- Top of page
Objective To examine birth outcomes and pregnancy complications in women with a history of anorexia nervosa.
Design Prospective cohort study.
Setting Nationwide study in Sweden.
Population All primiparous women—discharged from hospital with a diagnosis of anorexia nervosa during 1973 to 1996—who gave birth during 1983 to 2002 (n= 1000) were compared with all non-anorexia nervosa primiparous women who gave birth during the same years (n= 827 582).
Method Register study with data from Medical Birth Registry and National Patient Discharge Register.
Main outcome measures Pre-eclampsia, instrumental delivery, prematurity, small for gestational age, birthweight, Apgar score and perinatal mortality.
Results Main birth outcome measures in women with a history of anorexia nervosa were very similar to the main population. The only observed differences were a slightly lower mean birthweight and lower adjusted odds ratios for instrumental delivery in the anorexia nervosa group compared with the main population. Neither severity of the disease nor a shorter recovery phase after first hospitalisation was related to pregnancy complications or birth outcomes.
Conclusion A history of anorexia nervosa was not associated with negative birth outcomes. Thus, special obstetric monitoring of pregnant women with history of anorexia nervosa does not seem to be warranted in a country with a satisfactory maternity surveillance.
- Top of page
Anorexia nervosa is a severe psychiatric disease, which largely affects adolescent girls in young childbearing ages. It is characterised by a refusal to maintain minimally normal body weight (<85% expected weight), an intense fear of weight gain and a disturbance of the perception of bodyweight or shape.1 In postpubertal women, amenorrhoea (absence of at least three consecutive menstrual cycles) occurs.
Anorexia nervosa has obvious medical consequences due to malnutrition and social consequences associated with the psychiatric disturbance. Fertility is also affected with a reduced maternity rate as a consequence (A. Hjern et al., unpubl. obs.). However, after recovery from the active phase of the disease, or occasionally during this active phase, pregnancy may occur, awakening questions about health development not only for the mother but also for the fetus and for the newborn child. Except for the effects of poor nutrition on the growth of the fetus,2 starvation during the teens may also affect the development of the pelvis, thereby increasing obstetric risks.
Previous studies have focused on women during active phase or after recovery. Active episodes of eating disorders have been reported to increase risks of caesarean deliveries,3,4 lower Apgar scores and lower birthweights of infants,3,5 higher rates of preterm deliveries3,6 and perinatal mortality6 in comparably small clinical studies. This elevated risk of complications has been reported for women with a history of anorexia nervosa,6,7 irrespective of whether the anorexia nervosa was in a active phase or remitted.3 However, more favourable birth outcomes have also been reported in two studies of women who have recovered.4,5
In 2004, Sollid et al.7 presented findings from a register-based sample of 302 women with a history of eating disorders (anorexia nervosa, other specified feeding disturbances or other specified disturbances). A doubled risk of low birthweight and relative risks of 1.7–1.8 for preterm delivery and small for gestational age (SGA) were reported. The use of multiple diagnoses is, however, an important limitation with this study, since several studies indicate that women with bulimia nervosa have a greater risk of birth complications compared with other women with eating disorders, including anorexia nervosa.4,8–10
To summarise, there is some evidence, even if not unambiguous, supporting that a history of anorexia nervosa constitutes a risk for complications during pregnancy and delivery. However, inclusion of women with different kinds of eating disorders (and/or in different phases of the disease), questions about selection of cases and the use of small- or medium-sized samples imply some uncertainty as to how previous findings about the potential risks and their magnitude should be understood. In this study, we used data from the Swedish national registers to study birth outcomes in an entire national cohort covering over two decades of former inpatients with anorexia nervosa.
- Top of page
This study was based on national health databases held by the Swedish National Board of Health and Welfare linked through each individual’s unique personal identification number. Pregnancy and delivery outcome for all 828 582 primiparous women (and their newborn infants) who gave birth in Sweden during 1983–2002 were identified in the Swedish Medical Birth Register as well as the age of the mother, birth year and sex of the child and relevant perinatal and pregnancy indicators.11
The Swedish Hospital Discharge Register covers data on all psychiatric inpatient discharges and diagnoses in Sweden from 1974 onwards. From 1987, this register includes all inpatient care. Among the mothers defined above, we identified all women above 10 years of age in the register, discharged with a main diagnosis indicating anorexia nervosa during 1973–1996; 306.50 (International Classification of Diseases [ICD]-8) during 1973–86 and 307B (ICD-9) during 1987–96, in all 1000 individuals.
The perinatal variables identified in the register were gestational age (in completed gestational weeks based on ultrasound or, if such measurement was not performed, the last menstrual period), birthweight (in grams), birthweight for gestational age (in SDs below or above mean birthweight for gestational age according to the Swedish birthweight curve), Apgar score at 5 minutes and cephalhaematoma. Birthweight for gestational age was stratified into SGA (less than −2 SDs), appropriate for gestational age (between −2 and +2 SDs) and large for gestational age (greater than +2 SDs). Deaths were classified as stillbirths (death before delivery but after 28 weeks of gestation) and neonatal (death before 4 weeks of age). The following pregnancy complications were identified: preterm rupture of the membranes, pre-eclampsia and multiple pregnancies. Instrumental deliveries (caesarean section, forceps delivery and vacuum extraction) and anaesthetic procedures were also identified. Maternal smoking habits (smoker/nonsmoker), lone parenthood were recorded in early pregnancy, just as weight and height that were used to calculate body mass index. The latter was categorised according to World Health Organization standards into underweight (below 18.5 kg/m2), normal (18.5–25 kg/m2) and overweight (>25 kg/m2).
Multivariate analysis was conducted by logistic regression with individual variables as well as four dichotomised summarised outcomes (yes/no): instrumental deliveries, any perinatal complication, pregnancy complications and stillbirth/perinatal death. Year of delivery was entered as a continuous variable and maternal age at childbirth (year of birth of the mother) was entered as a category variable into the models. Birthweight was analysed in a linear regression model with adjustment for maternal age and year of birth. Statistical analyses were carried out using the SPSS 12.0 (SPSS Inc., Chicago, IL, USA) for Windows software package.
- Top of page
Birth outcomes and pregnancy complications for all 828 582 primiparous women (and their newborn infants) who gave birth in Sweden during 1983–2002 were investigated. Altogether 1000 of these women had been hospitalised at least once for anorexia nervosa in the period 1973–96.
Women in the anorexia nervosa group were slightly more often registered as lone parents and were less often overweight compared with non-anorexia nervosa mothers (Table 1).
Table 1. Maternal characteristics of the AN study population and the comparison group (non-AN)
| ||AN, n= 1000||Non-AN, n= 827 582|
|Maternal age at birth of child (years)|
|Missing||0|| ||1|| |
|Birth in gestational week|
|Missing||2|| ||1480|| |
|Other family situation||40||4.4||20 469||2.6|
|Missing||82|| ||52 329|| |
|Missing||83|| ||52 457|| |
|Overweight + obese||64||12.6||91 118||25.4|
|Missing||486|| ||469 322|| |
In Table 2, frequencies and odds ratios (OR) for pregnancy complications and birth outcomes are presented. The only child-related detected difference (not presented in table) was a significant lower mean birthweight in infants born to women in the anorexia nervosa group (mean 3387 g) compared with those born to women in the general population (mean 3431 g, P= 0.005; after adjustments for maternal age and year of birth). However, the proportion of SGA infants did not differ between the groups. When it comes to instrumental delivery, women in the anorexia nervosa group had lower risk of caesarean section and vacuum extraction, with a crude OR of 0.8 (95% CI 0.7–0.99) and 0.7 (0.5–0.8), respectively compared with mothers in the general population. There were no significant differences concerning use of analgesia (not presented in Table) during childbirth except for pudendal block (PDB) anaesthetics that had been used less often in the anorexia nervosa group OR 0.6 (0.5–0.7). However, when adjusting for year at birth (the use of PDB has decreased significantly from 64% in 1983 to 4% in 2002) the difference was erased, OR 0.9 (0.7–1.1).
Table 2. Frequencies and crude OR for pregnancy complication, instrumental deliver, perinatal complication and stillbirth/perinatal death
| ||AN, n= 1000||Non-AN, n= 827 582||OR (95% CI)|
|Yes||26||0.6||24 453||1.2||0.9 (0.6–1.3)|
|No||974||99.4||803 129||98.8|| |
|Premature rupture of membranes|
|No||994||97.4||817 812||97.0|| |
|No||992||99.2||820 279||99.1|| |
|Yes||113||11.3||114 485||13.8||0.8 (0.7–0.99)|
|No||887||88.7||713 097||86.2|| |
|No||993||99.3||820 940||99.2|| |
|Yes||81||8.1||98 560||11.9||0.7 (0.5–0.8)|
|No||919||91.9||729 022||88.1|| |
|Born in gestational week 32|
|Yes||10||1.0||10 382||1.3||0.8 (0.4–1.5)|
|No||990||99.0||817 200||98.7|| |
|Born in gestational week other than 37–41|
|Yes||64||6.4||62 546||7.6||0.8 (0.7–1.1)|
|No||936||93.6||765 036||92.4|| |
|Yes||37||3.7||29 910||3.6||1.0 (0.7–1.4)|
|No||942||94.2||784 888||94.8|| |
|Yes||13||1.3||15 541||1.9||0.7 (0.4–1.2)|
|No||987||98.7||812 041||98.1|| |
|No||990||99.0||819 122||99.0|| |
|Low Apgar score|
|No||987||98.7||819 643||99.0|| |
|Yes||13||1.3||15 000||1.8||0.7 (0.4–1.2)|
|No||987||98.7||812 582||98.2|| |
|No||1000||100||826 648||100|| |
|No||998||99.8||825 105||99.7|| |
In Table 3, results concerning composite groups of variables are presented: instrumental delivery, perinatal complication, pregnancy complication and perinatal death or stillbirth. In these multivariate analyses, we adjusted for the possible confounders such as year at birth of child, maternal age and cigarette smoking. However, these adjustments only marginally affected the OR. The only significant difference, not influenced by the adjustments, was a lower risk for instrumental delivery, OR 0.7 (0.6–0.8) in the anorexia nervosa group compared with the general population.
Table 3. Logistic regression (OR) for composite group of variables: pregnancy complication (pre-eclampsia, premature rupture of membranes or duplex), instrumental delivery (caesarean section, forceps delivery or vacuum extraction), perinatal complication (born in gestational week 22–32, SGA, fetal distress, low Apgar score or cephalhaematoma) and stillbirth or perinatal death among women hospitalised for anorexia nervosa in comparison with the general population
| ||Crude OR||Model 1*||Model 2**||Model 3***|
|Instrumental delivery||0.7 (0.6–0.8)||0.7 (0.6–0.8)||0.7 (0.6–0.8)||0.7 (0.6–0.8)|
|Perinatal complication||0.9 (0.7–1.1)||0.9 (0.7–1.2)||1.0 (0.8–1.2)||1.0 (0.7–1.2)|
|Pregnancy complication||0.9 (0.7–1.1)||0.9 (0.7–1.3)||0.9 (0.7–1.3)||1.0 (0.7–1.4)|
|Perinatal death or stillbirth||0.7 (0.2–2.7)||0.7 (0.2–2.9)||0.7 (0.2–2.9)||0.9 (0.2–3.4)|
To account for variations in severity of the anorexia disorder, the births of the 99 women with atleast 6 months in hospital care for anorexia were analysed separately. The mean birthweight in this group was 3406 g and perinatal complications such as SGA, low Apgar score and instrumental delivery were not more common than in other births of women with a history of anorexia nervosa. When the birth outcomes of the 58 mothers who gave birth within 3 years of the first hospital discharge with anorexia nervosa were analysed separately, a tendency towards a lower mean birthweight was observed, 3335 g, but perinatal complications were as rare as in other births of mothers with a history of anorexia nervosa.
- Top of page
In this Swedish national register study, we investigated pregnancy complications and birth outcomes in 1000 primiparous women with a history of hospital care due to anorexia nervosa in comparison with the general population of women with their first delivery during the years in focus. There were few differences between mothers with anorexia nervosa and the general population during pregnancy (with the exception that mothers with anorexia nervosa, not surprisingly, more often were underweight and less often overweight). In terms of obstetric complications, the only significant difference concerned was with instrumental delivery (valid also for caesarean section and vacuum extraction separately), which was less common in women with anorexia nervosa. Mean birthweight was 43 g lower in the children of women with anorexia nervosa, but otherwise there were no differences between the two groups concerning birth outcomes. A longer hospital care for anorexia nervosa (≥6 months), indicating a more severe disease, was not related to negative birth outcomes. A shorter recovery from anorexia nervosa (birth within 3 years after first hospitalisation) was only related to a tendency for lower birthweight.
Low birthweights have been reported previously in children born to mothers with anorexia nervosa, but the differences in comparisons with other children have usually been more pronounced than in our study. Bulik et al.,3 in a clinical follow-up sample of 66 anorexia nervosa cases, reported a difference of 198 g; Sollid et al.,7 in a register study of 302 mothers with eating disorders, reported a difference of 137 g (2004) and Kouba et al.,12 in a maternity care sample of woman with anorexia nervosa, reported a difference of 306 g (2005), compared with a difference of 43 g in our study. It should be noted that there was no difference between children born to mothers with anorexia nervosa and controls concerning SGA, which is a clinically more relevant parameter. This is contrary to recent findings from the Danish population of mothers with eating disorders, referred to above,7 but in line with the lack of increased risk for delivering a low birthweight infant reported in a sample of women with previous eating disorder in an Australian study.13 Although statistically significant, we believe that the slightly lowered mean weight of our study group is of minor or no clinical importance.
The results of this study, using a study group, which, to our knowledge, is the largest ever presented in research about anorexia nervosa and birth outcomes, contradict most previous studies that have reported increased risks for negative outcomes. Our findings may be a result of gradual improvement of the care process, involving both anorexia nervosa (psychiatric as well as medical) and maternity care and are in line with the improved health and social outcomes in women with anorexia nervosa in Sweden in recent decades reported by our group (F. Lindblad et al., unpubl. obs.). Previously, close monitoring of the pregnant woman with a history of anorexia nervosa or other eating disorder has been recommended by several authors.3,4,12 Such an approach is challenged by our findings and may supplement the usually more risk-oriented reports about anorexia nervosa mothers/child relations.14
There are potential weaknesses in the study. The use of data from anorexia nervosa hospital care obviously excludes women having received only outpatient care and individuals who have not even been identified as patients, since they have chosen not to seek professional help. This implies that our sample is most probably biased towards more severe cases. This presumption is supported by earlier studies from our group demonstrating that most fatal cases (possibly excluding cases due to starvation over decades) have received previous hospital care (F. Lindblad et al., unpubl. obs.). The more severe cases thus selected would rather imply a higher rate of delivery/perinatal complications in our study group than in a Swedish average anorexia nervosa comparison group. On the other hand, one may argue that individuals receiving no professional care would be at highest risk by denying and neglecting their medical state, even during pregnancy. However, many previous studies on anorexia nervosa and pregnancy have been performed with study groups who have been identified through their link to previous psychiatric care, which makes the findings of these studies comparable to ours.3,5,6 Register studies always have the limitations of incorrect classification and coding. Unfortunately, the ethical rules of the database used for this study does not make it possible to trace the medical records of the women in the study, to check the quality of the diagnoses.
To summarise, the relative dominance of more severe cases in our sample may imply an overestimation of negative birth outcomes when compared with the whole Swedish population of anorexia nervosa mothers of the same ages. On the other hand, the high quality Swedish medical monitoring of pregnant women may explain lower figures of complications than in similar samples from other countries. When comparing our results with previous studies, we also believe that specific selection factors in these studies may have contributed to some of the differences.
We had no data about the medical state of the women with previous anorexia nervosa just before or during pregnancy. Earlier studies seem to indicate that an active eating disorder, including anorexia nervosa, during pregnancy implies a worse outcome, but most of the results have not reached statistical significance, probably due to low numbers of investigated cases.3–5 It seems most plausible that an active stage of the disease would increase the risks, but possibly, the fertility problems related to anorexia nervosa (the absence of at least three consecutive menstrual cycles being one of the criteria in the Diagnostic and Statistical Manual of Mental Disorders) means that pregnancy will only occur in less severe cases with less risk for negative outcome. It is, however, important to emphasise that a history of anorexia nervosa could be related to a risk of never recovering fertility.
Strengths of this study are the use of a national cohort and the size of the study group, allowing for deepened statistical analyses. Moreover, only primiparas have been included with the purpose of facilitating comparisons with other groups.
We conclude that in a country with a satisfactory maternity surveillance, outcome of pregnancy and delivery may be just as good for women with a history of anorexia nervosa as for the general population without previous anorexia nervosa. Specialised maternity care may be called for only when there are current signs of anorexia nervosa, when the disease has not led to any earlier medical contacts and when there are signs of other kinds of psychopathology. In the future, investigation of birth outcomes in individuals with anorexia nervosa not previously diagnosed for anorexia nervosa and those treated as outpatients would complement our findings.