Data on birth outcome after exposure to azathioprine or mercaptopurine during pregnancy is sparse.
To examine the risk of adverse birth outcome among newborns of women exposed to azathioprine or mercaptopurine during pregnancy.
Data on drug use and births were obtained from Danish population registries. We included 76 exposed pregnancies in 69 women. Of these, we used 64 pregnancies exposed 30 days before conception or during the first trimester to examine the risk of congenital abnormalities, and 65 pregnancies exposed during the entire pregnancy to examine preterm birth and low birth weight at term. Their birth outcomes were compared with outcomes among women who did not fill prescriptions for azathioprine or mercaptopurine during pregnancy.
Azathioprine- or mercaptopurine-exposed women had a higher risk of adverse birth outcomes than unexposed controls. However, when the comparison was limited to newborns of women with the same types of underlying disease, relative risks for spontaneous and induced preterm birth, low birth weight at term, and congenital abnormalities were 1.1 (95% CI: 0.5–2.4), 4.0 (95% CI: 1.5–10.8), 1.7 (95% CI: 0.3–8.7) and 1.1 (95% CI: 0.5–2.9), respectively.
Our results suggest that adverse birth outcomes were caused by the underlying disease rather than by use of azathioprine or mercaptopurine.
The immunosuppressive drug azathioprine (AZA) is used to treat patients with severe inflammatory bowel disease (IBD), organ transplants and several autoimmune diseases.1, 2 Although these diseases often affect women of child-bearing age,1, 3 data on birth outcome among women exposed to AZA during pregnancy remain limited.
Azathioprine and its metabolite mercatopurine (MP), inhibit purine ribonucleotide and hence DNA synthesis.4, 5 Recent research has shown that the placenta forms a relative barrier to AZA and its pharmacologically active end-metabolites: the metabolite 6-thioguaninenucleotide (6-TGN) crosses the placenta, whereas 6-methylmercaptopurine does not.6 However, the role of specific AZA metabolites in inducing adverse effects during pregnancy is unknown.6 As AZA and mercaptopurine do not appear to differ substantially in efficacy and toxicity, the two drugs are usually studied together.7, 8
Azathioprine teratogenicity has been demonstrated in animals, with congenital abnormalities (CAs), such as cleft palate, limb abnormalities and ocular anomalies.5, 9 Human data on birth outcome after in utero exposure to AZA or MP are limited mostly to case reports or case series of women with organ transplants10–16 or women with IBD.17, 18 Only five controlled studies have reported birth outcome among AZA- or MP-exposed women, with conflicting findings.3, 7, 8, 19, 20 Only one small study8 estimated the relative risks of specific birth outcomes.
In our nationwide cohort study, we examined the risk of preterm birth, low birth weight at term and CAs among offspring of Danish women who received prescriptions for AZA or MP during their pregnancies.
Material and methods
This Danish nationwide cohort study was based on all women who filled prescriptions for AZA or MP during pregnancy and delivered live-born children between 1 January 1996 and 31 December 2001. Birth outcomes in this group were compared with outcomes in three cohorts of Danish women who did not fill prescriptions for AZA or MP during pregnancy. Data on drug use and birth outcomes were obtained from Danish population registries. We included only singleton births in our analyses to avoid potential confounding arising from multiple births, which are closely associated with an increased risk of preterm birth and low birth weight.21
We used the Fertility Database (FTDB), Statistics Denmark22 to obtain information on all women who gave birth between 1 January 1996 and 31 December 2001. This database encompasses all births in Denmark since 1 January 1973. Several key variables in the FTDB originate from the Danish Medical Birth Registry (MBR),23 including the civil registration number of the mother and child, date and place of birth, maternal age, gestational age, birth weight and parity.
Danish pharmacies are equipped with a computerized accounting system through which data are sent directly to the Danish Prescription database at the Danish Medicines Agency with key information on prescriptions for reimbursable drugs.24 The Prescription database, established in 1995, includes the patient's civil registration number, the type of drug prescribed [according to the anatomical therapeutic chemical (ATC) classification system] and the date of the prescription. The civil registration number was used to link the FTDB to the nationwide Danish Prescription database. This linkage permitted identification of all pregnancies in which prescriptions for AZA or MP [ATC codes L04AX01 (AZA) and L01BB02 (MP)] were filled in the period from 30 days before conception (i.e. the first day in the last menstruation) until birth. Both AZA and MP are available in Denmark only by prescription, and the national healthcare system provides partial reimbursement for their cost to the patient. From the Prescription database, we also obtained information on other drug prescriptions filled during pregnancy by AZA- or MP-exposed women who later gave birth to a child who was either preterm, had low birth weight at term or was born with CAs.
In the study, the women were classified according to exposure during two gestational stages: (i) the ‘early pregnancy’ group filled prescriptions for AZA or MP from 30 days before conception to the end of the first trimester; because organs are especially vulnerable to teratogenic exposure during this period, the ‘early pregnancy’ group was used to estimate the risk of CAs; (ii) The ‘entire pregnancy’ group filled prescriptions during the first through the third trimesters; this group was used to estimate the risk of preterm birth and low birth weight at term.
We excluded three births with missing data on gestational age or birth weight. The study also included only live births, as data on CAs among stillborn children were unavailable.
For each child born to a woman exposed to AZA or MP in ‘early pregnancy’ and for each child born to a woman exposed to AZA or MP during the ‘entire pregnancy’, 20 ‘comparison births’ were selected from the FTDB. We thus generated one comparison cohort for the ‘early pregnancy’ group and one comparison cohort for the ‘entire pregnancy’ group. The ‘comparison births’, matched by month and year of birth and the county in which the mother resided, were restricted to women who had not filled a prescription for any kind of reimbursable medication during the 3 months prior to conception (i.e. the first day of the last menstruation) and the ensuing pregnancy. If fewer than 20 ‘comparison births’ fulfilled the matching criteria, we used only those that were available. If more than 20 ‘comparison births’ were found eligible after matching, a subset of 20 was randomly selected.
The third comparison cohort was comprised of all children born between 1 January 1996 and 31 December 2001, to women who were treated with AZA or MP before pregnancy (defined as 3 months before conception), but not during pregnancy. This comparison cohort focused on children born to women with maternal diseases similar to those diagnosed in women in the drug-exposed cohort.8 As only 174 births could be identified for the third comparison cohort, they could not be matched 20 to 1 with exposed births. For this reason the third cohort was compared with both the ‘early pregnancy’ group and the ‘entire pregnancy’ group.
As with the exposed cohorts, all comparison cohorts included only live births, and excluded births with missing data on gestational age or birth weight.
Outcome data and potential confounders
The birth outcome data collected from the FTDB included preterm birth (birth before 37 completed weeks of pregnancy) and low birth weight at term (birth weight <2500 g with 37 or more weeks of pregnancy completed). Data on potential confounders included maternal age and parity. Information about CAs (including chromosomal abnormalities), diagnosed during the first year of life was obtained from the Danish Hospital Discharge Registry (HDR). This registry, which has recorded all discharge diagnoses from Danish hospitals since 1977 and all out-patient visits since 1995, includes civil registration numbers, dates of admission and discharge, and up to 20 discharge diagnoses coded according to the International Classification of Diseases (ICD-8 before 1994 and ICD-10 from 1994 onward25). The codes for CAs (including chromosomal abnormalities) were Q0.00 to Q99.9 in ICD-10. Diagnoses of congenital dislocation of the hip and undescended testis were excluded from the study because of their low validity.26
Information from the HDR was also used to categorize preterm births into two subtypes: induced preterm deliveries were identified from information about either induction or Caesarean section before labour onset; the remaining preterm births were considered spontaneous. For women in the exposed cohorts, we also obtained information about discharge diagnoses recorded in the HDR during the 5-year period preceding delivery.
Civil registration numbers, assigned to all live-born children at birth and new residents of Denmark and recorded in the Danish Civil Registration System27 permitted accurate linkage between registries. When data from the different registries were merged at Statistics Denmark, unique case numbers were substituted for civil registration numbers. Thus, the investigators did not have access to personally identifiable information.
For birth weight and maternal age, the median, mean and standard deviation were used as descriptive statistics for the ‘early pregnancy’ group, the ‘entire pregnancy’ group and the matched comparison cohorts for each exposed group. After stratifying for potential confounding factors [maternal age (<30 years, ≥30 years) and parity (1, 2+)]. We computed pooled Mantel-Haenszel estimates of relative risks and 95% confidence intervals (CI) for preterm birth, low birth weight at term and CAs associated with AZA or MP use. The final analyses were stratified only by maternal age, as parity did not change the risk estimates. For preterm births we repeated the analyses stratifying by type of delivery (spontaneous vs. induced) to examine whether this variable affected the risk estimates. We first computed Mantel-Haenszel estimates using the two initial comparison cohorts and then used the third comparison cohort. sas software version 8.2 was used for descriptive analyses and the statistical program episheet was used for the Mantel-Haenszel estimates.
The study was approved by the Danish Data Protection Agency (Record No. 2003-41-2833).
Women filled prescriptions for AZA or MP during 64 pregnancies in the ‘early pregnancy’ group and during 65 pregnancies in the ‘entire pregnancy’ group. The comparison cohorts for AZA- or MP-exposed births in the ‘early pregnancy’ and ‘entire pregnancy’ groups contained 1243 and 1274 births, respectively (Table 1). Seventy-six births from 69 different AZA- or MP-exposed women were included in both exposure periods. Women with ulcerative colitis or Crohn's disease delivered 41 children (53.9%); the remainder were delivered by women with chronic hepatitis or other liver diseases, renal transplant recipients and women with other systemic diseases (Table 2). Sixty-eight women had filled prescriptions for AZA and only one woman had filled a prescription for MP.
|Pregnancies exposed to AZA or MP in the first trimester or 30 days before conception (early pregnancy)* (N = 64)||Comparison births for the early pregnancy cohort† (N = 1243)||Pregnancies exposed to AZA or MP during the first to the third trimesters (entire pregnancy)‡ (N = 65)||Comparison births for the entire pregnancy cohort§ (N = 1274)|
|Mother's age (years)|
|Mean (s.d.)||28.7 (4.3)||29.3 (4.6)||28.8 (4.7)||29.2 (4.7)|
|Parity >1 (%)||32/64 (50.0)||680/1243 (54.7)||30/65 (46.2)||690/1274 (54.2)|
|Birth weight (g)|
|Mean (s.d.)||3055 (855)||3512 (567)||3031 (870)||3542 (554)|
|Gestational age (weeks)|
|Mean (s.d.)||37.7 (3.7)||39.5 (1.8)||37.6 (3.6)||39.6 (1.8)|
|Preterm birth (%)||–||–||17/65 (26.2)||59/1274 (4.6)|
|Low birth weight at term (%)||–||–||2/48 (4.2)||17/1215 (1.4)|
|Congenital abnormalities (%)||6/64 (9.4)||49/1243 (3.9)||–||–|
|Number of AZA- or mercaptopurine-exposed births according to maternal condition recorded in the National Hospital Discharge Registry||Preterm birth, N (%)||Low birth weight at term, N (%)||Congenital abnormalities, N (%)|
|Crohn's disease (N = 26)||3/26 (11.5)||0/23 (0.0)||1/26 (3.8)|
|Ulcerative colitis* (N = 6)||2/6 (33.3)||1/4 (25.0)||3/6 (50.0)|
|Crohn's disease or ulcerative colitis (N = 9)||0/9 (0.0)||0/9 (0.0)||0/9 (0.0)|
|Chronic hepatitis and other liver diseases (N = 11)||2/11 (18.2)||1/9 (11.1)||0/11 (0.0)|
|Renal transplant recipients (N = 8)||6/8 (75.0)||0/2 (0.0)||1/8 (12.5)|
|Interstitial lung disease (N = 3)||1/3 (33.3)||0/2 (0.0)||0/3 (0.0)|
|Myasthenia gravis (N = 2)||0/2 (0.0)||0/2 (0.0)||0/2 (0.0)|
|Glomerulonephritis (N = 2)||1/2 (50.0)||0/1 (0.0)||0/2 (0.0)|
|Systemic lupus erythematosus (N = 2)||1/2 (50.0)||0/1 (0.0)||0/2 (0.0)|
|Polyarteritis nodosa (N = 1)||0/1 (0.0)||0/1 (0.0)||1/1 (100.0)|
|Unknown† (N = 6)||1/6 (16.7)||0/5 (0.0)||0/6 (0.0)|
In the ‘entire pregnancy’ group there were 17 preterm deliveries (26.2%). Nine of these were induced (52.9%); in comparison only eight of 59 preterm deliveries were induced in the matched comparison cohort (13.6%). About 75.0% of all births from renal transplant recipients were preterm (six of eight children; Table 2). These six preterm births were all induced and included three elective Caesarean sections (one for a woman who also had diabetes and two for women who developed pre-eclampsia during pregnancy), two acutely indicated Caesarean sections for unknown reasons, and one induced delivery in a woman with hypertension.
There were two cases of low birth weight at term in the ‘entire pregnancy’ group. The diseases recorded in the HDR for the mothers included: (i) ulcerative colitis and chronic active hepatitis in one woman (treated with ciclosporin, prednisolone, hydrocortisone, sulfasalzine, propranolol and antibiotics in addition to AZA during pregnancy) and (ii) hepatic transplantation 3 years before pregnancy because of cirrhosis in the other (treated with prednisolone, folic acid and AZA during pregnancy).
In the ‘early pregnancy’ group, six children had CAs, three born to women with ulcerative colitis (Table 2). Of these, the first was a preterm newborn with stenosis of the pulmonal valve. (The mother had been treated with mesalazine, budesonide and AZA in early pregnancy.) The second had congenital hydronephrosis. (The mother had been treated with sulfazalazine and AZA in early pregnancy.) The third was born with a cyst at the ductus choledochus. (The mother had been treated with hydrocortisone, prednisolone and AZA in early pregnancy.) The fourth child was born to a mother with Crohn's disease and had an occipital encephalocele. In this case, the mother had been treated with budesonide and folic acid in addition to AZA in early pregnancy. The fifth, born to a woman with polyarteritis nodosa (treated with cimetidine, metoclopramide, prednisone, penicillin, paracetamol in combination with codeine and AZA in early pregnancy), had a malformation of the sternocleidomastoideus muscle. The sixth child was a preterm newborn with an asymmetrical face whose mother was a renal transplant recipient treated with prednisone, labetalol and nitrofurantoin in early pregnancy.
Relative risks of adverse birth outcome
Table 3 shows the risk estimates for preterm birth (overall, spontaneous and induced), for low birth weight at term and for CAs among AZA- or MP-exposed women, based on the first two comparison cohorts and stratified for maternal age. When the third comparison cohort, composed of women with diseases similar to those of the exposed women (i.e. women with lifetime use of AZA or MP, but not during the 3 months preceding conception or during pregnancy) was used, no associations were found, apart from an increased risk of preterm delivery that was entirely due to an increased rate of induced preterm deliveries (Table 4).
|AZA- or MP-exposed births||Control births||RRMH* (95% CI)|
|Preterm birth†||17/65 (26.2)||59/1274 (4.6)||5.6 (3.5–9.1)|
|Spontaneous||8/65 (12.3)||51/1274 (4.0)||3.0 (1.5–6.2)|
|Induced||9/65 (13.8)||8/1274 (0.6)||22.9 (9.3–56.3)|
|Low birth weight at term‡||2/48 (4.2)||17/1215 (1.4)||3.0 (0.7–12.4)|
|Congenital abnormalities§||6/64 (9.4)||49/1243 (3.9)||2.3 (1.0–5.2)|
|AZA- or MP-exposed births||Control births||RRMH* (95% CI)|
|Preterm birth†||17/65 (26.2)||25/174 (14.4)||1.9 (1.1–3.3)|
|Spontaneous||8/65 (12.3)||19/174 (10.9)||1.1 (0.5–2.6)|
|Induced||9/65 (13.8)||6/174 (3.4)||4.6 (1.7–12.0)|
|Low birth weight at term‡||2/48 (4.2)||4/149 (2.7)||1.7 (0.3–8.7)|
|Congenital abnormalities§||6/64 (9.4)||15/174 (8.6)||1.1 (0.5–2.9)|
We found an increased risk of preterm birth among women who filled prescriptions for AZA or MP during pregnancy, compared with women who were not prescribed any kind of reimbursable medication during the period extending from 3 months before conception to the end of pregnancy. This finding, however, applied primarily to induced preterm deliveries, and is likely to be a result of usual obstetric practice for these patients. Our data also indicated an increased risk of low birth weight at term and CAs for newborns of AZA- or MP-exposed women. However, these associations may be confounded by the underlying disease or concurrent use of other drugs. When we used the third comparison cohort in an attempt to take the underlying disease into consideration, the association became weaker for induced preterm birth and disappeared for spontaneous preterm birth, for low birth weight at term, and for CAs, implying that the adverse birth outcome were caused by the underlying disease and not by use of AZA or MP.
The main strengths of our study are its population-based design, made possible by Denmark's uniformly organized national healthcare system, and the minimization of selection bias through use of exposure data from a complete nationwide prescription database. Another strength is that measurement of exposure was based on prescriptions rather than self-reported use (which may lead to recall bias or under ascertainment28). Study weaknesses included a lack of information on patient compliance, as filling a prescription was used as a proxy for actual use of a drug. However, because the drugs are used for long-term treatment of severe chronic diseases, the likelihood of compliance is high. Still, if some women who filled prescriptions for AZA or MP stopped using the drug when they discovered they were pregnant, misclassification of exposure could ensue. It is also possible that some women in the third comparison cohort filled a prescription for AZA or MP earlier than 3 months before pregnancy, but actually used it during pregnancy. Another limitation is that we had no information on the prescribed daily dose of AZA or MP or on length of treatment. Furthermore, data on drugs administered during hospitalization, including AZA or MP, are not registered in the prescription database and therefore could not be included in the study. Any potential misclassification of exposure resulting from these limitations would tend to underestimate our risk estimates.
The quality of the outcome variables in the MBR (and thus in the FTDB) are high.29 However, its data on gestational age are subject to some misclassification, since the gestational age recorded in the Birth Registry is a week longer in some cases than that recorded in the medical records.29 This misclassification was more likely to affect women in the comparison cohorts, as women with severe chronic diseases who were treated with AZA or MP during pregnancy probably underwent more prenatal ultrasound examinations and therefore had more precise data on gestational age. Thus, differential misclassification of gestational age may have led to an overestimated risk of preterm birth for AZA- or MP-exposed women, compared with women who did not fill prescriptions during pregnancy.
With regard to possible confounders, we were able to consider the influence of maternal age and parity, but lacked data on maternal smoking and alcohol use, as well as sufficient data to study the influence of co-medication. If AZA- or MP-exposed women were more prone to smoke during pregnancy than were comparison women, the effect of AZA or MP on preterm birth could be overestimated, as smoking during pregnancy has been associated with an increased risk of preterm birth (OR = 1.4, 95% CI: 1.1–1.9).30 In fact, women with Crohn's disease who smoke are more likely to have severe disease31 and therefore may be more likely to be treated with AZA. With respect to CAs, tobacco use or moderate alcohol consumption during pregnancy has not been identified as independent risk factors for any of the CAs occurring among newborns of AZA- or MP-exposed women in our study.32, 33 However, concomitant use of other drugs during pregnancy could have confounded our effect estimates for CAs, as some of the drugs used by AZA-exposed women may have teratogenic effects themselves (i.e. prednisone34). Furthermore, aminosalicylates (i.e. mesalazine and sulfasalazine) may interact with AZA and lead to higher concentrations of the pharmacologically active metabolite, 6-TGN.35
Information on CAs was obtained from the HDR, whose data are of generally high quality with an 85% correct coding rate.26 However, we were not able to procure data on miscarriages or induced abortions. Selection bias could have occurred if women exposed to AZA or MP had more miscarriages or induced abortions related to fetal abnormalities than did comparison mothers. This would lead to underestimation of the risk of CAs. It is also important to consider that teratogens increase the rate of selected CAs but not all CAs.36 Cohort studies can detect only large increases in the risk of specific CAs, and are limited in their ability to provide an assurance of safety. The CAs observed among exposed newborns in the present study affected different organ systems, and none of the affected children had limb malformations, ocular anomalies or cleft palate, as found in animal studies.5, 9
Our finding of an overall increased risk of preterm birth following AZA or MP exposure in pregnancy corroborates two cohort studies of renal transplant recipients.19, 20 Sgro et al. reported an increased prevalence of preterm birth, stillbirth and low birth weight among 44 pregnancies in women who had undergone renal transplantation, and who had been treated with prednisone, together with AZA, ciclosporin, or both during pregnancy.19 Similarly, Bar et al. observed significantly more preterm births and children born with intrauterine growth restriction (IUGR), in a study of 38 renal transplant recipients who were treated with combinations of prednisone, AZA, ciclosporin and tacrolimus during pregnancy. In this study, the comparison cohort was composed of newborns of women with primary renal disease who were not treated with immunosuppressive drugs.20 Both studies did not find an increased prevalence of CAs among children of women exposed to immunosuppressive drugs during pregnancy.19, 20 However, neither study estimated relative risks, distinguished between spontaneous and induced preterm births, controlled for confounding, or specified whether children with adverse birth outcomes were born to women exposed to AZA, ciclosporin, or tacrolimus.
Based on a record review of 155 male and female patients who had conceived after developing IBD, Francella et al.7 concluded that use of MP before conception or during pregnancy appears to be safe. These data were compared with IBD patients who completed their pregnancies before taking MP. The authors reported an overall RR of 0.85 (95% CI: 0.47–1.55) for a successful pregnancy outcome for exposed parents, controlled for maternal age and parent gender. Similarly, another recent cohort study encompassing 101 pregnancies among women with IBD treated with AZA or MP found no overall association with poor pregnancy outcomes, after adjustment for maternal age.3 However, the relative risks of specific adverse birth outcomes were not estimated in either study.
To date, only one small cohort study has estimated the relative risks of specific adverse birth outcomes among children of AZA- or MP-exposed women.8 This study, conducted by some co-authors of this work, compared birth outcomes among 11 women with IBD or other diseases with those for 19 418 controls who did not receive a prescribed medication during pregnancy. Two CAs were documented among nine women exposed in early pregnancy (OR = 6.7, 95% CI: 1.4–32.4), and three preterm children with low birth weight were delivered by 10 women exposed during the entire pregnancy, corresponding to odds ratios of 6.6 (95% CI: 1.7–25.9) for preterm birth, and 3.8 (95% CI: 0.4–33.3) for low birth weight. However, no distinction was made between spontaneous and induced preterm births. The study included women who filled prescriptions in a Danish county between 1991 and 2000. Thus, some of their pregnancies were also included in the present study. Our nationwide study corroborates this regional study with respect to the overall increased risk of preterm birth.
A key problem complicating the interpretation of data in our study is confounding by indication. Because drug use is closely correlated with underlying disease and disease activity, it is difficult to distinguish between the influence of the disease itself on the risk of adverse birth outcome and possible adverse drug effects.
In conclusion, we found an overall increased risk of preterm birth and an indication of an increased risk of low birth weight at term and CAs among newborns of AZA- or MP-exposed women, compared with newborns of women who did not use any prescription drugs during pregnancy. These results are difficult to interpret, because we were unable to study the influence of disease activity and co-medication. However, when the comparison was limited to newborns of women with the same types of underlying disease, only the association for induced preterm birth remained elevated, implying that the adverse birth outcome were caused by the underlying disease and not by the use of AZA/MP.
This study received financial support from the Western Danish Research Forum for Health Sciences, from Ingeborg and Leo Dannins Foundation for Scientific Research, and from the Danish Pharmaceutical Association.