order to collect information regarding their treatment cycle and outcome, if treatment occurred.

Participants who were randomised to the trial but became spontaneously pregnant prior to embryo transfer were followed up in the same manner. However, the starting point for determining the timing of 3- and 6-month follow-ups was the date of the participant’s last menstrual period, rather than the date of egg collection (which did not go ahead in cases of spontaneous pregnancy).

Participants who underwent frozen embryo transfer (FET) and had received no previous fresh/frozen embryo transfers were followed up until a maximum of 6 weeks post partum, if the timelines of the trial allowed. Participants who commenced the stimulation phase of IVF but did not complete IVF for any reason were classed as having received their first cycle of IVF, and, therefore, their involvement in the trial was complete.

Safety outcomes

Adverse events and SAEs were collected at four follow-up time points, and were collected during the post-treatment cycle follow-up phase only if the participant had a positive pregnancy scan. Site staff were delegated the responsibility for recording AEs and SAEs electronically on the case report form (CRF). Unexpected SAEs were logged on the CRF and also sent by e-mail to Sheffield CTRU to expedite reporting.

Pre-funding preparation

All groups received the project plan and lay summary, and were asked about their perspective on the proposed research plan and research intervention, as well as the physiological and psychological impact of this new intervention on infertility treatment. The Chief Investigator and lead RN attended all meetings and were able to answer questions and resolve queries.

Feedback from PPI influenced the intervention design, data collection protocol and recruitment strategy, but it specifically influenced the follow-up of women who had failed to achieve a pregnancy or who experienced a miscarriage. PPI members felt that it was unreasonable and insensitive to complete questionnaires and be contacted by research staff to collect AE data if there was no ongoing pregnancy.

Post-funding preparatory work

The PPI lead arranged frequent communications with the PPI co-applicants, service users and advisory panel members (by e-mail, teleconference and face-to-face meetings) to develop study protocols and participant documentation. Both PPI co-applicants reviewed participant-related documentation (including, but not limited to, the poster, PIS/consent, questionnaires and patient introduction video). These documents and the patient introduction video were updated in the light of their comments.

In late 2015, the E-Freeze trial42 (funded by the National Institute for Health Research) and the HABSelect trial43 (which had required an extension to its recruitment timeline) were endeavouring to recruit the same patient population in many of the same research sites as was the Endometrial Scratch Trial. The Trial Management Groups (TMGs) of each trial requested input from the Reproductive Health Research Public Advisory Panel (Jessop Wing, Sheffield Teaching Hospitals NHS Foundation Trust) to help support a potential way forward to ensure the success of the three trials. The panel recommended the creation of an overarching leaflet for all potential participants to explain, in lay terms, the differences in the three trials in an unbiased manner. This was approved by the regulatory bodies and was used to facilitate recruitment. It was made clear that, at the time of writing this leaflet, no particular study was known to be superior to another in increasing the chances of a live birth.

Throughout progression of the internal pilot trial

Throughout the course of the internal pilot trial, changes to the study protocol and documentation were reviewed by the PPI co-applicants, service users and advisory panel members. Discussions among these groups also included recruitment progress and ideas to improve the process of collecting follow-up data. During the design of the qualitative substudy, PPI members were instrumental in supporting its design by, again, recommending changes to the recruitment techniques and the patient-facing materials, specifically the interview topic guide for participants.

Report writing, academic paper preparation and dissemination

All PPI co-applicants, service users and advisory panel members have been involved in the interpretation of the results, and they recommended that the trial team consider how certain aspects of the results were worded, in order not to provide undue positivity to the results of the trial. The PPI co-applicants, service users and advisory panel members also provided input to the presentation of results and dissemination materials, and the production of the final report to the funder (including preparation of the Plain English summary), and will continue to be involved in dissemination activities and the preparation of academic papers.

Intervention

Compliance with the intervention was ascertained through the record taken by the clinician or nurse of whether or not the participant had attended the clinic for the ES procedure and received the procedure per protocol. Any deviations from the ES procedure were noted and are reported.

Other

Any deviation from the trial protocol or Good Clinical Practice regulations44 was recorded by staff at the site or at the CTRU.

Trial Management Group

The TMG was responsible for the day-to-day management and oversight of the trial, and included the Chief Investigator, clinical co-applicants, a statistician, the trial manager, the lead RN and a PPI representative. The TMG reviewed blinded trial reports without any outcome data during trial management meetings.

Trial Steering Committee

The TSC provided overall oversight of the trial and comprised two independent gynaecologists and an independent statistician. Members of the TMG (including the trial manager and Chief Investigator) also attended meetings, which occurred at least once every 6 months. Blinded reports without any outcome data were provided by the data management team based at the CTRU prior to each meeting.

Data Monitoring and Ethics Committee

The DMEC was responsible for the safety of trial participants and the scientific integrity of the trial, and comprised two independent clinicians (gynaecologists) and one independent statistician. Even though the trial was not designed with formal stopping rules, blinded reports (for open-session discussions) and unblinded reports consisting of safety and outcome data (for closed-door discussions) were provided on a monthly basis and prior to each 6-monthly meeting.

Intention to treat

The primary analysis was based on an intention-to-treat (ITT) analysis population that included all randomised participants with informed consent regardless of circumstances after randomisation, except if they withdrew consent and explicitly stated that their data should not be used. Treatment allocation was as randomised even if participants switched treatments or did not receive their randomly allocated treatment. The Clinical effectiveness section of this chapter explains how missing data were addressed under the worst- and best-case scenarios.

Per protocol

The purpose of the per-protocol analysis was to explore the effect of the ES procedure among those women who were viewed to have met key protocol requirements. The per-protocol population included women who complied with key protocol requirements, defined as those who:

• met the inclusion criteria as stipulated in the protocol

• received the allocated treatment, that is, excluding those who were allocated to TAU but underwent the ES procedure within or outside the trial, those who were allocated to ES but received only IVF, and those who did not receive any trial intervention

• did not achieve a spontaneous pregnancy before the delivery of the interventions

• completed the fertility treatment cycle and successfully generated embryos

• used contraception before the ES procedure (i.e. excluding those who failed to use contraception before ES and whose procedure could not be rescheduled)

• received treatment using fresh embryos (i.e. excluding those who underwent FET)

• were treated using only the antagonist or long protocols (i.e. excludes other protocols such as ultra-long).

Based on the advice of the TMG, since the delivery of the ES procedure is simple and unlikely to vary significantly across fertility units, per-protocol analysis also included women who were randomised to ES but known to have received the ES procedure outside the trial (if applicable).

Analyses of the primary outcome (live birth rate)

The AD in LBRs was the primary measure of the treatment effect of interest. In the calculation of LBRs, we treated multiple live births per woman due to multiple pregnancies as a single event and the denominator was the number of women randomised. We estimated the 95% CI around the differences in LBRs using the normal approximation to the binomial distribution and calculated the associated p-value using Pearson’s chi-squared test. A 5% two-sided significance level was used for the hypothesis test. To aid interpretation, in line with the Consolidated Standards of Reporting Trials (CONSORT) guidance,48 we estimated the unadjusted OR with 95% CI using a simple logistic regression (binomial generalised linear model with a logit-link function) and the unadjusted RR with 95% CI using a simple binomial generalised linear model with a log-link function. We performed a sensitivity analysis by adjusting for fixed stratification factors (i.e. site and planned treatment protocol) and potential prognostic factors [i.e. history of pregnancy (yes or no), age, BMI, duration of infertility and smoking status (yes or no)]. The adjusted OR and 95% CI were estimated using a multiple logistic regression model and adjusted RR (95% CI) using a binomial generalised linear model with a log-link function. The adjusted AD and associated 95% CI were post estimated via margins using the delta method after fitting a binomial generalised linear model with a log-link function. 49

The primary approach to deal with missing data used the worst-case scenario by assuming that all women with an unknown live birth outcome, such as ‘lost to follow-up’, failed to achieve a live birth unless it was known otherwise (e.g. obtained from patients’ medical notes). We performed a sensitivity analysis using the best-case scenario by assuming that women without pregnancy and/or live birth outcome data were pregnant and delivered a live birth. Furthermore, we also performed a sensitivity analysis using a complete case that included only women with the known live birth outcome (see Complete case).

Per-protocol analysis on the primary outcome was performed assuming a worst-case scenario using the same methods as described for the primary ITT analysis based on a per-protocol population, as defined in Per protocol. Following the presentation of trial results to the TMG, TSC and DMEC on 18 May 2020, an unplanned analysis was requested to aid the interpretation of the per-protocol results. This involved the estimation of the posterior probability of the treatment effect being at least 10% AD of clinical importance (equivalent to an OR of ≈1.5 given the observed 40.5% LBR in the TAU arm) given the observed data. We fitted an equivalent Bayesian multiple logistic regression model with all covariates that were included in the primary ITT analysis with normal priors on parameters with a mean of 0 and variance of 105. The posterior distribution of the treatment effect was estimated via Markov chain Monte Carlo sampling with a sample size of 200,000 and burn-in of 500.

Analyses of secondary outcomes

We performed unadjusted analyses on all secondary clinical outcomes in the same manner as the primary outcome (see Analysis populations). This covered ectopic pregnancy, implantation, clinical pregnancy, miscarriage, multiple birth, preterm delivery and stillbirth. We repeated this analysis on secondary outcomes that also relate to safety signals (except implantation and clinical pregnancy) in women who achieved successful implantation (i.e. with a positive pregnancy test). There was no correction for multiple hypotheses testing, as this was exploratory and all tests were performed at 5% two-sided significance level.

Subgroup analyses: primary and secondary outcomes

The purpose of the subgroup analyses was to explore whether the estimated treatment effects were consistent or there was potential heterogeneity across prespecified subgroups. The adjusted OR with associated 95% CI and interaction p-value were estimated using a multiple logistic regression model that included an interaction term between treatment and subgroup as well as fixed stratification factors (site and planned treatment protocol) and potential prognostic factors [history of pregnancy (yes/no), age, BMI, duration of infertility, smoking status (yes/no)]. However, for subgroups relating to the treatment protocol followed and the history of miscarriage, planned treatment protocol and history of pregnancy were not included as covariates, respectively (because the subgroup and covariate were highly correlated). The same approach to estimate the adjusted RR with associated 95% CI and interaction p-value using a Poisson generalised linear model with log-link function and robust standard errors was used. We did not estimate the adjusted AD with associated 95% CI because the post-estimation of margins using the delta method failed to converge. 49

Subgroup analysis was only performed for the primary outcome (live birth) and selected secondary outcomes (clinical pregnancy and implantation). Other secondary outcomes with very small numbers of events were excluded because statistical models failed to converge; these were ectopic pregnancy, miscarriage, multiple birth, preterm delivery and stillbirth.

Participating women

We used the Wilson score method to estimate the 95% CI around the proportion of women who tolerated the ES procedure based only on women with tolerability data. 50 Sensitivity analysis was performed assuming that women who received ES but without tolerability data did not tolerate the ES procedure. The distribution of pain rating scores within 30 minutes and at 1 day and at 7 days was displayed using a multiple box plot and was also summarised using means (SD), median (IQR) and minimum and maximum scores.

The expected AEs that were recorded during the trial were summarised by the treatment group using the numbers and proportions of women who experienced at least one event. Unexpected AEs and SAEs that occurred between receiving the intervention and the end of the trial were summarised by the number and proportion of women who reported at least one event and a total number of repeated events per treatment group. The AE category, frequency, seriousness, intensity, outcome and its relationship with ES procedure (in those who received ES) were summarised in the same manner.

The numbers of repeated events per woman were analysed using a negative binomial regression model (accounting for follow-up period) to estimate the incidence rate (IR) per treatment group and incidence rate ratio (IRR) with associated 95% CI. We also performed a sensitivity analysis by including all unexpected AEs and SAEs reported at any point during the trial. In addition, events that occurred between receiving ES and IVF were also summarised in the ES group only.

Born babies

In women with a positive pregnancy test, safety outcomes experienced by born babies (see Chapter 2, Safety outcomes) were summarised using the numbers and proportions per treatment group as well as the unadjusted AD in proportions between treatment groups with 95% CI estimated using the normal approximation to the binomial distribution.

The UK World Health Organization Neonatal and Infant Close Monitoring Growth Charts51 were used to estimate the centiles of birthweight given gestational age and sex of baby using the hbgd R package [URL: https://hbgdki.github.io/hbgd/#growth-standards (accessed 4 February 2022)]. These centiles were used to classify babies at birth as low birthweight, very low birthweight, small for gestational age and large for gestational age, as described in Chapter 2, Safety outcomes.

Resource use

Resource use was collected as described in Chapter 2, Collection of health resource use and patient costs data. Unit costs were for 2018/19 and no discounting was applied as costs were presented over the time frame of the pregnancy, which was less than 1 year.

Primary cost-effectiveness outcome

We calculated the incremental cost-effectiveness ratio (ICER) as cost per percentage of successful pregnancies and the ICER as cost per successful pregnancies per 1000 population. To allow for uncertainty, a resampling method (known as bootstrapping) was used to obtain 95% bias-corrected CI around the ICER. 56 A total of 5000 simulations were used. Results are presented on the cost-effectiveness plane and on the cost-effectiveness acceptability curve.

Secondary cost-effectiveness outcomes

The cost-effectiveness of ES compared with TAU was examined for the following secondary outcomes: cost per clinical pregnancy, cost per ectopic pregnancy avoided, cost per miscarriage avoided, cost per multiple birth, cost per stillbirth avoided, cost per preterm delivery and cost per biochemical pregnancy. Analysis of variance was conducted to see whether or not there were any statistically significant differences in costs between the two treatment groups after allowing for the secondary outcome variable.

Prespecified subgroups

We explored the effectiveness differences in costs for different prespecified subgroups, as detailed in Chapter 3, Prespecified subgroups.

Sensitivity analysis

Resource use was collected at baseline, when participants were asked to recall their resource use over the previous 3 months. Analysis of covariance was undertaken to adjust for baseline costs and the ICER per successful live birth per 1000 population was calculated. A further analysis adjusting for baseline costs and any subgroups was also undertaken, as specified in Prespecified subgroups, and found differences in costs.

Triangulation based on labour and in vitro fertilisation resources reported by participants

The health resource use questionnaire was not designed to collect information on IVF or labour; this was collected directly from sites as part of the CRF. In the health resource use questionnaire, respondents were asked about visits to hospital, and some gave details of their IVF treatment and labour. Not all participants reported labour or IVF in the health resource use questionnaire; thus, if this questionnaire was used alone, without the CRF data, the costs of labour and IVF would be under-reported. We assumed that those who did not report these resources in the questionnaire were also under-reporting other resources that they accessed during the study in the same way. As a sensitivity analysis, a simple triangulation exercise was carried out assuming that the information reported on the CRF was accurate, as it was recorded from patient notes. The percentage of participants who reported undergoing IVF and labour on the patient resource use questionnaire was reported and costs were inflated to reflect this under-reporting.

NHS and wider perspective

As part of the study, participants were asked to complete a patient cost questionnaire that asked them about how they travelled to their last fertility appointment, the cost and time taken to travel and what they would have been doing had they not been at the appointment (lost time/earnings). The questionnaire also asked whether or not a companion accompanied them and about any lost time/earnings to their companion. Respondents indicated if they had travelled to their appointment by car, bus, train, taxi, foot or other mode of transport; they were also asked how far they travelled to their appointment. Respondents supplied details of the cost of car parking and any train, bus or taxi fares. The cost of car journeys was based on the UK Government’s mileage allowance payments for car travel of £0.45 per mile57 and multiplied by distance travelled to obtain a total cost for the journey.

Hourly rates of hours and earnings were taken from the ONS Earnings and Working Hours Survey for 2019. 58 It was assumed that any time away from usual activities had the same value for those in or not in paid employment, and the average hourly rate for the time spent at the appointment was applied to all participants. The cost of travel and time away from other activities was applied across all planned hospital visits plus seven additional visits for those who underwent IVF plus a further visit for those who became pregnant. These costs incurred by participants were added to the NHS costs to present a wider than NHS perspective in a sensitivity analysis.

Staff

Both RC and AP worked on the trial; RC managed the trial from the start and AP became involved just prior to the start of this qualitative study. A relationship between the staff and RC was therefore already formed, which would also have been true for the research assistant and the staff, but to a lesser extent. Staff would have been aware of the interviewers’ involvement in the trial (i.e. involvement in its design and delivery) and the reasons for the research.

Trial participants

The interviewers had no involvement in the clinical care of the trial participants, nor had they had previous contact with them. Participants would have been aware of the reasons for the research and may or may not have been aware of the interviewers’ involvement in the trial in which they participated.

Trial participants

Trial participants were initially approached by letter or e-mail; a copy of the PIS (see Participant information sheet for the qualitative substudy – trial participant version) for this qualitative substudy was included. If no response was received, trial participants were telephoned approximately 1 week after the letter was sent, to ascertain their interest in participating. If interested, the interview was scheduled at a mutually convenient date and time. A letter or e-mail was sent to the individual to confirm the appointment time.

Staff

Initial interest to be interviewed was sought by e-mail or telephone – a PIS (see Participant information sheet for the qualitative substudy – staff version) was sent to the individual. If interested, a mutually convenient date and time for the interview was scheduled.

Types of participants

Studies involving women undergoing IVF, with or without ICSI, for the first time were considered for the review.

Studies that include participants who are undergoing intrauterine insemination (IUI) or ovulation induction (or other treatments not classed as IVF) and/or their second or subsequent IVF cycle will be excluded from this review, unless separate outcome data can be extracted for the subset of participants who have undergone their first IVF cycle.

Types of interventions

Studies evaluating the effectiveness or efficacy of the ES procedure, or a similar intervention (endometrial biopsy), were considered for the review.

No restrictions were placed on the type of comparator (e.g. sham procedure, other intervention or usual treatment).

Types of studies

Reports of randomised feasibility trials, pilot RCTs and RCTs were included in this review, as long as the outcomes of IVF of interest were reported (see Outcome measures). No date limits were set.

We excluded reports published as meeting abstracts only and/or reports published in languages other than English if the methodological details reported in the abstract were insufficient to allow extraction of study characteristics.

Subgroups

When analysing the results of the included studies, subgroup analyses by the following prespecified attributes were attempted, where data allowed:

• age

• duration of infertility

• IVF protocol (antagonist/long).

Where possible, we also subgrouped analyses by the timing of the ES procedure within the menstrual cycle, which was not prespecified.

Outcome measures

No restrictions were placed on the types of outcome measures considered. However, the outcomes of interest were live birth, pregnancy, miscarriage, ectopic pregnancy, multiple birth, stillbirth, preterm delivery, pain of the procedure and safety events in the mother and newborn.

Electronic database sources

• MEDLINE(R) In-Process & Other Non-Indexed Citations and MEDLINE(R) (Ovid) 1948 to present

• EMBASE (Ovid) 1980 to present

• Cumulative Index to Nursing and Allied Health Literature (CINAHL) 1981 to present

• The Cochrane Library, including the Cochrane Database of Systematic Reviews (CDSR) 2005 to present, Cochrane Central Register of Controlled Trials (CENTRAL) 1898 to present, Health Technology Assessment (HTA) 1989–October 2016 (currently not being updated) and Database of Abstracts of Reviews of Effects (DARE) 1994–April 2015 (archive only)

Grey literature and internet sources

• ClinicalTrials.gov (www.clinicaltrials.gov/).

Data items for extraction

The following data items were extracted from the included studies:

• study characteristics, such as country of origin, recruitment period, number of treatment arms, number randomised, and ITT and per-protocol samples

• participant characteristics, such as average age, average duration of infertility, egg source (i.e. donor eggs) and IVF protocol used

• intervention and control group details, such as timing of ES, ES technique, device used, co-interventions, and control group condition

• outcome data: live birth, ongoing pregnancy, clinical pregnancy, biochemical pregnancy, implantation, miscarriage, ectopic pregnancy, multiple pregnancy, multiple birth, pain of ES, stillbirth, preterm delivery and AEs.

Endometrial scratch procedure and in vitro fertilisation in the endometrial scratch group and time to in vitro fertilisation after endometrial scratch procedure

Figure 3 summarises the uptake of the ES procedure and IVF among those randomised to the ES group, including reasons for failing for the 13.4% (70/523) of participants who did not receive the intervention. Of those randomised to ES, 86.6% (453/523) received the ES procedure within the trial. All participants who received ES received it as per protocol.

FIGURE 3.

Uptake of the ES procedure and IVF in the ES group. a, Investigator decisions (n = 9): deemed too risky to perform ES as the participant had a previous severe infection following IUI (n = 1); risk of perforation due to stenosis of the cervical canal (n = 1); lost to follow-up (n = 1); recently had surgery for adhesions (n = 1); had partial septate uterus on the scan (n = 1); identified adenomyosis within the uterus (n = 1); had leuprorelin acetate (Prostap, Takeda UK Ltd, London, UK) (n = 2); participant had a scratch procedure in the past year which was noted when an endometrial biopsy was taken during exploratory surgery (n = 1). b, Other reasons (n = 5): lost to follow-up (n = 1), never returned to commence treatment (n = 3) and BMI was too high to start treatment (n = 1).

Overall, 95.0% (497/523) of those randomised to ES received IVF; 85.9% (449/523) received IVF after receiving ES procedure and 9.2% (48/523) received IVF without undergoing ES procedure.

Among the 449 women in the ES group who received the ES and IVF, the median (IQR) time to the start of IVF post ES procedure was 15 (9–22) days, ranging from 2 to 126 days (Figure 4a). The mean time (SD) was 17.1 (11.5) days. In summary, the majority of women started their FSH drugs within 30 days of the ES procedure. It should be noted that the date when FSH drugs were started was used as a proxy for the start of IVF after the ES procedure. Figure 4b also displays the distribution of time from ES procedure to embryo transfer.

FIGURE 4.

Distribution of time from ES procedure to (a) start of IVF and (b) embryo transfer. a, Extreme outlier: participant developed a tubal ovarian abscess and required surgery before starting IVF.

In vitro fertilisation in the treatment-as-usual group

Of those allocated to TAU, 94.1% (494/525) received IVF (Figure 5). Only 5.9% (31/525) of participants did not receive IVF, for reasons that are detailed in Figure 5. Only 1.0% (5/494) of IVF recipients received ES outside the trial.

FIGURE 5.

Uptake of IVF in the TAU group.

Per-protocol analysis of the primary outcome (live birth)

We explored the effect of the ES intervention on achieving a live birth (primary outcome) among those who have complied with the key components of the protocol as defined in Chapter 3, Analysis populations.

Of those randomised, 79.4% (417/525) in the TAU group and 71.1% (372/523) in the ES group met the criteria for inclusion in the per-protocol analysis population, assuming few women with missing live birth outcome data had a negative pregnancy outcome (worst case). Thus, the proportion of women who met the per-protocol criteria was slightly higher in the TAU group than in the ES group. Table 8 reports both the unadjusted and adjusted effects of the ES procedure on live birth. The LBR was 40.5% (169/417) in the TAU group and 44.1% (164/372) in the ES group. This translates to only a 3.6% increase in live births in favour of ES compared with TAU (95% CI –3.3% to 10.5%; p = 0.312). In relative terms, the chance of achieving a live birth is only 1.09 times more (95% CI 0.92 to 1.28) in women who received ES procedure than in those who receive TAU. Although this effect is not statistically significant and could not rule out the lack of effect of ES in improving LBR, it could be of potential clinical importance, as the upper limit of the CI includes a 10% targeted AD. However, based on post hoc analysis [described in Chapter 3, Analyses of the primary outcome (live birth)], the observed effect (a 3.6% increase) suggests that such large differences in the LBRs are unlikely, as there is only a 2.9% probability of the effect being at least 10% given the observed 3.6% increase in the LBR. As a result, we did not find enough evidence to support the effect of ES in increasing the chances of achieving a live birth; however, we could not rule out the potential beneficial effect of ES in women who met per-protocol criteria. For comparability, Figure 7 displays the effect of ES on the live birth based on both ITT and per-protocol analysis populations.

FIGURE 7.

Effect of ES on live birth based on ITT and per protocol populations. a, Primary ITT results. PP, per protocol.

In summary, the adjusted and unadjusted results (see Table 8) are very similar, consistent with the fact that the characteristics of participants at baseline and treatment cycle were, on average, well matched between the treatment groups (see Tables 3–5).

Subgroup analysis results on the effect of ES on the primary outcome (live birth)

We explored the heterogeneity in the effect of the ES procedure on the primary outcome across prespecified subgroups detailed in Chapter 3, Prespecified subgroups. The interpretation of these subgroup results requires caution as the trial was not adequately powered to address this objective and the interaction tests were not controlled for multiple hypothesis tests. Therefore, these results should be used for hypothesis generation guided by biological plausibility and consistency. Figure 8 displays LBRs, the estimate of the treatment effect within each subgroup and the interaction test. Furthermore, we display the overall results for the primary outcome including per-protocol analysis for comparability.

FIGURE 8.

Forest plot of subgroup results relating to the primary outcome (live birth). Embryos transferred on day 6 were excluded because there were only two women in the TAU group and none in the ES group (see Table 5) and no live births were recorded. N/A, interaction test not applicable; PP, per protocol.

In summary, the effect of ES appears consistent across subgroups; however, we could not rule out potential benefits when embryos are transferred on day 5 or in women who had cycle programming performed using oral contraception, progestogens or oral oestrogen.

We found similar results in OR and RR measures of the treatment effect, respectively (see Appendices 14 and 15). Results on the AD scale are not presented as the models failed to converge owing to a small number of participants and events within subgroups.

Subgroup analysis of secondary outcomes

We explored heterogeneity in the treatment effect of the ES procedure on selected secondary outcomes (Tables 9 and 10), excluding those with a small number of events (i.e. ectopic pregnancy, stillbirth, multiple birth and preterm delivery). These were excluded because the numbers of events within each subgroup were insufficient for models to converge and for making a reasonable exploratory inference. It should be noted that implantation and clinical pregnancy outcomes are highly correlated.

Expected adverse events in participating women

Expected AEs that were reported at any point during the trial among women who received the study treatments are summarised in Table 11. In general, the prevalence of these expected AEs was very similar between the treatment groups. For instance, the proportion of women who reported at least one expected AE was 174 (32.4%) in the TAU group and 144 (31.4%) in the ES group. The most common expected AEs were abdominal pain, nausea and vaginal bleeding; however, these were only reported in < 14% of women.

Unexpected adverse events and serious adverse events in participating women

The incidences of unexpected AEs, expected SAEs and unexpected SAEs reported after the delivery of the interventions were similar between treatment groups (Table 12). For example, 16.9% (91/537) of participants in the TAU group and 19.0% (87/458) of participants in the ES group reported at least one unexpected AE. The total unexpected AEs including repeated events per participant were 120 in the TAU group and 114 in the ES group, and women were followed up for a total of 255.7 person-years in the TAU group and 248.6 person-years in the ES group. This translates to an incidence of 47 per 100 person-years in the TAU group and 48 per 100 person-years in the ES group. Thus, the incidences of unexpected AEs are 1.02 times (95% CI; 0.76 to 1.38) more in the ES group than in the TAU group.

In total, there were 9.3% (50/537) and 8.1% (37/458) SAEs with an IR of 22 per 100 person-years and 19 per person-years in the TAU group and ES group, respectively. This equates to an IR of 0.88 times (95% CI 0.58 to 1.34) in the ES group compared with the TAU group.

There were no deaths reported during the entire trial. Only 1 out of 458 (0.2%) SAEs in the ES group was viewed as possibly related to the ES procedure. Most of the SAEs resulted in inpatient or prolonged hospitalisation and were rated as isolated and of moderate intensity; however, these incidences were very similar between treatment groups. Only one vaginal bleeding incident was reported in the ES group, but this event did not occur within 48 hours of the ES procedure.

For sensitivity analysis, Appendix 19 is a replication of Table 12 that includes all unexpected AEs and AEs reported at any point during the trial (before and after receiving the interventions), as there is a possibility that down-regulation could have happened before the interventions in few unknown cases. In summary, there were very few events reported before the interventions so Table 12 and Appendix 19 are very similar.

Unexpected adverse events and serious adverse events in women between endometrial scratch and in vitro fertilisation

Only 11 (2.4%) women reported unexpected AEs that occurred after the delivery of ES but before receiving IVF procedure (Table 13). In addition, there was only one unexpected SAE, which was unlikely to be related to the ES procedure.

Serious adverse events in born babies

We present SAEs reported in born babies among women who had successful embryo implantation (with a positive pregnancy test) based on treatment as received. Successful implantation was reported in 270 and 226 women who received IVF only and ES procedure followed by IVF, respectively (Table 14). Three (1.1%) severe congenital abnormalities were reported, all in the TAU group. There were no neonatal deaths reported during the entire trial. Four other SAEs reported in the ES group only were congenital brain abnormality at 12 weeks (termination recommended as condition not compatible with life), lumbosacral myelomeningocele with an Arnold–Chiari malformation, hypoplastic left heart syndrome and full pulmonary cardiac resuscitation (baby recovered).

In summary, the rates of live births with low birthweight, very low birthweight and small for gestational age were higher in the TAU group than in the ES group, which appears to favour the ES procedure (see Table 14). For example, 14.1% (38/270) of mothers in the TAU group and 5.8% (13/226) in the ES group gave birth to at least one low birthweight baby resulting in an absolute reduction in proportions of 8.3% (95% CI 3.2% to 13.5%) in favour of the ES procedure. These results should be interpreted with caution because of the potential case-mix, as these results were not adjusted for potential prognostic factors relating to mothers. For example, there could be some differences between treatment groups concerning the characteristics of women who gave birth to babies with low birthweight, very low birthweight, or babies who were small for their gestational age that may influence these outcomes. In addition, low birthweight, very low birthweight, and small for their gestational age are highly correlated.

Visits to accident and emergency

Between 3% and 6% of the participants accessed accident and emergency (A&E) during the study with slightly more accessing A&E at 3 months and 6 weeks post partum in the ES group (Table 15). Those who accessed A&E accessed it between one and five times. Table 15 also shows the number of cases who went on to be admitted to hospital. Appendix 20 provides a list of reasons for admission in addition to the unit costs, reasons for admission included infections, bleeding, pain and ovarian hyperstimulation syndrome (OHSS).

The average cost of A&E visits was £26.63 in the TAU group and £34.21 in the ES group and the average cost of overnight stays after an A&E admission was also higher in the ES group (TAU mean £93.36; ES mean £130.94).

Ambulance services

Twenty-five participants used the ambulance service during the course of the study: 10 (1.9%) participants in the TAU arm and 15 (2.9%) participants in the ES arm. The cost of using the service was similar for both groups (TAU mean £7.89; ES mean £7.35).

Planned hospital visits

Planned hospital visits were higher post partum than at other time points and were slightly higher in the ES group than with TAU (see Table 15). Most planned visits were day cases, except at 6 weeks post partum when over half were overnight stays. Appendices 21 and 22 provide a list of reasons why participants were listed for day-case visits and overnight stays, respectively, and the unit costs involved. Participants’ most frequent reasons for day visits were IVF treatment, ultrasound scans, consultant appointments, blood tests and laparoscopy. Participants’ most frequent reasons for overnight visits were for giving birth: labour, induction and caesarean sections.

The average cost of planned admissions was slightly higher in the TAU group (mean TAU = £334.71; mean ES = £295.74).

Services

Many participants accessed GPs, dentists, pharmacists and practice nurses during the study, with around 80% accessing midwives at 3 months and at 6 weeks post partum and > 80% accessing health visitors post partum (see Table 15). Those in the ES group were more likely to access services; this was true at baseline as well as at 3 months and 6 weeks post partum. The mean cost of services was higher for the ES group than for the TAU group (mean TAU = £375.88; mean ES = £422.84).

Other services

Around 20% of participants accessed other services at baseline and 6 weeks post partum with a lower percentage (between 6% and 10%) accessing services at 3 months (see Table 15). The types of services ranged from scans to breastfeeding clinics, assistive conception units and acupuncture. Appendix 23 provides a list of reasons for admission in addition to the unit costs. The mean costs were slightly higher for the TAU group than for the ES group (TAU mean = £1178.62; ES mean = £930.83).

Ultrasound

Appendix 24 shows that the majority of participants had an ultrasound within 3 months of starting the study, which is as expected. A higher proportion in the ES group had scans later in their pregnancy. About one-third of participants had a scan privately. The average costs to the NHS were slightly higher for the ES group than for the TAU group (TAU mean = £597.95; ES mean = £621.61).

Treatment costs

The cost of IVF per cycle is £3218 and is incurred by all participants in the study undergoing IVF treatment; an additional study cost of £10 per participant is also included for the provision of ovulation kits. Those who underwent ES incurred an additional cost of £160.

Overall costs

The overall cost for TAU was on average £316 lower than for ES (TAU mean £6503; ES mean = £6819). Table 16 presents a summary of costs per treatment group.

Wider than NHS prospective (including travel costs and loss of earnings by participants and their companions)

Three-hundred and fourteen participants responded to the patient resource use questionnaire: 146 (27.8%) in the TAU group and 168 (32.1%) in the ES group. Table 20 presents the mode of transport and the median and IQR for total travel time in minutes, to and from appointments and the cost for each mode of transport (note that participants could report more than one mode, e.g. train and taxi).

The majority of participants travelled by car to the appointment: 93% of the TAU and 90% of the ES participants. The median distance from the appointment for both groups was 10 miles. The median time taken to travel to and from appointments was slightly higher for the TAU group (median = 80; IQR 50–120 minutes) than for the ES group (median = 70; IQR 80–100). However, overall costs of travel were similar for both groups (TAU mean = £20.42, 95% CI £17.34 to £25.37; ES mean = £19.00, 95% CI £16.87 to £23.81). Two participants were reimbursed for time at appointments; both were in the TAU group.

The majority of participants were accompanied to their appointment and most participants and their companions were in paid work with a few who would have been doing household activities, been in full-time or part-time education, on sick leave or unemployed (Table 21). Between 16% and 30% of participants and their companions reported lost earnings, with companions more likely to lose earnings than participants.

Appendix 25 summarises the list of occupations for participants and their companions.

The median number of hours lost was 3 (IQR 0–9) hours in the TAU group and 2 (IQR 0–5) hours in the ES group. Time lost was estimated regardless of whether or not someone was in paid employment, as a person’s time is valuable regardless of what they would have been occupied with. Table 22 presents the cost of lost time for participants, their companions and the combined cost of both, in addition to travel costs and overall non-NHS costs for one appointment. On average, the TAU group incurred £38 more in lost time and £40 more overall than did the ES group.

The cost incurred by participants per visit was applied to seven IVF fertility visits for those receiving IVF treatment, plus one additional visit for delivery for those with a positive pregnancy outcome. If a participant indicated other hospital visits on the resource use questionnaire, the costs for these visits were also included. The mean cost incurred by participants and their companions over the study period was slightly higher for the TAU group than the ES group (mean costs to participants: TAU = £2422, 95% CI £2223 to £2729; ES = £2145, 95% CI £1924 to £2656) (Table 23). Including participant costs resulted in a mean difference in costs between the ES and TAU groups of £40, with an ICER per successful live birth of £2.67 (95% CI £120 to £142).

What were the reasons for participants withdrawing from the intervention?

Seven participants were interviewed who, having been randomly allocated to receive the ES, did not receive it.

Five participants declined the ES because of reasons that were out of the control of their fertility centre: illness, holidays and work commitments.

However, two participants stated reasons for declining ES that could have been prevented. For P1-SITE5, the site appeared to have forgotten to check the participant’s records for a recent smear test result:

I, we were scheduled to start the IVF and then I got a call from, like the following cycle, I got a call from one of the nurses, not the research nurse one of the other nurses at the facility and she said ‘Oh I’ve just checked your records and it turns out you’re, you’re overdue a smear test, we can’t proceed with the IVF until you’ve had a smear test’ . . .

P1-SITE5

For P1-SITE1, the site appeared to have forgotten to tell the participant to refrain from sexual intercourse in the menstrual cycle in which the ES was being performed. This participant recommended that discussions around abstinence should be more prominent in the recruitment process:

I think that, that discussion [abstinence from sexual intercourse] needs to be kind of one of the leading things.

P1-SITE1

One participant discussed not feeling supported when she told the fertility unit staff that she did not want to receive the procedure:

I just felt that the lady was quite abrupt with me, once we’d decided that I wasn’t going to partake, I felt that the lady was a little bit abrupt. And didn’t really understand why I’d made that decision.

P1-SITE4

However, other participants felt well supported:

They were lovely about it all, they were really nice.

P2-SITE9

Of those women who did receive the intervention, what were their thoughts of the procedure and the information they received in preparation for the procedure?

What were recruited trial participants’ and research site staff’s perceptions of the recruitment process and how did research sites deal with recruiting to two large randomised trials; how can the recruitment process be improved in future trials?

What were recruited participants’ perceptions of the data collection tools used during the trial, specifically text messages to collect pain and electronic questionnaires?

Nature of trials and geographical coverage

Only three of these studies were multicentre RCTs. 14,20,35 All 12 studies were individually randomised and were conducted across 10 countries: Iran, Hong Kong, Egypt (two studies), China, USA, Belgium, Spain, Turkey, Brazil and the UK. One RCT was undertaken multinationally across five countries. 14 The number of participants included in each trial undergoing their first IVF cycle ranged from 18 to 1048.

Eleven studies were two-arm RCTs14,17,19–22,28,35,87–89 and one was a four-arm RCT. 18 Nine of the two-arm RCTs compared the ES procedure with usual care and two trials included a comparator involving a sham procedure. 21,87 The four-arm trial compared ES at two different time points with a sham procedure undertaken at the same two different time points in the menstrual cycle: proliferative and luteal. 18

Recruitment to three of the trials was prematurely ended owing to an unplanned futility analysis showing no differences in clinical pregnancy rates between intervention and control groups in one trial,21 a planned interim analysis identifying higher miscarriage rates in the intervention group in another trial,28 and identifying a significant benefit of the intervention during a planned interim analysis in the final trial. 87

Endometrial scratch procedure and timing

The method of undertaking ES was largely similar across studies. Most used a pipelle to invoke injury, except one trial that used an embryo transfer catheter,22 one that used a Novak Curette17 and another that used either a pipelle or Wallace endometrial sampler. 88 However, the timing of ES differed across trials. Two trials undertook ES during the IVF cycle, either on the day of egg collection17 or during ovarian stimulation. 28 Ten trials undertook ES in the menstrual cycle prior to IVF, with seven within the luteal phase19–22,35,88,89 and one during the early or mid-luteal phase. 87 Lensen et al. 14 undertook ES at any point between day 3 of the menstrual cycle prior to ES and day 3 of the cycle in which IVF was being undertaken. However, this trial reported that the median time (IQR) between ES and embryo transfer was 35 (22–39) days, and therefore it is likely that the majority of participants received ES in the menstrual cycle prior to IVF. Liu et al. 18 undertook ES either during the proliferative or luteal phases of the menstrual cycle; therefore, in this review, the two time points of delivery of ES or the sham procedure were combined, so that for each outcome there was one rate for the ES arm (both proliferative and luteal) and another for the sham arm (both proliferative and luteal).

We undertook an additional analysis in order to further assess the impact of the timing of ES procedure; we split trials into those that performed ES in the menstrual cycle prior to that in which IVF was being undertaken17,28 and those that undertook ES in the menstrual cycle during which IVF was also delivered. 14,18–22,35,87–89 Two trials provided hysteroscopy to all trial participants, prior to IVF. 20,89

Participant eligibility

Trials utilised different participant eligibility criteria. Ten trials had age restrictions with an upper limit of between 35 and 49 years. 14,17–21,28,35,87,88 Three trials restricted the BMI to an upper limit ranging from 30 to 35 kg/m². 17,19,35 Five trials selected participants who were deemed to have a good ovarian reserve, by allowing only those with a certain level of FSH to participate, with a maximum level of 10 IU/ml in two studies19,35 and 12 IU/ml in three studies. 17,18,20 At the point of assessing eligibility, two studies set requirements for the number of oocyctes collected or embryos replaced: ≥ 2 oocyctes replaced in one study20 and 4–14 oocyctes collected in another. 17 Only one trial stipulated that the embryos replaced had to be of a certain quality, with Mahran et al. 20 stipulating that the two or more embryos replaced had to be of ‘good’ quality. However, the exact method of defining a good-quality embryo was not reported. Two trials excluded women receiving donor eggs28,89 and one study only included those receiving donor eggs. 22

Trial outcomes, definitions and timing of end points

Ten trials reported LBRs. 14,18,20–22,28,35,87–89 Clinical pregnancy rates were reported in all trials. However, this was defined inconsistently, with variation in the time point at which this outcome was assessed: at 4 weeks post embryo transfer in two trials,19,20 5 weeks in one trial,17 6 weeks in four trials14,18,22,89 and 728 and 8 weeks35 in one trial each (and unknown in three trials21,87,88). Ongoing pregnancy rates were reported in four trials, the timing of which varied between 1214,17,22 and 2089 weeks post embryo transfer.

Implantation rates were reported in seven studies. 17,18,20,22,35,89,90 Six studies defined this similarly as the number of gestational sacs, divided by the number of embryos transferred, whereas in Pye et al. ,35 this was defined as the number of gestational sacs divided by the number of participants randomised to each arm (under ITT principles). Miscarriage rates per clinical pregnancy were reported in 11 trials,14,18–22,28,35,87–89 with the time point of data collection differing between 12 and 24 weeks of gestation, but were unclear in two trials. 14,28 Seven trials reported multiple pregnancy rates14,18–20,22,88,89 and multiple birth rates were reported in one study. 35 Two trials reported stillbirth rate,14,35 the definition of which was reported in only one trial,35 with one also reporting preterm delivery rates. 35 A subjective assessment of pain of the ES procedure on a numerical rating scale was reported in three trials,14,35,87 with four studies providing qualitative reports of pain. 18,20,28,89 Eight trials reported AEs and/or complications in the participating women,14,17–20,28,35,89 and four trials reported such events in the baby or neonate. 14,22,35,87

Risk-of-bias assessment

It was not possible to conduct a risk-of-bias analysis for Polanski et al. 88 The only information available was from a recent review, which used a previous version of the risk-of-bias tool and therefore the authors’ assessments could not be considered in this review. 16

The risk-of-bias summary can be seen in Figure 15. This relates to the review authors’ judgement for each category in the risk-of-bias tool. Domains 1–3 were consistent across all outcomes per study. Across all studies and outcomes, there was none that was considered to be at a high risk of bias.

FIGURE 15.

Risk-of-bias assessments of the included trials.

Domain 1 assesses whether or not the allocation sequence was random, if the allocation sequence was concealed and if there were baseline differences between the intervention groups. Nine trials used a computerised system to undertake randomisation,14,17,19,21,22,28,35,87,89 one trial used sealed envelopes,20 and another used a table of random numbers. 18 Four trials were assessed as ‘some concerns’ in domain 1. 17–19,22 Studies were assessed as having ‘some concerns’ due to there being no information on how the allocation sequence was concealed. In all trials, there were no substantial differences in baseline characteristics across treatment arms.

Domain 2 considered the risk of bias due to deviations from the intended interventions, which was deemed to be of ‘some concern’ in three trials. 17,22,28 The differing dropout rate between the intervention group (8.5%) and the control group (2.2%) in Izquierdo Rodriguez et al. 22 resulted in this assessment. Similarly, in Karimzade et al. 17 there were four patients excluded from the analysis in the intervention arm and none in the control arm.

Domain 3 considered the risk of bias due to missing outcome data. The rate of missing data was low across all trials, therefore all were considered to be at low risk of bias for this domain.

All studies were judged to be at low risk of bias for domain 4, ‘measurement of the outcome’. Given the nature of the outcomes being assessed, patient knowledge of the intervention was unlikely to have affected the analysis. Therefore, even though only 3 of the 12 included studies involved some form of blinding,18,21,87 this is not considered to affect the patient outcomes. There was also no blinding in most of the included trials; however, it is unlikely that participants’ or clinicians’ knowledge of the intervention could have biased outcome assessment owing to the included trials using objective outcome measures unlikely to be influenced by the placebo effect.

Domain 5 assessed the risk of bias in the selection of the reported result. For all outcomes, studies were deemed to be of ‘some concern’ if the outcome was not specified prior to the start of the trial (no protocol, trials registry or a retrospectively added trials registry), the timing of the outcome was not specified or the outcome specification in the paper did not match the protocol/registry.

Only Pye et al. ,35 Lensen et al. 14 and Nastri et al. 87 were considered to have a low risk of bias across all assessments.

Primary synthesis: live birth outcome

The overall effect from the 10 trials that reported live birth outcome did not provide evidence to support a significant effect of ES on increasing the LBR (Figure 16): OR 1.17 (95% CI 0.76 to 1.79). 14,18,20–22,28,35,87–89 However, this analysis should be interpreted with caution, as there was evidence of significant heterogeneity between trials. Over three-quarters of the between-trial variability was not attributable to chance alone (I² = 83%).

FIGURE 16.

Forest plot of all trials reporting LBRs. df, degrees of freedom; M–H, Mantel–Haenszel test.

Exclusion of Mahran et al. 20 from the analysis minimised heterogeneity (I² = 0%), but did not alter the conclusions of the analysis, which was consistent when both fixed-effects (see Appendix 27; OR 1.03, 95% CI 0.87 to 1.21) and random-effects (Figure 17; OR 1.03, 95% CI 0.87 to 1.22) models were used.

FIGURE 17.

Sensitivity analysis of LBRs excluding Mahran et al. 20 df, degrees of freedom; M–H, Mantel–Haenszel test.

Mahran et al. 20 is a clear outlier, being the only trial to report an extremely significant effect of undertaking ES on LBR. There are several potential explanations for this heterogeneity, however, uncertainty remains in the actual cause. First, this is the only trial that required participants to have two ‘good’ quality embryos transferred to be eligible to participate in the trial; including such a stipulation prior to randomisation (i.e. prior to egg collection) may have resulted in participants being randomised to the trial and excluded because of poor embryo quality, biasing the results of this trial in a positive direction. Second, all participants received hysteroscopy prior to IVF, a procedure that can have similar effects to ES. Finally, an average of three embryos were transferred, whereas all other trials included in this review averaged one or two embryos transferred.

A subgroup analysis was undertaken to explore the effect of the timing of ES (see Appendix 28); one trial reported LBR and undertook ES during the IVF cycle,28 and nine trials undertook ES in the menstrual cycle prior to ES. 14,18,20–22,35,87–89 The analysis identified a non-significant interaction between subgroups (p = 0.37), but with the point estimates (ORs) of the two subgroups suggesting that ES influenced LBR in different directions; the effect when ES was undertaken prior to IVF was 1.21 (95% CI 0.76 to 1.92) and during the IVF cycle it was 0.83 (95% CI 0.42 to 1.63). However, with only one trial in the ‘during IVF’ subgroup, there are insufficient data to draw robust conclusions, and the interaction test has insufficient power to detect meaningful differences between these subgroups even if such a difference exists.

Heterogeneity was substantial both in the ‘prior to IVF’ subgroup (I² = 84%) and across subgroups (I² = 83%). Evidence of both levels of heterogeneity was reduced to a minimal level (I² = 0) by the removal of Mahran et al. 20 This did not alter the conclusions of the analysis, which were consistent regardless of whether random- (see Appendix 29) or fixed-effects models were used (see Appendix 30).

Synthesis of secondary outcomes

Clinical pregnancy

Meta-analysis of all 12 trials that reported clinical pregnancy did not show evidence of a significant effect of ES in increasing clinical pregnancy rate when analysed using a random-effects model (see Figure 18; OR 1.18, 95% CI 0.82 to 1.72). 14,17–19,21,22,28,35,87–89 Owing to substantial evidence of heterogeneity (I² = 82%), sensitivity analysis using a fixed-effects model was not undertaken.

FIGURE 18.

Forest plot including all trials reporting clinical pregnancy rates. df, degrees of freedom; M–H, Mantel–Haenszel test.

Removal of Mahran et al. 20 reduced heterogeneity to a moderate level (I² = 46%), but the conclusion drawn from the analysis did not change, which was consistent across random-effects models (Figure 19; OR 1.06, 95% CI 0.84 to 1.35) and fixed-effects models (see Appendix 31; OR 1.08, 95% CI 0.93 to 1.26).

FIGURE 19.

Forest plot of clinical pregnancy rates, with Mahran et al. 20 excluded. df, degrees of freedom; M–H, Mantel–Haenszel test.

We undertook a subgroup analysis to assess the effect of the timing of ES, including two trials that undertook ES during the IVF cycle,17,28 and 10 trials undertaking ES prior to IVF (see Appendix 32). 14,18–22,35,87–89 The analysis identified a non-significant interaction between subgroups (p = 0.36), but with the point estimates (ORs) of the two subgroups suggesting that ES influenced LBR in different directions; the effect prior to IVF was 1.33 (95% CI 0.91 to 1.95) and during IVF it was 0.62 (95% CI 0.13 to 3.04). However, with only one trial in the ‘during IVF’ subgroup, there are insufficient data to draw robust conclusions, and the interaction test has insufficient power to detect meaningful differences between these subgroups even if such a difference exists. There was evidence of substantial heterogeneity within the subgroups themselves (prior to IVF: I² = 89%; during IVF: I² = 80%) and between trials (I² = 82%).

To reduce heterogeneity in the ‘prior to IVF’ subgroup, Mahran et al. 20 was excluded from this analysis, resulting in no evidence of heterogeneity within the ‘prior to IVF’ subgroup (I² = 0%), but substantial evidence of heterogeneity remained in the ‘during IVF’ subgroup (I² = 89%) and moderate evidence of heterogeneity between trials (I² = 46%) (see Appendix 33). This did not affect the conclusions of the analysis, identifying a non-significant interaction between subgroups (p = 0.46), but with the point estimates still suggesting that ES influenced LBR in two different directions, with an effect when ES was undertaken prior to IVF of 1.13 (95% CI 0.97 to 1.33), and the point estimate and CIs of the ‘during IVF’ subgroup remaining the same (OR 0.62, 95% CI 0.13 to 3.04). The conclusions were the same when a fixed-effects model was used (see Appendix 34).

Miscarriage

Ten trials were included in this meta-analysis,14,18–22,28,35,87,89 all undertaking ES in the menstrual cycle prior to IVF, which showed no evidence of a significant reduction when undertaking ES on this outcome when using a random-effects model (Figure 20; OR 0.96, 95% CI 0.57 to 1.63). There was evidence of moderate between-trial heterogeneity (I² = 46%). This conclusion was consistent with sensitivity analysis results from a fixed-effects model (see Appendix 37). Removal of Mahran et al. 20 from this analysis did not substantially reduce the evidence of heterogeneity (I² = 40%; see Appendix 38) and also did not alter the conclusions (OR 1.12, 95% CI 0.64 to 1.96).

FIGURE 20.

Forest plots of all trials reporting miscarriage rates. df, degrees of freedom; M–H, Mantel–Haenszel test.

However, this result should be interpreted with caution, as the method of calculating miscarriage rate varied substantially between studies. One trial collected this measure at 12 weeks,19 six trials between 20 and 24 weeks,14,20,28,35,87,89 and one trial collected both ‘early’ (prior to 12 weeks) and late (12–24 weeks) outcomes. 22 The timing of the collection of this outcome was unknown in two trials. 18,21 We therefore, undertook a subgroup analysis (Figure 21), separating those trials that collected miscarriages at an ‘early’ time point, and those that reported miscarriages at a ‘late’ time point. Izquierdo Rodriguez et al. 22 was included in both analyses by including both measures of miscarriage in the ‘late’ subgroup analysis (however, no late miscarriages were recorded). There was no evidence of a significant difference in miscarriage rates between subgroups when a random-effects model was used (p = 0.50), with no evidence of heterogeneity between pooled subgroup effects (I² = 0%). This result was consistent when a fixed-effects model was used (see Appendix 39).

FIGURE 21.

Subgroup analysis of miscarriages, by early and late miscarriage definitions. df, degrees of freedom; M–H, Mantel–Haenszel test.

Subgroup analyses

Analyses to assess the effects of the prespecified subgroups (age, duration of infertility and IVF protocol) could not be undertaken, as the included trials did not present outcome data individually for these subgroups.

Clinical effectiveness

We did not find evidence that performing the ES procedure before IVF increases the chance of achieving pregnancy and a live birth in women undergoing IVF for the first time. The observed LBRs were similar between treatment groups with 37.1% in the TAU group and 38.6% in the ES group; a very small increase in LBR of 1.5% (95% CI –4.4% to 7.4%; p = 0.621) was attributed to the ES procedure, which is not statistically significant and not of clinical importance. Sensitivity analyses under different assumptions concerning missing data and adjusting for prespecified covariates produced similar results. In the per-protocol analysis, ES increased LBR by only 3.6% (95% CI –3.3% to 10.5%; p = 0.312) and there was only a 2.9% probability that ES increases the LBR by at least the 10% targeted AD, given the observed data; therefore, this observed effect was not statistically significant and extremely unlikely to be of clinical importance.

Across secondary outcomes, we did not find evidence to support the beneficial effects of the ES procedure; rates were similar between groups. These secondary outcomes were implantation, clinical pregnancy, multiple birth, preterm delivery, miscarriage, ectopic pregnancy and stillbirth.

In general, the effect of ES on pregnancy and live birth appears consistent across prespecified subgroups, including protocol followed (long or antagonist), fertilisation method (IVF, ICSI or split IVF and ICSI), nature of embryo transfer (single or double embryo transfer), embryo transferred (fresh or frozen) and history of miscarriage (0–2 or ≥ 3). However, we could not rule out potential benefits when embryos are transferred on day 5, or in women who had cycle programming performed using oral contraception, progestogens or oral oestrogen. Readers should interpret these exploratory results with caution owing to the small numbers of participants and events within each subgroup.

Safety and tolerability

The ES procedure was found to be well tolerated, with high levels of reported tolerability (99.8%) and low pain levels, with a median pain score (IQR) of 4.0 (2.0–6.0) within 30 minutes post procedure, 1.0 (0.0–3.0) at 1 day and 0.0 (0.0–0.0) at 7 days post procedure. ES was found to be safe; rates of clinical outcomes that also relate to safety (ectopic pregnancy, miscarriages, preterm delivery, multiple births and stillbirth) and the incidences of AEs and SAEs were comparable between groups. There were slightly lower rates of babies born with low birthweight, very low birthweight, or who were small for their gestational age in the ES group than in the TAU group. However, these results should be interpreted with caution, owing to these events being highly correlated, based on small numbers and not adjusted for other potential confounding factors.

No deaths or severe congenital abnormalities were recorded. Only 2.4% (11/458) of women experienced at least one unexpected AE that occurred between ES and IVF (in the ES group).

Health economics

The cost-effectiveness analysis found that ES was, on average, £316 more expensive (£6819.38) than TAU (£6503). The ICER per successful live birth was calculated at £11.90 per woman (95% CI –£134 to £127), meaning that in order to achieve one extra live birth in 1000 women £11,900 would need to be spent. The ICER was similar across secondary outcomes, prespecified subgroups and sensitivity analyses.

Meta-analyses

To corroborate evidence of clinical effectiveness, we undertook a systematic review combining the results of this trial with other trials assessing the effectiveness of ES prior to the first IVF cycle, using meta-analysis. One trial contributed substantial heterogeneity to the analyses, possibly owing to an eligibility criteria stipulating that participants had to have two good-quality embryos transferred. 20 When this trial was excluded from the analyses, heterogeneity was reduced to minimal or moderate levels, finding no evidence that ES had a significant effect on LBR (nine trials: OR 1.03, 95% CI 0.87 to 1.22) and pregnancy rate (11 trials: OR 1.06, 95% CI 0.84 to 1.35), using random-effects models. We found no evidence of ES having a significant effect on ongoing pregnancy rate (four trials: OR 0.86, 95% CI 0.49 to 1.48), miscarriage rate (10 trials: OR 0.96, 95% CI 0.57 to 1.63) or multiple pregnancy rate (five trials: OR 1.09, 95% CI 0.68 to 1.74), also using random-effects models. Owing to a small number of trials (n = 5) reporting ectopic pregnancy rates, and the small number of events these trials reported, the results of a meta-analysis of ectopic pregnancy rates was highly uncertain (OR 0.66, 95% CI 0.17 to 2.51). Results were consistent across all outcomes when fixed-effects models were used. We did not undertake a meta-analysis of implantation rates because of concerns around this outcome being biased by the number of embryos transferred, which has been highlighted elsewhere. 91 Only two trials reported stillbirth rate, and one reported preterm delivery rate; therefore, a meta-analyses was not undertaken for these outcomes.

Only six of the trials reported pain post procedure;14,18,20,28,35,89 in those that did, few participants reported severe pain, with three studies reporting a numerical measure of pain, all three of which reported moderate pain post procedure. 14,35,87 Eight of the trials reported AEs;14,17–20,28,35,89 however, only our trial recorded such events in both ES and TAU arms of the trial, limiting the conclusions that can be made. Four trials reported AEs in the fetus or neonate;14,22,35,87 however, data were not obtained for three of these,14,22,87 also limiting the conclusions that can be drawn.

Qualitative substudy

To inform the design and conduct of future, related trials, we undertook a qualitative substudy to explore trial participants’ and site staff members’ experiences of being involved in the trial. Of the small proportion of trial participants (n = 27) and site staff (n = 7) who took part in the qualitative interviews, both participants and staff were generally happy with the recruitment process, and participants were well prepared to receive the ES procedure. Three participants discussed recruitment being too informal. Some participants and site staff felt that more information was required regarding the evidence base for or against ES. Eighteen of the interviewed participants discussed having positive preconceptions regarding the effect of the ES on the outcome of their IVF cycle, with site staff, in some instances, appearing to have contributed to these positive preconceptions. Some participants seemed to be unaware of the possibility that the ES could potentially reduce their chances of a live birth, instead choosing to focus on the potential positive impact that the intervention may have.

Participants’ positive preconceptions meant that the recruitment process was challenging for both staff and participants owing to some participants experiencing demoralisation when they were randomised to the control arm. Site staff developed mechanisms to assist participants in coming to terms with this. The potential for participation in the trial to delay the IVF cycle was also challenging. Five participants described the procedure as being more painful than they expected. Participants therefore recommended that more information regarding the potential painfulness of the ES is provided in trial recruitment materials.

Those participants who declined the ES procedure did so, on the whole, because of personal reasons that could not have been prevented (e.g. holidays and illness). However, the withdrawal of two participants could have been prevented with improved organisation by the fertility unit; a smear test was missed in one participant, and the other was not told to abstain from sexual intercourse prior to ES.

Strengths

This well-powered, pragmatic RCT recruited the targeted number of women undergoing IVF across 16 UK fertility units in both the private and the public sectors. To the best of our knowledge,16,92 this is the largest definitive trial evaluating the effect of ES procedure in women undergoing first-time IVF, thus bridging an important evidence gap on the benefits and safety of IVF ‘add-ons’. 2,8,93,94 Furthermore, the observed levels of clinical effectiveness and safety of the ES procedure are likely to be reflected in clinical practice because of the pragmatic nature of the trial,95 which mimicked the real world by recruiting participants from both NHS and private facilities across the UK, including both long and antagonist protocols, and by allowing pragmatic delivery of the ES.

Another major strength of this study is the specific focus on one particular population in order to minimise heterogeneity. The potential confounding effect of embryo quality was addressed by including only good responders who were likely to have a single embryo replacement. Indeed, our results show that the majority of participants responded well to stimulation and approximately 80% received a SET on day 5. We excluded patients with factors that could potentially influence endometrial quality (anovulation, BMI ≥ 35kg/m² and severe endometriosis).

There was a very high uptake (86.6%) of the ES procedure as per protocol in the ES group. The proportion of women who received IVF in both treatment groups was very high and similar (95.0% in the ES group and 94.1% in the TAU group). The contamination in the TAU group was negligible with only 1% (5/525) of women receiving ES procedure outside the trial, indicating high compliance and negligible contamination of the TAU group.

Last, we undertook a meta-analysis of all the available evidence within RCTs, including the results from our trial, which highlighted that, to our knowledge, this is the only trial that has comprehensively reported AE data both in participating women and in born babies.

Weaknesses

The study did not include a sham procedure in the control group. However, we do not believe that this would have influenced our results, as the objective fertility outcomes used in this study are unlikely to be influenced by a placebo effect.

Nearly 10% (9.2%; 48/523) of women who were randomised to the ES group did not receive the ES procedure, but they subsequently received IVF. These women technically received TAU although they were randomised to receive the ES procedure. As a result, this could have potentially diluted the effect of the ES procedure under the ITT principle. However, post hoc analysis that excluded women who did not receive the allocated treatment (ES plus IVF in the ES group or IVF in the TAU group) or those who were allocated to TAU but received the ES procedure outside the trial produced similar results to the main analysis, showing a non-significant effect of ES on live birth.

For ethical reasons, and under the advice from the PPI group, women who did not become pregnant or experienced a negative outcome, such as a loss of pregnancy, were not followed up further. 35 As a result, safety outcomes after these events were not collected in these participants, which may have resulted in some important events being missed, and the impact of interventions underestimated. Readers should bear this in mind when interpreting safety results.

A small proportion (0.9%; 10/1048) of women were found to be ineligible following randomisation, although these individuals remained in the trial and were followed up. Reasons for ineligibility included the identification of medical/gynaecological conditions (e.g. fibroids), slightly high BMI or age. However, this pragmatic trial mimicked routine practice where such issues occur and these numbers were very negligible relative to those randomised. Furthermore, sensitivity analyses that included per-protocol analysis were performed that excluded these women, and this produced consistent results.

The qualitative substudy included only participants who had participated in the trial, and, therefore, participants’ reflections on their experience of the recruitment process and delivery of the ES procedure may have been biased by the fact that they had agreed to participate (also known as survivorship bias). Those that withdrew from the trial were not approached for interview, meaning that the experience of those who could not tolerate the trial protocol, or who experienced particular demoralisation when randomised to the TAU arm, may have been under-represented. Interviews with fertility unit staff were undertaken by members of the central research team. The staff would have been aware of the interviewers’ involvement in the trial, which may have also biased how they reported their experience.

Owing to time constraints, we were unable to contact all authors within the systematic review to collect information to aid the assessment of the risk of bias. For one trial included in the review,88 we were unable to obtain the full-text article or correspond with the author in order to obtain data and, therefore, data for this trial were extracted from a recent systematic review. 16 As a result, a risk-of-bias assessment for this trial could not be undertaken.

In order to compare the quality of embryos between fertility units, we developed a method of converting between the three embryo grading techniques used across the trial. However, this was developed by one principal embryologist, and therefore consensus was not sought between fertility units. This method of comparing embryo quality was not validated and therefore may be open to bias.

Implications for health care

Despite a lack of evidence to support its use, some in the medical community have adopted ES, including for those undergoing their first IVF cycle. 27 Evidence from this trial concludes that the ES procedure does not significantly improve pregnancy outcomes for women undergoing their first IVF cycle. Furthermore, an updated meta-analysis of trials in this population, including unpublished data obtained from study authors, demonstrates a consistent overall effect, across all included trials and outcomes, of no benefit.

Implications for future research studies

Recruiting staff should be aware that participants may base their decision to participate on internet research, which may result in positive preconceptions of the potential benefits and a downplaying of the potential harms of receiving the novel intervention. Recruiting staff should try to introduce inclusion in the control group with similar emphasis to inclusion in the intervention group; however, once randomised to the control group discontent may still be felt, in which case we suggest three ways to lessen this, namely, explaining the importance of the control group, the lack of evidence to support the intervention, and the increased contact with patients that participating in research allows. Other studies have provided recommendations to guide ‘recruitment conversations’ in surgical trials; these guidelines are also applicable to studies in fertility. 107–109

The quantity of information provided to participants during the recruitment stages of fertility studies may need to be revised. Other methods of providing such information to potential participants should be explored, possibly by providing participants with a choice of levels of information, as suggested in previous studies. 101,110

Qualitative and quantitative evaluations of the experience of receiving ES highlighted the variability in pain experienced by patients, with some participants experiencing more severe pain post procedure. This information should be taken into account when undertaking ES in other populations.