PSYCHOLOGICAL TRAUMA AND EATING DISORDERS
Timothy D. Brewerton, M.D., D.F.A.P.A., F.A.E.D.
Department of Psychiatry and Behavioral Sciences Medical University of South Carolina Charleston, SC
Mailing Address:
2205 Middle Street, Suite 104 Sullivan’s Island, SC 29482
E-mail Address:
[email protected]
Annual Review of Eating Disorders (submitted)
2 Abstract
Objectives of Review: To review research published over the last two years on the relationship between traumatic experiences and the eating disorders (ED's). Summary of Recent Findings: Recent reviews and an 18-year longitudinal study confirm that CSA is a significant but nonspecific risk factor for the development of ED's, especially BN and other ED's with bulimic symptomatology. New cross-sectional research further demonstrates significant associations between sexual and other forms of abuse and bulimic disorders in adults, as well as adolescent girls and boys. The scope of abusive experiences has been widened to include bullying, racial discrimination, dating violence, date rape, physical neglect, emotional abuse, sexual harassment over the Internet, and Munchausen's syndrome by proxy. Future Directions: Further study is needed on the role of psychological trauma and PTSD on the long-term course and prognosis of the ED's. Randomized, controlled treatment trials in bulimic patients with abuse histories and trauma-related comorbidity, such as PTSD, are a logical next step. Finally, future research is likely to further elucidate the psychobiological underpinnings of abuse, ED's, and comorbidity, which may lead to better prevention and treatment strategies.
3 I. Introduction. The relationship between traumatic experiences, such as sexual, physical, and emotional abuse, and the eating disorders (ED's) continues to be of great scientific and clinical interest, and has been the subject of a number of research publications over the last two years, which will be the subject of this review. However, a brief summary of the knowledge base on this topic is in order before presenting these new data. The conclusions from the comprehensive review on the role of childhood sexual abuse (CSA) by Wonderlich and colleagues (1997) remain essentially uncontested to this day and can be briefly summarized as follows: 1.
CSA is associated with bulimia nervosa (BN).
2.
CSA is more common in BN than anorexia nervosa (AN).
3.
CSA is not a specific risk factor for ED's.
4.
CSA is not associated with greater severity of the ED.
5.
Particular features of CSA are associated with ED symptoms, including decreased
social competence, poor maternal relationships, unreliable parenting, the severity of the CSA, and the presence of lifetime post-traumatic stress disorder (PTSD) (Dansky et al., 1997). 6.
CSA is associated with psychiatric comorbidity in ED subjects. More recent research since this study was published has largely substantiated
these conclusions. In addition to CSA, other data have linked sexual assault and physical assault during adulthood, as well as resultant PTSD, to BN (Dansky et al., 1997). Other studies have reported that histories of childhood physical abuse or severe corporal
4 punishment are associated with bulimic symptomatology in children and adolescents. These data have been summarized elsewhere (Brewerton, 2002; 2004).
II. Literature Review. This literature review employed the following methodology. A MEDLINE search was conducted for all indexed articles published during the years 2000 and 2003 and during January 2004 using all possible combinations of two sets of keywords cross-referenced with each other. The first set of keywords included "eating disorders," " bulimia," "anorexia nervosa," "binge eating disorder," "eating disorder not otherwise specified" (EDNOS), while the second set of keywords included "sexual abuse," "sexual assault," "sexual harassment," "physical abuse," "physical assault," "emotional abuse," "child abuse," "neglect," "victimization," "trauma," "PTSD," and "dissociation." Particular attention was given to articles with new data and review articles, although a few case reports that illustrate new concepts or relationships have been included.
A. Reviews Four review articles have been published during the 2-year period of interest. However, sexual abuse was the primary focus in only one paper, while in the other 3 papers it was one of several factors examined. Smolak and Murnen (2002) reported their results of a meta-analysis of the relationship between CSA and ED's using 53 data-containing articles and book chapters. They focused on two specific objectives. The first was to assess the extent and consistency of the relationship between CSA and ED's, and the second was to examine methodological factors contributing to
5 the heterogeneity of this relationship. The studies examined fell into two basic types: 1) those comparing individuals with CSA versus those without CSA in terms of ED incidence (30 studies), and 2) those comparing individuals with an ED (usually BN) versus those without an ED in terms of CSA experiences (23 studies). A small, but statistically significant, positive relationship between CSA and ED emerged for both types of studies (r = 0.18, p < 0.001 and r = 0.12, p < 0.01 respectively). In the first type of study (CSA + ED) the positive relationship between CSA and ED's was marked by statistically significant heterogeneity (p < 0.01) in that the effect sizes were largest when the Eating Attitudes Test (EAT) or the Eating Disorders Inventory (EDI) was used as the primary ED measure (r = 0.284, p < 0.05), as opposed to a bulimic or other measure. Effect sizes also varied significantly by participant's age in these studies and were largest when participants were in their 20's (r = 0.24, p < 0.05) compared to when they were in their teens (r = 0.11) or > 30 years old (r = 0.12). In the second type of study (ED + CSA) there also was statistically significant heterogeneity (p < 0.01). Effect sizes were largest when nonclinical groups were compared with clinical samples (r = 0.2) as opposed to when both the control and the experimental (ED) groups were clinical samples (r = 0.08, p < 0.05). In addition, bulimic subjects had significantly larger r values (r = 0.08) when compared to subjects with a more general ED diagnosis (r = -0. 023, p < 0.05) in relation to their respective control groups. The age of participants did not predict a difference in r values between groupings in this subsample of studies. Across all studies from both groups, the overall effect size was calculated to be r = 0.1 (p < 0.01), thereby indicating a small effect but with significant heterogeneity. Interestingly, studies that used CSA status as the independent variable (CSA + ED) showed an effect size greater than twice that of studies using ED status as the independent variable (ED + CSA), and this disparity appeared to be primarily accounted for by important
6 methodological differences across studies. One of the most notable was the variation in definitions used to define an ED. The CSA focused studies, which had an effect size of r = 0.284, were much more likely to use the EAT or the EDI as ED measures than the ED focused studies. In addition, the nature of the comparison group was quite important in terms of explaining heterogeneity. Within ED focused studies, those that compared a clinical ED group to a nonclinical group had an r = 0.21, whereas those comparing a clinical ED group to other clinical patients yielded r = -0.12. The authors concluded that future models of CSA and ED need to more clearly specify what aspects of ED (e.g., body image or eating) are most influenced by which types of CSA. A reason not cited by the authors that might account for the difference in effect sizes between the two types of studies is a disturbance in memory. It is probably easier for already identified CSA subjects to remember and report whether they had an ED or ED symptoms than it is for eating disordered subjects to remember and report whether they were victims of CSA or not. In a large, representative sample of adult U.S. women, forgetting traumatic events was associated with not only CSA, but with BN and purging behavior independent of a BN diagnosis (Brewerton et al., 1999). Stice (2002) published a meta-analytic review using only prospective and experimental studies and concluded that several generally accepted risk factors for eating pathology, including sexual abuse, were not empirically supported. However, limiting studies to those only using prospective and experimental designs is extremely restrictive, and few studies in the trauma-ED field meet these stringent criteria. Stice cited the only longitudinal study published at the time of his review (Vogeltanz-Holm et al., 2000), which did not find a relationship between CSA and binge eating, dieting, or weight concerns in 790 U.S. women studied in 1991 and again in 1996. However, specific ED diagnoses were not obtained in the Vogeltanz-Holm study, and they did
7 not ask about purging behaviors unless respondents endorsed binge eating with loss of control. Unfortunately, Stice did not yet have access to the Johnson study (2002), the best longitudinal study to date, which is reviewed below. In addition, recent unpublished results from a large, representative sample of U.S. women indicate significant associations between sexual assault, PTSD, and purging behavior without bingeing (Brewerton et al., 2003). Jacobi and colleagues (2004) recently published a systematic and rigorous review of risk factors, including sexual abuse, for the onset of ED's, including AN, BN and BED. These investigators used Kraemer and colleagues (1997; 1999) recently described rigorous definitions of risk factors as the basis of their review. They found only one study that supported classifying sexual abuse as a retrospective correlate for the development of AN, and they called for more replication studies. However, there was much more data for BN, and they concluded that there is strong evidence to support the notion that sexual abuse is a nonspecific, retrospective correlate as well as a variable risk factor of medium potency for the development of BN. In terms of BED, they noted the paucity of studies to date, yet concluded that sexual abuse and physical neglect were nonspecific, variable risk factors for BED. In addition, perceived paternal neglect was noted to be a retrospective correlate of medium potency with BED. Klump and colleagues (2002) reviewed available data on the effects of nonshared environmental influences on the development of ED's. Nonshared environmental are experiences that are unique to siblings reared in the same family and include adverse life experiences, such as CSA and teasing. In their review of twin studies, they note that the reported variance in both AN and BN that can be accounted for by nonshared environmental factors is approximately 17%46%. The authors noted that although the number of studies directly examining these influences are limited, initial data indicate that differential paternal relationships, body weight teasing, peer
8 group experiences, and life events, including CSA, may account for the development of ED pathology in one sibling versus another. However, they noted that not all studies have found the link with CSA. In conclusion, the authors called for additional research to identify specific nonshared environmental influences on ED's such as differential parental and sibling treatment, disparate peer group characteristics, and differential experience of life events such as sexual and physical abuse.
B. New Longitudinal, Prospective Studies Johnson and colleagues (2002) published results from the most comprehensive longitudinal study done to date on the association between childhood adversities, including CSA, physical abuse, and neglect, and the subsequent later development of eating- or weight-related problems, including ED's. The authors studied a large community-based sample of mothers and their offspring (n = 782) followed over an 18-year period. Information about sexual abuse and physical neglect were obtained from structured interviews of both child and mother. Of the total sample of offspring 6.6% received a diagnosis of an ED during adolescence or adulthood, with female offspring having a rate of 11% and male offspring having a rate of 2% (odds ratio (OR) = 5.2). Individuals who had experienced sexual abuse or physical neglect during childhood were at elevated risk for subsequent ED's (OR: sexual abuse = 4.82; physical neglect = 5.11). The great majority of the ED's identified in this study were characterized by bulimic symptomatology (BN, BED, or EDNOS with purging); only one case of AN (in a boy) was identified. In addition, sexual abuse and physical neglect predicted recurrent fluctuations in weight, strict dieting, and self-induced vomiting, and physical neglect predicted obesity irrespective of the presence of an ED. It is notable that all of the significant relationships reported in this study were manifest after
9 controlling for age, difficult temperament, childhood eating problems, and parental psychiatric disorders. Another key finding that is important to emphasize is that offspring who had experienced three or more kinds of maladaptive paternal behaviors were 3 times more likely to have ED's. This lends credence to the hypothesis that repeated traumatic events are more specific to the development of ED's with bulimic symptomatology (see below). The results of this pivotal and impressive research indicate that maladaptive paternal behavior may play a more important role than maladaptive maternal behavior in the development of ED's in progeny. Previous work has focused much more on the mother's role and has relatively neglected the father's role. In terms of physical abuse and its derivatives, several findings were observed. Physical abuse predicted low body weight (OR = 4.71, p < 0.01), but not a diagnosis of AN. High peer aggression predicted use of medication to lose weight (OR = 3.85, p < 0.01), and harsh maternal punishment (OR = 4.82, p < 0.01) and loud arguments between parents (OR = 6.15, p < 0.01) predicted obesity (independent of an ED). Although sexual abuse and physical neglect were found to be non-specific and variable risk factors in this study, its longitudinal design clearly established these factors as significant contributors to the development of ED's and related eating problems, particularly those involving bulimic symptoms.
C. New Studies of Abuse and ED's in Adults Leonard and colleagues (2003) evaluated associations between childhood physical and sexual abuse and eating disturbances, psychiatric symptoms, and the probability of later abuse in adulthood. Fifty-one outpatient bulimic women and 25 age and BMI matched "normal eater" college women, who had no history of ED per the Eating Disorders Examination (EDE) and no psychiatric history, took part in the study. In addition to the EDE, the Eating Attitudes Test
10 (EAT), the Childhood Trauma Interview (CTI), the Trauma Assessment for Adults (TAA), the Dimensional Assessment of Personality Pathology Basic Questionnaire (DQPP-BQ), the Centre for Epidemiological Studies Depression Scale (CES-D), the Barrat Impulsivity Scale (BIS), and the Dissociative Experiences Scale (DES) were administered to all subjects in order to measure eating symptoms, comorbidity, and childhood and adulthood abuse. Bulimic women reported significantly higher levels of childhood sexual abuse, childhood physical abuse, and combined childhood sexual-physical abuse compared to the normal eater women. Bulimic women not only demonstrated more psychopathology than nonbulimic women, but they also showed an association between the presence and severity of trauma and the severity of concomitant psychopathologic symptoms. As might be expected, the results positively associated dissociation and submissiveness to the severity of prior abuse. An important finding from this study is that abuse during adulthood was almost always preceded by previous childhood abuse. Only one (6.7%) of the 15 bulimic women who reported abuse during adulthood did not report some form of previous childhood abuse. These findings suggest an association between certain psychopathologic traits and the likelihood of abuse (especially when occurring both in childhood and adulthood). Observed associations could implicate causal effects of childhood abuse on personality development, influences of personality traits in heightening the risk of abuse, or both. These data add to findings of other investigators that multiple episodes of abuse are particularly relevant to the development of BN. Striegel-Moore and colleagues (2002) reported results from an interview-based assessment of a community sample of 162 women with BED, 107 psychiatric comparison subjects and 251 healthy controls, all of who were matched for age, ethnicity, and level of education. Two questions were examined in this study: 1) whether sexual and physical
11 abuse, bullying by peers, and ethnicity-based discrimination were associated with increased risk for developing BED, and 2) whether any such increased risk is specific to BED. Caucasian women with BED reported significantly higher rates of all forms of abuse, including sexual and physical abuse, bullying, and racial discrimination when compared to healthy controls. Only discrimination rates were significantly higher in Caucasian women with BED compared to psychiatric comparison subjects. In black women with BED, all forms of abuse, except discrimination, were significantly higher in comparison to healthy controls. In addition, black women with BED reported sexual abuse rates that were significantly higher than those reported by psychiatric comparison subjects. These results are consistent with prior research that investigated patterns of risk for psychiatric disorder in that the authors discovered both ethnic similarities (physical abuse and bullying by peers) and differences (sexual abuse and discrimination) in the risk for BED.
D. New Studies of Abuse and ED's in Children and Adolescents Fonseca and colleagues (2002) examined a number of familial factors, including sexual abuse, in relationship to extreme weight control measures in a large group of adolescents (n = 9,042) using a comprehensive health survey of Connecticut students (7th, 9th, and 11th grades) administered in 1996. Extreme dieters, who intentionally vomited, took diet pills, laxatives, or diuretics to lose weight, were compared with adolescents reporting none of these behaviors (using logistic regression controlling for BMI and age). Extreme weight control behaviors were reported in almost 7% of youth. Risk factors for boys included sexual abuse history (OR = 2.8, p < 0.001) and high
12 parental supervision/monitoring, while protective factors included high parental expectations, maternal presence, and connectedness with friends and other adults. For girls, the only significant risk factor besides BMI (OR = 2.17, p < 0.002) was a history of sexual abuse (OR = 1.45, p < 0.001), while protective factors included family connectedness, positive family communication, parental supervision/monitoring, and maternal presence. These findings confirm an association between sexual abuse and extreme weight control behaviors, primarily involving purging, in both girls and boys, even when a formal ED diagnosis has not been made. The authors noted that sexual abuse as a predictive factor and connectedness to family members, other adults, and friends as a protective factor further establishes the importance of interrelationships between extreme weight control behaviors and interpersonal interactions at a family and social level. It is notable that purging, not bingeing, is what characterized the disordered eating group and it was these behaviors that were linked to sexual abuse. This link with purging rather than bingeing behaviors has been noted previously in a large representative sample of U.S. adult women (Dansky et al., 1997; Brewerton et al., 1999; 2003). Ackard and Neumark-Sztainer (2002) assessed the prevalence of date rape and date violence in adolescents as well as the associations between these events and disordered eating behaviors and psychopathology in a very large sample (n = 81,247) of boys and girls in 9th and 12th grades in Minnesota using the 1998 Minnesota Student Survey. Roughly 9% of girls and 6% of boys reported experiencing date rape or violence. More specifically, 4.2% of girls and 2.6% of boys reported date violence, while 1.4% of girls and 1.2% of boys reported date rape. Three percent of girls and 2.2% of boys reported both types of experiences on a date. Significant differences were found by grade for girls (but not boys), with girls in the 12th grade reporting the
13 highest rates of date-related violence (11.5%). Racial differences were also found, with white girls reporting the highest rates of date violence and combined date violence and date rape and Mexican American girls reporting the highest rates of date rape alone. The authors found that both date violence and rape was associated with higher rates of disordered eating behaviors in both girls and boys. Date violence and/or rape were also associated with suicidal thoughts and attempts, as well as lower scores on measures of self-esteem and emotional well being, especially in girls. Controlling for both age and race, those adolescents who experienced both date violence and rape were more likely to use laxatives (OR: girls = 5.76; boys = 28.22), vomit (OR: girls = 4.74; boys = 21.46), use diet pills (OR: girls = 5.08; boys = 16.33), binge eat (OR: girls = 2.15; boys = 5.80), and have suicidal thoughts or attempts (OR: girls = 5.78; boys = 6.66) than their nonabused peers. These odds remained significant, albeit weakened, after other abusive experiences by adults were controlled for. Of particular note is that a greater percentage of girls and boys who reported an abusive dating experience also reported repeat victimization (physical or sexual abuse perpetrated by an adult) when compared to their peers without an abusive dating experience. Because of the cross-sectional nature of this study, any statement about causality cannot be made with certainty, but nevertheless the authors speculated that normal developmental processes are likely to be disrupted by abusive experiences during dating relationships. Examples of such normal developmental processes that might be disrupted include the development of a stable self-concept and an integrated body image during adolescence. Fornari and Dancyger (2003) expounded eloquently on this important topic in a recent overview of psychosexual development and ED's. They noted potential adaptations might include anxiety about and avoidance of adult intimacy as well as sexually provocative acting out behaviors.
14 In another paper Ackard and Neumark-Sztainer (2003) used the same data set described above (Minnesota Student Survey) to examine associations between multiple forms of sexual abuse (including date rape, sexual abuse by an adult non-family member, and sexual abuse by an adult family member) and disordered eating behaviors and psychological health among adolescents in Minnesota. After controlling for grade and race, girls with multiple forms of abuse had significant odds ratios for the following behaviors: vomiting (OR = 4.1), laxative abuse (OR = 5.1), diet pill abuse (OR = 4.3), bingeing (OR = 2.2), fasting (OR = 2.3), and thinking about/attempting suicide (OR = 6.12). Boys with multiple forms of abuse had the following statistically significant OR's: vomiting (OR = 24.2), laxative abuse (OR = 29.2), diet pill abuse (OR = 17.3), bingeing (OR = 5.6), fasting (OR = 2.3), and thinking about/attempting suicide (OR = 9.5). Of particular interest in this study is that boys and girls with multiple forms of sexual abuse reported similar rates of bingeing (42.6% v. 41.1%), taking diet pills (22.3% v. 26.5%), and vomiting (18.7% v. 23.3%), but boys sexually abused in multiple ways had higher rates of laxative abuse than girls (22.4% v. 7.4%). The authors note that sexual abuse may be a particular risk factor for disordered eating behaviors among boys. It is also notable that the OR's for purging behaviors are much more robust that those for binge eating, thereby linking sexual abuse experiences more closely to purging than to bingeing, particularly when there are multiple forms or occurrences. These data also confirm links between sexual abuse experiences and comorbidity, such as suicidality, which are often associated with bulimic symptomatology. In another study by Ackard and colleagues (2003) using a different data set, the prevalence of adolescent dating violence and its associations with behavioral and mental health problems, including ED related behaviors, was investigated in a nationally representative sample of 3533 high school students in grades 9 through 12 using the Commonwealth Fund Survey of
15 the Health of Adolescent Boys and Girls. The authors also looked at the percentage of youth who stay in potentially harmful relationships because of fear of being hurt if they leave. The authors found that dating violence for both girls and boys was associated with dieting, binge eating, purging, as well as alcohol consumption, drug use, cigarette smoking, suicidal thoughts, depression and poor self-esteem. Girls and boys who endorsed both physical and sexual abuse reported higher rates of dieting (girls: 70.4% v. 56.4%; boys: 50.0% v. 22.5%) and binge-purge behavior (girls: 47.4% v. 13.7%; boys: 48.4% v. 4.7%) than their nonabused peers. These differences persisted after controlling for race, SES, and BMI. Among girls who had ever binged and purged, dating violence during adolescence was associated with frequencies of bingeing and purging at least several times a week. Notably, 100% of boys with sexual abuse and 95% of boys with both sexual and physical abuse reported bingeing and purging several times per week compared to 57% of non-abused boys. The parallel results for girls were not as striking but nevertheless higher in those reporting sexual abuse (63%) and combined sexual-physical abuse (64%) compared to girls denying abuse (45%). One major limitation of this study is that the inquiry about "bingeing and purging" was asked as one question rather than two separate questions such that respondents, who engaged in either bingeing or purging, but not both, would have to answer in the negative. Nevertheless, this study adds to the growing body of literature linking sexual and physical abuse to bulimia. These results also underscore the importance of abuse in the development of bulimic symptomatology in boys, an area that has often been overlooked in previous studies. Sherwood and colleagues (2002) reported their results from the administration of a 225item questionnaire to 5163 7th, 9th and 11th grade female public school students intended to examine factors associated with ED's among girls involved in weight-related sports. ED
16 symptoms were found in almost one third of girls involved in both weight-related and nonweight-related sports. However, after controlling for a number of factors, including grade, race, SES, and study design effect, girls in weight-related sports were 51% more likely to have ED symptoms than girls in non-weight related sports. Of note to this review is that girls in weightrelated sports who had ED's had more physical abuse (OR = 3.29) and sexual abuse (OR = 3.87) than girls in weight related sports without ED's. These results may be useful in future efforts to identify subpopulations of girls at higher risk for developing ED's. The uniqueness of the case report by Gati and colleagues (2002) involving sexual harassment over the Internet makes it worth mentioning. Prior to this report the authors could find no information about sexual abuse through the Internet as a precipitant to the development of ED's. In this case a 16 y/o adolescent female developed AN, restricting type, immediately following an incident of sexual harassment by a 51 y/o man who made contact with her via the Internet. This person had initially misrepresented himself as a 22 y/o man during the first few weeks of their romantic interaction, which soon evolved into sexual harassment. This case report of "on-line harassment suffered without off-line follow-up" is thought to have triggered a fairly rapid onset of AN, restrictive type. Longer-term follow-up of this case was not available; so it is not known if this young woman went on to develop bulimic symptoms. In a case report by Vanelli (2002) an 8 y/o girl was the victim of Munchausen's syndrome by proxy and was hospitalized for AN 3 years later. Using medical information obtained from a variety of health web sites her father contrived abnormally elevated glucose levels to feign a diagnosis of diabetes and obtain hospitalization. This child was subjected to a number of medical investigations, including endoscopy and magnetic resonance imaging, in at least three different countries. During the 6 months prior to her admission for AN, the child demonstrated
17 progressive food refusal, and her weight dropped from 33 kg to 21 kg. On psychological evaluation her father demonstrated obsessive-compulsive personality disorder involving an obsessive preoccupation with health and illness, as well as with obtaining medical treatment for nonexistent illnesses in his daughter. As with the previous case report, continued follow-up of this girl was not reported; it is therefore unknown if she went on to develop bulimic symptoms.
E. Intrauterine and Perinatal Trauma Studies looking at the impact of complications during pregnancy and the perinatal period on the development of ED symptoms during adolescence are scarce and worthy of mention. It can be argued that intrauterine and/or perinatal traumas may be the very first psychological traumas that could potentially influence the development of ED's. Feingold and colleagues (2002) attempted to address this question through an historical chart review of 84 infants born prematurely (