The prognostic profile of subfertile couples and treatment outcome after expectant management, intrauterine insemination and in vitro fertilisation: a study protocol for the meta-analysis of individual patient data

Authors


Drs NM van den Boogaard, Academic Medical Centre, Department of Obstetrics and Gynaecology, Room H4.140.1, PO Box 22660, 1100 DD Amsterdam, the Netherlands. Email n.m.vandenboogaard@amc.uva.nl

Abstract

Please cite this paper as: Boogaard N van den, Hompes P, Barnhart K, Bhattacharya S, Custers I, Coutifaris C, Goverde A, Guzick D, Litvak P, Steures P, Veen F van der, Bossuyt P, Mol B. The prognostic profile of subfertile couples and treatment outcome after expectant management, intrauterine insemination and in vitro fertilisation: a study protocol for the meta-analysis of individual patient data. BJOG 2012;119:953–957.

Objective  The current evidence concerning the best treatment option for couples with unexplained and male subfertility is inconclusive. Most studies that have evaluated the effectiveness of treatment options, such as expectant management (EM), intrauterine insemination (IUI), with or without controlled ovarian stimulation (COS), and in vitro fertilisation (IVF), have not taken the couples’ prognosis into account. It is very likely that the individual prognosis of the couple influences the effect of treatment. Individual patient data analyses allow us to take these prognostic factors into account, and to evaluate their effect on treatment outcome. This study aims to use anonymised data from relevant published trials to perform an individual patient data meta-analysis, evaluating the effect of couples’ prognosis on the effectiveness of EM, IUI, with or without COS, and IVF.

Methods  Based on earlier systematic reviews and an updated search, randomised controlled trials will be considered for inclusion. Untreated subfertile couples with unexplained or male subfertility included in trials comparing EM, IUI, with or without COS, and IVF are included. Authors of the included studies will be invited to share their original anonymised data. The data will be assessed on validity, quality and completeness. The prognosis of the individual couple will be calculated with existing prognostic models. The effect of the prognosis on treatment outcome will be analysed with marker-by-treatment predictiveness curves, illustrating the effect of prognosis on treatment outcome. This study is registered in PROSPERO (registration number CRD42011001832).

Conclusion  Ultimately, this study may help to select the appropriate fertility treatment, tailored to the needs of an individual couple.

Introduction

Subfertility affects at least 10% of couples trying to conceive.1,2 In approximately half of them, no major underlying cause is found.3 Although intrauterine insemination (IUI), with or without controlled ovarian stimulation (COS), is often the first step in the treatment algorithm for these couples, the evidence for the effectiveness of IUI over expectant management (EM) remains inconclusive.4,5 The results of trials comparing IUI alone, IUI with COS, or EM with each other have been pooled in several meta-analyses.6–9

A review on IUI for couples with unexplained subfertility showed a significant increase in pregnancy rates for treatment with both IUI and COS, separately. Data on multiple pregnancies and other adverse events after treatment with COS were insufficient to allow conclusions to be drawn.9 A review of IUI for male subfertility concluded that there was insufficient evidence to recommend IUI, with or without COS, above EM, or vice versa.6 A review including studies with unexplained, male and cervical subfertility found higher pregnancy rates for IUI and COS in couples with unexplained subfertility, but not in couples with good prospects of natural conception. In couples with cervical factor and male subfertility, IUI alone led to higher pregnancy rates.8 However, the quality of the included trials was poor, the sample sizes were small, and complications like multiple pregnancies were poorly reported.

In vitro fertilisation (IVF) and intracytoplasmic sperm injection (ICSI) were initially introduced to help couples with infertility caused by the inability of the male and female gametes to meet or the inability of the spermatozoa to penetrate the egg. Nowadays, IVF and ICSI are also used for couples in whom these conditions are not met, and who thus have a chance of natural conception. The effectiveness of IVF as primary treatment or after failed IUI in these couples is debatable. The pooled results of trials that compared IVF with EM or IUI, with or without COS, in couples with mainly unexplained or mild male subfertility are difficult to interpret, because of the heterogeneity of the studies and the lack of prognostic information about the couples in relation to treatment outcome (Pandian et al. 2010). In a cohort of newly referred subfertile couples, the contribution of IVF in couples with unexplained subfertility and ovulation disorders was extremely limited, with ongoing pregnancy rates of 13 and 4.5%, respectively, compared with patients with tubal factors, endometriosis and male factors, in whom pregnancy rates were 45, 45 and 37%, respectively.10 Here, just like in the IUI studies, there was limited information about the influence of prognostic factors on treatment outcome.

This information can be derived from prognostic models that estimate the chances of natural conception,11,12 or the chances of conception following fertility treatment.13–15 The use of these models can help to discriminate between couples who would benefit from intervention and those who would not.

When conventional meta-analyses are inconclusive or contradictory, individual patient data meta-analysis (IPD-MA) can have additional value, because it allows this very evaluation of prognosis on treatment effect. This is important, because it may well be that, if treatment is tailored to couples with a low chance of conceiving naturally, the treatment effect increases, and vice versa. Therefore, we plan to perform an IPD-MA of randomised controlled trials evaluating the effect of couples’ individual prognosis on the effectiveness of EM, IUI, with and without COS, and IVF.

Objectives of the study

The main goal of this study is to evaluate the effect of prognosis on the effectiveness of EM, IUI, with and without COS, and IVF in couples with mainly unexplained subfertility using IPD-MA. Mainly unexplained subfertility is defined as couples in which the gametes are able to meet, and includes couples with unexplained and male subfertility. The prognostic models used in our analyses are: Hunault’s11 model, which predicts the chance of natural conception within 12 months; Steures’14 model, which predicts the chance of pregnancy after IUI, with and without COS; and Templeton’s15 model, which predicts the chance of pregnancy after IVF. Those models were selected because they are the only models that performed well in the external validation.16

Hunault’s model includes the following predictors: female age; duration of subfertility; primary or secondary subfertility; the percentage of motile sperm; and the referral status. Steures’ model includes: female age; duration of subfertility; cervical factor; male factor; tuba pathology; uterus anomaly; endometriosis; and the use of clomifene or recombinant −FSH. Templeton’s model includes: female age; duration of subfertility; tubal subfertility; livebirth after IVF; livebirth that was not a result of IVF; a previous pregnancy after IVF that did not result in a livebirth; and a previous pregnancy not after IVF that did not result in a livebirth.

Methods

Literature search and data sharing request

Previously, systematic reviews of trials comparing the included treatment options for each diagnostic subgroup, i.e. unexplained, and male and cervical factor subfertility, have been performed.6–9,17 By means of these reviews we will identify studies for our IPD-MA. We will update the performed search strategies to include studies published up to date and we will check references and ask authors whether they are aware of unpublished studies in progress. In the case of a crossover design, the authors will be asked to supply the data from before the crossover separately. Readers of this protocol, who are familiar with studies performed in this field that are not integrated in the previous performed meta-analyses, are also invited to approach us.

Registration

The protocol is registered with PROSPERO.com (registration number CRD42011001832).

Data acquisition

We plan to contact the first authors of the studies included by email, and in the case of no response we will also email the last authors. When there is no reaction we will try to contact them by phone. We will ask them to send the complete anonymised data set so as to minimise their efforts in going through their data set to select appropriate variables. We accept any data format, provided that variables and categories are adequately labelled within the data set or with a separate dictionary. All participants will be identified by study number. Names and addresses will not be included in any of the data sets sent in by primary research groups. Anonymised data from relevant trials (the results of which have all been published in peer-reviewed journals), will be amalgamated and stored in a password-protected University of Amsterdam computer at the Amsterdam Medical Centre, Amsterdam, and the usual precautions regarding confidentiality and access will be observed. The data set will not be used for any other research apart from that described in the protocol. The data will be stored for 5 years beyond the life of this project (2011–2013).

Quality assessment

For each trial, information on the quality of the trial will be extracted based on a number of items: power calculation; adequate randomisation; concealment of allocation; parallel design; and exclusions after randomisation. If a trial does not adhere to these standards, its validity will be considered compromised. Depending on the level of the compromise, the randomised trial can be excluded in the analysis. If there is a large variety in quality and completeness of data between different trials, separate analysis will be made in high- and poor-quality trials, next to analysing all of the data as a whole. The same treatment arms of different trials will be assessed for comparability. Completeness of the data sets in terms of prognostic markers and outcomes will be reported. The consistency of data and the published article will be assessed. Incomplete data or major inconsistencies with published results will be discussed with the primary investigators, and may lead to exclusion. There is no strict limit set for the fraction of acceptable missing data, as this depends on which variable and the type of missing data. The decision to include or exclude a study will be discussed in the project group.

Analysis

All the different steps of this protocol are summarised in Figure 1. Before the start of the analysis, we aim to make the variable codes of all the acquired data compatible. Missing data will be imputed using multiple imputations within the original studies. All the prognostic factors will be used as predictors for the imputations. The prognosis of the patients will be calculated based on the prognostic models of Hunault, Steures and Templeton.11,14,15 If heterogeneity allows, the original data will be merged into a single set. A study identification variable will be added to reflect the stratified nature of the pooled data set.

Figure 1.

 Flow chart of the study protocol.

Initial analyses will be performed for the primary outcome livebirth or ongoing pregnancy per couple, depending on the availability of the data. Secondary outcomes are livebirth or ongoing pregnancy per cycle, and multiple pregnancies per couple. Livebirth is defined as the delivery of at least one living child beyond 22 weeks of gestation. Ongoing pregnancy is defined as the presence of fetal cardiac activity at ultrasound at 12 weeks of gestation or later. The baseline characteristics and the outcomes will be summarised in a table, per study and overall.

A marker by treatment predictiveness curve will be plotted per trial, and per prognostic marker, to illustrate the treatment effect as a function of the prognosis.18 The calculated prognostic variable will be used as the marker. If there is a differential benefit from treatment, these curves will show the treatment threshold for the prognostic marker. In that case, some patients benefit from treatment, whereas others do not, and the marker can be used to make the selection.

The effect of the prognostic marker-based treatment strategy will be evaluated by comparing the outcome of treating all patients without taking the prognostic marker threshold into account, versus treating couples according to the prognostic marker threshold.19 Subsequently, we will evaluate the proportion of patients for whom treatment recommendations would change after using the prognostic marker-based treatment strategy.20 Finally, we will compare the effects of the different prognostic marker-based treatment strategies on pregnancy outcomes.

If there is a relationship between the treatment effect and the prognosis, and the curves show a treatment threshold per prognosis, an algorithm will be developed for clinical practice. This algorithm will help the clinician to choose the best treatment strategy for the individual patient. based on their prognosis.

Collaboration

Meetings To have the opportunity to discuss the project with the co-authors, meetings at international fertility congresses will be organised. During these meetings the project in general and the practical, methodological and data-related aspects can be discussed.

Authorship We plan to provide one co-authorship for each contributor of individual patient data for the articles that we publish. In case the number of authors is limited by journal editors, we propose to publish under a collaborative group name. For articles explicitly focusing on methodological aspects of IPD-MA, where these data are used for illustratory purposes, contributors will be individually acknowledged in each article. The results of this project will be presented at international conferences and published in clinical and epidemiological journals.

Competing interests

Some of the authors are authors of the original trials upon which the proposed IPD-MA is based.

Disclosure of interests

There are no conflicting interests in this study.

Contribution to authorship

BMW is the principal investigator of this study. NB is responsible for the overall logistical aspects of this study, drafted the first version of this article and will perform the analysis, together with PB, BMW, FvdV and PH. BM and PB will supervise the whole process. SB, IC, AJG, DG and PS have agreed to share data, and all authors have read and approved the final article.

Details of ethical approval

We propose to use only anonymised data from published trials. In response to our query, the Ethics Committee of the Academic Medical Centre in Amsterdam did not feel that formal Ethics Approval was required.

Funding

This study is financially supported by the Academic Medical Centre and the Vrije Universiteit Medical Centre.

Ancillary