Types of studies

Published and unpublished randomised controlled trials (RCTs) that assess the effectiveness and safety of endometrial injection of embryo culture supernatant before embryo transfer in women undergoing IVF/ICSI. We will exclude pseudo‐randomised and crossover trials. We will not apply limitations in terms of country of origin and language.

Types of participants

Women and couples undergoing IVF/ICSI cycles (both fresh and frozen). We will exclude oocyte donation cycles.

Types of interventions

Endometrial injection of embryo culture supernatant before embryo transfer, during an artificial reproductive cycle with IVF or ICSI, versus any other intervention or no intervention (usual care). In case of the existence of data describing a sham or placebo‐type intervention, we will perform a stratified comparison as described in the ‘Data synthesis' section.

Types of outcome measures

Electronic searches

We will search the following databases from inception to present: the Cochrane Gynaecology and Fertility Group Specialised Register (Appendix 1), the Cochrane Central Register of Controlled Trials (CENTRAL CRSO) (Appendix 2), MEDLINE (Appendix 3), Embase (Appendix 4), and CINAHL Plus (Appendix 5). All searches will be carried out without any language or date restriction.

We will combine the MEDLINE search with the Cochrane highly sensitive search strategy for identifying RCTs that appears in the Cochrane Handbook of Systematic Reviews of Interventions (Chapter 6, 6.4.11; Lefebvre 2011). We will combine the Embase and CINAHL searches with trial filters developed by the Scottish Intercollegiate Guidelines Network (SIGN) www.sign.ac.uk/methodology/filters.html#random.

We will search the World Health Organization International Clincal Trials Registry Platform (http://apps.who.int/trialsearch/Default.aspx) and the ClinicalTrials.gov registry for ongoing and registered trials (Appendix 6). We will also search OpenGrey (www.opengrey.eu/) for grey literature. We will consult experienced clinicians for any ongoing or existing studies.

Searching other resources

We will examine the references lists of all studies (included and excluded) and relevant reviews in order to identify further relevant articles.

Selection of studies

Two review authors will independently screen the titles and abstracts of the publications identified by the literature search strategy. They will exclude studies that do not meet the inclusion criteria and will retrieve the full‐text of remaining publications. Three review authors will evaluate these independently to identify RCTs eligible for inclusion. Any potential disagreements related to study eligibility will be resolved by discussion with the first review author (CSS). We will list all studies excluded after full‐text assessment in the ‘Characteristics of excluded studies' tables, and will outline the study selection process in a PRISMA flow chart.

Data extraction and management

Two review authors will independently extract study characteristics and outcome data from the included studies using a pre‐designed data extraction form. We will seek detailed information on participants, interventions, comparators, outcomes, study design, funding sources and declarations of interest for the primary investigators. For studies with multiple publications, we will use the main RCT report as the reference and we will supplement it with additional data from the secondary publications.

With the aim of retrieving additional data or methodological details where necessary, we will contact authors of the included studies. We will contact the authors via e‐mail, including a reminder if needed (a second e‐mail 15 days after the first communication, if we receive no reply or insufficient data). We will resolve any potential disagreements through consensus involving the first review author or the statistician/methodologist review author. One review author will import data into RevMan 5, and a second review author will validate the imported values against the data extraction form.

Assessment of risk of bias in included studies

Two review authors will independently assess risk of bias in the included studies using the Cochrane ‘Risk of bias’ assessment tool for selection, performance, detection, attrition, reporting and other bias (Higgins 2011). We will resolve any disagreements through discussion with review author CSS. We will explicitly report our judgement of risk of bias in the ‘Risk of bias’ table in the ‘Characteristics of included studies’ section with relevant information supporting our assessment. If necessary, we will group multiple outcomes.

In case we cannot make reliable judgements for a whole study, we will examine the risk of bias due to lack of blinding and missing data separately for different outcomes. Regarding blinding of medical personnel and patients, we expect to come across a considerable degree of bias owing to the nature of the intervention when compared to no intervention. In this case we will proceed on meticulous evaluation of the methods employed in each study to the extent of the presented material but also in request for further specifications.

We will also investigate the possibility of selective reporting through comparison of the protocol outcomes (if data are available) with the published study outcomes or the outcomes listed in the methods section with the reported results.

Measures of treatment effect

All defined outcomes will be binary (dichotomous), and we will use the numbers of events in the control and intervention groups of each study to calculate Mantel‐Haenszel odds ratios (ORs). We will use Peto ORs for outcomes with low event rates, when needed, as described in the ‘Data synthesis' section. We will reverse the direction of effect of individual studies, if required, to ensure consistency across trials. We will present 95% confidence intervals for the ORs.

Unit of analysis issues

We expect that all studies will have the woman (or couple) as the unit of randomisation. When data are not reported per woman (or couple), for example data reported ‘per cycle’, then we will make every effort to extract the data from the text or retrieve them through correspondence with the study authors, or both. If we are unsuccessful, we will summarise data to the furthest possible detail in narrative and in additional tables as necessary.

We will count multiple live births (twins, triplets) as a single live birth event.

Dealing with missing data

We will evaluate included studies to determine whether missing data are randomly distributed. Where data are missing, we will contact trial authors to retrieve as much information as possible. Where this is unobtainable, we will undertake imputation of individual values for our primary effectiveness outcome live birth/ongoing pregnancy only: live birth/ongoing pregnancy will be assumed not to have occurred in participants not reporting this. For other outcomes, we will analyse the available data.

Assessment of heterogeneity

We will initially consider whether the clinical and methodological characteristics of the included studies are consistent enough to provide a clinically meaningful results through data pooling in a meta‐analysis.

We will assess statistical heterogeneity by the measure of the I² statistic. An I² statistic value of 30% to 60% will be considered as moderate heterogeneity, and a value of 60% to 90% as substantial heterogeneity across studies (Higgins 2011). In case of substantial heterogeneity with important clinical impact for a specific outcome, we will explore possible explanations through subgroup and sensitivity analyses (where data are available) and will consider not pooling the data.

Assessment of reporting biases

Through a thorough search for published and unpublished data, we aim to minimise the potential impact of publication bias and other reporting biases. We will use a funnel plot to explore publication bias when the number of included RCTs in the same analysis is at least 10.

Data synthesis

All outcomes are dichotomous (binary), and we will combine the data from similar RCTs using a fixed‐effect Mantel‐Haenszel model. We will report the pooled odds ratios with a 95% confidence interval. Where events are rare, and if all relevant criteria are fulfilled, we will consider the Peto method for pooling the data. An increase in the odds of the outcome will be displayed graphically in the meta‐analyses to the right of the centre‐line, and a decrease in the odds of an outcome to the left of the centre‐line.

We will carry out pooled analyses for the following comparisons: women/couples randomised receiving endometrial injection of embryo culture supernatant before the embryo transfer, during an artificial reproductive cycle with IVF or ICSI, versus women/couples randomised receiving either:

• any other intervention during an artificial reproductive cycle with IVF or ICSI; or

• usual care (no additional intervention) during an artificial reproductive cycle with IVF or ICSI.

Subgroup analysis and investigation of heterogeneity

Clinical heterogeneity, as the diversity in interventions, may affect the results. Differences in the cross‐talk between endometrium and embryo could affect the outcome of the endometrium culture injection at different times before embryo transfer. In addition, the quality of the embryo on day 3 or 5 may be different; the blastocyst is considered to be more viable with higher rates of successful implantation and the number of previous ART cycles reflects the potential of each woman to conceive (more unsuccessful attempts reduce the possibilities). Where there is substantial heterogeneity (I² statistic value > 60%), we plan to determine effects for the primary outcomes within the following subgroups, if data are sufficient for any meaningful analyses:

• age of the woman (< 37 years, 38 to 41 years, > 42 years);

• day of embryo transfer (3 or 5);

• type of cycle (frozen or fresh);

• time of endometrium culture injection before embryo transfer.

Sensitivity analysis

We will conduct sensitivity analyses for the primary outcomes in order to determine whether the conclusions are robust to arbitrary decisions made regarding the eligibility and analysis. These analyses will include consideration of whether the review conclusions would have differed if:

• we restricted eligibility to studies without high risk of bias (studies at low risk of bias with respect to randomisation methods and not at high risk of bias in any domain);

• we had adopted a random‐effects model;

• we have considered publication type (abstract vs. full text);

• miscarriage had been analysed per pregnancy rather than per woman;

• the primary outcome was live birth alone (not including ongoing pregnancy).