Perinatal outcomes and gestational weight gain in women with eating disorders: a population-based cohort study


  • N Micali,

    1.  Behavioural and Brain Sciences Unit, UCL Institute of Child Health, London, UK
    Search for more papers by this author
  • B De Stavola,

    1.  Department of Medical Statistics, London School of Hygiene and Tropical Medicine, London, UK
    Search for more papers by this author
  • I dos-Santos-Silva,

    1.  Department of Non-Communicable Disease Epidemiology, London School of Hygiene and Tropical Medicine, London, UK
    Search for more papers by this author
  • J Steenweg-de Graaff,

    1.  The Generation R Study Group, Erasmus Medical Centre, Rotterdam, The Netherlands
    2.  Department of Child and Adolescent Psychiatry and Psychology, Erasmus Medical Centre – Sophia Children’s Hospital, Rotterdam, The Netherlands
    Search for more papers by this author
  • PW Jansen,

    1.  Department of Child and Adolescent Psychiatry and Psychology, Erasmus Medical Centre – Sophia Children’s Hospital, Rotterdam, The Netherlands
    Search for more papers by this author
  • VWV Jaddoe,

    1.  The Generation R Study Group, Erasmus Medical Centre, Rotterdam, The Netherlands
    2.  Department of Epidemiology, Erasmus Medical Centre, Rotterdam, The Netherlands
    3.  Department of Paediatrics, Erasmus Medical Centre – Sophia Children’s Hospital, Rotterdam, The Netherlands
    Search for more papers by this author
  • A Hofman,

    1.  Department of Paediatrics, Erasmus Medical Centre – Sophia Children’s Hospital, Rotterdam, The Netherlands
    Search for more papers by this author
  • FC Verhulst,

    1.  Department of Child and Adolescent Psychiatry and Psychology, Erasmus Medical Centre – Sophia Children’s Hospital, Rotterdam, The Netherlands
    Search for more papers by this author
  • EAP Steegers,

    1.  Department of Obstetrics and Gynaecology, Erasmus Medical Centre, Rotterdam, The Netherlands
    Search for more papers by this author
  • H Tiemeier

    1.  Department of Child and Adolescent Psychiatry and Psychology, Erasmus Medical Centre – Sophia Children’s Hospital, Rotterdam, The Netherlands
    2.  Department of Epidemiology, Erasmus Medical Centre, Rotterdam, The Netherlands
    3.  Department of Psychiatry, Erasmus Medical Centre, Rotterdam, The Netherlands
    Search for more papers by this author

N Micali, Behavioural and Brain Sciences Unit, UCL Institute of Child Health, 30 Guilford Street, WC1N 1EH, London, UK. Email


Please cite this paper as: Micali N, De Stavola B, dos-Santos-Silva I, Steenweg-de Graaff J, Jansen P, Jaddoe V, Hofman A, Verhulst F, Steegers E, Tiemeier H. Perinatal outcomes and gestational weight gain in women with eating disorders: a population-based cohort study. BJOG 2012;119:1493–1502.

Objective  To investigate adverse perinatal outcomes and gestational weight gain trajectories in women with lifetime (current/past) eating disorders (ED: anorexia nervosa [AN] and bulimia nervosa [BN]).

Design  A longitudinal population-based birth cohort.

Setting  Rotterdam, the Netherlands.

Sample  Women who enrolled prenatally, had complete information on exposure (lifetime ED), and gave birth to a live singleton (n = 5256). Four groups of exposed women: lifetime AN (n = 129), lifetime BN (n = 209), lifetime AN + BN (n = 100), other lifetime psychiatric disorder (n = 1002) were compared with unexposed women (n = 3816).

Methods  Perinatal outcomes and gestational weight gain were obtained from obstetric and midwifery records, self-report and objective measurements. Exposed women were compared with unexposed women within the cohort using linear, logistic regression and mixed models.

Main outcome measures  Any pregnancy, delivery and postnatal complications. Birthweight adjusted for gestational age, prematurity (born <37 weeks), small-for-gestational age; maternal weight gain during pregnancy.

Results  Maternal AN was positively associated with suspected fetal distress. No differences were found in mean birthweight, prevalence of a small-for-gestational-age, or premature birth. Relative to unexposed women, women with AN had, on average, a lower body weight but a higher rate of weight gain subsequently; whereas women with BN had a higher body weight but a lower rate of weight gain.

Conclusions  Maternal lifetime ED is associated with few adverse perinatal outcomes in this sample. Differential gestational weight gain patterns in women with AN and BN are consistent with possible biological compensatory mechanisms aimed at protecting the fetus.


The literature on the effects of maternal eating disorders (ED) on pregnancy, perinatal complications and gestational weight gain is still limited.

Overall early smaller clinical studies suggested an increased risk of perinatal complications and delivery complications (gestational hypertension, hyperemesis, gestational diabetes, anaemia, caesarean section and instrumental deliveries) and prematurity in women with ED.1–7 Recent large studies in contrast have shown by and large a comparable prevalence of adverse perinatal outcomes in exposed and unexposed women.1,8,9 More evidence is available on the effect of maternal ED on birthweight: with a largely positive association between maternal anorexia nervosa (AN) (both active and past) and lower offspring mean birthweight compared with women with no history of ED, albeit with marked between-study variability in the magnitude of the estimate, perhaps because of differences in sample size and, hence, power.5,8–10 Inconsistencies might reflect the different populations studied, they might also reflect changed risk over time.

There is still uncertainty about biological mechanisms that might explain the increased risk for adverse perinatal outcomes in women with ED. The most plausible ones include poor weight gain in pregnancy, as well as low preconception body mass index (BMI).11 There is good evidence that low maternal BMI and poor gestational weight gain are associated with low birthweight and prematurity in the general population.12–14 We showed in a UK population-based study (the Avon Longitudinal Study of Parents and Children—ALSPAC) that low BMI preconception mediated, to a large extent, the risk of women with lifetime AN having lower birthweight babies.10 Gestational weight gain is particularly important because it might be partly responsible for the adverse obstetric outcomes seen in women with ED. Only one study to date has investigated gestational weight gain in a population-based sample finding that women with bulimia nervosa (BN) gained more weight compared with women without ED in the first and second trimesters of pregnancy.15 Gestational weight gain is a modifiable risk factor if adequately targeted during pregnancy.

In short, the available evidence suggests that both active and past maternal AN may be associated with a decreased birthweight and that maternal ED might be associated with pregnancy and postnatal complications. Methodological limitations of previous studies, i.e. small and unrepresentative clinical studies, small number of cases identified, low representation of ethnic minorities, highlight the need for replication and extension of findings. Replication is important given that: 1. heterogeneity of confounding structure for different exposures might be at play, therefore analysing samples with a different confounding structure allows unconfounded estimation of effect sizes; 2. many previous studies were carried out 10 or 20 years ago, analyses or recent data allow exploration of consistency over time as exposure patterns may have changed.

No previous study has investigated trajectories of gestational weight gain in women with lifetime ED.

The aim of this study in a large multi-ethnic population-based cohort was to determine if women with lifetime AN or BN were at higher risk of perinatal complications, and had different patterns of gestational weight gain compared with non-exposed women (and women with psychiatric disorders other than ED, to determine the specificity of effect of maternal ED on perinatal outcomes), to replicate existing findings in a sample with a different confounding structure compared with previous samples and extend them. We also sought to understand whether maternal underweight in women with ED was associated with low birthweight in the offspring.


Study design

Generation R is a prospective general population cohort study based in Rotterdam, the Netherlands at the Erasmus Medical Centre. Generation R was designed to identify early life risk factors for health and determinants of prenatal and postnatal growth in a multi-ethnic sample.16

Study population

All pregnant women living in the Rotterdam area were eligible for enrolment if they had a delivery date between April 2002 and January 2006. The study aimed to enrol women at the early stages of pregnancy (i.e. before 18 weeks of gestation); but enrolment was possible until the birth of the child. Estimated participation rate was approximately 61% of all liveborn children and parents living in the area at the time of recruitment. Characteristics of the overall sample and details about recruitment are given elsewhere.16

In total, 8880 women were recruited in pregnancy. Eligible for the present study were mothers enrolled during pregnancy, who gave birth to a live baby (104 had a fetal death) and gave consent for postnatal participation (6493 women, 74.6%).

Data on the exposure of interest (lifetime psychiatric disorders) were missing for 1176 women (18.1%); these women were therefore excluded from the current study, leaving 5317 women for analyses. Of these 5317 pregnancies, 61 were multiple births. Hence the final sample for this study comprised 5256 women and their babies.


Perinatal complications

These were obtained prospectively from midwifery and hospital registers and linked to the study participants (unless specified below), or via questionnaires administered to participating women.

Perinatal complications were grouped as:

  • 1 Pregnancy complications:
    •  Gestational diabetes (diagnosed according to Dutch obstetric guidelines, i.e. a random glucose level of ≥11.1 mmol/l or a fasting glucose level of >7.0 mmol/l in the absence of a pre-existing diabetes diagnosis)
    •  Pre-eclampsia, pregnancy-induced hypertension and HELLP (haemolysis, elevated liver enzymes, low platelet count) syndrome (defined according to the criteria given by the International Society for the Study of Hypertension in pregnancy17).
    •  Hospitalisation during pregnancy for >24 hours during the first two trimesters of pregnancy (obtained by questionnaire at 30 weeks of gestation)
    • Suspected fetal distress (diagnosed on the basis of a pH < 7.2 on a fetal blood sample or an abnormal cardiotocogram).18

  • 2 Delivery complications:
    •  Type of delivery; divided into five categories: spontaneous vaginal; instrumental (including forceps, vacuum extraction); elective caesarean section; emergency caesarean section; other: caesarean unknown and breech.
  • 3 Postnatal complications:
    •  Apgar score at 5 minutes: this was dichotomised into <7 or ≥7.
    •  Hospital admission: 2 months postpartum women were asked by questionnaire whether the baby had been hospitalised in the first week after delivery.

Birth outcomes

These included: birthweight obtained from obstetric records (recorded in grams); small-for-gestational-age (SGA) determined using z-scores of birthweight adjusted for gestational age (a z-score <−2 SD was considered indicative of SGA); premature birth: defined as <37 weeks of gestation. Gestational age was obtained by fetal ultrasound examination at the first face-to-face examination. All pregnancies were dated using pregnancy dating curves that were derived using women for whom ultrasound examinations at <25 weeks of gestation and reliable information on last menstrual period were available.19

Gestational weight gain

Objective weight and height were obtained from face-to-face assessments at three time-points in pregnancy: early, mid- and late pregnancy. The medians (95% range) of gestational age for the pregnancy measurements were 12.6 (9.6–16.9) weeks, 20.4 (18.6–22.5) weeks and 30.2 (28.5–32.5) weeks, respectively. Women were measured lightly clothed without shoes to obtain weight (kg) and height (cm).


Exposure was determined using data from a pregnancy questionnaire completed by the women at about 20 weeks of gestation. Women were asked whether they had suffered (ever and in the previous year) from a list of psychiatric disorders. For each specific disorder, a vignette was given clarifying with an example what was meant (available on request). All women were asked about having suffered from either AN or BN (ever and in the previous year): 112 women (2.1%) reported having ever suffered from AN, and 17 (0.3%) in the last year; 160 (3.0%) reported having ever suffered from BN and 49 (0.9%) in the last year; 93 (1.8%) reported having suffered from both ED (AN + BN) ever and seven (0.1%) in the last year. As there is evidence that this latter group might have a higher degree of ED severity, it was left as a separate exposure category.10 Given the relatively small number of women who reported an ED in the previous year, women with singleton births were grouped according to history of ED: lifetime AN (= 129); lifetime BN (= 209); lifetime AN + BN (= 100). A Dutch sub-sample drawn from the overall Generation R sample (= 928 women) was administered the Composite International Diagnostic Interview (CIDI)20 to diagnose lifetime and current mental health disorders. Although a small number of women received a lifetime diagnosis of ED (8, 0.9% of AN; 17, 1.8% of BN), self-reported lifetime AN had a sensitivity of 100% and specificity of 96%; self-reported BN had a sensitivity of 94% and specificity of 81%.

We also identified women who, in the same questionnaire, reported having ever suffered from at least one of these disorders: depression, anxiety, psychosis and manic episodes but no ED. This group ‘other psychiatric disorders’ comprised 1002 women (19.1%).

The remaining women formed the unexposed group (= 3816, 72.6%).


Information on women’s age, educational level, ethnicity, prepregnancy weight and height, and parity was obtained by questionnaire at enrolment. Educational level, defined as the highest schooling level attained, was divided into three categories: no education or primary only, secondary education, university degree or higher. Ethnicity was categorised as: Caucasian (to include Dutch or Western origin), non-Caucasian (which included Indonesian, Asian, Afro-Caribbean [Dutch Antilles, Suriname, African, Cape Verdian], Turkish, Middle Eastern and others). Parity was dichotomised as: primiparae versus multiparae. Family income was obtained at 30 weeks of gestation by questionnaire; and defined as net household monthly income categorised as: <1200 € (below social security level); 1200–2000 € and >2000 € (above national modal income).

Prepregnancy BMI was derived from self-reported weight (in kg) and height (in m) from the enrolment questionnaire (correlation between self-reported weight prepregnancy and weight measured at enrolment was 0.97; P < 0.01).

Information about smoking habits during the index pregnancy was obtained by postal questionnaires administered in the first, second and third trimester of pregnancy. This was collapsed to generate a summary variable categorised as: no smoker, ex-smoker (quit when woman knew about the pregnancy), continued to smoke during pregnancy.

Alcohol use was obtained by postal questionnaires in pregnancy as for smoking21 and the information was collapsed to generate a summary variable categorised as: no alcohol use, woman quit when she knew about the pregnancy, continued to drink during pregnancy.

Statistical analyses

Associations between history of ED and covariates were assessed using the chi-square test or the F test depending on the nature of the variable.

Normality assumptions were checked for all continuous outcome variables and, if these were met (possibly after appropriate transformation), mean and standard deviations (SD) were used for group comparisons, with linear regression models. Binary and categorical outcomes were analysed using logistic regression. All effects were also controlled for potential confounders, namely maternal age, parity, ethnicity, maternal education, smoking and alcohol use in pregnancy.

Reported results refer to women with complete data on the outcomes and the relevant confounders. Because of concerns regarding the pattern of missing outcome data (see Supplementary material, Appendix S1), the reported adjusted results include variables predictive of ‘missingness’ to deal with possible selection biases (see Supplementary information for details on attrition).

Longitudinal data analyses

Multilevel modelling of hierarchical data was used for the analysis of gestational weight gain during pregnancy while accounting for the correlations among the repeated observations for the same woman.22 The main purpose was to study whether weight gain trajectories over time—modelled in terms of initial size and rate of change between observations—were influenced by the exposure. Results obtained by additionally adjusting for maternal age, parity and smoking in pregnancy were compared with the unadjusted ones to assess confounding. Estimation was by restricted maximum likelihood and significance testing by Wald and likelihood ratio testing as appropriate.22

All statistical tests presented are two-sided, with a P < 0.05 used to define significance. All analyses were performed using STATA 11 (Stata Corp., Texas, USA).23


Sociodemographic characteristics and covariates

The distribution of sociodemographic variables and covariates across exposure groups is shown in Table 1.

Table 1.   Characteristics of the study population, by exposure groups
Maternal variablesAN lifetime (n = 129)BN lifetime (n = 209)AN + BN lifetime (n = 100)Other psychiatric (n = 1002)Unexposed (n = 3816)Statistic
Maternal ethnicity (%)
Caucasian (Dutch, European, of European origin)95 (73.6)147 (70.3)69 (69.0)669 (66.8)2433 (63.8)χ2 = 9.1
P = 0.06
Non-Caucasian (Indonesian, Asian, Dutch Antilles, Suriname, African, Cape Verdian, Turkish, other)33 (25.6)56 (26.8)30 (30.0)311 (31.0)1265 (33.1)
Missing1 (0.8)6 (2.8)1 (1.0)22 (2.2)118 (3.1)
Marital status (%)
Married50 (38.8)77 (36.8)44 (44.0)436 (43.5)1,872 (49.1)χ2 = 30.6
P < 0.0001
Unmarried77 (59.7)125 (59.8)52 (52.0)531 (53.0)1,750 (45.9)
Missing2 (1.5)7 (3.3)4 (4.0)35 (3.5)194 (5.1)
Maternal education (%)
None or primary only9 (7.0)7 (3.3)9 (9.0)58 (5.8)284 (7.4)χ2 = 14.6
P = 0.06
Secondary61 (47.3)88 (42.1)39 (39.0)432 (43.1)1,457 (38.6)
Higher57 (44.2)109 (52.1)50 (50.0)484 (48.3)1,889 (49.5)
Missing2 (1.5)5 (2.4)2 (2.0)28 (2.8)168 (4.4)
Net household income (%)
<1200 €/month16 (12.4)26 (12.4)14 (14.0)151 (15.1)507 (13.3)χ = 19.6
P = 0.01
1200–2000 €/month28 (21.7)52 (24.9)22 (22.0)249 (24.8)730 (19.1)
>2000 €/month79 (61.2)111 (53.1)50 (50.0)508 (50.7)2,107 (55.2)
Missing6 (4.6)20 (9.6)14 (14.0)94 (9.4)472 (12.4)
Parity (%)
Primiparae81 (62.8)125 (59.8)60 (60.0)611 (61.0)2,204 (57.8)χ2 = 4.2
P = 0.4
Multiparae48 (37.2)82 (39.2)40 (40.0)389 (38.8)1,593 (41.8)
Missing02 (1.0)02 (0.2)19 (0.5)
Smoking in pregnancy (%)
Never (no smoker)97 (75.2)146 (69.9)70 (70.0)708 (70.7)3,044 (79.8)χ2 = 65.2
P < 0.0001
Quit when knew about pregnancy (ex-smoker)9 (7.0)26 (12.4)8 (8.0)77 (7.7)285 (7.5)
Continued during pregnancy23 (17.8)37 (17.7)22 (22.0)217 (21.7)486 (12.7)
Missing00001 (0.03%)
Alcohol use in pregnancy (%)
Never53 (41.1)77 (36.8)37 (37.0)346 (34.5)1,469 (38.5)χ = 5.7
P = 0.7
Quit when knew about pregnancy17 (13.1)26 (12.4)11 (11.0)126 (12.6)466 (12.2)
Continued during pregnancy54 (41.9)93 (44.5)46 (46.0)465 (46.4)1,660 (43.5)
Missing5 (3.9)13 (6.2)6 (6.0)65 (6.5)221 (5.8)
Maternal age, Mean (SD)30.6 (5.0)30.5 (5.2)30.8 (5.1)31.1 (4.9)30.4 (4.9) F = 3.5
P = 0.007
Prepregnancy body mass index, mean (SD) 22.2 (3.5)24.4(5.4)22.5 (3.8)23.5 (4.3)23.4 (4.0) F = 6.1
P = 0.0001

Differences among groups were highlighted in relation to marital status, household income, smoking in pregnancy, maternal age at enrolment and, as expected, BMI prepregnancy (see Table 1).

Perinatal complications

Pregnancy complications overall occurred in similar percentages in women with ED and unexposed women (∼ 15%; see Table 2).

Table 2.   Pregnancy, delivery and obstetric complications by exposure status: numbers, percentages, crude and adjusted odds ratios (OR) and 95% confidence intervals (95%CI) from logistic regression
 AN lifetime (n = 129)BN lifetime (n = 209)AN + BN lifetime (n = 100)Other psychiatric (n = 1002)Unexposed (n = 3816)
  1. The adjusted OR are adjusted for maternal age, ethnicity, education, parity and alcohol use in pregnancy.

  2. In the logistic regression all groups are compared with the reference group (unexposed); *P ≤ 0.1; **P ≤ 0.05; ***P < 0.01

  3. ****ORs were not calculated for this variable due to small numbers

Pregnancy complications
Any pregnancy complications, n (%)21 (16.3)33 (15.8)16 (16.0)170 (16.0)530 (14.1)
 Crude OR (95% CI)1.2 (0.7–1.9)1.1 (0.8–1.7)1.1 (0.7–2.0)1.2 (0.9–1.4)1
 Adjusted OR (95% CI)1.1 (0.7–1.9)1.2 (0.8–1.7)1.1 (0.7–2.0)1.1 (0.9–1.3)1
Gestational diabetes,****n (%)02 (1)06 (0.6)32 (0.8)
Pre-eclampsia, pregnancy-induced hypertension, HELLP syndrome, n (%)8 (6.0)15 (7.0)4 (3.9)75 (7.0)221 (5.8)
 Crude OR (95% CI)1.0 (0.5–2.1)1.2 (0.7–2.1)0.7 (0.2–1.8)1.2 (0.9–1.6)1
 Adjusted OR (95% CI)0.9 (0.5–2.0)1.2 (0.7–2.1)0.6 (0.2–1.7)1.2 (0.9–1.5)1
Hospitalisation during pregnancy, n (%)2 (1.5)6 (2.8)4 (3.9)21 (2.0)70 (1.8)
 Crude OR (95% CI)0.8 (0.2–3.2)1.5 (0.6–3.6)2.2* (0.8–6.2)1.1 (0.6–1.7)1
 Adjusted OR (95% CI)0.8 (0.2–3.4)1.9 (0.8–4.5)2.7* (0.9–7.6)1.0 (0.6–1.7)1
Suspected fetal distress, n (%)15 (11.6)14 (6.7)9 (9.0)90 (8.5)273 (7.3)
 Crude OR (95% CI)1.7** (1.0–3.0)0.9 (0.5–1.6)1.3 (0.6–2.6)1.2 (0.9–1.5)1
 Adjusted OR (95% CI)1.8** (1.0–3.1)0.8 (0.4–1.5)1.2 (0.6–2.5)1.1 (0.9–1.5)1
Type of delivery
Spontaneous, n (%)92 (71.3)137 (65.5)67 (67.0)675 (63.7)2526 (67.2)
Instrumental, n (%)17 (13.2)20 (9.6)14 (14.0)158 (14.9)509 (13.5)
 Crude OR (95% CI)0.9 (0.5–1.6)0.7 (0.4–1.2)1.0 (0.6–1.9)1.2 (0.9–1.5)1
 Adjusted OR (95% CI)0.8 (0.4–1.4)0.6 (0.4–1.1)1.0 (0.5–1.8)1.2 (0.9–1.5)1
Elective caesarean, n (%)2 (1.5)10 (4.8)2 (2.0)50 (4.7)175 (4.7)
 Crude OR (95% CI)0.3 (0.1–1.3)1.1 (0.5–2.0)0.4 (0.1–1.8)1.1 (0.7–1.5)1
 Adjusted OR (95% CI)0.3* (0.1–1.2)1.0 (0.5–2.0)0.4 (0.1–1.4)1.0 (0.7–1.4)1
Emergency caesarean, n (%)7 (5.4)21 (10.0)4 (4.0)91 (8.6)246 (6.5)
 Crude OR (95% CI)0.8 (0.4–1.7)1.6* (1.0–2.5)0.6 (0.2–1.7)1.4** (1.0–1.8)1
 Adjusted OR (95% CI)0.7 (0.3–1.6)1.4 (0.8–2.3)0.6 (0.2–1.7)1.3*** (1.0–1.6)1
Caesarean (unknown), Breech, Other,****n (%)1 (0.8)002 (0.2)10 (0.3)
Postnatal complications
Any postnatal complications, n (%)14 (10.8)26 (12.4)9 (9)151 (14.2)498 (13.2)
 Crude OR (95% CI)0.8 (0.4–1.4)0.9 (0.6–1.4)0.6 (0.3–1.3)1.1 (0.9–1.3)1
 Adjusted OR (95% CI)0.7 (0.4–1.3)0.9 (0.6–1.3)0.6 (0.3–1.2)1.0 (0.8–1.2)1
Apgar score at 5 minutes (<7), n (%)2 (1.6)2 (1)08 (0.8)36 (0.9)
 Crude OR (95% CI)1.7 (0.4–7.0)1.0 (0.2–4.3)0.8 (0.4–1.8)1
 Adjusted OR (95% CI)1.7 (0.3–7.1)0.1 (0.2–4.1)0.7 (0.3–1.6)1
Hospital admission of the baby, n (%)14 (10.8)25 (12.0)9 (9)150 (14.1)484 (12.9)
 Crude OR (95% CI)0.8 (0.4–1.4)0.9 (0.6–1.4)0.7 (0.3–1.3)1.1 (0.9–1.4)1
 Adjusted OR (95% CI)0.8 (0.4–1.4)0.9 (0.6–1.4)0.6 (0.3–1.3)1.1 (0.9–1.3)1

Gestational diabetes was rare across the whole sample (0.8%). Pre-eclampsia, pregnancy-induced hypertension and HELLP syndrome, as well as hospitalisation during pregnancy, occurred in similar percentages in exposed and unexposed women (Table 2). There was borderline evidence that women with AN had a 80% higher odds of suspected fetal distress relative to unexposed women (OR 1.8; 95% CI 1.0–3.1; P ≤ 0.05 after adjustment for confounders), but not for the other exposure groups.

Women with lifetime AN + BN had increased odds of being hospitalised during pregnancy (OR 2.7; 95% CI 0.9–7.6; P = 0.06) in unadjusted and adjusted analyses; with a trend towards statistical significance.

Percentages of spontaneous delivery and postnatal complications were comparable across exposure groups before and after adjustment for confounders (see Table 2).

Gestational weight gain

As the distribution of the weight variables was positively skewed, these variables were transformed using a natural base logarithm (ln) in all analyses. Figure 1 shows the observed mean pregnancy weights across the three prenatal visits (in the first, second and third trimesters). This plot suggests a quadratic effect of time on weight so both a linear and quadratic effect of time on log-weight, both a linear and a quadratic term for gestational time (in weeks) were included in the mixed effects model. The model including random coefficients for the intercept and the linear and quadratic term of time was found to best fit the data (likelihood ratio test comparing the more complex model with the chosen one: P < 0.0001).

Figure 1.

 Mean gestational weight trajectories across exposure groups.

The results suggest a strong effect of gestational time (regression coefficient of the linear term = 0.005 (95% CI 0.004–0.006), and for the quadratic term = 0.00004 (95% CI 0.00003–0.00005), indicating an average increase of more than 0.5% in weight per week.

Maternal AN was associated with a lower weight compared with unexposed women (regression coefficient: −0.05; 95% CI −0.08 to −0.1), corresponding to a 5% lower weight on average. However, there was a positive interaction between maternal lifetime AN and the linear term for gestational time (regression coefficient: 0.0007; 95% CI 0.0002–0.001), corresponding to a 0.07% higher relative increase per week compared with unexposed women. In contrast women with lifetime BN had a higher weight (regression coefficient: 0.05; 95% CI 0.03–0.08); but a slower increase over time relative to unexposed women (regression coefficient: −0.0005; 95% CI −0.001 to −0.0001), corresponding to a 0.05% smaller weight gain per week compared with unexposed. These differences persisted after adjustment for confounders (Table 3).

Table 3.   Mixed effects models of gestational weight gain
 Natural base log-transformed gestational weight
Unadjusted b coefficients (95% CI)Adjusted†b coefficients (95% CI)
  1. SE, standard error; 95% CI, 95% confidence intervals.

  2. *P ≤ 0.05;**P ≤ 0.01; ***P ≤ 0.0001 for comparisons between exposed groups and unexposed.

  3. †Adjusted for maternal age, education, parity and ethnicity and smoking in pregnancy.

AN lifetime (n = 129)−0.05** (−0.08 to −0.1)−0.05** (−0.09 to −0.02)
BN lifetime (n = 209)0.05*** (0.03–0.08)0.05*** (0.02–0.08)
AN + BN lifetime (n = 100)−0.02 (−0.06 to 0.02)−0.03 (−0.06 to 0.01)
Other psychiatric lifetime (n = 1002)0.008 (−0.1 to0.02)0.006 (−0.007 to 0.02)
Unexposed (n = 3816)Ref.Ref.
Linear term for time (in gestational weeks)0.005*** (0.004–0.006)0.005*** (0.004–0.006)
Quadratic term for time0.00004*** (0.00003–00005)0.00004*** (0.00003–00005)
Interaction terms  
AN lifetime × time0.0007** (0.0002–0.001)0.0007** (0.0001–0.001)
BN lifetime × time−0.0005* (−0.001 to −0.0001)−0.0005* (−0.001 to −0.0001)
AN + BN lifetime × time−0.0001 (−0.0008 to 0.0005)0.00003 (−0.0006 to 0.0006)
Other psychiatric × time−0.0001 (−0.0003 to 0,0001)−0.00006 (−0.0003 to 0,0001)
Random effects Estimate (SE) Estimate (SE)
Standard deviation of Random slope0.002 (0.00004)0.002 (0.00004)
Standard deviation of Random intercept0.20 (0.002)0.19 (0.002)
Level 2 correlation coefficients−0.63 (0.01)−0.63 (0.01)
Level 1 residual standard deviation0.02 (0.0002)0.02 (0.0002)

In relation to weight gain until the last measurement, women with lifetime AN gained on average 9.1 kg (SD 3.2); those with lifetime BN 8.2 kg (SD 4.0), those with AN + BN 8.2 kg (SD 3.8); unexposed gained 8.3 kg (SD 3.6). There was evidence that women with lifetime AN gained more weight overall compared with unexposed (even after adjustment for maternal age, parity, education and ethnicity; b coefficient: 0.8; 95% CI 0.1–1.5; P = 0.02), the other groups had comparable weight gain to unexposed (data not shown).

Birthweight, SGA and prematurity

Mean birthweight was similar across exposure groups (∼ 3400 g) (see Supplementary material, Table S1) and comparable in babies of women with lifetime ED (and babies of women with other psychiatric disorders) and babies of unexposed women.

The prevalence of SGA was low in all exposure groups (between 0.8% and 4%; see Supplementary material, Table S1). There was no evidence that women with lifetime ED (or lifetime other psychiatric disorders) had increased odds of having SGA babies relative to unexposed women.

The frequency of premature births was comparable across exposure groups, except for a slightly higher percentage in women with AN + BN (7.8% versus 4.3% in unexposed women). There was no statistical evidence supporting either decreased or increased odds of prematurity for any of the exposure groups.

In an exploratory analysis the birthweight of offspring of women with ED underweight prepregnancy (n = 19) was compared with that of babies of normal-weight ED women (n = 269). The mean birthweight of babies born to underweight women with ED was 3275 g (SD 382.8); whereas that of babies born to normal-weight women with ED was 3488 g (SD 566.0). Formal comparisons showed a trend towards a lower birthweight in underweight ED women after adjusting for gestational age, gender of the baby, maternal age, education, ethnicity and parity (regression coefficient −156.6; 95% CI −354.3 to 41.0; P = 0.1).


Overall there was little evidence in this general population sample that maternal lifetime ED was associated with increased risk of perinatal, delivery and postnatal complications.

Few differences in pregnancy outcomes were highlighted in women with lifetime ED compared with unexposed: increased odds of suspected fetal distress in women with lifetime AN; and a trend for a two-fold increase in the odds of being hospitalised in pregnancy for women with lifetime AN + BN.

Women with ED had babies of comparable birthweight to unexposed women.

Evidence of a differential effect of exposure was found with regards to gestational weight gain patterns (Table 3). Women with lifetime AN had a significantly lower bodyweight and gained more during gestation compared with unexposed women. Women with lifetime BN however had a significantly higher body weight, but gained less across time compared with unexposed women.

Previous research on perinatal complications in relation to maternal ED has been inconclusive, with small clinical studies reporting increased risk of perinatal complications and larger general population studies finding no associations. The present study confirms findings from these larger studies in relation to perinatal complications and prematurity. The inconsistencies between clinical and large population-based studies is likely to be caused by three main reasons: (i) smaller clinical studies often sampled women who were likely to be actively ill with ED, therefore possibly more at risk of complications (although also more prone to bias); (ii) to be included in a cohort of pregnant women, those with ED have to be well enough to conceive; and (iii) the low prevalence of some perinatal complications means that even large population-based studies have limited power to detect differences between exposure groups.

Collective evidence from studies investigating birthweight in relation to maternal AN suggests an association with decreased birthweight.5,8–10 We failed to find an association between birthweight (or SGA) and maternal AN in the current study. Women from Generation R with lifetime AN had a mean prepregnancy BMI within the normal range (at 22.2 kg/m2); albeit lower than unexposed women; this might partly explain differences with our ALSPAC findings (in which women with lifetime AN had a mean prepregnancy BMI of 21.3 kg/m2).10 This might be a result of women with lifetime AN in Generation R being long-term weight-recovered or having achieved higher weight gain before planning a pregnancy. It is also possible that antenatal care for women with mental health problems in general is more adequate in the Netherlands than in the UK or that recognition of mental health disorders by professionals has improved in the 2000s compared with the 1990s, with subsequent less impact on childbirth outcomes. Gestational weight gain patterns, which suggest a protective pattern of more weight gain for AN women, might also explain the lack of differences in offspring birthweight.

Mean birthweight was lower in underweight ED women compared with normal-weight women with ED in this sample; however because of sample size and wide confidence intervals (crossing 0, which indicates no effect) we cannot be definitive about the effect of maternal underweight in women with lifetime ED as a risk for lower birthweight in the offspring in this sample.

There is a scarcity of evidence on gestational weight gain in women with ED. Earlier clinical studies did not find differences in gestational weight gain in women with ED and controls.1,4 The only large study to date to investigate gestational weight gain found that women with BN gained more weight at each trimester compared with unexposed women; they were also more likely to gain weight ‘excessively’ (according to the Institute of Medicine definition) compared with unexposed women.15 This contrasts with findings presented here that women with lifetime BN have a higher weight during pregnancy but gain less weight over time compared with unexposed women. Discrepancies might be a result of differences in case definition (lifetime disorder versus disorder in the 6 months before the pregnancy) and a less accurate weight measurement (self-report in the Siega-Riz study versus objective weight in this study).15

In the Siega-Riz study15 women with AN had a lower risk of gaining an inadequate amount of weight compared with non-ED women; our study confirms and extends this finding using more accurate weight data (based on objective weight measurements in each trimester) and focusing on trajectories rather than absolute weight gain. This is the first study to model gestational weight gain trajectories over pregnancy in women with lifetime ED. Future work should aim to determine what might cause higher gestational weight gain in women with lifetime ED, given the different components of gestational weight gain, i.e. fat and intravascular and extravascular fluids.

This study has several strengths but also some limitations. Its main strength is the use of a population-based, multi-ethnic cohort of pregnant women. The exposures under investigation are not common, therefore a cohort study is best suited to investigate their effect. Although the available numbers (especially for the smallest exposure group) precluded accurate examination of rare outcomes; the study was powered to detect a doubling in the prevalence of common outcomes (≥10%) in exposed versus unexposed women. Generation R is well suited to extend previous findings in the field, given that most previous studies relied on largely white samples, mostly of high socio-economic status.

Another important strength is the availability of objectively measured outcomes (pregnancy maternal weight and obstetric outcomes). Hence, these are unlikely to have been affected by information bias. If any misclassification of outcome has occurred this is likely to have been random (non-differential). In contrast, clinical studies have mostly relied on self-reported or retrospective ascertainment of outcomes and self-reported weight and height.

The main limitation of this study relates to ascertainment of exposure, obtained by self-report. However, validation of self-reported ED in a sub-sample of Generation R women yielded excellent sensitivity and specificity of self-report for lifetime AN and very good sensitivity and specificity for self-reported BN. Previous evidence has highlighted similar results for the diagnostic properties of self-reported ED.24 Moreover, the prevalence of self-reported lifetime ED in this cohort is well within known prevalence figures25 and consistent with previous cohort studies of pregnant women.10 It is also possible that a proportion of women classified as having had lifetime AN or BN might have suffered from an ‘ED not otherwise specified’ or a milder ED compared with clinical samples. As such, this sample is representative of women with ED who get pregnant.

Third, another limitation of the current study is lack of power to detect differences in rare outcomes between the various exposure groups. Despite the large size, we did not have enough power to detect differences for rare/uncommon outcomes between ED women and unexposed women. In fact, weak associations between maternal ED and some perinatal outcomes might be a result of lack of power to detect such small differences.


In summary, pregnancy and postnatal complications of the offspring were broadly speaking as common in women with lifetime ED as in unexposed women, although failure to find differences might be compounded by low prevalence of specific outcomes.

Women with lifetime AN gained more weight across pregnancy; whereas women with lifetime BN gained less weight across pregnancy compared with unexposed women. These weight gain patterns might be appropriate and protective considering prepregnancy weight across these two groups. The fact that birthweights were comparable with those of babies of unexposed women corroborates this hypothesis.

At a public health level it will be important to confirm whether all women with ED gain adequate weight in pregnancy, and, if not, which preventable risk factors predict poor gestational weight gain in this group of women.

Disclosure of interest


Contribution to authorship

NM had the original idea for this study, analysed data and wrote this manuscript. HT provided supervision and advice on data available, data analyses and interpretation of results. BDS and ISS provided supervision and guidance on data analyses and interpretation of results. JSD contributed to data cleaning and preparation. HT, BDS, ISS, PWJ, VWVJ, EAPS, AH and FCV contributed to the write-up of the study and revised the article critically.

Details of ethics approval

Ethical approval for the main study was given by the Medical Ethical Committee of the Erasmus Medical Centre in Rotterdam (MEC 198.782/2001/31). Further ethical approval for the secondary data analyses was given by the London School of Hygiene and Tropical Medicine ethical committee. Written consent was obtained from all participating mothers.


This work is produced by Dr Nadia Micali under the terms of a Clinician Scientist Award issued by the National Institute for Health Research. The views expressed in this publication are those of the author(s) and not necessarily those of the NHS, the National Institute for Health Research or the Department of Health. The general design of the Generation R Study was made possible by financial support from the Erasmus Medical Centre, Rotterdam, the Erasmus University Rotterdam, the Dutch Ministry of Health, Welfare and Sport, and the Netherlands Organisation for Health Research and Development (ZonMw). Additional support was received by a grant from ZonMw (no.10.000.1003). The work of HT is supported by a NWO-ZonMW VIDI grant (no. 017.106.370).


We are grateful to all the women who participated in the Generation R study. This study was submitted as a dissertation for completion of a MSc in Epidemiology at the London School of Hygiene and Tropical Medicine.