Regular Article INTRODUCTION

Transcription

1 Psychiatry and Clinical Neurosciences (2002), 56, Regular Article Psychological characteristics of eating disorders as evidenced by the combined administration of questionnaires and two projective methods: the Tree Drawing Test (Baum Test) and the Sentence Completion Test ICHIRO MIZUTA, md, YOICHI INOUE, md, phd, TOMOKO FUKUNAGA, phd, RYOHEI ISHI, md, phd, ASAO OGAWA, md AND MASATOSHI TAKEDA, md, phd Psychiatry, Department of Clinical Neuroscience, Osaka University Graduate School of Medicine, Osaka Japan Abstract The objective of this study is to examine psychological/psychopathological characteristics of eating disorders and their subtypes through a combined administration of questionnaires and projective tests. Three questionnaires (Eating Disorder Inventory 2, Social Adaptation Scale, Southern California University Eating Disorder Inventory Revised) and two projective tests (the Tree Drawing Test [TDT, Baum Test], and the Sentence Completion Test [SCT]) were administered to 126 female patients between the ages of 15 and 30 years, with eating disorders according to DSM-IV criteria at our outpatient clinic, and to 54 sex- and age-matched control subjects. The purging subtypes of eating disorders (anorexia nervosa binge-eating/purging type [ANBP] and bulimia nervosa purging type [BNP]) were clearly differentiated from the controls, both by the questionnaires and the projective tests. Compared with the controls, ANBP/BNP showed more problematic profiles across the three questionnaires, drew smaller and poorer trees in TDT to a more left location on the drawing paper, and gave fewer positive, and more negative responses in SCT. In contrast, few significant differences were found between anorexia nervosa restricting type (ANR) and the controls, and between ANBP and BNP. As a trend, however, ANR was consistently located between the controls and ANBP/BNP across the whole questionnaires and projective tests. Key words DSM-IV classification, eating disorders, projective tests, questionnaires, Sentence Completion Test, Tree Drawing Test. INTRODUCTION While many studies investigate psychological/psychopathological characteristics of eating disorders using questionnaires, only a few adopt projective methods. 1 This is noteworthy considering that they Correspondence address: Ichiro Mizuta, Psychiatry, Department of Clinical Neuroscience, Osaka University Graduate School of Medicine, D3, 2-2 Yamadaoka, Suita City, Osaka , Japan. mizuta@psy.med.osaka-u.ac.jp Received 15 February 2001; revised 28 May 2001; accepted 9 July are used almost ubiquitously in a variety of clinical settings. Questionnaires and projective methods are complementary to each other in that the former can illuminate the area where the latter cannot, and vice versa. More often than not, clinicians combine them to get more thorough understanding of patients. The relative neglect of projective tests in previous studies may be due to several factors, including a high cost with respect to time economy and staffing, as well as the essentially subjective nature of the interpretation of the results. These factors may have made it difficult to use projective tests actively in previous studies, which required relatively large numbers of

2 42 I. Mizuta et al. subjects, and assessment methods that were easy to administer and amenable to objective measurement. In other words, projective tests may have been kept at a distance from the standpoint of reliability and economy, but not their validity. Although the validity of projective tests itself has never been proven in an unequivocal, scientific way, it does not necessarily mean that they are not valid. Thus, as we feel is true in clinical practice, we may hope to obtain more valid, meaningful findings by incorporating projective methods into scientific studies than when we rely on questionnaires alone. Following this expectation, we introduced in the present study two kinds of projective tests: the Tree Drawing Test (TDT) and the Sentence Completion Test (SCT). The TDT is better known as the Baum Test ( baum meaning tree in German), and is a projective test devised by Koch in The SCT was originally devised as a means of measuring intelligence, but after around 1940, it came to be used as a projective test, and a variety of versions have been devised and developed. 3 Both TDT and SCT are noninvasive and relatively easy to administer. Through the tree drawing exercise in TDT, and a reaction to a probe in SCT, as an indirect means of expressing onself, subjects can express their latent, non-verbal state of mind projectively, with relatively little resistance. Thus, they are used to assess aspects of personality, self-image, and/or emotional states that may not be captured by questionnaires alone. Both TDT and SCT have been widely used in a variety of clinical settings, and their utility have been repeatedly demonstrated. For the present study, there are several reasons we chose TDT and SCT instead of other projective methods such as Rorschach test, House Tree Person (HTP), Kinetic Family Drawings, which are even more widely used in European and American countries. First, although no exact statistics are available, TDT and SCT seem to be at least as popular in Japan as other projective methods, if not more. Second, in spite of the fact that TDT and SCT have been extensively used in clinical settings, the number of studies, especially case-control studies, that adopted these instruments to examine eating disorders in general, and particularly, to compare their subtypes, is quite limited. Third, compared with other projective methods, TDT and SCT are relatively easy to administer and less costly. Finally, again compared to other projective methods, TDT and SCT seem more amenable to objective assessments in the interpretation of the results, and therefore, may be incorporated into scientific studies with less difficulty. METHODS Participants Out of 147 patients (143 females and four males) diagnosed with eating disorders according to DSM- IV criteria at our outpatient clinic between May 1996 and May 2000, we studied 126 female patients between the ages of 15 and 30 years. This inclusion criterion was set in order to maintain homogeneity in terms of the demographical characteristics of the subjects. The breakdown of the group according to diagnosis was as follows: 26 anorexia nervosa restricting type (ANR), 27 anorexia nervosa bingeeating/purging type (ANBP), 43 bulimia nervosa purging type (BNP), one bulimia nervosa nonpurging type (BNNP), 29 eating disorder not otherwise specified (EDNOS). Fifty-five volunteers were recruited as a control group through acquaintances of the staff of our department. One was eliminated from the study because a careful examination of questionnaires and other information raised questions about whether she was afflicted with an eating disorder. The remaining 54 volunteers served as a control group. Procedure The following questionnaires and projective methods were administered both to the patients and to the control subjects. The patients completed the test battery in a room adjoining to the consulting room at the time of their first visit to our clinic. For the controls, the test battery was mailed to their home along with the instructions. The patients were given their DSM-IV diagnosis based on the information acquired in the first interview. The questionnaires were: (i) Eating Disorder Inventory 2 (Japanese version); 4 (ii) Social Adaptation Scale (Japanese version; SAS) 5 and (iii) Southern California University Eating Disorder Inventory Revised (SCU-R). 6 The projective tests were (i) the Baum Test (The Tree Drawing Test [TDT]) 2 and the Sentence Completion Test for Adolescents (SCT). Questionnaires The EDI is a 91-item questionnaire to assess attitudes toward eating and other psychological tendencies characteristic of eating disorders, and consists of 11 factors: desire for thinness, bulimia, body dissatisfaction, ineffectiveness, perfectionism, interpersonal distrust, interoceptive awareness, maturity fears, asceticism, impulse regulation and social insecurity. In most cases, a higher score for any factor

3 Psychological characteristics of eating disorders 43 implies a stronger tendency of that factor. Exceptions are interoceptive awareness and impulse regulation, for which higher scores imply weaker tendencies. The SAS is a questionnaire for assessing a global level of social adaptation; a higher score indicating better social adaptation. The SCU-R is our revised version of the Southern California University Eating Disorder Inventory, which asks about a wide variety of areas including present and past history of the illness, medical conditions, eating-related behavior, and psychological problems associated with risk factors for eating disorders. Tree Drawing Test There are several ways to administer the TDT. In the present study, we used a sheet of A4-sized drawing paper and a 4B pencil, and gave the following instruction: Draw a tree which bears fruits in any way you like. The drawing was analyzed in terms of the (i) occupied area (size and location of the tree (Ichitani s method)), 7 (ii) global impression (energy, control, richness in elements (Aoki s method)), 8 (iii) fractal value calculated by a computer program and (iv) assessment of details of the drawing. Figure 1 shows how we measured the occupied area. The whole sheet of paper was first divided into four areas (upper right [A]; upper left [B]; lower left [C]; lower right [D]). Then each area was subdivided into 70 small squares, each of which constituted a unit. Area [O] in the figure represented the central 60 units. The height and width of the crown and trunk were measured directly in millimeter units according to the criteria shown in Fig. 2. A five-point Likert scale was used to score the global impression of the drawing, a higher score signifying higher energy, better control, and more richness in elements. Fractal value was obtained by an ad hoc computer program (i.e. a program written specifically for computing this value). This program was originally developed to assess the complexity of the lay of the land captured by an aerial photograph. The actual calculation was carried out after preliminary treatment of the drawing by a series of picture-processing programs. We predicted that this fractal value would positively correlate with energy and richness in elements assessed by the psychologist. Finally, we looked at how details of the drawing were processed, such as the use of an eraser, the strength of the lines, repeated drawing, the twodimensional display (i.e. lack of solidity of the drawing), transparency of the parts (e.g. a branch being seen through a fruit; horizon being seen Figure 1. The Tree Drawing Test: measurement of the occupied area. Figure 2. The Tree Drawing Test: measurement of the height and width of the crown and the trunk.

4 44 I. Mizuta et al. through the trunk), slanting of the trunk (whether the tree is standing upright, or slanting to the left or the right), the way of processing the base part of the trunk, security operations (i.e. the base part of the trunk widened and/or elements added to make the tree appear to be standing securely), the existence of a horizon and/or the surface of the earth, the number and shape of branches (i.e. whether the branches are drawn as one line or two lines), and finally, the number of leaves and fruits. Assessment of global impression and details of the drawing were performed by an experienced psychologist who was blind to the membership of the subjects. Sentence Completion Test for Adolescents In SCT, the subject is asked to finish a series of incomplete sentences with only the first part given as a probe. The responses of the subject are analyzed in terms of structure or contents, or both. A number of versions of SCT and methods of their analysis have been devised according to various objectives it was to be used for. In the present study, we employed the SCT for Adolescents, the SCT version developed by Shimisu M et al. (Shimizu M et al., unpubl. data, 1967). This version of SCT consists of 40 items covering six major areas of life: family, sex, personal relationships, school, self-concept, and anxiety-coping. The responses were analyzed item by item in terms of their contents and put into six categories ( positive, negative, neutral, ambivalent, avoidant, blank ). As in the TDT, the analysis was performed by an experienced psychologist who was blind to the subjects membership. Statistical analysis The results were analyzed in terms of the comparisons among the controls and the three subtypes of patients (ANR, ANBP, BNP). The types BNNP and EDNOS were eliminated from the comparison because there was only one patient with BNNP, and because patients with EDNOS were considered to be less homogeneous than the other four subtypes. One-way analysis of variance (anova) was conducted to analyze continuous variables. For continuous variables considered to form a group, manova was performed targeting those variables as a set. Tukey (when the assumption of homogeneity of variance was met) and Tamhane (when the assumption of homogeneity of variance was not met) methods were employed for post-hoc analysis of one-way anova and manova. For the analysis of categorical variables, c 2 tests (Fisher s exact tests) were used, and significance levels were lowered for post-hoc multiple comparisons in order to prevent chance elevations of the number of items with significant differences. Spearman s rank correlation was used to calculate correlation between the fractal value and the global impression in the TDT. RESULTS Table 1 shows the subjects mean age, height, bodyweight, per cent standard weight (ratio of the body weight to standard weight, which was defined as the body mass index [BMI] of 22), and EDI and SAS scores. In terms of height, there were no significant differences among the four groups (i.e. the controls and the three subgroups of the patients). Although there was a significant difference of age between ANR and ANBP, the difference was less than 3 years (i.e vs 22.78) and did not seem to have a large impact on the following results. As expected, the body weight and per cent standard weight of ANR and ANBP were significantly lower than those of the controls and BNP. In EDI, definite and consistent differences were found between the controls and ANBP/BNP, and also between ANR and ANBP/BNP. The ANBP/BNP scored significantly higher than the controls and ANR for almost all the factor scores. In contrast, there were no significant differences between the controls and ANR, although the latter tended to score higher for most factors. No significant differences were found between ANBP and BNP except for bulimia. In SAS, ANBP/BNP scored significantly lower than the controls. Anorexia nervosa restricting type tended to score lower than the controls, but higher than ANBP/BNP, although not significantly. Table 2 lists some of the SCU-R items relating to premorbid experiences and comorbid problems. Compared with the controls, BNP had been more overweight (I) and had gone on a diet more frequently (II). Anorexia nervosa binge-eating/purging type and BNP had problem behaviors much more frequently than the controls (III). Incidentally, it was confirmed, on further inquiry, that most of the violence here (III-c) was characterized as family violence. In terms of unwanted sexual experiences (IV), no significant difference was found among the four groups. Bulimia nervosa purging type had separation/loss experiences with people far more frequently than the controls (V). The relationships with their parents in childhood were significantly worse for ANBP/BNP than for the controls. Table 3 shows the findings of the TDT. Row A shows the mean size of the tree, the mean height and

5 Table 1. Age, height, bodyweight, % standard weight, questionnaires (EDI, SAS) Controls ANR ANBP BNP Controls versus ANR versus ANBP versus No ANR ANBP BNP ANBP BNP BNP Age (3.20) (3.27) (4.04) (3.28) * ns ns ns * ns ns Height (5.46) (5.12) (6.63) (6.42) ns ns ns ns ns ns ns Bodyweight (kg) (5.99) (4.19) (5.70) (7.95) *** *** *** ns ns *** *** % Standard weight 0.90 (0.08) 0.64 (0.06) 0.66 (0.08) 0.91 (0.13) *** *** *** ns ns *** *** EDI No Total (24.79) (41.63) (53.01) (50.55) *** ns *** *** ** *** ns Desire for Thinness 5.21 (4.72) 7.32 (5.80) (6.81) (5.27) *** ns *** *** * *** ns Bulimia 1.35 (1.86) 1.54 (3.95) 8.20 (6.91) (5.98) *** ns *** *** *** *** ** Body dissatisfaction (7.42) (6.01) (6.06) (7.68) ** ns ns * ns ** ns Ineffectiveness 5.51 (4.23) (7.85) (7.73) 14.8 (7.72) *** ns *** *** ns ns ns Perfectionism 3.13 (2.75) 3.11 (3.47) 7.90 (5.34) 6.91 (4.66) *** ns ** *** ** ** ns Interpersonal distrust 3.22 (2.93) 5.17 (3.93) 8.16 (4.85) 7.33 (4.39) *** ns *** *** ns ns ns Interoceptive awareness 2.50 (3.50) 5.82 (5.59) (8.72) (8.57) *** ns *** *** * *** ns Maturity fears 4.73 (2.39) 7.49 (5.28) (6.78) 9.68 (5.68) *** ns ** *** ns ns ns Asceticism 2.09 (2.70) 3.91 (3.53) 7.61 (5.66) 8.34 (5.57) *** ns *** *** * ** ns Impulse regulation 2.20 (3.55) 4.22 (4.83) (8.57) (8.56) *** ns *** *** * ** ns Social insecurity 4.44 (3.51) 6.96 (5.35) (5.67) (5.07) *** ns *** *** * ns ns SAS No Total (10.91) (12.06) (11.37) (9.95) *** ns *** *** ns ns ns Values shown are mean (SD). * P < 0.05; ** P < 0.01; *** P < EDI, Eating Disorders Inventory-2; SAS, Social Adaptation Scale; ANR, anorexia nervosa veswicting type; ANBP, anorexia nervosa binge-eating/purge type; BNP, bulimia nervosa-purging type; ns, not significant. Psychological characteristics of eating disorders 45

6 Table 2. SCU-R premorbid experiences and comorbid problems Controls ANR ANBP BNP Controls versus ANR versus ANBP versus No ANR ANBP BNP ANBP BNP BNP I. a Overweight (%) Ages 0 to ns Ages 7 to *** ns ns *** ns ns ns Age *** ns ns *** ns ** ns Age ns II. a Diet (%) Ages 0 to ns Ages 7 to ns Age ** ns ns ** ns ns ns Age ** ns ns ** ns ns ns III. a Problem Behaviors (%) a. School Refusal *** ns ** *** ns * ns b. Problem Behaviors at School * ns ns ns ns ns ns c. Violence *** ns ** *** ns ns ns d. Suicidal Idea *** ns *** *** ns ns ns e. Self Mutilation/Suicide Attempt *** ns *** *** ns ns ns f. Shoplifting/Stealing * ns * ns ns ns ns g. Alcohol Misuse *** ns ns ** ns ns ns h. Drug Misuse ns IV. a Unwanted Sexual Experiences (%) ns V. a Separation/Loss Experiences (%) a. Parental Separation/Divorce * b. Separation/Loss Experiences ** ns ns ** ns ns ns in Childhood VI. b Relationship Problems with Parents in Childhood a. With Mother 1.87 (0.99) 2.38 (1.20) 2.63 (1.41) 2.74 (1.24) ** ns ns * ns ns ns b. With Father 2.02 (1.00) 2.50 (1.32) 3.22 (1.28) 2.79 (1.39) ** ns * ns ns ns ns a Values shown are percentages. For initial four group comparison: * P < 0.05, ** P < 0.01; *** P < For post-hoc group comparison: * P < 0.05/6; ** P < 0.001/6; *** P < 0.001/6. b Values shown are mean (SD). Higher values mean worse relationship. * P < 0.05; ** P < 0.01; *** P < For abbreviations see Table I. Mizuta et al.

7 Table 3. The Tree Drawing Test (TDT) I: Size, location, global impression Controls ANR ANBP BNP Controls versus ANR versus ANBP versus No ANR ANBP BNP ANBP BNP BNP A Total (47.56) (48.85) (47.55) (62.77) *** ns *** *** ns ** ns Upper right (Area A) (14.53) (16.72) (17.42) (22.72) *** ns *** *** ns * ns Upper left (Area B) (15.36) (14.70) (16.05) (19.00) *** ns ** *** ns ** ns Lower left (Area C) (12.42) (10.54) (10.38) (14.76) *** ns ** *** ns * ns Lower right (Area D) (14.64) (13.61) (11.33) (14.34) *** ns *** *** ns ns ns Center (Area O) (7.91) (7.74) (9.59) (14.31) *** ns ns *** ns *** ns Height of crown (mm) (32.54) (39.36) (42.29) (54.52) *** ns ns *** ns * ns Height of trunk (mm) (35.16) (30.98) (31.66) (49.98) ns Width of crown (mm) (32.56) (36.35) (47.63) (53.95) *** ns ** *** ns ** ns Wiidth of trunk (mm) (17.77) (19.82) (15.54) (31.28) * ns ns ns ns ns ns B Upper right (A)/Total 0.31 (0.05) 0.29 (0.05) 0.27 (0.09) 0.25 (0.12) * ns ns ns ns ns ns Upper left (B)/Total 0.32 (0.07) 0.33 (0.06) 0.35 (0.07) 0.35 (0.18) ns Lower left (C)/Total 0.20 (0.05) 0.21 (0.05) 0.23 (0.11) 0.23 (0.15) ns Lower right (D)/Total 0.17 (0.07) 0.16 (0.07) 0.14 (0.07) 0.17 (0.15) ns Center (O)/Total 0.35 (0.12) 0.42 (0.17) 0.49 (0.17) 0.55 (0.27) *** ns ** *** ns ns ns (Left minus right)/total 0.04 (0.12) 0.10 (0.13) 0.17 (0.20) 0.15 (0.26) ** ns * ns ns ns ns (Upper minus lower)/total 0.26 (0.20) 0.25 (0.17) 0.25 (0.25) 0.20 (0.48) ns Width of crown (left)/(total) 0.48 (0.08) 0.50 (0.04) 0.50 (0.04) 0.53 (0.13) ns Height (crown)/(trunk) 1.64 (0.76) 1.97 (1.35) 1.71 (1.02) 1.38 (0.69) ns C Energy 3.50 (0.67) 3.20 (0.76) 2.87 (0.92) 2.60 (0.67) *** ns ** *** ns ** ns Control 2.50 (0.73) 2.48 (0.77) 2.74 (1.01) 2.40 (0.84) ns Richness in elements 3.31 (0.88) 2.96 (0.89) 2.96 (1.07) 2.73 (0.85) * ns ns * ns ns ns D Fractal 1.39 (0.16) 1.31 (0.10) 1.29 (0.11) 1.28 (0.15) ** ns * ** ns ns ns Values shown are mean (SD). * P < 0.05; ** P < 0.01; ***P < For abbreviations see Table 1. Psychological characteristics of eating disorders 47

8 48 I. Mizuta et al. width of the crown and the trunk. Compared with the controls, the size of the tree was significantly smaller for ANBP/BNP, both in total and each area (A, B, C, D, O). The height and width of the crown were also smaller for ANBP/BNP than the controls, all significantly, except for the height of the crown for ANBP. Comared with BNP, the ANR group drew trees significantly larger in size and the height and width of the crown. Although no significant differences were found between ANR and the controls, or between ANR and ANBP, there were consistent trends of ANR being located between the controls and ANBP/BNP. Row B shows the location of the tree, which is measured as the proportion of the tree size of the upper right (A), the upper left (B), the lower left (C), the lower right (D), the centre (O), left minus right ((B + C) (A + D)), upper minus lower ((A + B) (C + D)) to the total, the proportion of the left to the total width of the crown, and the proportion of the height of the crown to the trunk. The trees drawn by ANBP/BNP were located or tended to be located more to the left and to the centre than those of the controls. As in Row A, there were trends of ANR being positioned between the controls and ANBP/ BNP. No significant differences were found between ANBP and BNP. Row C shows the scores of global impression (i.e. energy, control, richness in elements). The ANBP/ BNP groups drew or tended to draw trees less energetically than those of the controls and ANR. The drawings of BNP were poorer in elements than those of the controls. Row D shows the fractal value. The fractal values of ANBP/BNP were significantly lower than those of the controls. The ANR group tended to be located between the controls and ANBP/BNP, although not significantly. Table 4 shows correlation between the fractal value calculated by computer and the global impression scored by an experienced psychologist who was blind to the membership of the subjects. As predicted, the fractal value correlated highly and Table 4. The Tree Drawing Test (TDT) II: Correlation between the fractal value and global impression Richness in Fractal Energy Control Elements Fractal 0.69* * Energy * Control 0.27* * P < positively with the energy (0.69) and the richness in elements (0.69). In terms of the details of the drawing, five out of 14 items showed significant differences. Three of them were explained by the differences between BNP and the controls. Compared with the controls, the drawings of BNP were more likely to be open at the base of the trunk, to lack security operations and horizons, and to bear fewer fruits. The remainig two differences were explained by the differences between ANR and the controls, and between BNP and ANBP, respectively. The ANR group drew one line only branches more frequently than the controls, and the drawings of BNP were more likely to lack horizons than those of ANBP. To summarize the findings of TDT, ANBP/BNP and the controls were clearly different in terms of the size, location, global impression, fractal value, and details of the drawing. Compared with the controls, the drawings of ANBP/BNP were smaller, located more to the left and to the centre, and were less energetic and poorer in elements and lower in fractal value. Also, the drawings of BNP were more likely than the controls to be open at the base of the trunk, lack security operations and horizons, and bear fewer fruits. There were consistent trends of ANR being located between the controls and ANBP/BNP. Finally, as predicted, the fractal value calculated by computer program correlated highly and positively with the energy and the richness in elements assessed by an experienced psychologist who was blind to the subjects membership. Table 5 shows the results of SCT. Across many factors, ANBP/BNP responded less positively, and more negatively than the controls. As in the questionnaires and TDT, there were consistent trends of ANR being located between the controls and ANBP/BNP. The only exceptions for them were blank responses, which were most frequently found in ANR, and for which ANBP/BNP came between the controls and ANR. DISCUSSION Major findings The present study documented that the purging subtypes of eating disorders by DSM-IV diagnostic criteria (ANBP, ANR) were clearly differentiated from the controls, both by the questionnaires (EDI, SAS, SCU-R) and the projective tests (TDT, SCT). In contrast, few significant differences were found between ANR and the controls, and between ANBP and BNP. As a trend, however, ANR was consistently

9 Table 5. The Sentence Completion Test for Adolescents (SCT) Controls ANR ANBP BNP Controls versus ANR versus ANBP versus No ANR ANBP BNP ANBP BNP BNP Total Positive (4.68) (5.21) (4.58) (4.24) *** ** *** *** ns ns ns Negative 8.44 (2.45) 9.50 (5.05) (3.71) (4.07) *** ns *** *** ns ns ns Neutral 3.98 (2.32) 3.12 (2.78) 3.63 (2.32) 3.08 (2.35) ns Ambivalent 7.24 (3.66) 5.85 (3.70) 6.04 (3.53) 7.74 (4.25) ns Avoidant 1.59 (1.55) 2.12 (2.05) 2.42 (2.67) 2.66 (3.31) ns Blank 0.35 (1.42) 4.62 (6.08) 1.75 (2.64) 2.5 (4.55) *** ** ns * ns ns ns I Family Positive 3.43 (1.85) 3.00 (1.55) 2.79 (1.89) 2.24 (1.75) * ns ns ** ns ns ns Negative 0.94 (1.12) 0.96 (1.34) 1.79 (1.79) 1.50 (1.47) * ns ns ns ns ns ns Neutral 0.89 (1.04) 1.00 (1.23) 0.96 (0.95) 0.74 (0.95) ns Ambivalent 1.59 (1.38) 1.62 (1.17) 1.13 (1.42) 2.13 (1.56) ns Avoidant 0.06 (0.23) 0.08 (0.27) 0.21 (0.51) 0.21 (0.53) ns Blank 0.09 (0.45) 0.35 (0.75) 0.13 (0.34) 0.18 (0.56) ns II Sex a Positive 3.13 (0.83) 2.81 (0.80) 2.96 (0.86) 2.68 (0.90) Negative 0.06 (0.23) 0.19 (0.40) 0.21 (0.41) 0.21 (0.47) Neutral 0.24 (0.47) 0.08 (0.39) 0.17 (0.38) 0.08 (0.27) Ambivalent 0.44 (0.60) 0.35 (0.49) 0.29 (0.69) 0.55 (0.50) Avoidant 0.11 (0.42) 0.27 (0.53) 0.25 (0.44) 0.32 (0.66) Blank 0.02 (0.14) 0.31 (0.55) 0.13 (0.45) 0.16 (0.44) III Positive 1.43 (0.86) 1.42 (0.86) 0.96 (0.75) 0.95 (0.84) * ns ns * ns ns ns Personal Negative 0.35 (0.52) 0.62 (0.75) 0.83 (0.82) 0.74 (0.83) * ns ns ns ns ns ns Relationshps Neutral 0.43 (0.66) 0.35 (0.63) 0.50 (0.66) 0.47 (0.65) ns Ambivalent 1.57 (0.84) 0.81 (0.85) 1.33 (0.70) 1.18 (0.90) ** *** ns ns ns ns ns Avoidant 0.17 (0.38) 0.38 (0.70) 0.08 (0.28) 0.29 (0.52) ns Blank 0.06 (0.30) 0.42 (0.90) 0.29 (0.46) 0.37 (0.82) * ns ns ns ns ns ns IV School Positive 1.07 (0.61) 0.77 (0.71) 0.75 (0.79) 0.39 (0.68) *** ns ns *** ns ns ns Negative 0.33 (0.51) 0.19 (0.49) 0.50 (0.59) 0.66 (0.85) * ns ns ns ns * ns Neutral 0.15 (0.36) 0.27 (0.60) 0.13 (0.34) 0.16 (0.37) ns Ambivalent 0.35 (0.52) 0.46 (0.65) 0.33 (0.56) 0.45 (0.69) ns Avoidant 0.06 (0.23) 0.12 (0.33) 0.00 (0.00) 0.13 (0.41) ns Blank 0.04 (0.19) 0.19 (0.40) 0.29 (0.69) 0.21 (0.47) ns V Positive 7.35 (1.78) 5.46 (2.49) 5.00 (2.02) 4.66 (2.07) *** *** *** *** ns ns ns Self-Concept Negative 3.11 (1.34) 4.04 (2.22) 4.71 (1.71) 4.71 (1.87) *** ns ** *** ns ns ns Neutral 0.61 (0.71) 0.23 (0.43) 0.50 (0.72) 0.53 (0.73) ns Ambivalent 2.11 (1.51) 1.77 (1.37) 1.79 (1.25) 2.05 (1.66) ns Avoidant 0.78 (0.96) 0.73 (1.15) 1.42 (1.69) 1.21 (1.61) ns Blank 0.04 (0.19) 1.77 (2.37) 0.58 (1.21) 0.84 (1.82) *** ** ns ns ns ns ns VI Anxiety Positive 1.98 (1.02) 1.35 (1.06) 1.42 (1.25) 1.26 (0.95) ** ns ns ** ns ns ns Coping Negative 3.65 (0.97) 3.5 (1.61) 4.25 (0.99) 4.03 (1.17) ns Neutral 1.67 (0.87) 1.19 (1.02) 1.38 (0.82) 1.11 (1.01) ns Ambivalent 1.17 (1.06) 0.85 (1.08) 1.17 (1.13) 1.37 (1.17) ns Avoidant 0.43 (0.66) 0.54 (0.95) 0.46 (0.88) 0.50 (0.76) ns Blank 0.11 (0.46) 1.58 (1.94) 0.33 (0.56) 0.74 (1.35) *** ** ns ns * ns ns Values shown are mean (SD). * P < 0.05; ** P < 0.01; *** P < a The results of comparisons are not shown in the table since manova revealed that this factor was not significant. For abbreviations see Table 1. Psychological characteristics of eating disorders 49

10 50 I. Mizuta et al. located between the controls and ANBP/BNP across the whole questionnaires and projective tests. The only exceptions for them were one line only branches in TDT and blank responses in SCT, which were most frequently found in ANR, and for which ANBP/BNP came between the controls and ANR. A brief review of previous studies Questionnaires In terms of the questionnaires, our findings are generally in agreement with most of the previous studies (e.g. see references 4, 6, 9 36). However, we will not discuss this issue further here since there are too many studies to be reviewed in this short section. Tree Drawing Test In spite of its wide use in clinical settings, only a limited number of studies have adopted TDT to examine psychological characteristics of eating disorders in general, or their subtypes This is especially true of case-control studies. Nonetheless, available studies indicate that patients with eating disorders tend to draw a smaller than average tree, in both the total size and the width of the trunk. Also, among the subgroups of patients, the anorexic subtype was found to draw a smaller tree than did the bulimic subtype. In terms of the location of the tree, patients trees tended to be located in the left area of the paper. Furthermore, the drawing by patients with eating disorders tended to lack horizon and security operations, and this tendency was more characteristic of anorexics than bulimics. Sentence Completion Test Similarly, the number of studies adopting SCT to examine eating disorders in general, and to compare their subtypes, is limited In reviewing these studies, a common trend emerges that patients were likely to respond in more negative and anxious ways than controls, in diverse areas including sex, maturation, affection, relations with parents and other people, appearances, talents, and internal emotions. They tended to attribute the cause of negativity and anxiety either to their concrete bodily features or inner problems, rather than to the environment. No study exists that specifically compares SCT among the DSM subgroups of eating disorders. In a related study comparing prepubescent patients with eating disorders in terms of their severity of symptoms and courses, the milder group responded consistently in a positive and superficial manner, while the more severe group revealed more of their inner states and stronger negative feelings toward their parents, especially toward the father. Another study compared stable and unstable types of eating disorders based on Selvini-Palazzoli s classification 45 and found that although both groups responded to SCT in a negative way, they were different in the area where they were particularly negative; that is, the stable group (i.e. with anorectic symptoms throughout the course of the illness) and the unstable group (i.e. with alternating anorectic and bulimic symptoms) tended to be especially negative about their outer (e.g. appearances and talents) and inner (e.g. personality, emotions) features, respectively. Possible interpretations of the findings of the present study, issues this study has left unresolved, and suggestions for future research Tree Drawing Test According to the calligraphy, cross-shape and space symbol theories, which are considered to be the representative interpretation theories of TDT, the size of the tree and the width of the trunk symbolize an amount of self-expression and psychic energy directed at the psychological space in which one lives and a sense of self-esteem, desire for self-expansion, and/or a level of psychic activity, while the location of a tree symbolizes how one perceives and relates to the psychological space and time in which one lives. 2,7,46 Thus, a tree of small size and narrow trunk, located on the left side of the paper reflects anxiety, low selfesteem, inhibition, inferiority complex, depression, helplessness, withdrawal, regression, dependency, and preoccupation with the past, introversion, passivity, withdrawal, respectively. Also, lack of horizon and security operations is related to psychic instability and difficulty in impulse control and reality coping. This type of tree is often found in the drawings of patients with depression and chronic schizophrenia The fact that the patients in the present study drew this type of tree more typically than the controls may suggest that the above-mentioned quality characterizes the psychology of patients with eating disorders as well. This is quite consistent with the findings of the previous studies, and also with our clinical impression of this group of patients. In terms of the differences between ANR and ANBP/BNP, however, the findings of the present study contradicted those of the previous studies; that is, while this type of tree was more characteristic of ANBP/BNP than ANR in our study, the opposite was

11 Psychological characteristics of eating disorders 51 true for the previous studies. One possible explanation for this disagreement may be that the previous studies differentiate those whose subdiagnosis changed during the course of illness and those who did not. Our diagnosis, however, was made based on the present symptoms at the time of the first visit. If we were to distinguish among the subjects according to the change of symptoms and diagnosis during the course of illness, different results might be obtained. In contrast, many studies using questionnaires and interviews indicate that depression and anxiety are more common in bulimics, or binge/purgers than in restricters. 6,12 The results of the present study are more in concordance with these findings than are those of the previous studies. However, given the dearth of studies examining this issue, we would rather defer any definite conclusion and merely suggest the need for further research on this matter. A difference between ANR and ANBP/BNP In the present study, ANR was consistently positioned closer to the controls across the whole questionnaires and the projective tests, and in appearance, looked milder in psychopathology than ANBP/BNP. The significance of this finding remains unclear. Many previous studies have suggested that ANR may be at least as severe, or more severe than the other groups of eating disorders, in terms of a variety of variables such as clinical symptoms, course and prognosis. The fact that our results pointed to the opposite may simply indicate that the questionnaires and the projective tests adopted in our study failed to capture adequately the severity of ANR. In this respect, it is often pointed out that the patients who do not binge and purge tend to deny the severity of their illness and physical condition more adamantly than the patients who binge and purge. If we assume that a tendency toward intense denial by ANR patients extends beyond the simple denial of the severity of the illness and into the patient s entire psychological makeup, our results may be a quite accurate reflection of this pervasive trend to denial in ANR. If so, it follows that the development of assessment methods, which can capture this psychological mechanism of denial in more direct ways, will be crucial in future research. Promising routes to this objective may be already shown in the present study, as in the number of blank responses in SCT (i.e. frequent blank responses in ANR), the details in TDT (i.e. frequent one line only branches in ANR). One way to check the validity of this denial hypothesis may be to examine these factors in connection with subtype shifts in the course of illness. Fractal method The significance of the fractal method for TDT may merit further pursuit. As mentioned, one reason that projective tests have not been widely used in previous studies of eating disorders may be attributed to the essentially subjective nature of the interpretation of the results. The fractal method, as an objective and quantifiable method of assessment, may play an important role in including and making full use of potential merits of TDT in future studies. Future study of eating disorders will benefit from clarifying the significance of fractal value (i.e. how can it be used as an index, and of what), and from discovering and developing other objective and quantifiable indexes of this and other projective tests. Other issues There are several limitations to the present study. First, the validity of the projective tests we adopted has never been proven or confirmed in an unequivocal, scientific way. The fact that, in this and other studies, it was possible to distinguish between the controls and patients with eating disorders in general, or among their subtypes lends some support to their validity as an instrument to assess psychological/psychopathological features of eating disorders. Still, it is clear that more studies are needed to confirm or reject this issue. Second, the fact that the present study is cross-sectional by design precludes examining the significance of our findings in relation to subtype shifts in the course of the illness. It is well known that a considerable proportion of patients with eating disorders show subtype shifts in the course of the illness. It is the task of future research to examine whether these stable and unstable groups may be differentiated by questionnaires and projective tests, such as the ones adopted here. Third, there is a confounding problem that the patients and the controls in our study were administered the test battery at different places; that is, the adjoining room to the consulting room in our clinic in the former, and at the home in the latter. The differences between the patients and the controls may be partly a reflection of this condition. For instance, the patients may have drawn a smaller tree in TDT than the controls because the former were less relaxed and more strained than the latter. Despite such limitations, we would assert that our findings at least suggest that patients with eating disorders and healthy controls may have very different psychological/psychopathological characteristics; that among the patients, ANR and ANBP/BNP may have

12 52 I. Mizuta et al. rather different, while ANBP and BNP may have rather similar psychological/psychopathological characteristics; and finally, that the projective tests we adopted (TDT, SCT) may be promising as old-new tools to deepen our understanding of the psychological/ psychopathological characteristics of eating disorders. ACKNOWLEDGMENTS The authors wish to thank Dr T. Tachi, Department of Psychiatry, Tokai University, and his colleagues, for permission to use Japanese version of EDI-2, and Drs T. Nakao and T. Kitamura, National Center of Neurology and Psychiatry, for permission to use Japanese version of SAS. The authors also wish to thank Dr E. Koga, Biwako Hospital, Dr K. Nakao and the members of the Study Group of Psychopathology at Psychiatry, Department of Neuroscience, Osaka University Graduate School of Medicine, for their precious comments and advice in preparing this manuscript. Finally, the authors wish to thank Dr Y. Kano, for his invaluable advice and help with statistical analysis. REFERENCES 1. Small AC. The contribution of psychodiagnostic test results toward understanding anorexia nervosa. Int. J. Eat. Disord. 1984; 3: Koch C. The Tree Test: the Tree Drawing Test as an Aid in Psychodiagnosis. (Hayashi K, Kuniyoshi M, Ichitani T. Baum Test Jumokuga ni yoru Jinkaku-Shindanho. Nihon Bunka-Kagakusha, Tokyo, 1970 [Japanese edition]). 3. Tsuji S, Fujii H, Yoshida M. SCT. In: Kaketa K, Tsujioka Y, Kataguchi Y, Yamaoka K, Itoh R, Tamai S, Ozawa M, Tsuji S, Fujii H, Yoshida Y (eds). Lectures on Abnormal Psychology, Vol. 2. Psychological Testing. Misuzu-shobo, 1966: Garner DM. Eating Disorder Inventory-2. Psychological Assessment Resources Inc, Odessa, Weissman MM, Bothwell S. Assessment of social adjustment by patient self-report. Arch. Gen. Psychiatry 1976; 33: Powers PS, Fernandez RC. Current Treatment of Anorexia Nervosa and Bulimia. Karger, Basel, Ichitani T, Kobayashi T, Tsuda K et al. A developmental research on the vicissitudes of the Baum Test through the life cycle (II) The relationship between the use of space and aging from middle to old age. Bull. Kyoto University Education 1987; 71: (in Japanese). 8. Aoki K. Baum Test. Rinsyo Byoga Kenkyu 1986; 1: (in Japanese). 9. Wonderlich SA. Personality and eating disorders. In: Brownell KD, Fairbairn CG (eds). Eating Disorders and Obesity: A Comprehensive Handbook. The Guilford Press, New York, 1995; Bruch H. The Golden Cage. Harvard University Press, Cambridge, Rosen AM, Murkofsky CA, Steckler NM, Skolnick NJ. A comparison of psychological and depressive symptoms among restricting anorexic, bulimic anorexic, and normal-weight bulimic patients. Int. J. Eat. Disord. 1989; 8: DaCosta M, Halmi KA. Subtyping anorexia nervosa. In: Widiger TA, Frances AJ, Pincus HA, Ross R, First MB, Davis W (eds). DSM-IV Sourcebook, Vol. 3. The American Psychiatric Press, Washington DC, 1997; Geist R, Davis R, Heinmaa M. Binge/purge symptoms and comorbidity in adolescents with eating disorders. Can. J. Psychiatry 1998; 43: Mickalide AD, Andersen AE. Subgroups of anorexia nervosa and bulimia: Validity and utility. J. Psychiatr. Res. 1985; 19: Cachelin FM, Maher BA. Restricters who purge. Implications of purging behavior for psychopathology and classification of anorexia nervosa. Eating Disorders: J. Treatment Prevention 1998; 6: Sunday SR, Halmi KA, Werdann L, Levey C. Comparison of body size estimation and eating disorder inventory scores in anorexia and bulimia patients with obese, and restrained and unrestrained controls. Int. J. Eat. Disord. 1992; 11: Cooper Z. The development and maintenance of eating disorders. In: Brownell KD, Fairburn CG (eds). Eating Disorders and Obesity: A Comprehensive Handbook. The Guilford Press, New York, 1995; Mizuta I. School refusal and psychiatric disorders: When to consider consulting a psychiatrist. J. Counseling Osaka Private High Schools. 1994; 6: (in Japanese). 19. Suzuki Y. Surviving a Long-Drawn-Out War with Anorexia and Bulimia. Joshi Eiyo Daigaku Shuppan, Tokyo, 1997 (in Japanese). 20. Minuchin S, Rosman BL, Baker L. Psychosomatic Families: Anorexia Nervosa in Context. Harvard University Press, Cambridge, Zerbe KJ. The Body Betrayed: Women, Eating Disorders, and Treatment. American Psychiatric Press, Washington DC, Baum A, Goldner EM. The relationship between stealing and eating disorders: A review. Harv. Rev. Psychiatry 1995; 3: Norton KR, Crisp AH, Bhat AV. Why do some anorexics steal? Personal, social, and illness factors. J. Psychiatr. Res. 1985; 19: Lee S. The heterogeneity of stealing behaviors in Chinese patients with anorexia nervosa in Hong Kong. J. Nerv. Ment. Dis. 1994; 182: Ziolko HU. Bulimia and kleptomania: Psychodynamics of compulsive eating and stealing. In: Schwartz HJ (ed.). Bulimia: Psychoanalytic Treatment and Theory. International University Press, Madison, 1988;

13 Psychological characteristics of eating disorders McElroy SL, Hudson JI, Pope HG, Keck PE. Kleptomania: Clinical characteristics and associated psychopathology. Psychol. Med. 1991; 21: Christenson GA, Faber RJ, dezwaan M et al. Compulsive buying: Descriptive characteristics and psychiatric comorbidity. J. Clin. Psychiatry 1994; 55: Vanderlinden J, Vandereycken W. Trauma, Dissociation and Impulse Dyscontrol in Eating Disorders. Brunner/Mazel, Bristol, Wilson GT. Eating disorders and addictive disorders. In: Brownell KD, Fairburn CG (eds). Eating Disorders and Obesity: A Comprehensive Handbook. The Guilford Press, New York, 1995; Walsh BW, Rosen PM. Self-Mutilation. Guilford Press, New York, Favazza AR, DeRosear L, Conterio K. Self-mutilation and eating disorders. Suicide Life-Threatening Behav. 1989; 19: Garfinkel PE. Classification and diagnosis of eating disorders. In: Brownell KD, Fairburn CG (eds). Eating Disorders and Obesity: A Comprehensive Handbook The Guilford Press, New York, 1995; Schwartz MF, Cohn L. Sexual Abuse and Eating Disorders. Brunner/Mazel, New York, Berger D, Saito S, Ono Y et al. Dissociation and child abuse histories in an eating disorder cohort in Japan. Acta Psychiatr. Scand. 1994; 90: Nagata T, Kiriike N, Iketani T, Kawarada Y, Tanaka H. History of childhood sexual or physical abuse in Japanese patients with eating disorders: Relationship with dissociation and impulsive behaviors. Psychol. Med. 1999; 29: Fairburn CG, Cooper Z, Doll HA, Welch SL. Risk factors for anorexia nervosa: Three integrated casecontrol comparisons. Arch. Gen. Psychiatry 1999; 56: Azuma Y, Nakamura K. Eating disorder. Shinshin-Iryo 1990; 2: (in Japanese). 38. Nakamura K, Takeuchi K. Psychological characteristics of anorexia nervosa seen from the findings of the Baum Test.J. Psychometry 1987; 23: (in Japanese). 39. Akama T, Aruga C, Seki N. Psychological assessment and psychotherapy for eating disorders. Shinshin-Igaku 1994; 34: (in Japanese). 40. Kubo R, Komai S, Taniguchi K, Makihara H, Horino K. A consideration on the Baum Test of patients with eating disorders. Jpn. J. Child Adolescent Psychiatry 1997; 38: (in Japanese). 41. Osone A. A study of personality characteristics of patients with eating disorders using K-SCT. Seishinigaku 1997; 39: (in Japanese). 42. Koga E, Kobayashi T, Kanai H, Nakajima T. A study of SCT on interpersonal characteristics of patients with eating disorders. Shinshinigaku 1992; 32: (in Japanese). 43. Shirai H, Saito K, Suzuki M et al. A study of Rorschach Test and SCT on anorexia nervosa with onset before the age of twelve. Jpn. J. Child Adolescent Psychiatry 1998; 39: 67 (in Japanese). 44. Kamitani E. A comparison by SCT of an object- and a self-image between patients with eating disorders and healthy female college students. Shinririnshogakkai 1995; 14: (in Japanese). 45. Selvini-Palazzoli MP. Self-Starvation. Chaucer Publishing Co., London, Takahashi M, Takahashi Y. The Tree Drawing Test. Bunkyo Shoin, Tokyo, Hayashi K, Ichitani T. A Clinical Research on the Baum Test. Nihon-Bunka-Kagakusha, Tokyo, 1973 (in Japanese). 48. Ide M, Iwabuchi J. Brief report of size and position on tree drawing A study of clinical drawing method (1). Ann. Report Asahikawa Med. College 1991; 12: