21. Impact of pregnancy on bulimia nervosa, The British Journal of Psychiatry
(C) 1999 The Royal College of Psychiatrists
Volume 174(2), February 1999, pp 135-140
Impact of pregnancy on bulimia nervosa
MORGAN, JOHN FARNILL; LACEY, J. HUBERT; SEDGWICK, PHILIP M.
JOHN F. MORGAN, MRCPsych, J. HUBERT LACEY, MD, PHILIP M. SEDGWICK, BSc,
Department of General Pscyhiatry, St George's Hospital Medical School, London
Correspondence: Dr Morgan, Department of General Psychiatry, St George's
Hospital Medical School, Cranmer Terrace, London SW17 ORE; e-mail: firstname.lastname@example.org
(First received 30 October 1998, accepted 17 November 1998)
Background Bulimia nervosa affects women at a peak age of reproductive
functioning, but few studies have examined the impact of pregnancy on bulimia.
Aim To examine the impact of pregnancy symptoms of bulimia nervosa and
Method Women actively suffering from bulimia nervosa during pregnancy (n=94)
were interviewed using the eating disorder examination (12th edn) and structured
clinical interview for DSM-III-R, with additional structured questions.
Behaviours were recorded at conception, each trimester and postnatally. Relative
risks were calculated for prognostic factors.
Results Bulimic symptoms improved throughout pregnancy.After delivery, 57% had
worse symptoms than pre-pregnancy, but 34% were no longer bulimic. Relapse was
predicted by behavioural severity and persistence, previous anorexia nervosa
('Type II' bulimia), gestational diabetes and 'unplanned' pregnancy. Unplanned
pregnancies were the norm, usually resulting from mistaken beliefs about
fertility. 'Postnatal depression' was suggested in one-third of the sample, and
in two-thirds of those with 'Type II' bulimia, and was predicted by alcohol
misuse, symptom severity and persistence.
Conclusions Postnatal treatment intervention should focus on women 'at risk' of
relapse, but all women with bulimia should be assessed for postnatal depression.
Bulimia nervosa affects women at a peak age of reproductive functioning. Its
prevalence in pregnant women is unknown, but eating disorder psychopathology is
increased in the general population postnatally (Stein & Fairburn, 1996) .
Few studies have examined the impact of pregnancy on bulimia, and none has
explored post-puerperal psychiatric disorders in women with bulimia nervosa.
Lacey & Smith's (1987)  small prospective cohort study of pregnant women
with bulimia suggested that most of the women reduced bingeing during pregnancy,
a minority experienced pregnancy as 'curative', but half were worse after
delivery than before conception. Case reports are less consistent (Price et al,
1986; Stewart et al, 1987), [15,22] and other studies have not differentiated
anorexia nervosa from bulimia nervosa (Lemberg & Phillips, 1989) . The
present study aimed to examine patterns of bulimic behaviours throughout
pregnancy, and to determine predictors of relapse and post-puerperal mental
illness, by means of a large retrospective analysis based on standardised
The sample consisted of patients who had previously been assessed at the St
George's Hospital eating disorder unit for bulimia nervosa between 1988 and
1994. This unit is one of the largest treatment centres for bulimia nervosa in
the UK. Subjects were recruited retrospectively, and time between pregnancy and
study interview was recorded. All subjects were actively suffering from fully
fledge bulimia nervosa at the point of conception and fulfilled DSM-IV criteria
(American Psychiatric Association, 1994) . Subjects whose pregnancies had not
progressed to term were excluded. Data concerning all pregnancies for each
subject were obtained at assessment, but only information about first pregnancies
was entered into the study, because different pregnancies for the same
individual were not independent of each other. We identified 113 subjects who
satisfied the study's inclusion criteria.
The interview schedule
Participants were interviewed by the first author using a semi-structured
interview schedule based on behavioural components of the eating disorder
examination, 12th edition (Fairburn & Cooper, 1993),  items concerning
affective disorders in the postnatal period from the structured clinical
interview for DSM-III-R (SCID; Spitzer et al, 1990)  and additional
questions concerning demographic and social status at conception, alcohol
consumption, menstrual history, previous history of anorexia nervosa ('Type II'
bulimia in Lacey, 1984),  presence of gestational diabetes, planning of
pregnancy and decision to breast-feed. Behaviours relating to bulimia nervosa
were estimated at five of stages: at the point of conception, in each of the
three trimesters of pregnancy, and for 12 months of the postnatal period.
Severity of each behaviour was recorded in terms of weekly frequency; and also
dichotomously in terms of 'presence' or 'absence', to minimise recall bias.
Questions pertaining to eating disorder psychopathology were not used, except to
establish diagnosis, because we regard them as being more susceptible to recall
bias than behavioural measures. Corroborative information was sought from
subject's families, medical records and general practitioners, when permitted by
subjects. Descriptive comments pertaining to the experience of pregnancy were
also recorded. 'Postnatal depression' was defined as "non-psychotic major
depressive disorder with onset within four weeks post-partum", in concordance
with the DSM-IV category of "major depressive disorder-with post-partum onset".
No assessment was made concerning episodes of depression unrelated to childbearing.
Descriptive statistics were recorded for all of the above, and subjects were
classified in terms of patterns of change in bingeing behaviour. Statistical
significance was set at the 5% level. Relative risks, with their associated 95%
confidence intervals (CIs), were estimated for a series of prognostic factors
concerning "relapse into bulimia nervosa following delivery" and "postnatal
depression." Relative risks were significant at the 5% level where the 95%
confidence interval did not include unity. The Mann-Whitney test (U) was used to
compare means for variables associated with relapse, "postnatal depression" and
"history of anorexia nervosa". Associations between categorical variables were
tested using the chi-squared test.
Characteristics of the sample
Of 113 women approached, 94 (83%) agreed to participate in the structured
interview and 18 refused. One subject was excluded from the study because she
continued to suffer from a psychotic illness with onset following childbirth,
characterised by delusions relating to her baby and to food. Mean age at
conception for the sample was 25.2 years (s.d. 4.53, range 17-38), with 17
subjects over the age of 30. Mean time between pregnancy and study interview was
4.76 years (s.d. 1.78, range 2-10). Eighty-eight subjects were Caucasian, four
were British-Asian and two were British-Afro-Caribbean. Seventy-eight were
married or cohabiting at the point of conception; the remaining 16 were single
or separated from their partner. Twenty-one subjects gave a clear previous
history of anorexia nervosa.
Bulimic behaviours and sequelae at conception
At conception, mean body-mass index (BMI) was 22.1 kg/m2 (median 22.0, s.d.
1.60, range 19.3-26.4). Fifteen subjects felt unable to estimate their BMI with
any accuracy. Mean weekly number of objective 'binge episodes' was 22.2 (median
21.0, s.d. 7.80, range 12-58), with 18.5 episodes of vomiting (median 16.8, s.d.
6.23, range 10-45), and 5.1 units of alcohol use (median 4.0, s.d. 4.83, range
By definition, all subjects satisfied DSM-IV criteria for bulimia nervosa.
Bingeing 30 or more times a week took place in 5% of the sample, while 18% of
the sample fulfilled DSM-IV criteria for alcohol abuse, 18% were misusing
laxatives and 5% were misusing slimming pills or amphetamines for the purpose of
Regular menstruation occurred in 71% of the sample, and 28% suffered oligomenorrhoea.
There were no cases of amenorrhoea. Three-quarters of the pregnancies were
unplanned, of which 25% had arisen from a failure of contraception and 75%
because menstrual irregularities were falsely believed to imply infertility. The
relative risk estimate of unwanted pregnancy in the presence of oligomenorrhoea
was 1.35 (95% CI 1.11-1.64) - in other words, subjects were 1.35 times more
likely to have an unwanted pregnancy if oligomenorrhoea was present.
Complications of pregnancy
One-third of the sample developed 'postnatal depression', in concordance with
the DSM-IV category of "major depressive disorder - with postpartum onset" and
defined as non-psychotic major depressive disorder with onset within four weeks
postpartum. Two-thirds of the sample with a previous history of anorexia nervosa
suffered from postnatal depression.
Gestational diabetes during the pregnancy was reported by 17% of the total
sample. This Figure shouldbe treated with caution as a possible underestimate,
because many subjects expressed uncertainty in their responses to this question
and were therefore recorded 'negative'.
Changes in bulimic behaviours during pregnancy
Changes in mean weekly frequency of bingeing and vomiting before, during and
after gestation are shown in Figure 1.
Figure 1. Mean weekly frequency of bingeing and of vomiting in women with
bulimia during each stage of first pregnancies proceeding to term.
The percentage of subjects with any bingeing behaviour is illustrated in Figure
2, which shows that a third of the sample were no longer bingeing in the
postnatal period. A similar pattern of change is shown for laxative misuse,
alcohol intake and slimming-pill misuse.
Figure 2. Percentages of pregnant women with bulimia continuing to binge,
misusing laxatives, drinking more than 13 units of alcohol per week or misusing
slimming pills during each stage of first pregnancies proceeding to term.
Patterns of change and risk of relapse
When comparing the situation at conception with the postnatal period, four
patterns of change in bingeing behaviour were identified, namely 'worse',
'unchanged', 'improved' and 'cured'. Thirty-four per cent were 'cured' and 4%
were 'improved' by pregnancy; 58% were 'worse' and 4% 'unchanged'.
Differences in impact of pregnancy on bingeing were noted where there had been a
previous history of anorexia nervosa ('Type II' bulimia) and where patients
developed postnatal depression. Eighty-six per cent of patients with a previous
history of anorexia nervosa relapsed postnatally, compared with 62% without; 94%
developing postnatal depression relapsed into bulimia nervosa, compared to 52%
without postnatal depression.
There were differences in BMI, age and binge frequencies at conception and
postnatally for patients with a previous history of anorexia nervosa (Type II
bulimia nervosa). Those with Type II bulimia had lower BMIs (mean 21.4 v. 22.2
kg/m2, U=542.5, P=0.042) and higher age at conception (mean 27.6 v. 24.7 years,
U=475.0, P=0.008), with greater weekly binge frequency at conception (mean 27.6
v. 20.7, U=549.5, P=0.040) and postnatally (mean 33.8 v. 16.7, U=411.5,
P=0.001). They were also less likely to breast-feed than those with Type I
bulimia (chi squared1=22.8, P=0.000).
The estimated relative risks of relapsing into bulimia postnatally were
calculated for a series of prognostic factors, shown in Table 1. The Table
showsthat a range of behavioural parameters of severity are associated with an
elevated risk of bulimic relapse. Individuals were 1.53 times more likely to
relapse if they had a high frequency of bingeing at conception (> 30 episodes
per week), 1.39 times more likely with a previous history of anorexia nervosa,
1.56 times with comorbid gestational diabetes and 1.73 times with postnatal
Table 1. Estimated relative risks for factors associated with relapse into
bulimia nervosa after delivery
There were differences between subjects who subsequently relapsed into bulimia
and those who did not in BMI, binge frequency and alcohol consumption at
conception, but not age. Relapsers had lower BMIs (mean 21.6 v. 23.0 kg/m2,
U=405.0, P=0.000), greater weekly binge frequency at conception (mean 24.0 v.
18.6, U=454.0, P=0.000) and greater weekly alcohol consumption (mean 6.2 v. 2.9
units, U=564.0, P=0.001).
Risk of postnatal depression
The estimated relative risks of developing postnatal depression were calculated
for the variables shown in Table 2. Again, it can be seen that a range of
patient characteristics and behavioural parameters of severity are associated
with an increased risk of postnatal depression. Individuals were 3.42 times more
likely to develop postnatal depression if they misused alcohol at conception and
2.87 times more likely if they had a previous history of anorexia nervosa.
Table 2. Estimated relative risks for factors associated with postnatal
There was no statistical difference in age between those with and without
postnatal depression (mean 25.9 v. 25.1 years, U=858.0, P=0.339). Patients
developing postnatal depression had a lower BMI at conception (mean 21.5 v. 22.3
kg/m2, U=654.0, P=0.009), a greater mean weekly frequency of bingeing postnatally
(mean 31.9 v. 15.0, U=434.0, P=0.000) and a higher weekly alcohol intake at
conception (mean 7.7 v. 3.8 units, U=701.5, P=0.036).
A majority of subjects described their pregnancy in positive terms, of which the
following quotations are representative:
"I liked feeling close to my child. I liked it when it moved in my tummy."
"What happened to my body seemed so much more natural than anything my body had
"I think I stopped wanting to binge even before I knew I was pregnant."
"I'd never felt so close to anything or anyone before."
"I was surprised that I didn't mind getting big and fat. It didn't feel like my
fault. I thought I'd feel disgusted but I wasn't."
"I really liked the way my body was growing. I didn't mind at all."
"Being a mother was something I could handle better than being someone's
"It was comforting to breast-feed. It was nice to feel so close to my baby and
to be useful to it."
For most subjects tolerance of changes in their body was very quickly lost after
"I immediately felt like a great big fat blob again."
"There was this ugly fat body in the mirror. I couldn't believe it was me."
Subjects with a previous history of anorexia nervosa seemed to experience
pregnancy very differently:
"This thing inside me making me fat."
"The worst thing about it was that I couldn't stop what was happening to my body
without hurting the baby, which made me resent it. I really hated it sometimes."
"I tried to breast-feed but it was disgusting . . . this thing just clamped
there, sucking away. I didn't know when it had had enough. It seemed so chubby."
This study benefits from a sample size almost five times greater than its
predecessors, which has allowed a detailed and meaningful analysis of specific
risk factors and subgroups. To achieve this, a retrospective study design was
employed; it is unlikely that a prospective study could ever raise such a large
sample of pregnant women with bulimia. The possibility of recall bias is always
present in retrospective research. We minimised this risk by looking at
quantifiable behaviours rather than psychopathology, by employing dichotomous
measures, by seeking corroborative information, and by using a structured
interview rather than screening questionnaires, which have been shown to
overestimate levels of eating disorders in previous studies (Fairburn et al,
1991) . As with any comparable study in this field, social desirability bias
may lead to an underestimation of the overall prevalence of symptoms in
pregnancy, but would not explain the changes in these symptoms during different
Impact of pregnancy on bulimic symptoms
The study demonstrates that all behavioural symptoms of bulimia nervosa improve
with each passing trimester of pregnancy. Relapse occurred in a majority of this
sample, over half presenting with worse symptoms than prior to pregnancy, but a
third of our sample experienced pregnancy as curative. It is our impression that
these subjects were motivated to cease bulimic behaviours in order to avoid
harming the unborn child, and also in recognition of the responsibilities of
motherhood. Virtually all the sample ceased bingeing in the third trimester, and
this is likely to be due to the physical constraints of the enlarged uterus upon
gastric volume; but, as the majority of these individuals subsequently relapsed,
diminution of gastric volume would not explain the long-term improvements of the
minority. Although a reduction in gastric volume was sufficient to break the
bingeing-purging cycle temporarily, this was insufficient to sustain recovery.
The phenomenological descriptions of pregnancy provided by this sample suggest
an alleviation of a sense of responsibility for body-weight and shape during
pregnancy. The thin female ideal, incompatible with female biological drive, is
relinquished during pregnancy, providing a culturally endorsed license to feed,
compatible with the prevailing physiological milieu. This relaxation of
restraint appears to be abruptly curtailed following delivery, particularly in
subjects with previous anorexia nervosa. A minority of subjects described a
sense of reduced bulimic impulses before they were even aware of being pregnant.
Although this could be explained by recall bias, the possibility of an early
physiological influence on feeding behaviour in pregnancy remains, associated
with either the changes in taste and smell reported by Fairburn et al (1992),
 or changes in satiety associated with physiological parameters such as
altered leptin levels in pregnancy (Butte et al, 1997) .
By contrast with previous studies (Lacey & Smith, 1987)  we were able to
differentiate relapsers from non-relapsers, with clear clinical implications.
Postnatal treatment intervention should particularly focus on those women with
more severe symptomatology at conception, those whose symptoms persist into the
second trimester, those with a previous history of anorexia nervosa, and those
whose pregnancy was unplanned. Interestingly, gestational diabetes also appears
to predict relapse. Although studies indicate an absence of altered insulin
tolerance in bulimia nervosa (Raphael et al, 1995),  our findings suggest
that a subgroup with more severe bulimia may have an increased vulnerability to
Any retrospective diagnosis needs to be interpreted with caution, and cannot be
regarded as entirely reliable. Although not validated for retrospective use, the
SCID was chosen for its inclusion of detailed and precise questions, in an
attempt to minimise recall bias. Despite high rates of comorbidity for affective
disorders in bulimia nervosa, there have been no previous studies exploring
puerperal psychiatric disorders in women with bulimia nervosa. Childbirth is a
major life-event and raises issues of mutable gender and social roles,
attachment and separation, which have been well documented in pregnancy (Bailey
& Hailey, 1987)  and which are considered fundamental in understanding eating
disorders (Johnson, 1991) . It would seem probable that pregnancy has a major
impact on the psychological equilibrium of women with bulimia nervosa.
In the present study, a third of women with active bulimia nervosa at the point
of conception appeared to develop postnatal depression, and this rises to
two-thirds in Type II bulimia, contrasting with estimated rates in the general
population within the range 10-15%, which may be no higher than a matched
non-pregnant population (O'Hara, 1997) . This crude comparison with general
population statistics suggests a raised vulnerability to postnatal depression,
in line with the association between bulimia and depression in general. However,
in the absence of a matched control, statistical significance was not established.
In addition, because the prevalence of depression unrelated to childbearing was
not established in this sample, it is not possible to comment on the specific
risk of postnatal depression in bulimia nervosa, relative to the general risk of
affective disorders. None the less, to our knowledge, an association between
postnatal depression and bulimia nervosa has not been documented before, and
such an association has important implications in understanding parenting skills
of patients with eating disorders, in which the quality of parent-child
interactions may be altered (Stein, 1995) .
The retrospective study design and absence of a control preclude firm conclusions
on the association between bulimia nervosa and postnatal depression. However,
the apparent trend is strong enough to suggest that all women with bulimia
should be specifically assessed for postnatal depression following delivery,
whether or not they are in treatment for bulimia nervosa. Particular attention
should be given to women with bulimic symptoms of greater severity at conception,
for example bingeing more than 30 times per week, and greater persistence during
the first two trimesters. Alcohol abuse before, during and after pregnancy is an
important associate of postnatal depression. For example, the risk of developing
postnatal depression appears to be increased more than threefold by the presence
of alcohol abuse at the point of conception.
Previous history of anorexia nervosa: 'Type I' and 'Type II' bulimia
The features distinguishing between those with bulimia with and without a
previous history of anorexia nervosa support the notion of categorisation into
'Type I' and 'Type II'. The greater age of the subjects with Type II bulimia at
conception can be explained simply in terms of the lost years of fertility. At a
descriptive level, those with Type II bulimia experienced pregnancy as intrusive
and alien, and felt uncomfortable with the intimate nature of breast-feeding.
This is consistent with our findings that the vast majority of these patients
relapsed postnatally, that very few engaged in breast-feeding and that the risk
of postnatal depression was elevated almost threefold.
Family planning: advice to patients with bulimia
Three-quarters of all the pregnancies studied were unplanned, of which the
majority resulted from mistaken beliefs about fertility in the presence of an
irregular menstrual cycle. It is important for clinicians to advise their
patients fully about the implications of their eating disorder for their
fertility, and to give contraceptive advice where necessary, particularly where
the oral contraceptive is taken in conjunction with regular vomiting. However,
the obstetric consequences of persistent bulimia remain unknown, making it
difficult for doctors to offer informed advice. Of patients attending the St
George's bulimia clinic, 9% were motivated to seek help through a fear of the
impact of the disorder upon their fertility, and concern for the potential harm
that their disordered eating might cause the unborn foetus (Lacey & Smith, 1987)
. This concern is supported by the limited available literature, which
raises the possibility of increased rates of miscarriage, foetal abnormalities
and smaller birth weights (Fahy & Treasure, 1989; Ford & Dolan, 1989; Mitchell
et al, 1991; Stein & Fairburn, 1989; Stewart et al, 1987; Woodside & Shekter-Wolfson,
1990; Stein et al, 1996) [4,8,13,19,22,23,21].
Pregnancy alters the course of bulimia nervosa in a variety of ways that appear
to be predictable. Clinicians should be aware of predictors of relapse and of
postnatal depression. Further research is needed to establish levels of eating
disorders in the general population during pregnancy, based on interviews rather
than screening questionnaires, and also to examine the obstetric complications
of bulimia nervosa, so that patients can be properly advised. Postnatal
depression appears to be massively increased in bulimia nervosa. This finding
requires confirmation by prospective, controlled studies and should also be
considered in any future studies of the effects of bulimia nervosa on parenting
- Postnatal treatment intervention for bulimia nervosa should focus on women
with more severe or persistent symptoms, previous anorexia nervosa, unplanned
pregnancy and gestational diabetes, all of which predict relapse.
- All women with bulimia should be specifically assessed for postnatal
depression, but particularly where their eating disorder is severe, where they
misuse alcohol or where there is a previous history of anorexia nervosa.
- Women with bulimia are at risk of unwanted pregnancies because of mistaken
beliefs about fertility in the presence of oligomenorrhoea, and should be
- The retrospective design employed to achieve a large sample size may have
generated recall bias, particularly in the diagnosis of postnatal depression.
- The sample was drawn from a tertiary referral centre, which may not reflect
women with eating disorders in the general population.
- Subjects uncertain about gestational diabetes were coded negative, which may
We gratefully acknowledge the contribution of Mrs Sarah Reed, research
secretary, in the preparation of this manuscript. We are indebted to the
patients and ex-patients who participated in this study, and from whom we
continue to learn.