Introduction

In 2010, the European Crohn’s Colitis Organisation issued a consensus report summarizing the existing evidence for links between inflammatory bowel disease (IBD) and adverse reproductive outcomes [1]. According to the report, “data on the frequency of complications of pregnancy and labour are very inconsistent, precluding a meaningful conclusion” [1]. Studies of preterm delivery and low birth weight have yielded the most consistent results, with most studies finding that women with IBD are at increased risk of both outcomes [1,2]. Pregnancy loss, pre-eclampsia, placental problems and most neonatal outcomes have been studied very little [1]. Congenital abnormalities have received much attention but published results diverge greatly [1,2]. Whether disease activity and/or medication use during pregnancy affect the risks of adverse pregnancy outcomes is also unclear. Contributing to the difficulty in addressing these issues are the relatively low prevalence of IBD and the infrequency of outcomes such as congenital abnormalities, which together can limit the power of even the most well-conducted studies (e.g. [3]).

Previously, Ajslev and colleagues studied the link between IBD and low birth weight in the Danish National Birth Cohort (DNBC) [4]. Our study also took advantage of the DNBC’s 700 women diagnosed with IBD before pregnancy and its information on potential confounders and early pregnancy losses. We linked DNBC information and Danish national health register data to investigate potential associations between IBD and pregnancy loss, pre-eclampsia, preterm delivery, Apgar score, and congenital abnormalities, taking IBD-related medication use into account.

Methods

Results

Pre-eclampsia

The pre-eclampsia cohort comprised 86,792 women, including 666 diagnosed with IBD before the end of the study pregnancy (CD, 278; UC, 388). The overall pre-eclampsia rate in women with IBD did not differ significantly from that in women without IBD (HR 1.21, 95% CI 0.76–1.95) (Table 1). However, rates of severe pre-eclampsia were more than 2-fold greater in women with IBD than in IBD-free women (Table 1). Estimates for CD and UC were not statistically significantly different from one another (p_{severe pre-eclampsia} = 0.07, p_{early-onset severe pre-eclampsia} = 0.13).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Associations between inflammatory bowel disease, medication use and rates of pre-eclampsia in the Danish National Birth Cohort, 1996–2003

https://doi.org/10.1371/journal.pone.0129567.t001

When we considered IBD status in combination with medication use, our numbers were sufficient only to differentiate between users and non-users of OCSs. Our results suggested that the increased rates of severe pre-eclampsia associated with IBD were driven by tremendous increases in rates among OCS users, whereas there was little suggestion of an increase in pre-eclampsia rates in women with IBD who did not use OCSs while pregnant (Table 1).

Preterm delivery

There were 86,591 women included in the PTD cohort, including 666 diagnosed with IBD before the end of the study pregnancy (CD, 278; UC, 388). Overall, women with IBD had rates of PTD that were 2-fold greater than those observed in women without IBD (HR 2.20, 95% CI 1.71–2.84). This association appeared stronger for UC than for CD (Table 2), but the estimates were not significantly different from one another (p = 0.29). Women with IBD who used OCSs during pregnancy were at particularly high risk of delivering prematurely (HR 6.32, 95% CI 3.13–12.7) (Table 2); interestingly, 13 of the 14 OCS-using women with IBD who delivered preterm had disease activity registered during pregnancy. Women who used other IBD medications during pregnancy also had increased rates of preterm delivery, although to a much more modest extent (<80% increase in rates compared with women without IBD, Table 2), and even women who did not use medication while pregnant had moderately increased rates of PTD (Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Associations between inflammatory bowel disease, medication use during pregnancy and overall rates of preterm delivery in the Danish National Birth Cohort, 1996–2003.

https://doi.org/10.1371/journal.pone.0129567.t002

When we examined PTD subtypes, we found that IBD was associated with a 2-fold increase in the rate of PPROM, an increase that was driven by women with UC, as very few women had both CD and PPROM (Table 3). Women with IBD who used OCSs while pregnant had a 24-fold increase in the rate of PPROM compared with women without IBD, while pregnant women with IBD using other medications had more than triple the PPROM rate (Table 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Associations between inflammatory bowel disease, medication use during pregnancy and rates of preterm delivery (PTD) subtypes in the Danish National Birth Cohort, 1996–2003.

https://doi.org/10.1371/journal.pone.0129567.t003

IBD was also associated with a 2-fold increase in the rate of medically-indicated PTD, with estimates for CD and UC having similar magnitudes (Table 3). This association was strongest among OCS users, whose rate of medically-indicated PTD was more than 7 times that of women without IBD (Table 3). However, our results were also consistent with an almost 2-fold increased rate of medically-indicated PTD among women with no registered medication use (Table 3).

Our results also suggested that women with IBD might have a modestly increased risk of spontaneous PTD (Table 3). There were too few women with both medication use during pregnancy and spontaneous PTD to estimate the effect of medication use on the rate of spontaneous PTD with any confidence. However, women with IBD who did not use medication during pregnancy appeared to have a 70% increase in spontaneous PTD rate, compared with women without IBD; for women with UC, the spontaneous PTD rate increased 2.6-fold (Table 3).

There were too few women (n = 243) with complication-related PTD to allow for meaningful analyses for this PTD subtype.

Birth outcomes

The Apgar score cohort included 85,445 women, 654 of whom were diagnosed with IBD before the end of the study pregnancy (CD, 276; UC, 378). Among the cohort women, 629 (0.7%) delivered a living singleton whose 5-minute Apgar score was <7. Women with IBD were more than 1.5 times as likely as women without IBD to deliver a child with a low Apgar score, although this increase in risk was not statistically significant (Table 4). Restricting our analyses to women who delivered their babies at ≥37 weeks gestation (since the Apgar score is of limited use in preterm infants [8]) yielded a stronger, statistically significant association between IBD and low Apgar score (Table 4). Women with CD had a more than 3-fold increased risk of delivering a child with a low Apgar score, whereas there appeared to be no association between UC and Apgar score (Table 4); however, the two estimates were not significantly different (p = 0.17). Since there were so few women with IBD who delivered infants with low Apgar scores, we could not assess the impact of IBD medication use on the risk of low Apgar score.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Associations between inflammatory bowel disease and risks of low 5-minute Apgar score and congenital abnormalities in living singletons born to Danish National Birth Cohort mothers, 1996–2003.

https://doi.org/10.1371/journal.pone.0129567.t004

The congenital abnormalities cohort included 86,221 women, 663 with an IBD diagnosis registered before the end of the study pregnancy (CD, 279; UC, 384). Of these women, 6,067 (7.0%) delivered a child with a congenital abnormality; 2,305 (2.7%) of the offspring had a major abnormality. The risk of delivering a living singleton with a congenital abnormality did not differ for women with and without IBD, both overall and for major abnormalities alone (Table 4). However, women with CD had a moderately increased risk of delivering a child with a major congenital abnormality (RR 1.85, 95% CI 1.06–3.21). In contrast, our findings did not support an increased risk of major congenital abnormalities in the offspring associated with UC (RR 0.71, 95% CI 0.32–1.56; p_{CD vs. UC} = 0.05). The same pattern was observed when we adjusted for major maternal congenital abnormalities (along with the potential confounders identified a priori) instead of periconceptional folic acid use (RR_IBD 1.17, 95% CI 0.75–1.83; RR_CD 1.72, 95% CI 0.99–2.99; RR_UC 0.76, 95% CI 0.36–1.58). (Adjustment for maternal abnormalities and folic acid use in the same model proved impossible.)

Major abnormalities in the offspring of women with CD included choanal atresia (3 children), congenital hip dislocation (2), spina bifida (2), talipes equinovarus (2), cleft lip and palate (1), cystic kidney disease (in a child with talipes), hydrocephalus (in a child with choanal atresia), hypospadias (1), and polydactyly (1). Major abnormalities in the offspring of women with UC included congenital hip dislocation (3), cleft palate ± cleft lip (2), Hirschsprung disease (1) and talipes calcaneovarus (1). When we looked at specific major abnormalities affecting ≥2 children born to IBD mothers, we found that CD was very strongly associated with spina bifida (RR 20.2, 95% CI 4.80–84.9) and choanal atresia (RR 125, 95% CI 32.7–478). Confidence intervals for cleft palate ± cleft lip, congenital hip dislocation/subluxation and talipes were too wide to allow us to draw meaningful conclusions (S1 Table).

Of the 19 children with major abnormalities born to women with IBD, seven were exposed to medication in utero. No child was exposed to OCSs; the child with talipes born to a woman with UC was exposed to LCSs. Six children were born to women who took 5-ASA while pregnant (3 children with congenital hip dislocation, and 1 each with cleft palate, hypospadias and polydactyly).

Discussion

We investigated associations between IBD, IBD-specific medication use and adverse pregnancy and birth outcomes in the Danish National Birth Cohort. We found no evidence of an association between IBD and pregnancy loss. In contrast, IBD was strongly associated with severe pre-eclampsia, medically-indicated PTD, PPROM and low 5-minute Apgar score, and modestly associated with spontaneous PTD. CD was also moderately associated with delivering a child with a major congenital abnormality.

When we examined the effect of medication use for IBD, it became clear that the observed increase in risk of severe pre-eclampsia associated with IBD was limited to women who used OCSs during pregnancy, and that the associations with PPROM and medically-indicated PTD were also largely driven by OCS users. However, we also found associations with PPROM in users of medications other than OCSs, and possibly between untreated IBD and spontaneous PTD. Women with IBD and no medication use during pregnancy also appeared to have an increased risk of medically-indicated PTD. Furthermore, of the women with IBD whose children had a major congenital abnormality, only 37% had used medication while pregnant, and none was an OCS user.

We found no excess of pregnancy losses among women with IBD, which is consistent with the findings of Stephansson et al. and Cornish et al. for stillbirth and of Bortoli et al. for miscarriage [3,9,10]. An important caveat, however, is that we can say nothing about very early pregnancy losses in women with IBD. Women were recruited into the DNBC between weeks 6 and 12, when they visited their general practitioner to confirm their pregnancy. Women suffering losses occurring before this visit were therefore never included in the cohort. On the other hand, although we had limited power to definitively state that IBD does not increase the pregnancy loss rate and to examine the impact of IBD medication use on pregnancy loss, our results for later losses (miscarriage in physician-confirmed pregnancies and stillbirth) are reassuring nonetheless.

Our pre-eclampsia findings are inconsistent with the results of two previous studies showing no association between IBD [11] or CD [10] and pre-eclampsia. However, Bush et al. included only 116 women with IBD in their study and the determination of no association was made using a simple χ² test [11]. While the study by Stephansson et al. was considerably better powered, they only considered pre-eclampsia in the aggregate (any severity and timing of onset) [10], which is problematic since pre-eclampsia is not a single entity but a syndrome encompassing multiple conditions with similar symptoms. Moderate pre-eclampsia is attributed predominantly to maternal factors such as obesity while severe preeclampsia is characterized by placental pathology, endothelial dysfunction and inflammation. Therefore, considering pre-eclampsia as a single entity could mask associations with pre-eclampsia subtypes, particularly since moderate pre-eclampsia occurs more frequently than severe pre-eclampsia. Indeed, we found no association between IBD and pre-eclampsia overall; however, we observed a strong association between IBD and severe pre-eclampsia. Active inflammatory processes in women with IBD may interfere with proper placentation or with the placenta’s ability to meet increasing fetal demands, leading to severe pre-eclampsia as early as mid-second trimester. Our finding that the increased risk of pre-eclampsia appeared to be limited to OCS users is consistent with this hypothesis, since only women with severe, active IBD are treated with OCSs while pregnant. Consequently, in general, women with IBD need not worry about severe pre-eclampsia. Furthermore, even women requiring treatment with OCSs while pregnant—less than 7% of the women with IBD in our study—need not be unduly concerned: although the risk of severe pre-eclampsia in OCS-using women with IBD was 17 times that of women without IBD, the absolute number of affected women was small (n = 4, 8.8% of OCS-using women), since severe pre-eclampsia occurs infrequently in the general pregnant population. While a recommendation that women with IBD who require systemic steroids while pregnant should be followed more closely than the average pregnant woman for signs of pre-eclampsia seems warranted, women with IBD disease activity and/or those using steroids while pregnant (for whatever reason) are likely already subject to intense medical supervision.

Most [9,10], but not all [3,12], previous studies have found an association between IBD and PTD, typically reporting up to a doubling of PTD risk. However, whether the association only exists for women with active disease/using IBD medications—as has been suggested [1,2,11,13,14]—is unclear. Furthermore, despite many studies showing that women with IBD have increased rates of Caesarian section and induction [1], only one study has looked specifically at rates of medically indicated PTD and other PTD subtypes; Stephansson et al. found a strong association between CD and induced PTD (prevalence odds ratio [POR] = 2.59) and a moderate association (POR = 1.52) with spontaneous PTD [10].

Our spontaneous PTD findings were of similar magnitude to those reported by Stephansson et al. [10], but were driven by women with UC and no medication use. Many of the biologic mechanisms that are upregulated or altered when parturition is initiated (e.g. Th1/Th2 balance, specific cytokine cascades, expression of toll-like receptors, NOD-like receptors and prostaglandins) are also dysregulated in IBD, suggesting that the mechanisms producing IBD pathology can contribute to spontaneous PTD, even in women with quiescent disease [15].

Our findings with respect to medically indicated PTD were consistent with those of Stephansson et al. [10], but were of greater magnitude and applied to both UC and CD. The associations were strongest in women using OCSs while pregnant and were almost certainly driven at least in part by severe pre-eclampsia, which in our cohort was strongly associated with OCS use in women with IBD and which often necessitates early delivery. However, OCS use while pregnant is also a sign of severe, active IBD, which can itself be an indication for early delivery if the condition of mother or fetus warrants it [1,2].

Our finding of strong associations between IBD and PPROM in both women who used OCSs while pregnant and women who used other medications are novel and biologically plausible. Expansion of the fetal membranes as pregnancy progresses and membrane weakening as labor approaches are controlled processes mediated by matrix metalloproteinases (MMPs) and caspases [16]. Systemic inflammation may promote premature upregulation of MMP and caspase activity via proinflammatory cytokine cascades, leading to excessive collagen turnover, extracellular matrix degradation and premature membrane rupture. Medication use during pregnancy, OCS use in particular, in women with IBD is an indicator of active disease with upregulation of inflammatory pathways, potentially including MMPs and caspase. On the other hand, glucocorticoids, the corticosteroids most commonly used to treat IBD flare-ups, may themselves contribute to the initiation of parturition by causing upregulation of fetal prostaglandin synthesis; depending on the timing of exposure, glucocorticoid use may therefore contribute directly to PTD, PPROM in particular [15].

With the exception of low birth weight and possibly congenital abnormalities, the term offspring of women with and without IBD are thought to have similar birth outcomes [1]. However, contrary to the results of Stephansson et al. [10], we found a strong association between IBD and low 5-minute Apgar score in term infants, suggesting that some infants born to women with IBD may be in greater distress, at least in the period immediately after birth, than their counterparts with IBD-free mothers. Whether this is a consequence of pregnancy complications (e.g. preeclampsia), delivery complications, fetal morbidity or some combination of the three needs to be investigated further. However, while the association between IBD and low 5-minute Apgar score was strong in relative terms (HR>3), the absolute number of affected infants was small, indicating that while the midwife/physician overseeing an IBD patient’s delivery should be aware of the greater-than-usual risk of fetal distress, the chance that any given delivery will be affected is still slight.

Whether IBD is linked to congenital abnormalities remains controversial despite the large number of studies that have addressed this issue. Studies have shown UC to be strongly associated with congenital abnormalities overall [9,14] and with limb deficiencies, obstructive urinary anomalies and multiple anomalies in particular [17]; however, other studies have shown no association with UC, overall [3,18] or for the specific abnormalities mentioned [18]. Similarly, studies of CD show no association [3], an overall association [9], and an association with a specific abnormality (limb reduction) [10]. The inconsistencies may stem from differences in the abnormalities included for analysis as well as variations in study power; we saw no association between IBD and congenital abnormalities when both minor and major abnormalities were included in the analysis but a moderate association between CD and major congenital abnormalities as a group. We also saw strong links with specific groups of major abnormalities for CD (spina bifida, choanal atresia) but these did not overlap with previously reported links. However, these results should be interpreted with caution. Our defect-specific findings stem from post hoc analyses, rather than a planned, methodical examination of specific groups of abnormalities. Furthermore, the results are based on very small numbers of exposed outcomes, and even if the increases in risk are real, the absolute risks are still very small. Also worthy of note is that OCS use did not appear to be linked with an increased risk of major congenital abnormalities; six of the seven children with major abnormalities born to mothers who used IBD-related medication while pregnant were exposed only to 5-ASA.

One of our study’s great strengths was its use of the Danish National Birth Cohort, with its large number of women with IBD diagnosed prior to the study pregnancy and minimal loss to follow-up. Use of the DNBC also permitted us to adjust our estimates for potential confounders not available to solely register-based studies. Linkage to the Register of Medicinal Product Statistics allowed us to incorporate information on medication use and thereby differentiate between women with active disease and those whose disease was presumably quiescent.

Although our cohort included a large number of women with IBD, the absolute numbers of adverse pregnancy events and adverse birth outcomes among these women were often small. Consequently, we frequently did not have the power to show differences between the experiences of women with CD and those with UC (if such differences exist), despite finding strong, statistically significant associations between IBD overall and many outcomes. Another potential limitation was the use of filled corticosteroid prescriptions as a proxy for disease activity, under the assumption that corticosteroids are prescribed as sparingly as possible during pregnancy. However, this step meant that we are unable to comment on whether the overwhelming associations observed in OCS users reflect a direct adverse effect of corticosteroids themselves or illustrate the consequences of out-of-control inflammation during pregnancy.

Biologic therapy was introduced in Denmark in 1999; none of the women in our study used these medications. In addition, only six women were treated with azathioprine. It might therefore be argued that it could be difficult to generalize some of the study results to current IBD patients. On the other hand, the extent to which these newer therapies are used in pregnant women is likely limited in many settings. (Likewise, the degree to which these therapies are the current standard of care in the general IBD patient population probably also varies widely). In Denmark in the period 1999–2012, only 8% of all IBD patients were treated with tumor necrosis factor-α antagonists [19], and the use among pregnant women is assumed to have been even more limited. Similarly, while 21.9% of all IBD patients 15 years of age or older were treated with azathioprine during the same period [19], current Danish product labelling recommends that azathioprine not be used by pregnant women. We therefore contend that a study of pregnant patients using therapies other than biologics and azathioprine still represents the experience of contemporary pregnant women with IBD. Furthermore, our findings provide an important baseline against which to compare the pregnancy experiences of IBD patients using newer therapies. Such baseline data on pregnancy outcomes in women using older therapies is currently lacking, making it difficult to evaluate whether biologic therapies are better options for pregnant patients.

Despite its potential limitations, our study is one of the hitherto largest to examine associations between IBD and adverse pregnancy and birth outcomes with detailed information on potential lifestyle confounders and information on IBD-related medication use during pregnancy. We showed that IBD was strongly associated with the rates of severe pre-eclampsia, PPROM and medically induced PTD, associations that were largely limited to OCS users. However, additional associations between IBD and PPROM and spontaneous PTD, and between CD and major congenital abnormalities, all among women not using OCSs, suggest that some feature of even quiescent IBD plays a major role in several adverse pregnancy and birth outcomes.

Supporting Information