Patients and methods

A regularly updated database of AIH patients attending the Institute of Liver Studies at King's College Hospital London since 1983 was interrogated. Up to the end of 1998, 162 females who had been regularly followed up during this study period were identified who fulfilled criteria for definite AIH as defined by the International Autoimmune Hepatitis Group.2 ,4 Of these, 83 were aged less than 50 years and 68 less than 45 years. At accession, all had moderate to severe interface hepatitis on liver biopsy. Eighteen had reported being pregnant at any time and these comprised the present study cohort. Sixteen of these had type 1 AIH, with antinuclear (ANA) and/or smooth muscle (SMA) autoantibody titres ranging from 1:40 to 1:2560 (median 1:160) and two (patients Nos 5 and 6, table 1) had type 2 disease, both with anti-liver-kidney microsomal (anti-LKM1) antibody titres of 1:2560 at presentation. Seven had biopsy proven cirrhosis and 11 were non-cirrhotic at conception. Median age at diagnosis of AIH was 19 years (range 3–33) and median duration of follow up since diagnosis was 14 years (range 4–27). There were no significant differences between cirrhotic and non-cirrhotic patients with respect to age at presentation (median 18.5 v 20.5 years, respectively) or duration of disease (median 14.5v 13.5 years). The records of these women were systematically reviewed with respect to their progress during their pregnancies, including therapy, development of complications, and maternal and fetal outcomes.

The remaining 50 patients aged less than 45 years (24 with cirrhosis) served as a control group. Forty five had type 1 AIH with the same median and range of ANA and/or SMA titres at presentation as the pregnancy group and five had type 2 AIH (median anti-LKM1 1:320; range 1:160–1:1280). Median age at diagnosis in the control group was 28 years (range 9–44) and median duration of follow up was 11 years (range 3–28). There were no significant differences from the pregnancy group with respect to these parameters.

For the purpose of this study, a flare of hepatitis was defined as a twofold increase in serum aspartate aminotransferase (AST) activity above the upper normal limit (⩽50 IU/l) or a lesser increase in AST in conjunction with increased serum globulin concentration (upper normal limit 35 g/l) and the re-emergence of symptoms, as defined previously.11 The study was undertaken with informed consent of patients and under the supervision of the local ethics committee in accordance with the 1975 Helsinki Declaration guidelines.

Results

Thirty five pregnancies were reported by 18 women. Median age at conception (all pregnancies) was 28 years (range 18–36). In two patients (Nos 15 and 18, table 1), AIH was diagnosed de novo during their pregnancies. In the remainder, median duration of disease at the time of pregnancy was eight years (range 1–27). One patient reported termination of a pregnancy at six weeks' gestation for reasons unconnected with her disease. Two others reported early miscarriages, one within the first trimester and the other at 14 weeks' gestation. Details of the remaining 32 pregnancies are summarised in table 1. Thirty one live births resulted from these 32 pregnancies, including one successful twin pregnancy in a cirrhotic patient.

Conception was achieved by in vitro fertilisation in three pregnancies, without prior discussion with a hepatologist. One was in a woman with cirrhosis (patient No 11) who had a full term delivery without any complications or deterioration in liver function throughout the pregnancy. The other two were in one (patient No 18) of the two cases diagnosed de novo and were associated with major complications which are discussed below. Apart from this one patient, where emergency caesarean section was required in her first pregnancy (see below), delivery was vaginal in all completed pregnancies (including the twins).

Discussion

Reduced fertility has been associated with AIH12 ,13but conception is not uncommon in women whose disease is in remission. However, the risk of an unsuccessful pregnancy has been reported to be high in these patients, with a fetal death rate of 50% in one series and of 33% of 36 cases pooled from the literature.12 ,13In the largest previous study, conducted 20 years ago, Stevenet al reported 30 pregnancies in 16 women.13 Fetal loss was high (23%) due to spontaneous abortions in three instances and elective termination of the pregnancies in four because of the belief that the pregnancy would adversely affect the course of the liver disease.13 Among the remaining 23 pregnancies, there were two stillbirths (9%) and the neonatal death rate was also 9%. From pooled data, largely prior to 1975, Steven and colleagues13 noted that the abortion rate reported from 36 pregnancies was 11%, termination rate was 6%, perinatal death rate 20% (from 30 completed pregnancies), and a maternal death rate of 14% was described, although they reported no maternal deaths in their own series. These previous studies were of women diagnosed as having “chronic active hepatitis” and may have included some patients with liver disease of aetiologies other than AIH.

In the present study of a relatively homogeneous group of patients with definite AIH according to international criteria,2 ,4 no terminations for clinical management reasons were recorded but there were two maternal deaths (11%). However, one of these deaths was due to acute variceal bleeding six months post partum in a patient who had been advised against pregnancy because of her history of severe portal hypertension and previous bleeding from gastro-oesophageal varices, which had been treated by injection sclerotherapy. The other, due to severe pulmonary hypertension, was unexpected (given that the patient had had two uncomplicated deliveries previously) and it is difficult to know whether her liver disease might have contributed to this. There were no perinatal deaths but it is worth noting that the fetal loss rate (including one termination) of 14.3% is less than that (33.0%) reported by Stevens and colleagues13 for AIH patients and than the comparable rate (24.0–29.3%) reported for diabetic patients in the USA30 and north west England.31 Two of the fetal deaths in the present study were spontaneous abortions prior to 20 weeks' gestation. This early fetal loss rate may be an underestimate because women with AIH (with or without cirrhosis) tend to become accustomed to erratic menstrual cycles and, certainly in our experience, sometimes may not realise that they have conceived. In earlier series, prematurity was reported to be as high as 30% and the caesarean section rate up to 26%,13 although the latter was thought to reflect attempts to salvage threatened pregnancies. In the present series, both the frequency of prematurity (2/32, 6%) and the caesarean section rate (3%) were low. Comparative data for the general population of England in 1997 are as follows: terminations (all causes) 21.5%, caesarean sections 17.0%, still births 5.4%, perinatal mortality 0.8%, infant mortality 0.6%.32

Other complications during pregnancy were surprisingly few but it is also important to recognise that pregnancy can apparently precipitate previously unsuspected AIH, as evidenced by the de novo diagnosis of the disease in two of the patients in the present series. Indeed, one of these presented with ascites and decompensated liver disease which contributed largely to the fetal complication rate in this study, with the birth of a premature (physically and mentally handicapped) infant by caesarean section and a subsequent fetal death after 20 weeks' gestation (although the latter was unrelated to the patient's initial presentation).

Exacerbations of AIH during pregnancy were relatively infrequent (12.5%) and similar to that reported previously.13However, relapses occurred post partum in four pregnancies (12.5%) in three patients whose disease had apparently been well controlled during their pregnancies, although in one case this was due to a documented instance of non-compliance with therapy. Remission was rapidly achieved in all of these cases by reintroducing or increasing the doses of steroids but it is important to recognise that rebounds post partum do occur and that, such is the diversity of the disease in patients who are well controlled during pregnancy, it is difficult to predict such relapses. Why some patients have exacerbation of their AIH during pregnancy while others maintain in remission only to relapse post partum is unclear. These events have been documented not only in AIH16 but also in other autoimmune diseases.33 ,34 It might be expected that the natural mild immunosuppression in the mother that protects the fetus during gestation would favour remission but it is now well recognised that the immunological responses in autoimmune diseases are profoundly affected by the relative concentrations of oestrogen, prolactin, progesterone, and testosterone.35 Thus the different outcomes in AIH patients during and after pregnancy may be related to differential variations in complex hormonal profiles between individuals.

The appropriateness of azathioprine in pregnant patients with AIH is still controversial. The drug has been found to have teratogenic effects in mice and rabbits but not in rats.17 ,18 ,36 The reasons for these species and strain differences are unknown but may be related to differences in metabolism of azathioprine, differentials in tissue sensitivity to the drug or its metabolites, dose related effects, or a combination of these factors. Although azathioprine crosses the placenta,24 the lack of teratogenicity recorded in newborns suggests that the human fetus is not overtly affected by the drug.37 Neither is there evidence of permanent genomic or gonadal damage. Lymphopenia, hypogammaglobulinaemia, and thymic hypoplasia have been reported in children of mothers who have taken azathioprine in addition to prednisolone in pregnancy but these changes have been regarded as reversible.29 In the present series, we have found little to suggest that this drug or its metabolites are toxic in pregnancy. Although formal haematological or immunological evaluation was not performed in the children born to the mothers described here, there have been no reports of defective immune responses or other abnormalities attributable to azathioprine in any of the children.

As individuals who are sensitive to azathioprine usually exhibit side effects of the drug within the first few months of commencing therapy, in de novo presentations of AIH during pregnancy it may be prudent to initially achieve remission using steroid monotherapy. However, in patients who become pregnant while taking azathioprine, there is little evidence to suggest that there should be either a dose reduction or drug withdrawal. This is particularly pertinent for patients with established disease in whom azathioprine is critical for the maintenance of remission, because the risk to the fetus of a relapse of AIH during pregnancy is probably greater than that associated with azathioprine therapy.38 The doses used to maintain remission, either 1 or 2 mg/kg/day, appear not to be harmful to the fetus in either early or late pregnancy. Thus the present observations suggest that termination of pregnancy purely as a prophylactic measure is not mandatory for women who conceive on the drug. Moreover, as patients have often been established on the drug for many years (median 7.2 years in the present cohort) it may be inadvisable to suddenly stop this agent. If discontinuation of azathioprine therapy is contemplated, it should be remembered that the drug is used for its “steroid sparing” effect in AIH and that a compensatory increase in steroid dosage may be required to sustain remission.

Given that no factors could be identified in the present study to predict which patients were likely to experience activation of AIH during their pregnancies, the need for careful surveillance and disease control during pregnancy is emphasised. To use the analogy with another high risk condition such as diabetic patients undergoing pregnancy, in which live births can be expected in up to three quarters of pregnancies,31 ,32 it has been reported that close attention to euglycaemia significantly reduces complications in the newborn.31 Thus it could be argued that a similar paradigm should exist with respect to AIH and that, in view of the long experience with drugs such as azathioprine and prednisolone, immunosuppression withdrawal is probably unnecessary and may be detrimental to mother and child.