Types of studies

Seven relevant systematic reviews (including four meta-analyses) and 18 primary studies were identified. The following information was abstracted from reviews: the number of included studies, outcomes and conclusions of the reviewers. Information abstracted from primary studies included: number of subjects, method of IA assessment, IA procedure type, study design, type of control group, confounding variables controlled for and the study findings ( Tables 1–8 ). Where appropriate, odds ratios (ORs) with 95% confidence intervals are presented in Tables 1–8 .

Five of the seven systematic reviews described the study designs of the included primary research papers, which included 59 cohort and 52 case-control studies. In their review of the literature, the Royal College of Obstetricians and Gynaecologists (RCOG) focussed on one particular review along with 14 other studies. Details of study design were only mentioned for six studies (four cohorts and two case-control studies) published after 2000. The Atrash and Hogue review did not provide any information of study design; its aim was to examine previous reviews and update their findings using data from new studies. The 18 papers in this review included one prospective and 12 retrospective cohort studies and five case-control studies.

Secondary infertility

Except for one study from Greece, where abortion is illegal, none of the studies included in the Atrash and Hogue review in 1990 suggested an increased risk of infertility following an IA. The Thorpe et al. review in 2000 identified seven studies on the association between abortion and subsequent infertility. Only two studies from Greece noted a positive association. Thus, there is little evidence to suggest that IA is associated with secondary infertility.

Effects on early pregnancy loss

The probability of pregnancy loss in a subsequent pregnancy following IA was examined by two primary studies and three systematic reviews.

Miscarriage

Earlier studies have described an association between miscarriage and IA performed by D&E while a recent study by Sun and co-workers reported a higher risk of early pregnancy losses after vacuum aspiration. The Atrash and Hogue review suggested that D&E (but not vacuum aspiration) carried an increased risk of subsequent miscarriage. A more recent definitive review by Thorpe and colleagues, based on five studies (two cohort and three case-control studies), was unable to demonstrate any significant associations between IA and subsequent miscarriage. As the literature is relatively sparse and conflicting in nature, no clear-cut conclusions about the association between IA and miscarriage can be made.

Ectopic pregnancy

The Atrash and Hogue review based on a few small case-control studies does not support an overall relationship between IA and ectopic pregnancy. Of nine studies included by Thorpe et al., two case-control studies found an association; Parazzini et al. showed a strong association, OR = 2.9 (1.6–5.3); and Tharaux-Deneux et al. showed a significant effect with a dose–response effect: one IA, OR = 1.4 (1.0–2.0); two or more, OR = 1.9 (1.0–3.7). The other seven cohort studies were unable to show a link between IA and subsequent ectopic pregnancy.

Bouyer et al. have described a dose-dependent relationship between surgical IA and ectopic pregnancy in a previous study (included in the Thorpe review) and suggested that this could be the result of uterine injury or infection. This finding was not confirmed by their more recent study, where they only found an association with medically IA. However, they surmise that infection could still be the common association. Antibiotic prophylaxis is more routinely given in cases of surgical abortion than medical abortion, and unreported infection after a medical abortion may account for the surprising finding in their more recent study. Further studies that concentrate on specific populations are needed to understand if there is an actual association between IA and ectopic pregnancy.

Obstetric complications following IA

The probability of obstetric complications occurring in a continuing pregnancy following IA was examined by several primary studies and systematic reviews.

Placenta previa

Systematic reviews with meta-analyses by Ananth and colleagues (six studies included) and Fraiz and Ananth (eight studies included) have reported a strong association between IA and placenta previa. Both reviews noted substantial heterogeneity in the primary studies, which was addressed by random-effects analysis.

The Thorpe et al. review found three studies, one cohort and two case-controls, exploring IA and placenta previa (one of these studies had been included in the Ananth study, which is cited in the Thorpe review). The three studies found a positive association between IA and placenta previa.

Johnson and colleagues investigated two surgical methods of IA, vacuum aspiration and dilatation and curettage, and their association with placenta previa. While vacuum aspiration did not increase the risk of placenta previa, there was a trend towards an association with dilatation and curettage. They concluded that the risk of placenta previa in subsequent pregnancy may be increased in a dose–response fashion by multiple sharp curettage abortions. The Hung et al. study also found a positive association along with a dose-response effect. Thus, the available literature suggests that there is an association between IA and placenta previa.

Placental abruption

Koifman et al investigated pregnancy outcomes in 2010 women with vaginal bleeding during the second half of their pregnancy (both placenta praevia and abruption). Multivariate analysis indicated that repeated IA was a risk factor for antepartum haemorrhage (OR 1.8 (1.1–2.1)).

Hung et al. examined risk factors for placental abruption in a retrospective review of 37 245 women. They excluded pregnancies complicated by placenta previa, multiple gestations and foetal anomalies. They found no association between placenta abruption and one or more IAs.

Zhu et al. examined placental complications in subsequent pregnancies after an initial first trimester medical abortion and found that nearly twice as many women with a prior induced medical abortion (22/4620) experienced abruption, compared with women who had not had a prior induced medical abortion (12/4620). The Zhu et al. research does suggest that women with a previous IA have higher rates of placenta abruption; however, after taking confounders into account the difference was not found to be statistically significant. These authors suggest that the connection between placenta abruption and IA may be associated with factors such as gestational age >6 weeks at abortion, curettage after abortion and a longer inter-pregnancy interval rather than the medical abortion itself.

Pre-eclampsia

Beck reviewed case records of 220 women who had had an abortion (137 spontaneous, 83 induced), 220 women in their first pregnancy and 220 women pregnant for the second time after a full-term pregnancy. They found that the incidence of pre-eclampsia was significantly lower when there had been a previous IA as compared with no previous abortion or previous spontaneous abortion. Any type of abortion conferred some degree of protection; although this protection was not as high as the protective effect of a first-term pregnancy.

Seidman et al. found that women who had undergone a prior IA were at a decreased risk of pre-eclampsia than those who were pregnant for the first time. However, the protective effect was not as strong as the protection following a full-term pregnancy.

Eras et al., using obstetric records from 2739 births reported a significant reduction in the risk of pre-eclampsia in women who had experienced one spontaneous abortion or one IA as compared with women in their first pregnancy.

Trogstad et al. examined how a previous abortion impacts on the risk of pre-eclampsia in a subsequent pregnancy in 20 846 women. They found a slightly reduced risk of pre-eclampsia in women with one IA (adjusted OR (AOR) 0.84, (0.69–1.02)), and in women with two or more IAs the risk of pre-eclampsia was reduced by >60% – AOR 0.36 (0.18–0.73) than in women without IAs. The protective effect of two IAs is similar to the protective effect of one prior birth.

There seems to be some evidence to support the hypothesis that normal pregnancies that are terminated in early pregnancy may stimulate immunological changes that reduce the risk of pre-eclampsia in a subsequent pregnancy. This reduction in risk seems to have a dose–response effect. However, the protection does not appear to be as high as that conferred by a full-term pregnancy.

Perinatal complications

The probability of perinatal complications occurring in a continuing pregnancy following IA was examined by several primary studies and systematic reviews.

Low birth weight (LBW)

Shah and Zao’s meta-analysis reviewed 18 studies investigating IA and LBW in subsequent pregnancies. They found increased unadjusted and adjusted odds of LBW among women with a history of IA; the risk increasing with the number of IAs.

Atrash ang Hogue’s review found no significantly increased risk of LBW following vacuum aspiration, but some association with D&E. The effect was strongest when compared with women with previous full-term pregnancies (G2, P1). Yimin and co-workers compared live births in 12 995 women who either had a previous medical abortion (4385), a previous surgical abortion (4358) or no history of abortion (4252). They found no significant differences among these groups.

An Italian study investigated IA as a risk factor for small for gestational age (SGA) births but failed to find an association. They concluded that any association between IA and low birth weight may be confounded by the occurrence of pre-term birth (PTB). Reime et al. examined the medical records of 8857 adolescents –7845 nulliparous, 801 previous live birth and 211 one previous abortion. They found no association between IA and PTB, but found an association between IA and very low birth weight (<1500 g).

There appears to be an association between D&E and LBW in subsequent pregnancies, but not with vacuum aspiration. Very few studies have looked at the relationship between medical IA and LBW, but those that have, have failed to find any association.

Pre-term birth

Atrash and Hogue’s 1990 review states that there is no increased risk of premature delivery following an IA, independent of type of procedure and gestational age at the time of the procedure. The Thorpe et al. review included 24 studies of which half found an association (two for LBW and 10 for PTB). Of the 12 studies, seven identified a dose–response relationship between the number of IAs and the risk of PTB. Moreover, seven studies demonstrate an association with early PTB (<32 weeks). This review did not identify studies of medical abortion with long follow-up, so associations are between surgical IA procedures (unspecified) and PTB.

A South Australian study found an association between IA and PTB with the risk rising with increasing numbers of IAs. The Voigt et al. cohort study found a trend towards PTB and early PTB (<32 weeks) in the IA cohort compared with the no IA controls for all age groups, but these results were only statistically significant for the 28–30 years age group.

Jackson et al. explored the relationship between mid-trimester D&E with laminaria and subsequent pregnancy complications, including PTB. They found no association with any of the outcomes of interest; however, the sample size was relatively low (85 cohort subjects compared with 170 controls) and may not have had sufficient power to detect a statistical difference.

Two recent well-designed systematic reviews with meta-analyses found an association between IA and subsequent PTB. Swingle et al. included 12 studies, of which nine showed an increased risk of PTB following IA. Five studies described a dose–response effect. Four of the studies included data on early PTB (delivery between 28 and 32 weeks), which resulted in a common adjusted OR of 1.64 (1.38–1.91). Shah and Zao reviewed 22 studies and found a clear association along with a dose–response relationship. Both these meta-analyses noted clinical heterogeneities among the studies included in their reviews, but despite the limitations inherent in meta-analyses of observational data these studies strongly suggest that a previous IA is associated with an increased link of PTB in a subsequent pregnancy.

Apart from the Atrash and Hogue review and another study investigating a surgical IA, none of these studies made a distinction between types of surgical and medical procedures or gestational age at abortion. More studies investigating the effect of different procedures of IA and IA at different gestational ages are needed to identify the safest method of IA and to better understand the association between IA and PTB.

Although the evidence for an association between PTB and IA is conflicting, recent systematic reviews as well as primary studies suggest that women who have undergone a previous IA are at higher risk for PTB in subsequent pregnancies. A dose response has also been demonstrated. In addition, evidence linking IA with early PTB (<32 weeks) has been found in a number of studies. A major problem with some of these studies is their inability to discriminate between spontaneous and induced PTB. Induced PTB may be the result of clinical decisions based on maternal complications, which themselves could be linked to IA. Only two original papers performed a sub-analysis distinguishing spontaneous from induced PTB. This sub-analysis did not change the results of either study. Owing to lack of appropriate information, the other studies and reviews could not differentiate between spontaneous and induced PTB. Analysing spontaneous and induced PTB together in the same group weakens the strength of the any association found.

Limitation of existing studies

Interpretation of available data should take into account methodological problems associated with studying long-term obstetric consequences of IA. First, the choice of an appropriate comparison group is problematic. Women pregnant after having an IA in their first pregnancy are gravida 2 and parity 0 (G2P0). It is impossible to control for both gravidity and parity (unless those in the control group have had a prior miscarriage; yet women with a previous miscarriage are at higher risk for subsequent pregnancy complications and are therefore not ideal). Other comparison groups can be either women in their first ongoing pregnancies (gravidity 1 parity 0) or in their second ongoing pregnancies after a previous delivery (gravidity 2 parity 1). Ideally, both comparison groups should be used and the results interpreted with caution.

Out of the 18 papers examined, only five studies compared the effects of IA to G2P1 and G1P0 by analysing them separately. Five of the studies only compared the effects of IA to G1P0; the remaining eight studies compared G2P1 and G1P0 together as a single group.

Second, abortion is not an exposure that someone can be allocated to, so subjects cannot be randomised into groups to minimise bias. Observational studies are therefore the method of choice for investigation in this area but results need to be interpreted with care as they are prone to bias.

A major bias that needs to be considered when investigating obstetric performance after an IA is evaluating true exposure status. Even in countries where abortion has been legal for many years there is evidence of under-reporting, so selective recall can be a problem if data are collected retrospectively. In addition, women who have suffered early complications of pregnancy may be more likely to report a prior abortion than women who have not suffered any complications. Out of the 18 papers examined, eight studies used self-report methods, nine used medical records and one used both self-report and medical records.

A third problem surrounds data collection on outcomes and confounders. Women pregnant after an IA differ from comparison groups in a number of ways that are related to an increased risk of adverse pregnancy outcomes. Factors relating to the abortion itself, which may confound results, such as complications of the IA, infections after the IA are often not available. The 18 original papers in this review controlled for different confounders, age was the only confounder controlled for in all studies. An important issue concerns the distinction between spontaneous and iatrogenic PTB, which are often considered together as a single outcome.