| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
ORIGINAL ARTICLES |
From the Department of Mental Health, St George’s Hospital Medical School, London, United Kingdom.
Address correspondence and reprint requests to Dr. John Morgan, Consultant Psychiatrist, Yorkshire Centre for Eating Disorders, Seacroft Hospital, Leeds LS14 6UH, UK. E-mail: john.morgan{at}leedsmh.nhs.uk
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
Methods: This was a retrospective case-control comparison of obstetric complications in primigravidae previously treated for bulimia in a specialist eating disorder service. A cohort of 122 women with active bulimia during pregnancy was contrasted against 82 with quiescent bulimia, using structured interviews comprising the Eating Disorders Examination, Structured Clinical Interview for DSM-III-R, and systematic questions addressing obstetric complications.
Results: Odds ratios (ORs) for postnatal depression, miscarriage, and preterm delivery were 2.8 (95% confidence interval [CI], 1.2–6.2), 2.6 (95% CI, 1.2–5.6) and 3.3 (95% CI, 1.3–8.8) respectively. Risk of unplanned pregnancy was markedly elevated (OR, 30.0; 95% CI, 12.8–68.7). Risk estimates were not explained by differences in adiposity, demographics, alcohol/substance/laxative misuse, smoking, or year of birth, but relative contributions of bulimic behaviors were not discerned.
Conclusions: Active bulimia during pregnancy is associated with postnatal depression, miscarriage, and preterm delivery. Bulimia may be a treatable cause of adverse obstetric outcome.
Key Words: bulimia nervosa • postnatal depression • miscarriage • prematurity • hyperemesis gravidarum • gestational diabetes
Abbreviations: BN = bulimia nervosa; CI = confidence interval; DSM-IV = Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition; EDNOS = eating disorder not otherwise specified; GP = general practitioner; OR = odds ratio; PCOS = polycystic ovary syndrome; SCID = Structured Clinical Interview for DSM-III-R; SD = standard deviation.
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
Design
This was a retrospective case-controlled comparison of women treated for BN at the St. George’s Hospital Eating Disorders Unit between 1988 and 1994, using structured clinical interviews carried out between September 1996 and October 1997. The study was approved by the LREC (Local Research and Ethics Committee).
Participants
The sample consisted of patients who had previously presented at the St. George’s Hospital Eating Disorders Unit. Participants were recruited retrospectively by written invitation to all patients treated for bulimia between 1988 and 1994 to participate if they had conceived following treatment. Some were included in the study of women actively suffering from bulimia during pregnancy previously published by Morgan et al. (6). All participants had fulfilled Diagnostic and Statistical Manual of Mental Disorders, Third Edition, Revised (DSM-III-R) criteria for BN when first treated, and had later conceived (11). Two hundred thirty-three women fulfilling these criteria were identified. Participants were classified as "cases" if they met Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV) criteria for bulimia at some stage during their pregnancy. Participants were classified as "controls" if they did not have any eating disorder during the pregnancy, including eating disorder not otherwise specified (EDNOS), despite a history of BN. This control group was selected rather than one from the normal population because a lifetime presence of bulimia has been associated with factors that may affect pregnancy outcome, including polycystic ovary syndrome (PCOS) (12,13). Participants were excluded from the study if they were unable to give a valid retrospective history, or if they suffered from a subclinical eating disorder or EDNOS, to permit a comparison of definite bulimia against definite nonbulimia, especially given the possibility of recall bias in a retrospective study design. Only information about first pregnancies was used because different pregnancies for the same individual were not independent of each other.
Main Outcome Measures
The first author interviewed participants to establish eating disorder symptoms and diagnoses, using a structured interview schedule generating DSM-IV diagnoses based on behavioral components of the Eating Disorder Examination (EDE) 12th edition (14), with additional questions concerning demographic and social status at conception, alcohol consumption, menstrual history, history of anorexia nervosa, planning of pregnancy, and items concerning affective disorders in the postnatal period from the Structured Clinical Interview for DSM-III-R (SCID) (15). The SCID for DSM-IV was not yet available at the time of commencement of the study.
A structured questionnaire examining complications of pregnancy was designed for this study with input from a senior academic obstetrician, including fetal growth and development, complications of early pregnancy, antenatal disorders, antepartum hemorrhage, abnormal labor, and neonatal complications.
Corroborative information was sought from the subjects’ general practitioners (GPs), medical records, and families, where permitted. "Postnatal depression" was defined as nonpsychotic major depressive disorder with onset within 4 weeks postpartum; "miscarriage" as any report of clinically recognized miscarriage after a positive pregnancy test that occurred before the 24th week of pregnancy, and "preterm delivery" as delivery before 36 weeks (24–32 weeks, "extreme"; and 33–36 weeks, "moderate"). Other definitions are available from the authors on request.
Statistical Analysis
Continuous variables were summarized by the mean, standard deviation (SD), and range. Descriptive statistics were recorded for all of the above, and active BN cases were compared with quiescent BN controls for all variables. Univariate comparisons of dichotomous data were performed using the 
2 test (>5 observations in all cells) or Fisher’s exact test (
5 observations in 
1 cell). The p values for all hypothesis tests were two-sided, and 0.05 was the significance level. Odds ratios (ORs) with their associated 95% confidence intervals (CIs) were obtained for factors relating to presence or absence of bulimia. The Mann-Whitney test (U) was used to compare means for continuous variables. All statistical analyses were performed using version 9.0 of SPSS for Windows.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
|
Demographic characteristics of the study population are shown in Table 1. There were no significant differences in any demographic variables between active BN cases and quiescent BN controls.
|
Active BN cases were significantly more likely than quiescent BN controls to report a history of anorexia nervosa before initial presentation with bulimia, with 23% of cases and 10% of controls giving a positive history (2 (1) = 5.87, p = .015).
Each subject was asked if they would allow us to contact their GP, obstetrician or midwife, or close family member to validate the information supplied. Only 38 of the 204 cases and controls consented to our contacting someone for corroboration (21 cases versus 17 controls), of which 20 (11 cases, 9 controls) only consented to our contacting family members. Eighteen consented to our contacting a family member or GP; however, despite this, only 15 GPs were willing to cooperate (eight cases, seven controls.) There was no statistical evidence of bias toward refusing corroboration. Although it is disappointing that the majority of patients refused corroboration, despite the perhaps high response rate for those agreeing to participate in the study, this is in our experience the norm for the population being studied and the phenomenology of bulimia, in which a sense of shame, secrecy, and denial is central.
Characteristics of Participants at Conception
Active BN cases were more likely to report unplanned pregnancies (76% versus 10%; OR, 30.0; 95% CI, 12.8–68.7) and to have conceived with oligomenorrheic menstrual status (27% versus 7%, 2 (1) = 12.3, p < .001). There were no significant differences in estimated body-mass index at conception (mean [SD], 22.2 kg/m2 [1.49] versus 21.9 [1.68], U = 4316, Z = –1.66, p = .10).
By definition, all active BN cases were bingeing at least three times per week during the pregnancy, and no quiescent BN controls reported binge episodes. Ninety-five percent of cases (0 controls) regularly induced vomiting, 6 cases misused slimming pills (0 controls), and 20 cases and 1 control gave a history of laxative misuse (2 (1) = 12.2, p < .001), but cases were no more likely than controls to misuse alcohol (18% versus 18%, 2 (1) = 0.0022, p = .96), illicit substances (four cases versus three controls) or to smoke (39% versus 33%, 2 (1) = 0.87, p = .35).
Pregnancy Outcome
Twenty-six percent of active BN cases reported "miscarriage," compared with 12% of quiescent BN controls (2 (1) = 5.91, p = .015). This difference remained significant when nonlaxative abusing cases were compared with nonlaxative abusing controls (29% versus 12%, 2 (1) = 7.70, p = .0055) and likewise for substance misuse (26% versus 11%, 2 (1) = 6.47, p = .011). Rates of miscarriage were more than twice those of women with quiescent bulimia (OR, 2.6; 95% CI, 1.2–5.6).
Thus, 90 out of 122 active BN cases and 72 out of 82 quiescent BN controls proceeded to delivery. Reported rates of a variety of obstetric complications are shown in Table 2.
|
Active BN cases were significantly more likely to experience postnatal depression, preterm delivery, hyperemesis gravidarum, and gestational diabetes, and there was a nonsignificant trend toward an association between cleft lip/palate. Rates of postnatal depression (OR, 2.8; 95% CI, 1.2–6.2) and preterm delivery (OR, 3.3; 95% CI, 1.3–8.8) were three times greater than in quiescent BN controls. Data for gestational diabetes indicated rates more than five times (OR, 5.7; 95% CI, 1.2–26.6) those of quiescent BN controls but are insufficiently robust given dependence on women’s interpretation, testing policy, and possible increased frequency of testing in a subbulimic population, despite careful questioning at interview. Rates of hyperemesis gravidarum also appeared significantly greater among active BN cases than quiescent BN controls, but this should be interpreted with caution, given the small numbers involved. Questions concerning maternal weight gain and length of stay in neonatal unit were not answered with sufficient confidence to be included in the study.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
Strengths and Limitations
This study benefits from a sample size more than six times bigger than its predecessors, application of interview-based measures, use of an appropriate control, and consideration of confounding factors. Nonetheless, it has two principal limitations. First, in common with most eating disorder research, the sample was drawn from a clinic population that may not be representative of bulimia in the general community. Second, statistical power was achieved at the expense of a retrospective study design, and attempts to corroborate clinical data were thwarted by participants’ fears of stigmatization. To date, only two prospective studies (4,7) have achieved sufficient statistical power to generate clinically meaningful data but did so in heterogeneous samples of 49 women with symptoms of both anorexia and bulimia, with a particular risk of type II errors. The possibility of recall bias, particularly response acquiescence bias, should be considered throughout our study. We have minimized this risk by excluding women suffering from subclinical eating disorders or EDNOS, and by looking at quantifiable behavior rather than psychopathology, by employing dichotomous measures and by using structured interview rather than questionnaires, which have been shown to overestimate levels of disorder in previous studies (17). Dichotomous outcomes such as postnatal depression, miscarriage, preterm delivery, and fetal abnormalities are still subject to significant recall bias, affected by active BN during pregnancy or by women’s recall of eating-disordered symptoms during pregnancy in light of outcome, but the necessarily large prospective studies are unlikely to take place without the justification of warts-and-all retrospective studies such as this. A further limitation associated with the retrospective study design is the possibility of recruitment bias. Some women with less positive pregnancy outcomes may have sought entry in a search for meaning, whereas others may have declined. Only 18 out of the 233 women approached actually refused to participate (8%). Interestingly, but perhaps unsurprisingly, no women came forward to participate who had elected for medical or other reasons to terminate their pregnancies. This bias was somewhat minimized by the comparison of active bulimia against quiescent bulimia, as determined by structured clinical interview.
Postnatal Depression
This study provides further support to the possibility of an increased risk of postnatal depression in women actively bulimic during pregnancy, as previously suggested by Abraham (18), Franko et al. (7), and our previous study (6). However, this apparent increased risk requires confirmation by robust, interview-based prospective studies. BN is strongly associated with comorbidity for affective disorders, although whether these are depressive symptoms secondary to bulimia itself remains unclear (18–20). The SCID was chosen as a diagnostic tool for its inclusion of detailed questions in order to minimize recall bias.
Rates of postnatal depression in the general population are probably no higher than a matched nonpregnant population, falling within the range of 10% to 15% (21). Thus, this study’s findings do not suggest a special link between bulimia and postnatal depression, but are consistent with the association between bulimia nervosa and depression in general (22). In this sample, a slightly higher than expected proportion, 21%, of even those with quiescent BN during pregnancy was found to have postnatal depression. The association between postnatal depression and BN has not been extensively documented, although predictors of postnatal depression have been delineated (6) and, with perinatal depression, warrants further investigation. Pregnancy and childbirth are major life events and raise issues of mutable gender and social roles, attachment, and separation (23), which are also regarded as fundamental in understanding eating disorders (24).
The association of postnatal depression and active BN in pregnancy has important implications in understanding parenting skills of patients with eating disorders, in which the quality of parent-child interactions may be altered (25–27). In addition to the distress postnatal depression causes to the sufferers and their families, there is a significant body of evidence from longitudinal studies that postnatal depression adversely affects child emotional and cognitive development (28), one study still finding effects evident at 11 years (29). Stein et al. (26) compared maternal controlling behavior with 1-year-old infants in women with eating disorders, women with postnatal depression, and a healthy comparison group and found that, although there was significant depressive symptomatology in the eating disorders group, the mothers with eating disorders used more verbal control, especially strong control. All women with bulimia should be specifically assessed for depression following childbirth, and further research predicting risk, exploring causality and positive postnatal interventions is called for.
Obstetric Complications
A substantial proportion of nonpregnant women seek treatment for their bulimia because they desire a child but fear obstetric complication, and there is a tendency for women to become pregnant at the end of treatment (10). The impact of pregnancy on preexisting bulimia is predictable (6), and fear of harming the fetus motivates some women to desist from bingeing and vomiting during pregnancy. The majority of women with active or quiescent bulimia had normal pregnancies and healthy babies. However, this study also demonstrates a clear association between active bulimia and elevated risks of miscarriage and preterm delivery and suggests that active BN cases were also more likely to experience hyperemesis gravidarum and gestational diabetes, with a nonsignificant trend toward an association between cleft lip or palate. These suggestions of a potential association between active bulimia in pregnancy, hyperemesis gravidarum, gestational diabetes, and possibly cleft lip or palate are consistent with our previous findings, though these studies also lacked sufficient power to demonstrate clear association (3,6). These findings contribute to information we can offer to our patients on the risks of persistent bulimia during pregnancy, in order to answer their questions, and to provide motivation for recovery.
It is unclear if bulimic behaviors precipitate miscarriage and prematurity or if a common risk factor is shared. When there are multiple variables involved, a randomized controlled study is the best to give a creditable answer. Stress-diathesis (30), PCOS (13,31), leptin (32), and apoptosis (33) provide potential mechanistic models that are not necessarily mutually exclusive and provide the theoretical possibility that bulimia could manifest in miscarriage and preterm delivery, as demonstrated in this study, as well as cleft lip or palate and gestational diabetes, as suggested in this study.
The association between hyperemesis gravidarum and bulimia only just reached statistical significance and may be explained by chance in the context of repeated statistical tests. Furthermore, the distinction between self-induced vomiting in bulimia and spontaneous vomiting in hyperemesis depends on issues such as denial and awareness of conscious motivation. Nonetheless, almost 1 in 10 cases had features of hyperemesis compared with prevalence figures in the general population of at most one in a hundred (34), and an association was also suggested by Abraham (18), although this lacked statistical power. Our findings generate the hypothesis that some cases of hyperemesis mask bulimia, and interview-based prospective studies should be carried out to test this hypothesis.
Future Research
Limited by the methodological expediencies of comparing true cases against true noncases and use of retrospective design, naturalistic prospective studies in an antenatal setting are now justified. These should examine impact of subclinical bulimic behaviors, more common in the general population. Further retrospective analysis of eating behaviors among mothers of children with fetal defects and among women with repeated miscarriage is more realistic than a prospective design, given the low incidence of fetal defects and the logistics of clinical presentation with miscarriage.
Clinical Implications
This study demonstrates a clear association between active bulimia, miscarriage, and preterm delivery, suggesting the former may be a treatable cause of the latter. Pregnancy offers a window of opportunity to engage bulimic women in treatment, and issues of fertility are important motivational factors. Women with bulimia are at risk of unplanned pregnancy and should be counseled accordingly. Persistent severe bulimia during pregnancy requires access to treatment and even possible hospital admission to stabilize eating patterns, whereas cases of hyperemesis gravidarum should be screened for cryptic eating disorders. All pregnant women with eating disorders should be considered at increased risk of postnatal depression and screened accordingly. For this to work, obstetricians and midwives need help in case identification through liaison psychiatric input, simpler screening questionnaires (35), and availability of eating-disorder treatment facilities. Advances in knowledge of reproductive physiology of eating disorders serve to emphasize that cycles of feast and famine modulate fertility far more profoundly than simply through adiposity.
We thank Professor S. Arulkumaran, professor of obstetrics and gynaecology at St George’s Hospital Medical School, for advice in the preparation of this manuscript. We gratefully acknowledge the advice of Professor S. Campbell and Dr M. Atherton in the original planning of this study. We are indebted to the women who participated.
|
NOTES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
Received for publication February 28, 2005; revision received January 28, 2006.
DOI:10.1097/01.psy.0000221265.43407.89
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
This article has been cited by other articles:
 |
|
 |
|
  A. M. Siega-Riz, M. Haugen, H. M Meltzer, A. Von Holle, R. Hamer, L. Torgersen, C. Knopf-Berg, T. Reichborn-Kjennerud, and C. M Bulik Nutrient and food group intakes of women with and without bulimia nervosa and binge eating disorder during pregnancy Am. J. Clinical Nutrition, May 1, 2008; 87(5): 1346 - 1355. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 G. C. Patton, C. Coffey, J. B. Carlin, L. Sanci, and S. Sawyer Prognosis of adolescent partial syndromes of eating disorder The British Journal of Psychiatry, April 1, 2008; 192(4): 294 - 299. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 V. B. Ward Eating disorders in pregnancy BMJ, January 12, 2008; 336(7635): 93 - 96. [Full Text] [PDF]
|
|
|
|
|
|
N. Micali, E. Simonoff, and J. Treasure Risk of major adverse perinatal outcomes in women with eating disorders The British Journal of Psychiatry, March 1, 2007; 190(3): 255 - 259. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
H. C Murphy and J. F Morgan Society's advice on low weight and IVF was ignored by media BMJ, September 23, 2006; 333(7569): 654 - 654. [Full Text]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
