Personality Dimensions in Eating Disorders and Their Relevance for Subtyping REGINA C. CASPER, M.D., DONALD HEDEKER, PH.D., AND JOEL F. McCLOUGH, B.A.
Abstract. Personality dimensions and psychopathological symptoms were assessed in 50 female patients hospitalized for the treatment of anorexia nervosa or bulimia nervosa and in 19 healthy female controls of similar age. Restricting anorexia nervosa patients, who had lost weight by consistently reducing their food intake, reported significantly greater self-control, inhibition of emotionality, and conscientiousness than controls or bulimia nervosa patients, before and after the data were corrected for depressive and eating pathology. Both nonbulimic and bulimic anorexia nervosa patients expressed stronger than normal conformance to moral and family values. On the impulsivity dimension, bulimia nervosa patients scored in the high normal range, whereas bulimic anorexia nervosa patients rated in the low normal range. The results suggest that a personality disposition toward overcontrol and reserve might constitute a risk factor for the restricting type of anorexia nervosa through fostering restrictive behavior toward food and avoidance of personal relationships. J. Am. Acad. Child Adolesc. Psychiatry, 1992,31, 5:83Q-.840. Key Words: personality traits, anorexia nervosa, restricting and bulimic type, bulimia nervosa. Historical accounts (Bell, 1985), case reports (Bruch, 1974; Janet, 1903), and systematic investigations (Casper et al., 1980; Garfinkel et al., 1980; Strober, 1980) suggest that anorexia nervosa may be a heterogeneous disorder. Earlier investigations distinguished between a primary or typical and a secondary or atypical form of anorexia nervosa (Bruch, 1974; Dally, 1969; King, 1963; Selvini-Palazzoli, 1978) and applied psychiatrically derived personality characteristics, such as the schizoid, obsessional, or histrionic personality (Dally, 1969; Dally and Gomez, 1979; Sours, 1980), to describe the characteristics of each type. A more recent approach stayed closer to the clinical impression these patients created and classified patients by their eating pattern. These studies (Beumont et al., 1976; Casper et al., 1980; Garfinkel et al., 1980; Haimes and Katz, 1988) have provided evidence for the notion that anorexia nervosa patients who consistently restrict their food intake display certain differences in psychopathological, emotional, and behavioral features from vomiting or bulimic patients with anorexia nervosa. Systematic studies into the personality associated with the two types of anorexia nervosa have given mixed results. Beumont (1977) reported obsessional symptoms and traits in both "dieting" and "vomiting" patients with anorexia nervosa when patients were asked to respond to three selfreport scales, the Eysenck Personality Inventory (Eysenck and Eysenck, 1964), Cattell's 16 Personality Factors Questionnaire (Cattell and Eber, 1962), and the Leyton Obsessional Inventory (Cooper, 1970); nevertheless, greater
Accepted February 28, 1992. Dr. Casper is Professor of Psychiatry, The University of Chicago; Dr. Hedeker is Associate Professor, Department of Public Health, The University of illinois; and Mr. McClough is with the Michael Reese Hospital, Chicago, 1L. This work was supported by N1MH Grant #1 R01 MH 35585-03 and by the Nathan and Emily Blum Fund (Dr. Casper). Reprint requests to Dr. Casper, Department of Psychiatry, The University of Chicago, Box 411,5841 S. Maryland Ave., Chicago.Ll. 60637. 0890-8567/92/3105-0830$03.00/0©1992 by the American Academy of Child and Adolescent Psychiatry.
830
extroversion distinguished vomiting patients from dieting patients. Ben-Tovim et aI. (1979) failed to find personality differences between 12 abstaining and 9 habitually vomiting patients with anorexia nervosa on the basis of the Eysenck Personality Inventory and Manual of Hostility and Direction of Hostility Questionnaire (Philip, 1973). By contrast, Strober (1980) reported that, in adolescents newly hospitalized with anorexia nervosa, restricting patients rated themselves less sociable and hostile and more seriously minded and rigid than bulimic adolescents on the California Psychological Inventory. However, when patients were retested after weight gain, these differences were no longer significant. In another study (Strober, 1981), anorectic patients portrayed themselves as more rigid and more conforming compared with healthy control or depressed patients. Bulimia nervosa patients have been mostly evaluated for personality disorders (Gartner et al., 1989; Kennedy et al., 1990; Pope et al., 1987; Yates et al., 1989; Yager et al., 1989; Zanarini et al., 1990). and little work has been done on the personality features of this population. A contrast between outpatients with bulimia nervosa and outpatients who suffered from a depressive disorder reported greater than normal egocentricity and more negativity in bulimic compared with depressed women (Weisberg et al., 1987). Heilbrun and Bloomfield (1986) noted poorer impulse control in normal young women with bulimic tendencies who were compared with young women without bulimic tendencies. These clinical and psychometric studies provide tentative evidence in support of the notion that the different eating patterns in anorexia nervosa may be associated with particular personality tendencies and that individual personality attributes may influence the expression of a particular eating pattern (Swift and Wonderlich, 1988). The present study was designed to test these propositions by assessing personality dimensions in patients who had different types of eating disorders with focus on the adaptive rather than the dysfunctional aspects of the personality. We hypothesized that patients with bulimic anorexia nervosa or bulimia nervosa would be characterized by greater impulsivity (Eckert et al., 1979) and show greater flexibility and J. Am.Acad.ChildAdolesc. Psychiatry, 31:5,September 1992
PERSONALITY DIMENSIONS IN EATING DISORDERS
sociability than restricting patients with anorexia nervosa. Conversely, anorexia nervosa patients who consistently restricted their food intake and thus successfully suppressed their appetite would show a greater overall tendency toward self-control, internal control , and restraint when compared with healthy subjects or patients who show bulimic behavior. To test the control/impulsivity hypothesis, we used the Multidimensional Personality Questionnaire (MPQ) by Tellegen (1982), which consists of three higher order dimensions, positive emotionality, negative emotionality, and constraint. In addition, the Self-control/Social Systems Control Scale developed by Reid and Ware (1974) and another established personality inventory, the California Psychological Inventory (Gough, 1957) were administered to assess flexibility and sociability. Method
Study Sample
The clinical sample was composed of 50 consecutively admitted female patients with eating disorders, either anorexia nervosa or bulimia nervosa. They were diagnosed according to DSM-III-R criteria (American Psychiatric Association, 1987) independently by two psychiatrists using body weight on the day of admission. Patients with anorexia nervosa were divided on the basis of their eating patterns into a restricting group (N = 12, age range 12 to 34 years old, because they lacked present or past evidence of binge eating or pathological vomiting behavior, as defined in DSM-III-R, and a bulimic group (N = 19, age range 12 to 39 years old). Past bulimic behavior differed in frequency from three times per week to several times daily. Bulimia nervosa patients (N = 19) ranged in age from 14 to 34 years; all except for one patient who did not vomit, but used laxatives, had engaged in binge-eating-vomiting behavior, at least twice daily. One among the bulimia nervosa patients had a history of anorexia nervosa. The control group conTABLE
sisted of 19 normal female college and medical students whose ages ranged from 17 to 35 years. The control group was screened with the Diagnostic Interview Schedule (Robbins and Helzer, 1985), and subjects who qualified for any psychiatric disorder were excluded. Other details of the demographic and clinical characteristics of the patient and control sample are given in Table 1. Instruments
The three personality questionnaires were administered in random order along with other assessment measures during the first 4 days after hospital admission. All patients signed a written consent form before participation in the study that had been approved by the Human Subjects Investigational Committee. Control subjects completed all forms in the research offices after they had consented to the study in writing. All subjects knew that code letters were used on all questionnaires to maintain anonymity. Subjects were instructed to describe their usual way of acting and feeling on the personality questionnaires. The Multidimensional (formerly Differential) Personality Questionnaire (MPQ) (Tellegen, 1982; 1985) is a 300-item factor-analytically developed self-report instrument. Its scales describe 11 primary personality dimensions: wellbeing, social potency, achievement, social closeness, stress reaction, alienation, aggression, impulsivity/control, danger seeking/harm avoidance, traditionalism, and absorption. Three higher order factors are entitled' 'positive emotionality" Goy, vigor-calm, engagement), "negative emotionality " (anger, anxiety-depressive disengagement), and "constraint." In addition, three independent personality dimensions: novelty seeking, harm avoidance, and reward dependence recently proposed by Cloninger (1987), which can be derived from the MPQ scales, were calculated for all groups. The California Psychological Inventory (CPI) (Gough,
1. Demographic and Clinical Characteristics of In-Hospital Patients with Anorexia Nervosa and Bulimia Nervosa
Age in Years (SD)
Duration of Illness in Years X (SD)
Restricting anorexia nervosa patients (N = 12) 20.4 B D (6.6) Bulimic anorexia nervosa patients (N = 19)
25.1
Bulimia nervosa patients (N = 19) Control subjects (N = 19)
Body Weight (% ideal) (SD)
Body Mass Index in kglm2 X (SD)
Eating Attitudes Test X (SD)
Beck Depression Inventory X (SD)
2.5 (3.2)
7 I.I A* B* D* (8.2)
14.7 A* B* D* (2.2)
54.3 D* (26.6)
25.1 D* (12.8)
(6.9)
2.8 (2.1)
88.7 C F*
(20.8)
18.4 C F*
(3.9)
49.1 F* (26.2)
17.9 F* (11.7)
21.4 E
(5.0)
2.8 (2.1)
99.2
(14.3)
21.2
(3.1)
54.2 E* (24.6)
25.3 E* (11.6)
25.6
(4.6)
103.7
(8.1)
21.0
(1.6)
X
X
9.4
(5.4)
2.5
(2.8)
ANOVA post-hoc LSD comparison: p < 0.05, *p < 0.01; A = anorexia nervosa restricters versus bulimia nervosa; B = anorexia nervosa restricters versus bulimic anorexia nervosa; C = bulimia nervosa versus bulimic anorexia nervosa; D = anorexia nervosa restricters versus control subjects; E = bulimia nervosa versus control subjects; F = bulimic anorexia nervosa versus control subjects.
J. Am. Acad. Child Adolesc. Psychiatry, 31:5, September 1992
831
CASPER ET AL.
1957) is a standardized 480-item, true/false self-report scale, describing scale scores for 18 descriptive personality attributes. Factor analysis of the CPI results in three major dimensions, called person orientation, value orientation, and flexibility/control. A third scale (Reid and Ware, 1974) was a newer 34-item questionnaire of personal beliefs, a modified version of Rotter's scale (Rotter, 1966) of internal versus external control expectancies. This scale has three factors, a factor describing social political influence or social systems control, a factor in which luck or fate is believed to influence one's life called fatalism, and internal control. This scale was included to test Bruch's (Bruch, 1980) hypothesis that anorexia nervosa patients initiated actions primarily in response to external cues rather than in response to their own wishes and needs. To obtain information concerning the patients' thinking and behavior about weight and eating, the Eating Attitudes Test (EAT) (Garner and Garfinkel, 1979), an 40-item selfreport questionnaire, and the Eating Disorders Inventory (EDI) (Garner et aI., 1983) were used. For evaluating psychopathology, the Beck Depression Inventory (BDI) (Beck, 1978), a 23-item scale with statements describing dimensions of depression on a 4-point scale of severity, and the Hopkins Symptom Checklist (HCSL-58) (Derogatis et aI., 1974), composed of 58 items on a 5-point scale of severity yielding scores for five dimensions of psychopathology, were administered. Additionally, the Minnesota Multiphasic Personality Inventory (MMPI) (Hathaway and McKinley, 1951) profiles were compared. Patients were weighed in the fasting state each morning after voiding. Each patient wore a gown. Body weights reported are those taken on the same day as the personality measures were completed. Control subjects were clothed, but removed shoes or jackets before weighing on the same calibrated balance scale. Body weight was calculated as percentage of ideal weight for age and height (Hamill et aI., 1979, National Center for Health Statistics Growth Charts) for adolescents or based on the Metropolitan Life Insurance Tables (1964) for adults and expressed as body mass index (kg/m'). Statistical Analysis
Comparisons on the demographic variables between patients and controls were calculated using one-way analysis of variance (ANOVA). We used MANOV A for calculating differences for each questionnaire and for each set of derived higher order factors. Pairwise group comparisons were performed only for those subscales with significant univariate group differences from questionnaires exhibiting significant multivariate group differences. All pairwise group comparisons were performed using Fisher's protected least significant difference (LSD) procedure (Kirk, 1982). With equal levels of Type I errors, several studies reveal the enhanced power of the LSD procedure, relative to other post-hoc tests (Carmer and Swanson, 1973; Bernhardson, 1975). Because the literature suggests that even mild depression can influence personality dimensions (Liebowitz et aI., 1979) and because it is conceivable that abnormal eating attitudes may interact with certain personality traits, we performed multivariate analyses of covariance (MANCO VA) 832
to determine which group differences remained significant on the personality scales and the MMPI once age, total EAT scores, and total Beck Depression scores were introduced into the model. For these MANCOV As, the same procedure outlined above for the MANOVAs was followed. Namely, the univariate tests (of group effects) associated with each subscale of a questionnaire were examined only in the presence of a significant multivariate group effect for the questionnaire as a whole. Then, the Fisher LSD procedure was used to examine all pairwise group comparisons, according to the method outlined in Huitema (1980). Finally, duration of illness was also considered as a potential covariate, however, it was not used because it did not pertain to the controls, and for the patients, was highly associated with age (correlation = 0.70, p < 0.001), a variable which was being used as a covariate. To construct the higher order factor scores, we used the loadings reported by Nichols and Schnell (1963) for the CPI and the loadings reported by Tellegen (1985) for the MPQ. We used the coefficients reported by Cloninger (1987) to construct his three dimensions from the MPQ: harm avoidance, reward dependence, and novelty seeking. Results THE DEMOGRAPHIC AND CLINICAL DESCRIPTION OF THE GROUPS STUDIED
Table 1 shows the clinical and demographic characteristics for the sample. Restricting anorexia nervosa patients and patients with bulimia nervosa were comparable in age. Both were younger than the bulimic anorexia nervosa patients (ANOVA F = 3.31; P < 0.03). Because of these age differences, age was introduced as a covariate. The average duration of the illness was similar for the groups, between 2 and 3 years, with a wide range in the length of illness, from 6 months to 13 years. Ideal body weights and body mass indices showed significant between group differences (ANOVA F = 14.8; p < 0.001). Body weights were lowest in restricting anorexia nervosa patients, lower than normal in bulimic anorexia nervosa patients and within the normal range for bulimia nervosa patients. The scores obtained from the Beck Depression Inventory suggested clinical depression of about equal severity in each patient group, whereas control subjects were free of depressive symptoms (ANOVA F 19.3; p < 0.001). EATING ATTITUDES AND BEHAVIOR
Table 1 indicates that patients held markedly abnormal attitudes and behavior toward weight and food as measured on the EAT, in contrast to the control subjects who showed no weight or food preoccupation (MANOV A F = 4.22; df = 24.171; p < 0.001). Table 2 presents the EDI scores and the group comparisons. All patient groups displayed significant symptomatology on each subscale compared to normal controls. As would be expected, bulimic behavior was absent in restricting anorexia nervosa patients and bulimic patients reported more bulimic behavior than either restricting anorexia nervosa patients or controls. Post-hoc comparisons revealed no differences between restricting and J. Am. Acad. Child Adolesc. Psychiatry, 31:5, September 1992
PERSONALITY DIMENSIONS IN EATING DISORDERS TABLE
2. Comparisons of Eating Disorders Inventory and Hopkins Symptom Checklist (58) Scores for Patients with Acute Anorexia Nervosa (Restricting or Bulimic Type), Bulimia Nervosa, and Control Subjects Anorexia Nervosa Restricting (N = 12) X (SD)
Eating Disorders Inventory Drive for thinness Interoceptive awareness Bulimia Body dissatisfaction Ineffectiveness Maturity fears Perfectionism Distrust
Bulimic Anorexia Nervosa (N = 19) X (SD)
Bulimia Nervosa (N= 19) X (SD)
Control Subjects (N = 19) X (SD)
ANOVA P
Abstract. Personality dimensions and psychopathological symptoms were assessed in 50 female patients hospitalized for the treatment of anorexia nervosa or bulimia nervosa and in 19 healthy female controls of similar age. Restricting anorexia nervosa patients, who had lost weight by consistently reducing their food intake, reported significantly greater self-control, inhibition of emotionality, and conscientiousness than controls or bulimia nervosa patients, before and after the data were corrected for depressive and eating pathology. Both nonbulimic and bulimic anorexia nervosa patients expressed stronger than normal conformance to moral and family values. On the impulsivity dimension, bulimia nervosa patients scored in the high normal range, whereas bulimic anorexia nervosa patients rated in the low normal range. The results suggest that a personality disposition toward overcontrol and reserve might constitute a risk factor for the restricting type of anorexia nervosa through fostering restrictive behavior toward food and avoidance of personal relationships. J. Am. Acad. Child Adolesc. Psychiatry, 1992,31, 5:83Q-.840. Key Words: personality traits, anorexia nervosa, restricting and bulimic type, bulimia nervosa. Historical accounts (Bell, 1985), case reports (Bruch, 1974; Janet, 1903), and systematic investigations (Casper et al., 1980; Garfinkel et al., 1980; Strober, 1980) suggest that anorexia nervosa may be a heterogeneous disorder. Earlier investigations distinguished between a primary or typical and a secondary or atypical form of anorexia nervosa (Bruch, 1974; Dally, 1969; King, 1963; Selvini-Palazzoli, 1978) and applied psychiatrically derived personality characteristics, such as the schizoid, obsessional, or histrionic personality (Dally, 1969; Dally and Gomez, 1979; Sours, 1980), to describe the characteristics of each type. A more recent approach stayed closer to the clinical impression these patients created and classified patients by their eating pattern. These studies (Beumont et al., 1976; Casper et al., 1980; Garfinkel et al., 1980; Haimes and Katz, 1988) have provided evidence for the notion that anorexia nervosa patients who consistently restrict their food intake display certain differences in psychopathological, emotional, and behavioral features from vomiting or bulimic patients with anorexia nervosa. Systematic studies into the personality associated with the two types of anorexia nervosa have given mixed results. Beumont (1977) reported obsessional symptoms and traits in both "dieting" and "vomiting" patients with anorexia nervosa when patients were asked to respond to three selfreport scales, the Eysenck Personality Inventory (Eysenck and Eysenck, 1964), Cattell's 16 Personality Factors Questionnaire (Cattell and Eber, 1962), and the Leyton Obsessional Inventory (Cooper, 1970); nevertheless, greater
Accepted February 28, 1992. Dr. Casper is Professor of Psychiatry, The University of Chicago; Dr. Hedeker is Associate Professor, Department of Public Health, The University of illinois; and Mr. McClough is with the Michael Reese Hospital, Chicago, 1L. This work was supported by N1MH Grant #1 R01 MH 35585-03 and by the Nathan and Emily Blum Fund (Dr. Casper). Reprint requests to Dr. Casper, Department of Psychiatry, The University of Chicago, Box 411,5841 S. Maryland Ave., Chicago.Ll. 60637. 0890-8567/92/3105-0830$03.00/0©1992 by the American Academy of Child and Adolescent Psychiatry.
830
extroversion distinguished vomiting patients from dieting patients. Ben-Tovim et aI. (1979) failed to find personality differences between 12 abstaining and 9 habitually vomiting patients with anorexia nervosa on the basis of the Eysenck Personality Inventory and Manual of Hostility and Direction of Hostility Questionnaire (Philip, 1973). By contrast, Strober (1980) reported that, in adolescents newly hospitalized with anorexia nervosa, restricting patients rated themselves less sociable and hostile and more seriously minded and rigid than bulimic adolescents on the California Psychological Inventory. However, when patients were retested after weight gain, these differences were no longer significant. In another study (Strober, 1981), anorectic patients portrayed themselves as more rigid and more conforming compared with healthy control or depressed patients. Bulimia nervosa patients have been mostly evaluated for personality disorders (Gartner et al., 1989; Kennedy et al., 1990; Pope et al., 1987; Yates et al., 1989; Yager et al., 1989; Zanarini et al., 1990). and little work has been done on the personality features of this population. A contrast between outpatients with bulimia nervosa and outpatients who suffered from a depressive disorder reported greater than normal egocentricity and more negativity in bulimic compared with depressed women (Weisberg et al., 1987). Heilbrun and Bloomfield (1986) noted poorer impulse control in normal young women with bulimic tendencies who were compared with young women without bulimic tendencies. These clinical and psychometric studies provide tentative evidence in support of the notion that the different eating patterns in anorexia nervosa may be associated with particular personality tendencies and that individual personality attributes may influence the expression of a particular eating pattern (Swift and Wonderlich, 1988). The present study was designed to test these propositions by assessing personality dimensions in patients who had different types of eating disorders with focus on the adaptive rather than the dysfunctional aspects of the personality. We hypothesized that patients with bulimic anorexia nervosa or bulimia nervosa would be characterized by greater impulsivity (Eckert et al., 1979) and show greater flexibility and J. Am.Acad.ChildAdolesc. Psychiatry, 31:5,September 1992
PERSONALITY DIMENSIONS IN EATING DISORDERS
sociability than restricting patients with anorexia nervosa. Conversely, anorexia nervosa patients who consistently restricted their food intake and thus successfully suppressed their appetite would show a greater overall tendency toward self-control, internal control , and restraint when compared with healthy subjects or patients who show bulimic behavior. To test the control/impulsivity hypothesis, we used the Multidimensional Personality Questionnaire (MPQ) by Tellegen (1982), which consists of three higher order dimensions, positive emotionality, negative emotionality, and constraint. In addition, the Self-control/Social Systems Control Scale developed by Reid and Ware (1974) and another established personality inventory, the California Psychological Inventory (Gough, 1957) were administered to assess flexibility and sociability. Method
Study Sample
The clinical sample was composed of 50 consecutively admitted female patients with eating disorders, either anorexia nervosa or bulimia nervosa. They were diagnosed according to DSM-III-R criteria (American Psychiatric Association, 1987) independently by two psychiatrists using body weight on the day of admission. Patients with anorexia nervosa were divided on the basis of their eating patterns into a restricting group (N = 12, age range 12 to 34 years old, because they lacked present or past evidence of binge eating or pathological vomiting behavior, as defined in DSM-III-R, and a bulimic group (N = 19, age range 12 to 39 years old). Past bulimic behavior differed in frequency from three times per week to several times daily. Bulimia nervosa patients (N = 19) ranged in age from 14 to 34 years; all except for one patient who did not vomit, but used laxatives, had engaged in binge-eating-vomiting behavior, at least twice daily. One among the bulimia nervosa patients had a history of anorexia nervosa. The control group conTABLE
sisted of 19 normal female college and medical students whose ages ranged from 17 to 35 years. The control group was screened with the Diagnostic Interview Schedule (Robbins and Helzer, 1985), and subjects who qualified for any psychiatric disorder were excluded. Other details of the demographic and clinical characteristics of the patient and control sample are given in Table 1. Instruments
The three personality questionnaires were administered in random order along with other assessment measures during the first 4 days after hospital admission. All patients signed a written consent form before participation in the study that had been approved by the Human Subjects Investigational Committee. Control subjects completed all forms in the research offices after they had consented to the study in writing. All subjects knew that code letters were used on all questionnaires to maintain anonymity. Subjects were instructed to describe their usual way of acting and feeling on the personality questionnaires. The Multidimensional (formerly Differential) Personality Questionnaire (MPQ) (Tellegen, 1982; 1985) is a 300-item factor-analytically developed self-report instrument. Its scales describe 11 primary personality dimensions: wellbeing, social potency, achievement, social closeness, stress reaction, alienation, aggression, impulsivity/control, danger seeking/harm avoidance, traditionalism, and absorption. Three higher order factors are entitled' 'positive emotionality" Goy, vigor-calm, engagement), "negative emotionality " (anger, anxiety-depressive disengagement), and "constraint." In addition, three independent personality dimensions: novelty seeking, harm avoidance, and reward dependence recently proposed by Cloninger (1987), which can be derived from the MPQ scales, were calculated for all groups. The California Psychological Inventory (CPI) (Gough,
1. Demographic and Clinical Characteristics of In-Hospital Patients with Anorexia Nervosa and Bulimia Nervosa
Age in Years (SD)
Duration of Illness in Years X (SD)
Restricting anorexia nervosa patients (N = 12) 20.4 B D (6.6) Bulimic anorexia nervosa patients (N = 19)
25.1
Bulimia nervosa patients (N = 19) Control subjects (N = 19)
Body Weight (% ideal) (SD)
Body Mass Index in kglm2 X (SD)
Eating Attitudes Test X (SD)
Beck Depression Inventory X (SD)
2.5 (3.2)
7 I.I A* B* D* (8.2)
14.7 A* B* D* (2.2)
54.3 D* (26.6)
25.1 D* (12.8)
(6.9)
2.8 (2.1)
88.7 C F*
(20.8)
18.4 C F*
(3.9)
49.1 F* (26.2)
17.9 F* (11.7)
21.4 E
(5.0)
2.8 (2.1)
99.2
(14.3)
21.2
(3.1)
54.2 E* (24.6)
25.3 E* (11.6)
25.6
(4.6)
103.7
(8.1)
21.0
(1.6)
X
X
9.4
(5.4)
2.5
(2.8)
ANOVA post-hoc LSD comparison: p < 0.05, *p < 0.01; A = anorexia nervosa restricters versus bulimia nervosa; B = anorexia nervosa restricters versus bulimic anorexia nervosa; C = bulimia nervosa versus bulimic anorexia nervosa; D = anorexia nervosa restricters versus control subjects; E = bulimia nervosa versus control subjects; F = bulimic anorexia nervosa versus control subjects.
J. Am. Acad. Child Adolesc. Psychiatry, 31:5, September 1992
831
CASPER ET AL.
1957) is a standardized 480-item, true/false self-report scale, describing scale scores for 18 descriptive personality attributes. Factor analysis of the CPI results in three major dimensions, called person orientation, value orientation, and flexibility/control. A third scale (Reid and Ware, 1974) was a newer 34-item questionnaire of personal beliefs, a modified version of Rotter's scale (Rotter, 1966) of internal versus external control expectancies. This scale has three factors, a factor describing social political influence or social systems control, a factor in which luck or fate is believed to influence one's life called fatalism, and internal control. This scale was included to test Bruch's (Bruch, 1980) hypothesis that anorexia nervosa patients initiated actions primarily in response to external cues rather than in response to their own wishes and needs. To obtain information concerning the patients' thinking and behavior about weight and eating, the Eating Attitudes Test (EAT) (Garner and Garfinkel, 1979), an 40-item selfreport questionnaire, and the Eating Disorders Inventory (EDI) (Garner et aI., 1983) were used. For evaluating psychopathology, the Beck Depression Inventory (BDI) (Beck, 1978), a 23-item scale with statements describing dimensions of depression on a 4-point scale of severity, and the Hopkins Symptom Checklist (HCSL-58) (Derogatis et aI., 1974), composed of 58 items on a 5-point scale of severity yielding scores for five dimensions of psychopathology, were administered. Additionally, the Minnesota Multiphasic Personality Inventory (MMPI) (Hathaway and McKinley, 1951) profiles were compared. Patients were weighed in the fasting state each morning after voiding. Each patient wore a gown. Body weights reported are those taken on the same day as the personality measures were completed. Control subjects were clothed, but removed shoes or jackets before weighing on the same calibrated balance scale. Body weight was calculated as percentage of ideal weight for age and height (Hamill et aI., 1979, National Center for Health Statistics Growth Charts) for adolescents or based on the Metropolitan Life Insurance Tables (1964) for adults and expressed as body mass index (kg/m'). Statistical Analysis
Comparisons on the demographic variables between patients and controls were calculated using one-way analysis of variance (ANOVA). We used MANOV A for calculating differences for each questionnaire and for each set of derived higher order factors. Pairwise group comparisons were performed only for those subscales with significant univariate group differences from questionnaires exhibiting significant multivariate group differences. All pairwise group comparisons were performed using Fisher's protected least significant difference (LSD) procedure (Kirk, 1982). With equal levels of Type I errors, several studies reveal the enhanced power of the LSD procedure, relative to other post-hoc tests (Carmer and Swanson, 1973; Bernhardson, 1975). Because the literature suggests that even mild depression can influence personality dimensions (Liebowitz et aI., 1979) and because it is conceivable that abnormal eating attitudes may interact with certain personality traits, we performed multivariate analyses of covariance (MANCO VA) 832
to determine which group differences remained significant on the personality scales and the MMPI once age, total EAT scores, and total Beck Depression scores were introduced into the model. For these MANCOV As, the same procedure outlined above for the MANOVAs was followed. Namely, the univariate tests (of group effects) associated with each subscale of a questionnaire were examined only in the presence of a significant multivariate group effect for the questionnaire as a whole. Then, the Fisher LSD procedure was used to examine all pairwise group comparisons, according to the method outlined in Huitema (1980). Finally, duration of illness was also considered as a potential covariate, however, it was not used because it did not pertain to the controls, and for the patients, was highly associated with age (correlation = 0.70, p < 0.001), a variable which was being used as a covariate. To construct the higher order factor scores, we used the loadings reported by Nichols and Schnell (1963) for the CPI and the loadings reported by Tellegen (1985) for the MPQ. We used the coefficients reported by Cloninger (1987) to construct his three dimensions from the MPQ: harm avoidance, reward dependence, and novelty seeking. Results THE DEMOGRAPHIC AND CLINICAL DESCRIPTION OF THE GROUPS STUDIED
Table 1 shows the clinical and demographic characteristics for the sample. Restricting anorexia nervosa patients and patients with bulimia nervosa were comparable in age. Both were younger than the bulimic anorexia nervosa patients (ANOVA F = 3.31; P < 0.03). Because of these age differences, age was introduced as a covariate. The average duration of the illness was similar for the groups, between 2 and 3 years, with a wide range in the length of illness, from 6 months to 13 years. Ideal body weights and body mass indices showed significant between group differences (ANOVA F = 14.8; p < 0.001). Body weights were lowest in restricting anorexia nervosa patients, lower than normal in bulimic anorexia nervosa patients and within the normal range for bulimia nervosa patients. The scores obtained from the Beck Depression Inventory suggested clinical depression of about equal severity in each patient group, whereas control subjects were free of depressive symptoms (ANOVA F 19.3; p < 0.001). EATING ATTITUDES AND BEHAVIOR
Table 1 indicates that patients held markedly abnormal attitudes and behavior toward weight and food as measured on the EAT, in contrast to the control subjects who showed no weight or food preoccupation (MANOV A F = 4.22; df = 24.171; p < 0.001). Table 2 presents the EDI scores and the group comparisons. All patient groups displayed significant symptomatology on each subscale compared to normal controls. As would be expected, bulimic behavior was absent in restricting anorexia nervosa patients and bulimic patients reported more bulimic behavior than either restricting anorexia nervosa patients or controls. Post-hoc comparisons revealed no differences between restricting and J. Am. Acad. Child Adolesc. Psychiatry, 31:5, September 1992
PERSONALITY DIMENSIONS IN EATING DISORDERS TABLE
2. Comparisons of Eating Disorders Inventory and Hopkins Symptom Checklist (58) Scores for Patients with Acute Anorexia Nervosa (Restricting or Bulimic Type), Bulimia Nervosa, and Control Subjects Anorexia Nervosa Restricting (N = 12) X (SD)
Eating Disorders Inventory Drive for thinness Interoceptive awareness Bulimia Body dissatisfaction Ineffectiveness Maturity fears Perfectionism Distrust
Bulimic Anorexia Nervosa (N = 19) X (SD)
Bulimia Nervosa (N= 19) X (SD)
Control Subjects (N = 19) X (SD)
ANOVA P
