Presence of antiphospholipid antibodies is associated with increased implantation failure following in vitro fertilization technique and embryo transfer: A systematic review and meta-analysis

Purpose A systematic review and meta-analysis was conducted comparing the presence of anti-phospholipid (anti-PL) antibodies between women of reproductive age, without diagnosis of antiphospholipid syndrome, who experienced at least two implantation failures following in vitro fertilization and embryo transfer (IVF-ET), and either women who had a successful implantation after IVF-ET or women with at least one successful spontaneous pregnancy or unselected healthy fertile women with no history of IVF-ET. Methods Systematic search of the literature and meta-analysis of the relevant studies studying presence of antiphospholipid antibodies in women experiencing at least two implantation failures in IVF-ET as compared to either women who had a successful implantation after IVF-ET or/and women with at least one successful spontaneous pregnancy or unselected healthy fertile women with no history of IVF-ET. Six hundred ninety-four published reports were retrieved; 17 of them fulfilled the inclusion criteria set. Results Presence of either any type of anti-phospholipid or anticardiolipin antibodies or lupus-anticoagulant in women experiencing at least two implantation failures in IVF-ET was associated with increased implantation failure compared to women who had a successful implantation after IVF-ET (relative risk, RR: 3.06, 5.06 and 5.81, respectively). Presence of either anticardiolipin or lupus-anticoagulant or anti-beta2 glycoprotein-I or anti-phosphatidylserine antibodies in women experiencing at least two implantation failures in IVF-EΤ was associated with increased implantation failure compared to unselected healthy fertile women with no history of IVF-ET (RR:13.92, 6.37, 15.04 and 164.58, respectively). Conclusion The prevalence of antiphospholipid antibodies, particularly that of anti-beta2 glycoprotein-I and anti-phosphatidylserine antibodies, in women experiencing at least two implantation failures in IVF-ET without diagnosis of antiphospholipid syndrome is significantly greater than either in women who had a successful implantation after IVF-ET or women with at least one successful spontaneous pregnancy or unselected healthy fertile women with no history of IVF-ET. Trial registration number PROSPERO ID: CRD42018081458


Introduction
Infertility is a public health problem which affects 1:10 women of reproductive age [1,2]. Its estimated world prevalence is 186 million people [1,2]. Assisted reproductive technology (ART) has led to a significant rise in live births following the introduction of in vitro fertilization (IVF) [3]. Over 1,250,000 ART cycles, resulting in birth of over 225,000 babies, were reported by 2,419 clinics globally in 2007. The availability of ART varies by country, from 12 to 4,140 treatments per million population [4].
Rheumatic diseases can affect quality of life and reproduction. Pregnancy complications are increased in patients with systemic lupus erythematosus and antiphospholipid syndrome (APS). The latter is an autoimmune acquired thrombophilia, which occurs either alone or in combination with other autoimmune diseases, mainly with systemic lupus erythematosus [4]. Antiphospholipid antibodies represent a heterogeneous group of antibodies, which recognize various phospholipids, phospholipid-binding proteins, and phospholipid protein complexes. Clinical manifestations of APS include fertility problems and pregnancy complications (such as repeated miscarriages) as well as venous or arterial thrombosis [5]. Evaluation of circulating anti-phospholipid (anti-PL) antibodies is part of the serological work-up following miscarriage. When circulating anti-PL antibodies are positive at initial diagnosis, testing should be repeated at least 12 weeks later to confirm diagnosis of APS [5]. According to revised Sapporo criteria, diagnosis of APS takes into account lupus anticoagulant (LA), anti-cardiolipin (anti-CL) antibodies or anti-β 2 glycoprotein I (anti-β 2 GP I) antibodies of either IgG or IgM isotype.
The relationship between presence of anti-PL antibodies (without diagnosis of APS) and implantation failure has been examined by several original studies which suggested that presence of anti-PL antibodies, even without diagnosis of APS, impairs implantation. Antiphospholipid antibodies, especially anti-beta 2 glycoprotein I (anti-β 2 GPI) antibodies, in pregnancy, appear to act directly on trophoblasts by activating pro-apoptotic and pro-inflammatory mechanisms [6]. At the same time, thrombosis of placental chorionic arteries and Study outcomes extracted for the meta-analysis. The primary outcome extracted from the selected studies was presence or not of any type of anti-PL antibodies. Secondary outcomes extracted from the selected studies were presence or not of: anticardiolipin (anti-CL), lupus anticoagulant (LA) and anti-β2GPI antibodies (all three representing aPL included in the Sapporo criteria for APS diagnosis), as well as anti-phosphatidylserine (anti-PS), anti-phosphatidylcholine (anti-PC), anti-phosphatidylethanolamin (anti-PE), anti-phosphatidylinositol (anti-PI), anti-phosphatidylglycerol (anti-PG) and anti-phosphatidic acid (anti-PA) antibodies which have gained importance in recent literature for APS diagnosis.

Risk of bias assessment
Risk of bias was assessed by two authors independently (E.P. and A.G.M.) by employing the Newcastle-Ottawa Scale [25]. Disagreement was resolved by discussion or adjudication by a third investigator if necessary (G.M.). In line with the previously submitted protocol of the meta-analysis, a publication bias analysis (Funnel plot and Egger's) was not performed because comparisons consisted of less than 25 eligible studies render such analysis less informative [26].

Statistical analyses
Heterogeneity among selected studies was evaluated in each analysis, using I 2 statistic [26,27]. According to Cochrane handbook: I 2 � 75%, I 2 between 75% and 50% or I 2 � 50% reflect substantial, significant or non-significant heterogeneity among the selected studies, respectively [27]. When I 2 was � 75%, possible causes of heterogeneity were investigated by performing pre-specified subgroup analyses. Meta-analysis was not performed in case of substantial heterogeneity which could not be resolved by subgroup analysis. In these cases, findings were reported narratively.
All outcomes were dichotomous and were analyzed by calculating relative risks (RR) and 95% confidence intervals (CI). For data synthesis the random effect model was employed as significant clinical and methodological heterogeneity among the included studies was anticipated. In addition, a priori specified subgroup analyses were conducted to explore significant or substantial heterogeneity and to further evaluate the soundness of results.
In a pre-specified sensitivity analysis, meta-analyses were repeated using the fixed effects model. All analyses were performed using Review Manager 5.3 Software [28].
Ethical approval. This article does not contain any studies with human participants or animals performed by any of the authors.

Search strategy for the systematic review
Search strategy identified 621 and 73 references in Medline and Cochrane Library, respectively (Fig 1). After removal of duplicate records between the two databases, a total of 629 studies were screened by title and abstract. Of these references, 40 were deemed potentially eligible because they reported studies regarding healthy women of reproductive age with implantation failure after IVF-ET and were assessed by full-text. Twenty-three of these 40 studies were excluded because either the outcome was not precisely reported or their design did not fulfill the inclusion criteria set. In the remaining 17 studies included in the present systematic review and meta-analysis (ten and seven studies were retrospective and prospective cohort studies, respectively), 4,075 healthy women of reproductive age were evaluated. Selection process is described in Fig 1, while the characteristics of each of the included studies are reported in the online Appendix 2 in S1 Appendix.

Characteristics of the women in the selected studies
The women with implantation failure [absence of positive pregnancy tests based on beta human chorionic gonadotropin (hCG) evaluation; in none of the studies the timing of hCG measurement was reported] after IVF-ET included in all studies had at least two implantation failures (range of failures: 2-6) following transfer of good quality embryos. Three studies included confirmation of a gestational sac by ultrasound three weeks after embryo transfer [9,16,18]. All women included in these studies were not receiving any additional treatment.
Subgroup A of selected studies. Studies of this subgroup, included women referred for IVF-ET due to similar indications [9][10][11][12][13][14]. Unexplained infertility was the predominant indication, followed by tubal factor -for the majority of them. Age, duration and type of infertility did not differ between population index and controls studied in this subgroup.
Subgroup B of selected studies. In these studies, the majority of patients received IVF for unexplained infertility. In studies of this subgroup, other, not frequently encountered, indications for IVF-ET were endometriosis, ovulation disorders or mixed infertility.

Characteristics of the included studies in the systematic review
Ovarian stimulation protocols employed. Four studies reported that a standard protocol for ovarian stimulation was followed (a combined regimen of GnRH agonists and human menopausal gonadotrophins) [9][10][11]18]. The remaining studies did not report the specific ovarian stimulation protocol employed.
Assays employed for the antibodies evaluation. All selected studies employed enzymelinked immunosorbent assays (ELISA) for presence of any type of anti-PL antibodies. Lupus anticoagulant was evaluated either by the kaolin cephalin clotting time utilizing sensitive reagents or by the dilute Russell's viper venom time with a neutralization procedure using frozen-thawed platelets or by both techniques. Results for anti-PL and LA antibodies were expressed as positive or negative. Diagnostic cut-offs for the antibodies were reported in: Qublan H.  [12].
In six studies positivity was based on optical density measurement exceeding the 99 th or the 95 th percentile of measurements established for each phospholipid in healthy individuals of reproductive age [11,14,17,19,22,23]. Six studies did not report diagnostic cut-offs [3,9,10,15,16,24]. Manufacturers of the assays are reported in all studies except two [13,20].
Qublan et al. and Khizroeva et al. evaluated women with implantation failures after IVF-ET compared with both control groups. Therefore, these studies are included in both subgroups A and B of selected studies [9,13].
Data regarding antibodies studied only in one study in each subgroup were not included in a meta-analysis. Thus, neither data on anti-P, anti-PC, anti-PE, anti-PA, anti-PG and anti-PS antibodies reported only in the study by Kaider

Risk of bias assessment
Based on the Newcastle Ottawa scale all studies of subgroup A were rated as of low risk in all assessed domains: selection bias of population index and control, performance bias, detection bias and attrition bias.
Ten out of thirteen studies of subgroup B (77%) were rated as having low risk and three (23%) as having unclear risk of bias regarding selection of population index. Regarding selection of controls, five out of thirteen studies of subgroup B (39%) were rated with unclear risk of bias. All studies of subgroup B were rated with low risk of bias regarding performance bias, detection bias and attrition bias. Detailed assessment of the risk of bias is available in online Appendix 3 in S1 Appendix.

Outcomes of systematic review and meta-analysis
The reported results are the outcome of separate random effect analyses of subgroup A and subgroup B. The sensitivity analyses using fixed effect methods showed similar results.
The majority of selected studies supported the association between presence of any type of anti-PL antibodies and infertility. Studies included in subgroup A were fairly homogenous (I 2 range: 0-15%) and provided quite reliable results, whereas studies included in subgroup B were less homogeneous.
Outcomes of subgroup A of selected studies (Table 1). Twenty-nine percent and 9.6% of population index and controls respectively, reported presence of any type of anti-PL antibodies among anti-CL, anti-β 2 GPI, anti-PS, anti-PC, anti-PE, anti-PI, anti-PA and anti-PG antibodies of IgG, IgM or IgA isotypes (six studies) [3,[9][10][11][12][13]. Population index showed a RR for the presence of any type of the above mentioned anti-PL antibodies of 3.06 for implantation failure (95% CI: 1.97, 4.77, I 2 = 15%) compared to controls (Fig 2). Because the anti-PC and anti-PE antibodies have lost part of their importance in the recent literature, a meta-analysis has been also performed without including the only study which evaluated these antiphospholipid antibodies. The obtained RR did not change substantially [RR: 2.89 (95% CI: 1.73, 4.81), I 2 = 22%]. In addition, 8.6% and 1.5% of population index and controls, respectively, Table 1

PLOS ONE
had present anti-CL antibodies (four studies) [9,[11][12][13]. Population index showed a RR for the presence of anti-CL antibodies of 5.06 for implantation failure (95% CI: 2.14, 11.95; I 2 = 0%) compared to controls (Fig 3). Moreover, 16% and 2.7% of population index and controls, respectively, had LA antibodies present (two studies) [9,12]. Population index showed a RR  for presence of LA antibodies of 5.81 for implantation failure (95% CI: 2.66, 12.71; I 2 = 0%) compared to controls (Fig 4). Table 2). Three studies out of six in subgroup B point out two distinct aCL isotypes (aCL/Ig-G, aCL/Ig-M) in their reporting of aCL evaluation. However, meta-analysis was conducted only for the aCL/Ig-G isotype because, for this isotype, a non-significant heterogeneity among the selected studies was observed, whereas for the aCL/Ig-M isotype the heterogeneity observed was substantial. Thus, in line with the methodology and guidance of Cochrane handbook, a meta-analysis was not meaningful for the aCL/Ig-M isotype. Nineteen percent and 1% of population index and controls, respectively had anti-CL-IgG antibodies present (three studies) [18,20,21]. Presence of anti-CL antibodies of the IgG isotype was more strongly associated with implantation failure after IVF-ET compared to that of the IgM isotype. Specifically, population index showed a RR for the presence anti-CL antibodies of the IgG isotype of 13.92 for implantation failure (95%CI: 6.21, 31.21; I 2 = 0%) compared to controls (Fig 5).

Outcomes of subgroup B of selected studies (
Fourteen percent and 1.3% of population index and controls, respectively, had LA antibodies present (four studies) [9,12,15,20]. Population index showed a RR for the presence of LA antibodies of 6.37 for implantation failure (95% CI: 2.25, 18.04; 2 = 0%) compared to controls (Fig 6).
Two studies assessed anti-PS antibodies [18,21]. Forty-two percent of population index had anti-PS antibodies present, as compared 0% of controls. Population index showed a RR for the presence of anti-PS of 164.58 for implantation failure (95% CI: 23.31, 1162.26; I 2 = 0%), compared to controls (Fig 8).
Two studies of this subgroup did not show any association between the presence of any type of anti-PL or anti-CL antibodies and IVF-ET outcome [3,21]. However, the results of both studies should be examined with caution as Bellver et al. were based on a very limited study population, while it is unclear how Steinvil et al. selected their control group.

Discussion
Among 629 references that this systematic search yielded from Medline and Cochrane Library, a limited number of 17 studies, involving 4,075 women of reproductive age, were included in this systematic review and meta-analysis. The addition of good studies in the future will improve the accuracy of the deduced conclusions. All included studies involved women with at least two implantation failures in IVF-ET vs. either women with one successful IVF-ET or women with at least one successful spontaneous pregnancy or unselected healthy fertile women with no history of IVF-ET. When a specific type of anti-PL antibodies was evaluated in heterogeneous or single studies, the reported data were not included in the meta-analysis.
We found, in this meta-analysis, that in women experiencing at least two implantation failures in IVF-ET, presence of either any type of anti-PL antibodies or anti-CL antibodies only or LA antibodies is associated with a significant 3.06, 5.06 and 5.81 RR for impaired implantation rate, respectively, as compared to women experiencing one successful IVF-ET. In addition, in women experiencing at least two implantation failures in IVF-ET, presence of either anti-CL or LA or anti-β 2 GPI or anti-PS antibodies is associated with a significant 13.92, 3.37, 15.04 and 164.58 RR for impaired implantation rate, respectively, as compared to women with at least one successful spontaneous pregnancy or unselected healthy fertile women with no history of IVF-ET. Most importantly, presence of either anti-β 2 GPI or the rarely evaluated anti-PS antibodies in 13% and 41%, respectively, of all women experiencing at least two implantation failures in IVF-ET and in virtually none of control subjects suggests that these may be very accurate biomarkers (more accurate than the more frequently measured anti-PL antibodies) and urges further evaluation of their potential clinical use in infertility, as well as in APS in general. This is the first meta-analysis evaluating presence of any type of anti-PL antibodies in women experiencing implantation failure in IVF-ET. The results from the 17 selected studies were consistent, of strong methodological quality and support an association between recurrent implantation failure in IVF-ET and presence of anti-PL antibodies. This meta-analysis looked into multiple anti-PL antibodies, including newer markers (i.e. anti-PS antibodies) aiming at reporting a quantitative result based on the homogeneity and similarity in the findings of the included studies. Both in the retrospective and prospective studies included in the present meta-analysis the compared population indices and control groups are well defined. Thus, the accuracy and reliability of the extracted results cannot be limited in either case. The main limitation of this meta-analysis is the lack of relevant prospective, controlled studies and the resulting small number of included studies. Despite the small number of included studies, the conclusions of the present meta-analysis can be considered scientifically valid due to the  lege artis mathematical approach followed in this meta-analysis and the homogeneity of the included studies.
Except the above mentioned anti-PL antibodies, which were evaluated by meta-analysis, this systematic review pointed towards other types of anti-PL antibodies, which seem to impair implantation and consequently fertility. Presence of specific types of anti-PL antibodies (i.e. anti-PS, anti-PC, anti-PI and anti-PA antibodies) was significantly increased in women with at least two implantation failures in IVF-ET and was associated with increased implantation failure rates, although specific types of anti-PL antibodies are in debate in the literature. Of note, these antibodies were not present in the respective control groups (Tables 1 and 2). The prevalence of these specific types of anti-PL antibodies reaches 40.4% among women with at least two implantation failures in IVF-ET compared to controls (range of prevalence among population index 2.4%-40.4%). Prevalence of anti-PE and anti-PG antibodies was also increased in women with at least two implantation failures in IVF-ET (range among population index 7.1%-19.6%) compared to their respective controls (range among controls: 0-4.8%) (Tables 1 and 2). It is noteworthy that the aforementioned types of anti-PL (anti-PS, anti-PC, anti-PI and anti-PA anti-PE and anti-PG antibodies) are rarely evaluated in clinical practice. One might suspect that they might be positive where other more often evaluated types of anti-PL such as anti-CL, LA or anti-β 2 GPI antibodies are negative. Of note, in this meta-analysis, were included not only studies reporting anti-PL antibodies included in the revised Sapporo criteria (LA, anti-CL, anti-β 2 GP I) but also anti-PL antibodies (anti-PS, anti-PC, anti-PE, anti-PI, anti-PG, anti-PA) which have gained importance for APS diagnosis in recent literature. Thus, it should be investigated whether these antibodies, not routinely measured, are involved as well in pathophysiologic aspects of implantation. Thus, questions emerge about their involvement in the underlying pathophysiological mechanisms in implantation.
The possible association of anti-PL antibodies with female infertility has been suggested since 1980s. Women with APS and women with anti-PL antibodies may present with impaired ovarian follicles reserve and more frequently with premature ovarian failure [29][30][31]. Α retrospective study suggested that the presence of APS or just anti-PL antibodies are frequently encountered acquired risk factors for recurrent pregnancy loss and that they are associated with increased risk for ischemic placental dysfunction, such as fetal growth restriction, preeclampsia, premature birth and intrauterine death [32]. Till now, little is known about the biological mechanisms involved in the recurrent ET failures observed in presence of anti-PL antibodies as well as in APS. Anti-PS antibodies bind to human trophoblast in a dose-dependent way affecting thus, trophoblast invasiveness and differentiation of cytotrophoblast into a syncytium. It is also shown that, specifically anti-PS and not anti-CL antibodies are responsible for the decrease of hCG production by the placenta [33]. In vitro and in vivo studies, have suggested that anti-PL antibodies might affect negatively conception, implantation as well as early and recurrent miscarriages. They can impair spontaneous as well as IVF-ET implantation as they are directed against negatively charged phospholipids located in the blood vessels of the uterine mucous membrane, or on the surface of oocytes, or they can affect the early embryo at the initial implantation process [32]. The latter and the maintenance of pregnancy in its early stages could be affected by inhibition of prostaglandin synthesis caused by anti-PL antibodies [32]. The latter affect maternal blood vessels, decidua and trophoblasts. They have been suggested to target tissue plasminogen activator, plasmin, annexin A2 and thrombin [34]. Prothrombin and β 2 GPI mediate binding of anti-PL antibodies to target cells such as endothelial cells, monocytes, platelets and trophoblasts leading to thrombosis of placental vessels and fetal loss [35,36]. Anti-PL antibodies and particularly β 2 GPI-mediated anti-PL antibodies bind to trophoblast monolayers and can induce direct cellular injury, inhibition of proliferation and syncytia formation, apoptosis, and defective invasiveness [35,36]. In mice, β 2 GPI is essential for a successful pregnancy and for optimal placental development [12]. In vitro studies indicate that anti-CL antibodies inhibit trophoblast proliferation possibly by a prostacyclin-thromboxane A2 imbalance [37,38]. Patients' anti-PS antibodies when co-cultured with rats embryos delay the development of rat yolk sacs [35]. These antibodies have been shown to affect negatively implantation in rats while they bind to human trophoblast in a dose dependent way affecting thus, trophoblast invasiveness and differentiation of cytotrophoblast into a syncytium [33]. To our knowledge, the presence of anti-PC, anti-PE, anti-PI, anti-PG, or anti-PA antibodies has not as yet been associated directly with implantation failure.
In summary, this meta-analysis has shown that presence of any type of anti-PL antibodies is associated with impaired implantation among women experiencing at least two implantation failures in IVF-ET and not suffering from APS compared to either women experiencing one successful IVF-ET or women with at least one successful spontaneous pregnancy or unselected healthy fertile women with no history of IVF-ET. Presence of anti-β 2 GPI and anti-PS antibodies suggests an excessive risk. Importantly, types of anti-PL, not frequently measured in daily medical practice (anti-PS, anti-PC, anti-PE, anti-PI, anti-PA and anti-PG antibodies), seem to be stronger predictors of implantation failure in IVF-ET. In women not suffering from APS, presence of antibodies (lupus anticoagulant, anti-cardiolipin antibodies and anti-beta 2 -glycoprotein I antibodies) included in the definition of APS, seem to be strongly associated with implantation failure in IVF-ET. In guidelines, it is suggested to evaluate anti-PL antibodies in women suffering from recurrent miscarriages [5]. Similarly, in women presenting multiple implantation failures in IVF-ET, it could be suggested to measure these antibodies, in order to investigate causality and, eventually, suggest treatment when additional studies will be available.
Well designed randomized controlled trials are needed in order to understand the impact of different types of anti-PL on implantation and consequently on infertility, before using them in everyday clinical management of infertile women with recurrent ET failures. Moreover, cost-effectiveness studies should be conducted to evaluate benefits and costs of this approach. Well designed interventional studies might confirm presence of anti-PL antibodies as predictive markers of implantation failure, but also target them pharmacologically in women suffering from infertility or subfertility. The present meta-analysis highlights the importance of the presence of anti-β 2 GPI and anti-PS antibodies regarding the risk for implantation failure. Thus, it would be useful to insist on measuring them in cases of infertility at least concurrently with the more frequently measured aPL antibodies.