A total of 485 records were identified in the search, of which 422 records were excluded during title/abstract screening. On the basis of a review of the full texts of 63 studies, an additional 39 studies were excluded, primarily for inappropriate comparison (i.e., no comparison between different treatments) and for not reporting the outcomes of interest. Twenty-four studies met eligibility criteria and were included for data extraction. The PRISMA flow diagram is presented in Fig. 1.
There were 18 randomized controlled trials (RCTs), two prospective non-randomized studies, two prospective observational studies, and two retrospective observational studies, with publication dates ranging between 2000 and 2017. Study populations included women with recurrent pregnancy loss (15 studies), history of placental vascular complications (five studies), recurrent in vitro fertilization failure (two studies), and thromboprophylaxis (two studies). The characteristics of studies and reported outcomes are summarized in the Electronic Supplementary Material (Table S5).
Twenty publications [22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41], pertaining to 19 clinical trials [22,23,24,25,26,27,28,29,30,31,32,33, 35,36,37,38,39,40,41] were included in the Cochrane risk of bias assessment. A summary of the assessment for the randomized trials is presented (Table S3 in the Electronic Supplementary Material). The risk of bias across the included studies was generally low and limited to the blinding of participants and personnel (performance bias) and the blinding of outcome assessment (detection bias): 14 studies were judged to be at low risk for the random sequence generation item in the selection bias domain [23,24,25,26,27,28,29,30,31,32,33, 38,39,40], three studies were judged as unclear, as the random sequence generation methods were not clearly described [22, 35, 40], and four studies were considered high risk as they were non-randomized [36, 37, 42, 43]. Regarding allocation concealment, 14 studies were considered to be at low risk of bias [22,23,24,25, 27, 28, 30,31,32, 35, 36, 38, 40, 41], four studies were considered unclear as methods for allocation concealment were not provided or not clearly described, one study was considered as high risk as randomization was not carried out and the groups were made aware of the treatment they received upon enrollment . Few studies presented adequate blinding methods to reduce performance bias: only three studies were considered as low risk for sufficiently packaging their treatments to disguise their identities to patients receiving them at all points in the study [28, 38, 41], eight studies were considered unclear as blinding methods were not provided, ten studies were open-label for participants and personnel and were thus considered to be at high risk for performance bias [22, 24, 25, 27, 30,31,32,33, 37, 40]. Similarly, for detection bias, only five studies were considered low risk [28, 30, 38, 39, 41], six studies were of unclear risk as the study was not described as either open-label or blinded [23, 29, 35, 36, 42, 43], and nine studies were considered high risk because of an open-label study design [22, 24,25,26,27, 31, 33, 37, 40]. All studies were considered as low risk in the domains of incomplete outcome data (attrition bias) and selective reporting (reporting bias). Most studies used an intent-to-treat analysis in their primary outcome analysis, and where dropouts were present, the numbers were few and similar between groups. All studies reported on their prespecified primary outcome, as mentioned in the publication. Finally, no other sources of bias other than the ones previously described were detected across the studies.
Four studies were included in the NOS quality assessment, all of which were cohort studies [42,43,44,45]. A summary of the assessment is presented in the Electronic Supplementary Material (Table S4). One study had a score of 8 and three studies scored the maximum score of 9. All studies scored the maximum score of 4 stars in the selection category, indicating that the data used was representative of the targeted community. For the comparability category, one study scored only one star because the authors did not describe any additional baseline characteristics that were well matched between cohorts  and the other three studies scored the maximum score of 2 stars [43,44,45]. Finally, in the outcome category, all studies scored the maximum of 3 stars.
Thromboembolic Events: Efficacy and Safety
Only two studies looked at enoxaparin in the context of thromboprophylaxis; a pooled meta-analysis was therefore not possible.
In terms of safety of enoxaparin, a pooled meta-analysis for thromboembolic events was not feasible because of the low number of reported cases across treatment comparisons. Data reported in the studies are summarized in Table S6 (Electronic Supplementary Material). Two cases of DVT were reported in patients with recurrent pregnancy loss who received enoxaparin monotherapy compared to four cases in patients who received no intervention (RR 0.5; 95% CI 0.09–2.69) . No thromboembolic events were reported in patients receiving combined therapy with enoxaparin plus aspirin. In one study that compared enoxaparin plus aspirin versus aspirin in women with a history of preeclampsia, one case of superficial venous thrombosis was reported in patients who received aspirin .
Results of the pooled meta-analyses for bleeding are presented in Table 1 and Fig. 2. The incidence of bleeding events with enoxaparin monotherapy was 35% higher compared to placebo/no treatment controls. When compared to aspirin, risk of bleeding was 7% lower with enoxaparin monotherapy and 5% lower with enoxaparin plus aspirin. These results were not statistically significant, and heterogeneity between the studies was low. Enoxaparin was compared to tinzaparin in one study in women with recurrent pregnancy loss . No bleeding events were reported in either treatment group. In another study in women with recurrent pregnancy loss who also tested positive for antiphospholipid antibodies, three bleeding events were reported in patients receiving enoxaparin plus aspirin compared to two events in those who received UFH plus aspirin . Bleeding event data are summarized in Table S7 (Electronic Supplementary Material).
Results of the pooled meta-analyses for pregnancy loss are presented in Table 1 and Fig. 3. Risk of pregnancy loss was reduced by 58% when enoxaparin monotherapy was compared to control groups, and significant heterogeneity was observed (Fig. 3a). Though there was a large numerical difference, reduction in pregnancy loss for enoxaparin monotherapy compared to aspirin did not achieve statistical significance (Fig. 3b). A sensitivity analysis removing one study with no events in the enoxaparin group did not change the result. Pregnancy loss was significantly reduced by 42% with enoxaparin plus aspirin compared to aspirin alone (p < 0.0001) (Fig. 3c). In a sensitivity analysis based on the study design, a lower risk ratio was found for pregnancy loss in the observational data than the RCT data. Enoxaparin plus aspirin was still statistically favored over aspirin and the study design did not appear to influence this result. A non-statistically significant reduction of 66% was found when enoxaparin plus aspirin was compared to UFH plus aspirin (Table 1). Pregnancy loss reported in the studies is summarized in Table S8 (Electronic Supplementary Material).
Few cases of thrombocytopenia occurred in the studies (Table S9 in the Electronic Supplementary Material). In two RCTs in women with recurrent pregnancy loss, ten events were reported for enoxaparin and three events for placebo [22, 38]. Thrombocytopenia with enoxaparin plus aspirin was reported in one study, which reported two cases for enoxaparin plus aspirin and four cases for aspirin, in women with a history of preeclampsia .
Eight cases of congenital malformations were reported for enoxaparin, two cases for aspirin, and five cases for placebo in three RCTs in women with recurrent pregnancy loss [24, 26, 38]. In an observational study in women with a history of obstetric and/or thromboembolic complications, single cases of congenital malformations were reported in patients receiving enoxaparin plus aspirin in the first trimester or aspirin only. There were no cases observed with second trimester enoxaparin plus aspirin (Table S10 in the Electronic Supplementary Material).
Maternal mortality was reported in one study and no deaths were reported in either of enoxaparin plus aspirin and aspirin groups during the study.
Allergy or skin reactions were reported in three studies. One of the studies reported no allergy, and the remaining two studies reported around one to four more cases of allergic reaction when compared to the standard care or aspirin groups.
Study withdrawal was reported in eight studies. However, five reported zero discontinuations in the remaining three studies, and each of them reported a unique treatment comparison involving enoxaparin.