Ethics approval

This study was an analysis of published data, which did not require ethics committee approval.

Search strategy

A systematic review was performed according to the guidelines established by the Meta-analysis of Observational Studies and Epidemiology Group (MOOSE criteria) [18]. The databases searched (January 1998 to May 2013) were MedLine, EMBASE, the Cochrane Library, and the homepages of the FDA and the European Medicines Agency (EMA), using the combinations of the following Medical Subject Heading (MeSH) search terms: (inflammatory bowel disease OR inflammatory bowel diseases OR Crohn’s disease OR ulcerative colitis) AND biologics OR biologic products OR infliximab OR adalimumab OR golimumab OR certolizumab. This was again combined with either pregnancy OR newborn OR lactation OR breast feeding OR infant. No authors were contacted.

Selection criteria

We wanted to include all randomized controlled trials, observational studies, case series, and case reports evaluating pregnancy outcomes of women with IBD treated with TNF-α inhibitors (IFX, ADA, GLM, or CZP) during any trimester of pregnancy, including the 90 day periconceptional period. Outcome assessment included miscarriages (spontaneous abortions), preterm delivery, stillbirth, low birth weight, and/or congenital malformations, and if data were available, it also included assessment of infections in the offspring. Only abstracts or articles published in English were considered. Figure 1 shows the flowchart of the study screening process.

Data extraction

Two authors (OHN and TJ) independently identified candidate articles from the results of the initial search, based on the title and abstract. Subsequently, both authors independently reviewed the full text of candidate articles to identify interventions and assess study quality. In cases of any discrepancies between the independent searchers, these were resolved in consensus with the third author (EVL).

The reference lists of relevant articles were hand-searched to identify additional studies. Further, abstracts from the meetings of American Gastroenterological Association (Digestive Disease Week), American College of Gastroenterology, and the European Crohn’s and Colitis Organisation from 1998 to May 2013 were searched manually for relevant abstracts based on the aforementioned search terms.

Summarization of data

Infliximab

In the early days of biologic treatment for IBD, IFX was reported to be associated with adverse pregnancy outcomes in some case reports [20, 21], but not in others [22–27], and in case series where IFX was administered throughout pregnancy to maintain remission in IBD, no harm to the fetus/child was reported [16, 19, 28–34]. Furthermore, larger and subsequent studies did not report any increased risk for adverse events compared with unexposed IBD pregnancies [12, 17, 35–49]. In a study on pregnancies exposed to IFX, Zelinkova et al. observed one adverse event (polydactyly), actually reported on two occasions [12, 50]. In this particular case, the mother of the infant had also received methotrexate (MTX) 2 months prior to conception [12, 50]. MTX is a drug classified as ‘X’ by the FDA; that is, a drug for which studies in animals or humans have shown fetal abnormalities and/or for which there is positive evidence of human fetal risk, based on adverse reaction data from investigational or marketing experience, and for which the risks involved in use of the drug in pregnant women clearly outweigh the potential benefits [9]. However, polydactyly is an uncommon disorder; it is not usually seen as the only complication to MTX treatment, but is more often present together with other abnormalities [50].

The first series studying intentional IFX administration throughout pregnancy examined outcomes in 10 women with active CD, and all ended in live births [32]. Of these 10 births, there were 3 cases of premature birth and 1 case with low birth weight, findings that were not unexpected in women with CD that was sufficiently severe to require IFX treatment. No congenital abnormalities were observed 6 months post-partum [32].

The four largest studies providing data on the safety of IFX in an IBD population are from the Therapy Resource Evaluation Assessment Tool (TREAT) registry [36], a pregnancy safety database maintained by the manufacturer of IFX [49], the Pregnancy in Inflammatory Bowel Disease and Neonatal Outcomes (PIANO) registry [51] and the Groupe d’Etude Thérapeutique des Affections Inflammatoires du Tube Digestif (GETAID) collaboration [52].

The TREAT registry is a prospective study of North American patients with CD. As of February 2010, there were 6,273 patients in the registry, and 142 pregnancies with IFX exposure were reported. The risk of fetal malformations was no higher than in the general population [36], and a more detailed analysis of pregnancies in the TREAT registry is under way [36].

The IFX Safety Database is a retrospective data-collection instrument set up by the manufacturer of IFX (Centocor, now Janssen Biotech) to study women with rheumatoid arthritis or CD exposed to IFX before or during pregnancy. There were 96 women with direct exposure to IFX who gave birth to 100 children; however, in most case, IFX treatment was discontinued in the first trimester of pregnancy. The rates of adverse outcomes were not different from that expected in the general population [49].

Preliminary data from the US national prospective PIANO registry has recently been disclosed [53]. A total of 1,232 women with IBD were enrolled as of April 1, 2013, of whom 357 women received biologics, including during the third trimester, and an additional 109 women received a combination of thiopurines and biologics [53]. Adverse pregnancy outcomes such as spontaneous abortions, preterm birth, low birth weight, and congenital abnormalities were evaluated. A slightly higher relative risk (RR) for preterm birth was found for women on combination therapy (2.4; 95% confidence interval (CI) 1.3 to 4.3), which was, however, not found for those on biologics only (0.8; 95% CI, 0.5 to 1.3). No elevation in the risk of spontaneous abortion, low birth weight, or congenital abnormalities was observed for biologics, either alone or in combination with thiopurines. Earlier reports (May 2012) had indicated that there was a higher rate of infections in the offspring [51], but the RR at month 12, adjusted for premature birth, was shown in April 2013 to be 0.9 (95% CI 0.7 to 1.1) for biologics alone, and 1.0 (95% CI 0.7 to 1.3) for those on combined therapy. When results for CZP were excluded, the values were 0.9 (95% CI 0.8 to 1.2) and 1.0 (95% CI 0.7 to 1.3), respectively [53]. In addition, breast feeding was not associated with any risk of infection or impairment of height or weight. The limitation on the preliminary data was that not all infants had reached the age of 12 months, and the observation will continue until the age of 4 years for all of the offspring [53].

The French GETAID consortium prospectively recorded pregnant women with IBD during a 2 year period (2009 to 2010), and preliminary data analysis of 127 pregnancies in 120 women was reported in 2010 [52]. Of these 120 women, 54 women had received biologic therapy for their IBD. Birth outcomes of the women on biologics did not differ from those in the unexposed IBD population, suggesting an absence of any excess risk linked to anti-TNF-α therapy.

Schnitzler et al. [54] conducted an observational study of 35 pregnancies under direct exposure to IFX until week 20 of pregnancy, 53 pregnancies with indirect exposure to IFX (women treated with IFX prior to pregnancy), 7 pregnancies with direct exposure to ADA, and 56 matched pregnancies in healthy women. Several patients experienced a flare of disease in the third trimester. However, exposure to anti-TNF-α preparations was not associated with a higher incidence of adverse pregnancy outcomes [54]. The paper concluded that pregnancy outcomes after exposure to TNF inhibitors did not differ from those of women with IBD who were not exposed to anti-TNF treatment, but were worse than in healthy women [54]. Further, a literature review including women with rheumatic disorders or CD with direct exposure to IFX at any time during pregnancy did not find any increase in undesirable pregnancy or fetal outcome [55].

However, in contrast to studies providing evidence of the safety of anti-TNF-α preparations during pregnancy, a review of more than 120,000 adverse events voluntarily reported to the FDA after drug exposure revealed 61 congenital abnormalities in 41 children of 40 mothers exposed to TNF-α blockers for various indications, including IBD [56]. Of these, 22 mothers were exposed to etanercept (a biologic agent not efficacious for IBD [57]), and 19 were exposed to IFX. Of the 41 children, 24 (59%) had abnormalities thought to be part of the VACTERL (vertebral abnormalities, anal atresia, cardiac defects, tracheoesophageal, renal and limb abnormalities) spectrum. The findings led the authors to conclude that the use of TNF-α blockers should be avoided during pregnancy. However, this study was subsequently criticized for significant methodological flaws [58, 59], including selection bias, lack of a control group, and the fact that only one of the infants (the etanercept group only) had a clustering of three or more abnormalities included in the VACTERL spectrum [60]. Furthermore, the review did not take into account the total number of women exposed to anti-TNF-α agents or any confounding by severity of the underlying disease. Finally, the most common defects reported in this review were cardiac defects, which are observed in the general population as well [58, 59].

Like other IgG antibodies, IFX is transferred across the placenta with the help of an Fc receptor neonatal (FcRn) molecule responsible for the transfer of immunoglobulins from the mother to the neonate [11]. The transfer occurs partly in the second, and mainly in the third trimester [31]. This finding has led many clinicians to discontinue IFX in the third trimester when the highest level of transfer occurs [61], in the belief that detectable serum IFX in the neonate might lead to clinically relevant immunosuppression. Moreover, the use of biologics in the second and third trimesters has been questioned [12]. However, Zelinkova et al. [12, 50] showed that even though two patients discontinued IFX treatment as early as in week 26 of pregnancy, the infants still had detectable amounts of IFX in their serum samples after delivery. Low levels of IFX have been detected up to 6 months post-partum in the serum of infants whose mothers have received this drug up to delivery [16]. IFX was undetectable in the infant at delivery only in one case, where the mother discontinued drug therapy at week 21, suggesting that discontinuation of drug therapy might be considered earlier than the beginning of the third trimester [50].

Because of this presence of IFX in the circulation after delivery, the use of live-virus vaccines (for example, varicella, measles, mumps, rubella, rotavirus, intranasal influenza, and bacillus Calmette-Guérin; BCG) is contraindicated in patients receiving biologic therapy and their children [62]. A case report described a fatal incident of an infant who was exposed to IFX during gestation and who received a BCG vaccine at 3 months; this infant died of disseminated BCG infection [63]. Thus, the most recent recommendation from the World Congress of Gastroenterology consensus statement on vaccinations in infants exposed to biologic therapy in utero recommend delay of all live-virus vaccines until after biologic molecules are no longer detectable in the child’s blood [62].

Adalimumab

The clinical data for the safety of ADA during pregnancy in women with IBD are more limited than for IFX, but in animals, it appears that ADA does not increase obstetric risks and has no teratogenic effects [70]. In humans, data on ADA and pregnancy primarily concerns patients with diseases other than IBD, such as rheumatoid arthritis and psoriasis.

For IBD, 21 case reports and series with more than 300 children exposed showed no increased risk of adverse pregnancy outcome or congenital malformations associated with ADA treatment during pregnancy compared with pregnancies in unexposed women with IBD [12, 17, 19, 35, 37, 39, 43, 46],[48, 51, 52, 54, 71–80], even if ADA was administered in the third trimester [17, 19, 35, 37, 39, 43, 46, 48],[51, 52, 71–74, 76, 77, 79, 80]. The Organization of Teratology Information Specialists reported on a group of 34 women treated with ADA for various indications in a prospective study, and another 133 ADA-exposed women in a case series. There was no difference in preterm deliveries, stillbirths, spontaneous abortions or congenital malformations in the group treated with ADA compared with either the general population or a control group with the same disease but not exposed to ADA [62, 81]. In line with these observations, Schnitzler et al. did not find a higher incidence of adverse pregnancy outcomes in seven women treated with ADA compared with unexposed women with IBD, although the power of that study was limited [54].

An ongoing cohort study evaluating ADA in pregnancy compared 80 live offspring born to 94 ADA-exposed women with CD, 53 live offspring of 58 women with CD not exposed to ADA, and 78 of 87 women from a non-diseased comparison group. The frequency of congenital abnormality in the three groups was 9.6%, 5.4%, and 5.0%, respectively [72] (P>0.05 between all groups).

Like IFX, ADA is transferred across the placenta in the third trimester of pregnancy [31, 76], and accordingly, it has been recommended at many IBD centers that ADA should be discontinued in the third trimester (that is, week 30), even though data, as mentioned before, do not support this theory [82].

Osting et al. carried out a review of TNF-α blockers (ADA and eternacept only) administered during pregnancy from 1999 to 2009 using the Organization of Tetratology Information Specialists (OTIS) prospective registry, and concluded that 7 to 10% of children had congenital abnormalities [83]. However, the underlying disease activity might be a confounding factor with greater effect on the risk of congenital abnormalities than the biologic treatment per se.

Golimumab

Because GLM was approved only recently (May 2013) by the FDA for treatment of the UC, data are not yet available for this drug in relation to IBD pregnancies.

Certolizumab pegol

Because CZP lacks the Fc portion of the antibody, which is crucial for the main transfer of immunoglobulins across the placenta, the transfer of CZP to the offspring is low, but a small transfer of Fab fragments is seen [31]. In a group of 10 women receiving CZP during pregnancy, with the final dose 1 to 4 weeks before delivery, CZP was detectable at minimal levels in the infants’ circulation [17]. Overall, CZP is transferred across the placenta to a much lower degree than IFX or ADA [17].

Compared with the other two TNF inhibitors approved for CD, the human data on pregnancy outcomes in mothers treated with CZP during pregnancy are more limited. In three case reports where CZP injections were administered during the third trimester, healthy infants were delivered at full term [84–86], and in a case series of 16 pregnancies of CZP-treated mothers, all children were healthy [31]. In a recent case series with 139 pregnancies, 17 were in mothers with rheumatoid arthritis; of these 139 pregnancies, 103 ended in live birth (2 with congenital abnormalities; see Table 1), 21 in spontaneous miscarriage, and 15 in elective termination [87], results that are similar to those reported in the general population [87]. Based on these findings and the low placental transfer of CZP, the drug appears to be safe to use throughout pregnancy [17]. One of the reports observed uterine bleeding in the mother at 7 days post-partum; whether this can be ascribed to the CZP treatment remains to be established [86].

Evaluation of the identified studies

In the studies included in this systematic review, more than 1,426 live births occurred among all the women exposed to IFX, ADA, or CZP. The number of congenital abnormalities associated with live births reported did not point to any pattern of specific birth defects (Table 1). There were 128 miscarriages, 81 elected abortions (presumably ‘under-reported’), 12 stillbirths, and 33 preterm births reported (Table 1).

Owing to considerable variation in study designs, selection of controls, and the definition and selection of outcomes in the identified studies, we were unable to conduct a meta-analysis. However, it appears clear from Table 2 that for all outcomes except congenital abnormalities, the OR or RR did not point systematically towards a negative or positive effect of TNF-α inhibitors. A slightly increased risk of congenital abnormalities in women with IBD exposed to TNF-α inhibitors compared with unexposed pregnant women with IBD seemed to be suggested from the few studies with controls (Table 2), but the CIs were very broad. Moreover, this observation should be weighed against the fact that the vast majority of case reports did not report congenital abnormalities, although it is thought that such ‘positive’ cases are more likely to be described and published than negative ones. Lastly, the meta-analysis by Cornish et al. [1] including 3,907 women with IBD, showed that an underlying increased risk of the studied outcomes already exists (stillbirths: OR 1.48 (95% CI 0.89 to 2.47); preterm delivery: OR 1.87 (1.52 to 2.31); low birth weight: OR 2.10 (1.38 to 3.19); and congenital abnormalities: OR 2.37 (1.47 to 3.82)) independent of treatment, compared with 320,531 pregnant non-IBD controls.

Limitations of the study

Limitations of studies identified in this systematic review included small sample sizes and paucity of studies with control groups. Future studies assessing the outcome of pregnancies exposed to IFX or other biologics during conception and pregnancy should be prospective in nature, and should assess comparable women with IBD with or without the relevant exposure as a control group to draw firm conclusions. Data on potential confounders, especially maternal age, disease activity, and co-medications, should also be recorded. Furthermore, long-term follow-up studies of children exposed to biologic medications in the uterus should be carried out, especially as no long-term data exist on outcomes in children related to the development of diseases such as asthma or autoimmune diseases.