Aspirin for in vitro fertilisation

Types of studies

All parallel‐design randomised controlled trials (RCTs), published or unpublished, which addressed the objectives of the review. Non‐randomised and quasi‐randomised trials were excluded as they are not randomised and are therefore more prone to bias. From crossover trials, only data from the first phase were included in meta‐analyses.

Types of participants

Women undergoing in vitro fertilisation (IVF) or intra‐cytoplasmic sperm injection (ICSI), or both, and their partners.

Types of interventions

Aspirin, versus placebo, any other active intervention or no treatment, taken either pre‐conceptually or at different stages of the treatment cycle (for example during down‐regulation, during stimulation of ovulation, after oocyte collection or after confirmation of pregnancy by a pregnancy test or ultrasound). Aspirin treatment given for varying lengths of time was considered (for example up to different points in the treatment cycle; until confirmation of pregnancy; up to 12 weeks' gestation, 34 weeks' gestation or delivery). We defined low dose as a dose of 150 mg or less taken once per day. Trials reporting on the synchronous use of aspirin with any other treatment were excluded from the review.

Types of outcome measures

These were divided into primary and secondary outcomes.

Electronic searches

All reports which appeared to describe RCTs of aspirin for women undergoing IVF were obtained using the following search strategies (from the inception of the databases to 9 May 2016):

(1) The Cochrane Gynaecology and Fertility Group Specialised Register of controlled trials was searched for any trials (Appendix 1).

(2) The Cochrane Central Register of Controlled Trials (CENTRAL CRSO) was searched in all fields (Appendix 2).

(3) The electronic databases MEDLINE (Appendix 3), Embase (Appendix 4), and PsycINFO (Appendix 5) were searched using search terms that included "(aspirin OR acetylsalicylic acid) AND (in‐vitro fertilisation OR intracytoplasmic sperm injection)". The MEDLINE search was combined with the Cochrane Menstrual Disorders and Subfertility Group's Highly Sensitive Search Strategy for identifying randomised controlled trials which appears in the Cochrane Handbook for Systematic Reviews of Interventions (Version 5.0.2 chapter 6, 6.4.11).

(4) Published or ongoing trials in the trial registers for ongoing and registered trials: 'ClinicalTrials.gov', a service of the US National Institutes of Health (clinicaltrials.gov/ct2/home) and www.controlled‐trials.com, as well as the World Health Organization International Trials Registry Platform search portal (who.int/trialsearch/Default.aspx) (Appendix 6).

The latest search was run on 9 May 2016. All searches have been date limited from 1 January 2014 to 9 May 2016. following the same search strategy as mentioned above.

Searching other resources

The citation lists of any relevant publications identified (including review articles and included studies) were searched for further potentially eligible studies; as were articles in books identified through the citation lists of any relevant publications together with internal reports and conference proceedings.

Selection of studies

During the update of this review, the search was largely supported by the Gynaecology and Fertility Group Information Specialist. C Siristatidis and P Vogiatzi independently selected trials among all potentially eligible studies and disagreements were resolved by discussion. The same two authors independently examined these full‐text articles for compliance with the inclusion criteria and selected studies eligible for inclusion in the update of the review. Further information was sought from the authors of studies where insufficient information was included in the papers to make a decision about eligibility; e‐mails were sent seeking the necessary clarifications. In order to allow an intention‐to‐treat analysis for binary outcome measures, data were sought on the number of participants with each outcome event by allocated treatment group, irrespective of compliance and whether or not the participant was later thought to be ineligible or otherwise excluded from treatment or follow‐up.

During the initial version, from the trials identified by the above search strategies, V Poustie and A Drakeley independently selected trials which were eligible for inclusion based on the criteria for considering studies for the review (as detailed above). Disagreements were resolved by discussion or through arbitration by S Dodd. As a large number of studies were identified, V Poustie and A Drakeley independently screened the titles and discarded those studies that were clearly ineligible. The abstracts of the remaining studies were reviewed independently by V Poustie and A Drakeley. The full papers of any studies which appeared to be eligible, or for which we could not confirm eligibility, were obtained and reviewed independently. Further information was sought from the authors of studies where insufficient information was included in the papers to make a decision about eligibility. During the current update of the review, the same strategy was followed, performed by C Siristatidis and P Vogiatzi.

Data extraction and management

Data were extracted independently by C Siristatidis and P Vogiatzi using a data extraction form designed and pilot‐tested by the authors and disagreements between the two aforementioned authors were resolved by consensus with G Basios. Where studies had multiple publications, the authors collated multiple reports of the same study, so that each study rather than each report is the unit of interest in the review, and such studies have a single study ID with multiple references. We sought additional information on trial methodology or actual original trial data from the principal author of trials which appeared to meet the eligibility criteria when aspects of methodology were unclear, or where the data were unpublished or in a form unsuitable for meta‐analysis. The following quality criteria and methodological details are presented in the table 'Characteristics of included studies' and help provide a context for discussing the reliability of the results.

Analysis was carried out using Review Manager 5 software.

Where rates were reported, such that women may be counted more than once (for example 'per embryo transfer' or 'per cycle'), or reporting in percentages, rates 'per woman or couple' were calculated and reported if sufficient data were available, either extracted from papers or obtained from the trial authors. Binary outcome measures for each study were expressed and pooled using relative risks (RR) with 95% confidence intervals (CI). We presented relative risks in preference to odds ratios (OR) as they are more easily understood where event rates are high, as may have been the case in these studies.

Assessment of risk of bias in included studies

The risk of bias of all studies that were deemed eligible for the review was assessed independently by C Siristatidis and G Basios.

The Cochrane 'Risk of bias' assessment tool (handbook.cochrane.org) was used to assess: selection bias (random sequence generation and allocation concealment); performance bias (blinding of participants and personnel); detection bias (blinding of outcome assessors); attrition bias (incomplete outcome data); reporting bias (selective reporting); and other bias. Disagreements were resolved by discussion. All judgements were fully described and presented in the conclusions through the 'Risk of bias' table.

Measures of treatment effect

Risk ratios (RR) and 95% confidence intervals (CIs) were calculated for all outcomes.

Unit of analysis issues

The primary analysis was per woman/couple randomised. Multiple live births (e.g. twins or triplets) were counted as one live birth event.

Only first‐phase data from crossover trials were included.

Dealing with missing data

Only the available data were analysed. In the cases that missing data were considered significant, the original investigators were contacted by e‐mail to request missing data or information. Live births were assumed not to have occurred in participants without a reported outcome. For other outcomes, we analysed only the available data.

Assessment of heterogeneity

Statistical heterogeneity between the results of different studies was examined by inspecting the scatter in the data points and the overlap in their confidence intervals in the forest plot and, more formally, by checking the results of the Chi² test for heterogeneity and the I² statistic.

Assessment of reporting biases

Funnel plots were used to explore the possibility of small study effects. A funnel plot was not presented due to the small number of studies reporting the primary outcome.

Data synthesis

Low‐dose aspirin was compared to placebo or no treatment and data for each outcome were pooled using a fixed‐effect model.

Subgroup analysis and investigation of heterogeneity

If sufficient studies had been identified (five or more), we would have performed subgroup analyses on the results to look at the possible contribution of:

1. Differences in duration and timing of treatment.

2. Couples with/without infertility.

3. Aspirin dose ‒ higher or lower than 150 mg.

If heterogeneity was evident, it was investigated by carrying out the pre‐planned subgroup and sensitivity analyses stated below. If unexplained heterogeneity remained, it was incorporated by using random‐effects modelling. An I² value greater than 50% was predefined as the cut‐off for further exploration.
 Exploratory subgroup analyses were undertaken to assess the effect of timing of treatment (started pre‐conceptually or at first confirmation of pregnancy). Further subgroup analysis would have been undertaken to assess the effects of duration of treatment, had sufficient studies been identified to allow this.

Sensitivity analysis

If sufficient studies had been identified (five or more), we intended to perform sensitivity analysis to determine whether the conclusions are robust to arbitrary decisions made regarding the eligibility and analysis, according to study risk of bias, including only studies at low risk of bias related to:

(1) sequence generation;
 (2) allocation concealment;
 (3) loss to follow‐up;
 (4) blinding.

Results of the search

The search retrieved 1450 articles. Eighty‐six studies were potentially eligible and were retrieved in full text. Fifty‐six studies met our inclusion criteria. We excluded 41 studies, there were no studies awaiting further assessment and two studies are ongoing. Complete characteristics of the studies can be found in the respective tables, Characteristics of included studies, Characteristics of excluded studies, Characteristics of ongoing studies, while a complete flow diagram of study selection process is presented in Figure 1.

Included studies

Thirteen studies involving a total of 2653 women met the criteria for inclusion in the review (Bordes 2003; Check 1998; Dirckx 2009; Duvan 2006; Haapsamo 2010; Hung Lok 2004; Lambers 2009a; Lentini 2003; Moini 2007; Pakilla 2005; Rubinstein 1997‐1999; Urman 2000; Van Dooren 2004).

Excluded studies

Forty‐one studies were excluded from the review. Twelve of these investigated the use of aspirin alongside prednisone, heparin or terbutaline therapy (Geva 2000; Grandone 2014; Guan 2007; Hanevik 2012; Mollo 2002; Revelli 2008; Sher 1994; Stern 2001; Stern 2003; Strehler 2002; Ubaldi 2002; Zhu 2013), thirteen were not randomised studies (Akhtar 2013; Frattarelli 2008; Gizzo 2014; Hatasaka 2000; Hurst 2005; Kuo 1997; Kutteh 1997; Sher 1998; Stoval 1999; Villani 2015; Wada 1994; Wallace 2003; Ying 2012), four were quasi‐randomised, while in two studies participants were allocated to treatment group according to which day of the week they attended the clinic (Waldenström 2004); or on a weekly basis (Pergolini 2013). In another two, allocation was based on the paired or unpaired status of participants' social security number (Várnagy 2010; Weckstein 1997). Rabiee 2011 compared different regimens of aspirin in IVF. Haapsamo 2009 measured uterine artery haemodynamics after aspirin treatment; Lambers 2009 described the results after a questionnaire given to participants of the original study; and two studies did not refer to IVF or ICSI (Hsieh 2000; Kosar 2011). Another study was excluded because researchers investigated the impact of low‐dose aspirin on patients undergoing ovulation induction with a Chinese patent medicine and, thus, the outcomes were not relevant to IVF (Zhao 2014). Four studies were excluded as the authors failed to respond to our request for additional information concerning their study (Matassa 2001; Madani 2014; Lee 2001; Vandana 2014); two studies were conducted in the framework of an oocyte donation programme (Rubinstein 1999; Weckstein 1997); while in Zolghadr 2011 the outcomes investigated were number of follicles, serum estradiol and ovarian size, thus greatly differing from the outcomes of this analysis.

Two ongoing studies have been registered as clinical trials: one has completed recruitment with no results published to date (Gourabi 2012), and the second was due in December 2014 (Vandana 2013); however, to date, the authors have not provided their results. The authors of both studies have been contacted in order to provide some preliminary results to assess whether these are eligible to be included in the analysis, but no information is available to date.

Allocation

Blinding

Nine out of 13 studies were double blind, with both the participants and treatment providers reported as blinded (Dirckx 2009; Duvan 2006; Haapsamo 2010; Hung Lok 2004; Lambers 2009a; Pakilla 2005; Rubinstein 1997‐1999; Van Dooren 2004; Moini 2007); one was at least single blind as the assessors were blinded (Check 1998); one reported to be double blind but no further information was given (Bordes 2003); and one study was single blind as the treatment providers were unaware of which group the participants belonged to (Urman 2000).

Eight trials were reported as placebo controlled (Bordes 2003; Haapsamo 2010; Hung Lok 2004; Lambers 2009a; Moini 2007; Pakilla 2005; Rubinstein 1997‐1999; Van Dooren 2004); in one study tablets of aspirin or placebo of identical appearance were given (Haapsamo 2010); in another a similar dose of 100 mg of placebo was reported (Lambers 2009a).

Incomplete outcome data

Three studies were assessed as at low risk of attrition bias, as they provided rates of the primary outcome (Dirckx 2009; Haapsamo 2010; Pakilla 2005). Three others provided pregnancy rates as a percentage rather than actual numbers (Bordes 2003; Lentini 2003; Moini 2007). These studies did not state how many women completed the study or withdrew from treatment, so that it is unknown whether the rates were calculated per woman, per transfer or per cycle. They have not been included in the analysis.

Intention‐to‐treat analysis was clearly stated in three studies (Haapsamo 2010; Lambers 2009a; Rubinstein 1997‐1999). Three of the studies did not specifically state that an intention‐to‐treat analysis was undertaken; however they presented results on all recruited participants and reported no withdrawals (Check 1998; Duvan 2006; Rubinstein 1997‐1999).

Withdrawals were reported in eight studies: they were stated either as none (Rubinstein 1997‐1999); or they represented less than 10% of the study population (Dirckx 2009; Haapsamo 2010; Hung Lok 2004; Lambers 2009a; Pakilla 2005; Urman 2000), while in one study cancellation rates were stated as more than this percentage at 16.7% and 23.3% of the study and control groups, respectively (Moini 2007).

Selective reporting

Four out of 13 studies were at low risk of reporting bias (Dirckx 2009; Haapsamo 2010; Lambers 2009a; Pakilla 2005), while the other nine were at unclear risk (Bordes 2003; Check 1998; Duvan 2006; Hung Lok 2004; Lentini 2003; Moini 2007; Rubinstein 1997‐1999; Urman 2000; Van Dooren 2004).

Other potential sources of bias

Inclusion and exclusion criteria were clearly described and treatment and control group participants appeared to be comparable in eight studies (Dirckx 2009; Check 1998; Duvan 2006; Haapsamo 2010; Hung Lok 2004; Lambers 2009a; Pakilla 2005; Urman 2000). The remaining studies provided insufficient detail to reach a decision on these factors. In most of the studies the general care programmes in the treatment and control groups were identical, although one study failed to provide this information (Lentini 2003).

Compliance to aspirin treatment was not assessed in all but one of the studies (Hung Lok 2004)

We found no other potential sources of within‐study bias in the 13 included studies.

Primary outcome measure

Secondary outcome measures