- Email: [email protected]
An attachment insecurity model of negative affect among women seeking treatment for an eating disorder
Eating Behaviors 7 (2006) 252 – 257
An attachment insecurity model of negative affect among women seeking treatment for an eating disorder Giorgio A. Tasca a,b,c,d,*, John Kowal a,c, Louise Balfour a,b,c, Kerri Ritchie a,b, Barbara Virley a,b,1, Hany Bissada a,b,c a
c
University of Ottawa, Canada b Ottawa Hospital, Canada Ottawa Health Research Institute, Canada d Carleton University, Canada
Received 1 April 2005; received in revised form 2 September 2005; accepted 22 September 2005
Abstract The purpose of this study was to propose and test a model of attachment insecurity in a clinical sample of 268 eating disordered women. Structural relationships among attachment insecurity, BMI, perceived pressure to diet, body dissatisfaction, restrained eating, and negative affect were assessed. A heterogeneous sample of treatment seeking women with a diagnosed eating disorder completed psychometric tests prior to receiving treatment. The data were analysed using structural equation modeling. Fit indices indicated that the hypothesized model fit adequately to the data. Although cross-sectional in nature, the data suggested that attachment insecurity may lead to negative affect. As well, attachment insecurity may lead to body dissatisfaction, which in turn may lead to restrained eating among women with eating disorders. Attachment insecurity could be a possible vulnerability factor for the development of eating disorder symptoms among women. D 2005 Elsevier Ltd. All rights reserved. Keywords: Eating disorders; Attachment theory; Structural equation modeling
1. Introduction Researchers have investigated the role of negative affect and dietary restraint as common pathways to bulimic symptoms (e.g., Stice, 2001). Such models indicate that social pressure to be thin and an internalization of a thin ideal body lead to body dissatisfaction. In turn, body dissatisfaction leads to negative affect and dietary restraint, both of which result in bulimic symptoms. Longitudinal data have supported such models (see Stice, 2001 for a review). In a recent study, Duemm, Adams, and Keating (2003) added sociotropy, a measure of interpersonal dependency, to such a model. These authors found a significant relationship between sociotropy and both negative affect and ideal body internalization. Although these structural models have been tested with college women and adolescent girls (Duemm, et al., 2003; Stice, Shaw, & Nemeroff, 1998), no studies appear to have tested similar models in a clinical sample. * Corresponding author. Ottawa Hosptial — General Campus, 501 Smyth Road, Room 7300, Ottawa, Ontario, Canada, K1H 8L6. Tel.: +1 613 737 8035; fax: +1 613 737 8895. E-mail addresses: [email protected] (G.A. Tasca), [email protected] (J. Kowal), [email protected] (L. Balfour), [email protected] (K. Ritchie), [email protected] (B. Virley), [email protected] (H. Bissada). 1 Mailing address: 430 Gilmour Street, Suite 302, Ottawa, ON, Canada, K2P 0R8, Tel.: +1 613 798 3014. 1471-0153/$ - see front matter D 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.eatbeh.2005.09.004
G.A. Tasca et al. / Eating Behaviors 7 (2006) 252–257
253
The goal of this study was to assess whether attachment insecurity was related to negative affect and restrained eating in a structural model among women with a diagnosed eating disorder. Attachment insecurity (Bowlby, 1988) (i.e., attachment anxiety and attachment avoidance) was hypothesized to play a role in the development and maintenance of eating disorder symptoms, although the nature of this role remains largely unknown due to a paucity of research (Ward, Ramsey, & Treasure, 2000). In one study in this area, Ward, Ramsey, Turnbull, Benedettini, and Treasure (2000) found that individuals with eating disorders scored higher than a comparison group on measures of attachment anxiety and attachment avoidance, but no differences were found among individuals with different eating disorder diagnoses. For the current investigation, it was hypothesized that attachment insecurity would influence the development of eating disorder symptoms through negative affect. That is, attachment insecurity influences negative affect which, in turn, influences eating disordered behavior. Research has supported the relationship between negative affect and eating disorder symptoms (e.g., Stice, 2001). Avoidant and anxious attachment styles have been associated with binge eating under stress (Brennan & Shaver, 1995), and binge eating has been conceptualized as a maladaptive means of coping with stress (Wonderlich, Mitchell, Peterson, & Crow, 2001). Further, attachment avoidance was associated with dietary restraint and a lack of interoceptive awareness, both of which were associated with anorexic symptoms (Brennan & Shaver, 1995). A second path by which attachment insecurity could lead to eating disorder symptoms (e.g., dietary restraint) is through body dissatisfaction. Duemm et al. (2003) found a relationship between interpersonal dependence and ideal body internalization, which is known to be correlated with body dissatisfaction. Hence, in the current study, it was predicted that attachment insecurity, which encompasses aspects of interpersonal dependence, would be related to body dissatisfaction. A number of other directional relationships (i.e., paths) were hypothesized based on previous research with non-clinical samples (e.g., Duemm, et al., 2003; Stice, 2001). The hypothesized model tested in the present study is presented in Fig. 1. 2. Method 2.1. Participants Participants were 268 women seeking treatment at a center for eating disorders of a general hospital. Of these participants: 30 were diagnosed with Anorexia Nervosa–Restricting subtype (ANR) (mean age = 26.4 F 10.4; mean BMI = 15.3 F 1.7); 43 were diagnosed with Anorexia Nervosa–Binging or Purging subtype (ANB) (mean age = 29.8 F 10.8; mean BMI = 16.3 F 1.6); 57 were diagnosed with Bulimia Nervosa–Purging subtype (BN) (mean age = 29.0 F 8.0; mean BMI = 25.6 F 6.0); 115 were diagnosed with Binge Eating Disorder (BED) (mean age = 42.3 F 10.8; mean BMI = 40.5 F 8.7); and 23 were diagnosed with an Eating Disorder Not Otherwise Specified (EDNOS) (mean age = 26.2 F 8.6; mean BMI = 22.4 F 6.0). Most women (96%) were Caucasian, and the median family income was 40 to 49 thousand Canadian dollars. 2.2. Procedure Participants were referred by a physician for an assessment of an eating disorder and for possible treatment. Eating disorder diagnoses were based on a semi-structured interview conducted by a psychologist or psychiatrist using DSMIV (APA, 1994) criteria. To test for reliability, 10% of all assessment reports with the diagnosis removed were reviewed by an independent psychologist. Agreement between the original diagnosis and the independent review was high (Cohen’s n = .76). In all cases, the original diagnosis was retained. All participants completed questionnaires used in this study and provided informed consent prior to beginning treatment. The research was approved by the institution’s research ethics board. 2.3. Factors, variables and measures The Attachment Insecurity factor was comprised of the Anxious and Avoidant attachment scales from the Attachment Styles Questionnaire (Feeney, Noller, & Hanrahan, 1994). The Avoidant Attachment scale has 16 items (possible range 16–96), and the Anxious Attachment scale has 13 items (possible range 13–78) (J. Feeney,
254
G.A. Tasca et al. / Eating Behaviors 7 (2006) 252–257
Avoidant 0.76
Anxious
Depression
0.82
0.81
.74***
Attachment insecurity
Authority 0.70
Current 0.87
** * .59
0.77
0.11
Perceived pressure
.76***
BMI
.18*
Restrained eating -.99***
.23 **
.67
***
0.74
Drive for thinness
.34***
0.68
Restraint
Body dissatisfaction
0.98
BMI
0.88
Negative affect
.17 *
Family
Ineffectiveness
0.67
Dissatisfaction
0.86
Appearance
0.82
Weight
0.34
Fig. 1. Measurement model and structural model with standardized path coefficients. Notes. N=268. Measurement model: pb.001 for all parameters; v 2 (df=50)=125.22, pb.001, GFI=.93, CFI=.96, RMSEA=.08 (90%CI=.06,.09). Structural model: *** pb.001, ** p=.003, * p=.01; v 2 (df=55)=150.87, pb.001, GFI=.92, CFI=.95, RMSEA=.08 (90%CI=.07,.10). Elipses represent latent variables or factors, rectangles represent indicator or measured variables, small circles represent error terms of indicators (shown only for Body Dissatisfaction and Drive for Thinness), double headed arrows represent correlated error terms. Standardized path coefficients (single headed arrows between elipses) appear above arrows and indicate direction and strength of relationships between factors.
November 3, 2002, personal communication). In the current sample, coefficient alpha for the Avoidant Attachment scale was .76, and coefficient alpha for the Anxious Attachment scale was .85. Perceived social pressure to diet was assessed by three scales derived from the Diagnostic Survey of Eating Disorders–Revised (Johnson & Connors, 1987). Items were scored on a six-point Likert scale (range: 1 = never to 6 = always) and mean item scores were calculated. The items asked: bRate the degree to which any of the following people encouraged you to dietQ. The Pressure to Diet by Family of Origin scale had four items for the patient’s mother, father, sister, and brother. The Pressure to Diet by Current Relationships scale consisted of two items for the patient’s spouse/boyfriend/girlfriend and children. The Pressure to Diet by Authority scale consisted of two items for the patient’s physician and teacher/coach. Because of the small number of items in these scales, mean inter item correlation were assessed as the appropriate measure of internal consistency (Clark & Watson, 1995). Mean inter item correlations of the three scales were .49, .22, and .25, respectively, demonstrating adequate internal consistency. Body Mass Index (BMI; kg/m2) was a single indicator variable based on patient self report at the time of assessment. The correlation between self-reported and actual body weight was very high for adolescents (r = .97) (Attie & Brooks-Gunn, 1989), and for obese adults with BED (r = .99) (Masheb & Grilo, 2001). The Body Dissatisfaction factor was made up of three scales. The Body Dissatisfaction scale was used from the Eating Disorders Inventory (EDI; Garner & Olmsted, 1984). Coefficient alpha for this scale was .95. The Body Esteem – Appearance (BEA) and the Body Esteem – Weight (BEW) scales were taken from the Body Esteem Scale for Adolescents and Adults (Mendelson, Mendelson, & White, 2001). For the current sample, coefficient alphas were .87 and .85, respectively. Scores on BEA and BEW were reversed so that high scores on the Body Dissatisfaction Factor represented greater body dissatisfaction.
G.A. Tasca et al. / Eating Behaviors 7 (2006) 252–257
255
The Dietary Restraint factor consisted of two scales. The Drive for Thinness scale was taken from the EDI (Garner & Olmsted, 1984). Coefficient alpha was .95. The Restraint scale from the Eating Disorders Examination Questionnaire (Fairburn & Beglin, 1994) was also used. Coefficient alpha for this scale was .71. One of the two scales comprising the Negative Affect factor was the Center for Epidemiological StudiesDepression scale (Radloff, 1977). Coefficient alpha for the current sample was .76. The Ineffectiveness scale from the EDI (Garner & Olmsted, 1984) was also used. Coefficient alpha was .84. 2.4. Overview of structural equation modeling analyses Structural equation modeling with EQS (Bentler, 1995) was used to test the hypothesized model (Fig. 1). The data were analyzed in two steps. First, a confirmatory factor analysis (CFA) was conducted to evaluate the measurement model. This allowed for an assessment of the relationships (loadings) between indicators (observed variables) and factors (latent variables). Because the BMI factor had a single indicator, the error term was set to one minus the reliability, and this was multiplied by the indicator variance (Bollen, 1989). Because the Body Dissatisfaction and the Restrained Eating factors each had one indicator taken from the EDI, the error variances of these two scales were correlated (see Fig. 1). Second, the hypothesized directional relationships among latent variables were tested (Fig. 1). The exogenous (independent) variables were Attachment Insecurity and BMI. Endogenous (dependent) variables were Perceived Pressure to Diet, Body Dissatisfaction, Restrained Eating, and Negative Affect. A maximum likelihood method of estimation was used for all analyses. Several goodness-of-fit indices were calculated. For the comparative fit index (CFI) and the goodness-of-fit index (GFI), values greater than .90 are indicative of a well fitting model (Bentler, 1992; Bentler & Bonnett, 1980). For the root-mean-square error of approximation (RMSEA), values less than .08 represent an acceptable fit, values between .08 and .10 indicate a moderate fit, and values above .10 indicate a poor fit (Browne & Cudeck, 1993). 3. Results The following variables had significantly skewed distributions: Pressure to Diet by Family of Origin, Pressure to Diet by Current Relationships, Pressure to Diet by Authority, Body Dissatisfaction, BEA, BEW, and Drive for Thinness (all zs z 3.3). These variables were transformed using logarithmic or square root transformations, resulting in normally distributed data. The original sample (N = 271) had three multivariate outliers that were removed from the Table 1 Correlations among indicator variables, including their univariate statistics for the total eating disordered sample (N = 268) Indicator Variable 1. Attachment avoidant 2. Attachment anxiety 3. Pressure — family 4. Pressure — authority 5. Pressure — current 6. Body mass index 7. Body dissatisfaction 8. Body esteem appear 9. Body esteem weight 10. Drive for thinness 11. Dietary restraint 12. Depression 13. Ineffectiveness Mean Standard deviation Skewness Kurtosis
1
2
.62 .20 .14 .25 .27 .23 .31 .18 .31 .25 .59 .58 59.95 11.58 0.20 0.24
.20 .17 .16 .25 .31 .39 .22 .40 .29 .59 .66 54.92 10.15 0.42 0.07
3
4
5
.61 .65 .58 .17 .17 .26 .18 .29 .17 .06 2.26 1.29 0.18 1.40
.60 .48 .14 .16 .27 .18 .27 .21 .08 1.81 0.98 0.53 1.17
.67 .27 .29 .42 .13 .35 .18 .03 2.23 1.07 0.02 1.23
6
7
.29 .28 .41 .36 .54 .28 .17 29.14 12.44 0.59 0.60
.59 .59 .39 .05 .29 .39 21.11 7.02 0.65 0.84
8
9
.72 .32 .01 .38 .52 2.26 0.59 0.71 0.30
.22 .11 .24 .34 2.25 0.63 0.21 0.61
10
.52 .42 .43 13.54 5.79 0.64 0.64
11
.33 .25 4.38 1.78 0.32 0.98
12
13
.71 30.80 12.57 0.14 0.85
12.30 7.98 0.29 0.85
Item means are reported for pressure — family, pressure — authority, pressure — current, body esteem — appear, body esteem — weight, and dietary restraint. Total score means are reported for all other scales. Data for pressure — family, pressure — authority, pressure — current, body dissatisfaction, body esteem — appearance, and body esteem — weight were transformed due to univariate non-normality. The means and standard deviations are for non-transformed data. All other values are based on the transformed data.
256
G.A. Tasca et al. / Eating Behaviors 7 (2006) 252–257
subsequent analyses. This resulted in the final sample of 268. Pearson correlations among indicator variables and univariate statistics are presented in Table 1. 3.1. Confirmatory factor analysis and structural model Results of the CFA demonstrated significant factor loadings of all indicator variables onto their respective latent variables ( ps b .001) (Fig. 1). Fit indices indicated that the measurement model provided an adequate fit of the data, CFI = .96, GFI = .93, and RMSEA= .08 (90% CI = .06 to .09). Results of the structural model also appear in Fig. 1. In terms of significant direct effects, attachment insecurity was positively related to body dissatisfaction and to negative affect ( ps b .001). BMI was positively associated with perceived pressure and body dissatisfaction, but negatively related to restrained eating ( ps b .001). Perceived pressure to diet was positively associated with body dissatisfaction ( p = .003), but not to restrained eating ( p = .313). Body dissatisfaction was positively related to both restrained eating ( p b .001) and negative affect ( p = .01). Finally, restrained eating was positively related to negative affect ( p = .01). The hypothesized model adequately fit the data, CFI = .95, GFI = .92, and RMSEA= .08 (90% CI = .07 to .10). The structural model accounted for 58% of variance in perceived pressure, .64% of variance in body dissatisfaction, 66% of variance in restrained eating, and 87% of variance in negative affect. 4. Discussion Results indicated that attachment insecurity was related to body dissatisfaction and negative affect for women with a diagnosed eating disorder. These results are consistent with attachment theory as well as research demonstrating an association between attachment insecurity and negative affect (Cole-Detke & Kobak, 1996; Kobak & Sceery, 1988). The current study extended attachment research to an eating disordered sample. Specifically, these results suggested that attachment insecurity may be a vulnerability factor for the development of an eating disorder, and attachment insecurity may lead to eating disorder symptoms through its relationship with body dissatisfaction and negative affect. Consistent with previous studies (Stice et al., 1998; Stice, 2001), body dissatisfaction was directly related to negative affect, however the relationship observed in the current study was modest. The nature of the relationship among body dissatisfaction, restrained eating and negative affect suggested that body dissatisfaction might be indirectly associated with negative affect through its influence on restrained eating. It is likely that the addition of attachment insecurity as an independent variable accounted for much of the variance in the negative affect factor. Thus, attachment insecurity may have reduced the relationship between body dissatisfaction and negative affect. It is possible that the development of negative affect among women with a diagnosed eating disorder may, in part, have its origin in attachment insecurity, which is likely rooted in early negative family interactions (Ward, et al., 2000). Although a number of therapeutic interventions have targeted body dissatisfaction in the treatment of individuals with eating disorders (Fairburn, Marcus, & Wilson, 1993), the present findings suggest that attachment insecurity and problems in interpersonal relationships could be targeted to address negative affect associated with eating disorders. This is consistent with interpersonal psychotherapies in the treatment of individuals with eating disorders (Tasca et al., in press; Wilfley et al., 1993). Results of this investigation may provide a model for the processes of change among eating disordered individuals who benefit from interpersonally based therapies. This study had several limitations. First, the eating disordered sample was heterogeneous in terms of diagnoses, which likely increased error variance in the model. The tested model also assumed that underlying relationships between insecure attachment and other latent variables were similar across diagnostic categories, but this may not be the case. The dual pathway model (Stice et al., 1998) has been typically discussed in terms of bulimic symptoms, but not all patients in this study had these symptoms. Future studies could focus on more homogeneous groups of individuals with an eating disorder and could include an eating disorders symptom factor. Second, there was a high correlation between anxious and avoidant attachment scales and a single attachment insecurity factor emerged. Attachment research has demonstrated differences between anxious and avoidant attachment styles in terms of coping and negative affect (Wei, Heppner, & Mallinckrodt, 2003). Anxious and avoidant attachment styles may play differing and complementary roles in the development of eating disorder symptoms, and this could be evaluated in future studies. Third, the current study was cross-sectional in nature. Longitudinal studies examining the consistency of attachment styles over time, as well as the role played by attachment insecurity in the development of eating disorder symptoms are needed.
G.A. Tasca et al. / Eating Behaviors 7 (2006) 252–257
257
Acknowledgements Thanks go to Ann Barber, Dr. Alan MacDonald, and Dr. Amy Silverman for their help with this manuscript. This study was partially supported by a grant from the Ontario Mental Health Foundation. References American Psychiatric Association. (1994). Diagnostic and statistical manual of mental disorders (4th ed.). Washington7 Author. Attie, I., & Brooks-Gunn, J. (1989). Development of eating problems in adolescent girls: A longitudinal study. Developmental Psychology, 25, 70 – 79. Bentler, P. M. (1992). On the fit models to covariance methodology in the Bulletin. Psychological Bulletin, 112, 400 – 404. Bentler, P. M. (1995). Structural equation program manual. Encino, CA7 Multivariate Software. Bentler, P. M., & Bonnett, D. G. (1980). Significance tests and goodness of fit in the analysis of covariance structures. Social and Behavioral Sciences, 19, 16. Bollen, K. A. (1989). Structural equations with latent variables. New York7 Wiley and Sons. Bowlby, J. (1988). A secure base. New York7 Basic Books. Brennan, K. A., & Shaver, P. R. (1995). Dimensions of adult attachment, affect regulation, and romantic relationship functioning. Personality and Social Psychology Bulletin, 21, 267 – 283. Browne, M. W., & Cudeck, R. (1993). Alternative ways of assessing model fit. In K. A. Bollen, & J. S. Long (Eds.), Testing structural equation models (pp. 445 – 455). Newbury Pk, C.A7 Sage. Clark, L. A., & Watson, D. (1995). Constructing validity: Basic issues in objective scale development. Psychological Assessment, 7, 309 – 319. Cole-Detke, H., & Kobak, R. (1996). Attachment processes in eating disorders and depression. Journal of Consulting and Clinical Psychology, 64, 282. Duemm, I., Adams, G. R., & Keating, L. (2003). The addition of sociotropy to the dual pathway model of bulimia. Canadian Journal of Behavioral Sciences, 35, 281 – 291. Fairburn, C. G., & Beglin, S. J. (1994). Assessment of eating disorders: Interview or self-report questionnaire? International Journal of Eating Disorders, 16, 363 – 370. Fairburn, C. G., Marcus, M. D., & Wilson, G. T. (1993). Cognitive-behavior therapy for binge eating and bulimia nervosa: A comprehensive treatment manual. In C. G. Fairburn, & G. T. Wilson (Eds.), Binge eating: Nature, assessment and treatment (pp. 361 – 404). NY7 Guilford. Feeney, J. A., Noller, P., & Hanrahan, M. (1994). Assessing adult attachment. In M. B. Sperling, & W. H. Berman (Eds.), Attachment in adults (pp. 128 – 151). New York7 Guilford. Garner, D. M., & Olmsted, M. P. (1984). The eating disorders inventory. Odessa, FL7 Psychological Assessment Resources. Johnson, C., & Connors, M. E. (1987). The etiology and treatment of bulimia nervosa: A biopsychosocial perspective. New York7 Basic. Kobak, R. R., & Sceery, A. (1988). Attachment in late adolescence: Working models, affect regulation, and representations of self and others. Child Development, 59, 135 – 146. Masheb, R. M., & Grilo, C. M. (2001). Accuracy of self reported weight in patients with Binge Eating Disorder. International Journal of Eating Disorders, 29, 29 – 36. Mendelson, B. K., Mendelson, M. J., & White, D. R. (2001). Body-Esteem Scale for adolescents and adults. Journal of Personality Assessment, 76, 90 – 106. Radloff, L. S. (1977). The CES-D scale: A self-report depression scale for research in the general population. Applied Psychological Measurement, 1, 385 – 401. Stice, E. (2001). A prospective test of the dual-pathway model of bulimic pathology: Mediating effects of dieting and negative affect. Journal of Abnormal Psychology, 110, 124 – 135. Stice, E., Shaw, H., & Nemeroff, C. (1998). Dual pathway model of bulimia nervosa: Longitudinal support for dietary restraint and affect-regulation mechanisms. Journal of Social and Clinical Psychology, 17, 129 – 149. Tasca, G. A., Ritchie, K., Conrad, G., Balfour, L., Gayton, J., Daigle, V. & Bissada, H. (in press). Attachment scales predict outcome in a randomized controlled trial of two group therapies for Binge Eating Disorder: An aptitude by treatment interaction. Psychotherapy Research. Ward, A., Ramsay, R., & Treasure, J. (2000). Attachment research in eating disorders. British Journal of Medical Psychology, 73, 35 – 51. Ward, A., Ramsey, R., Turnbull, S., Benedettini, M., & Treasure, J. (2000). Attachment patterns in eating disorders: The past in the present. International Journal of Eating Disorders, 28, 370 – 376. Wei, M., Heppner, P. P., & Mallinckrodt, B. (2003). Perceived coping as a mediator between attachment and psychological distress: A structural equation modeling approach. Journal of Counseling Psychology, 50, 438 – 447. Wilfley, D. E., Agras, W. S., Telch, C. F., Rossiter, E., Schneider, J. A., Cole, A. G., Sifford, K., & Raeburn, S. (1993). Group cognitive-behavioral therapy and group interpersonal therapy for the nonpurging bulimic individual: A controlled comparison. Journal of Consulting and Clinical Psychology, 61, 296 – 305. Wonderlich, S. A., Mitchell, J. E., Peterson, C. B., & Crow, S. (2001). Integrative cognitive therapy for bulimic behavior. In R. H. Striegel-Moore, & L. Smolak (Eds.), Eating Disorders: Innovative directions in research and practice (pp. 173 – 195). Washington7 American Psychological Association.
An attachment insecurity model of negative affect among women seeking treatment for an eating disorder Giorgio A. Tasca a,b,c,d,*, John Kowal a,c, Louise Balfour a,b,c, Kerri Ritchie a,b, Barbara Virley a,b,1, Hany Bissada a,b,c a
c
University of Ottawa, Canada b Ottawa Hospital, Canada Ottawa Health Research Institute, Canada d Carleton University, Canada
Received 1 April 2005; received in revised form 2 September 2005; accepted 22 September 2005
Abstract The purpose of this study was to propose and test a model of attachment insecurity in a clinical sample of 268 eating disordered women. Structural relationships among attachment insecurity, BMI, perceived pressure to diet, body dissatisfaction, restrained eating, and negative affect were assessed. A heterogeneous sample of treatment seeking women with a diagnosed eating disorder completed psychometric tests prior to receiving treatment. The data were analysed using structural equation modeling. Fit indices indicated that the hypothesized model fit adequately to the data. Although cross-sectional in nature, the data suggested that attachment insecurity may lead to negative affect. As well, attachment insecurity may lead to body dissatisfaction, which in turn may lead to restrained eating among women with eating disorders. Attachment insecurity could be a possible vulnerability factor for the development of eating disorder symptoms among women. D 2005 Elsevier Ltd. All rights reserved. Keywords: Eating disorders; Attachment theory; Structural equation modeling
1. Introduction Researchers have investigated the role of negative affect and dietary restraint as common pathways to bulimic symptoms (e.g., Stice, 2001). Such models indicate that social pressure to be thin and an internalization of a thin ideal body lead to body dissatisfaction. In turn, body dissatisfaction leads to negative affect and dietary restraint, both of which result in bulimic symptoms. Longitudinal data have supported such models (see Stice, 2001 for a review). In a recent study, Duemm, Adams, and Keating (2003) added sociotropy, a measure of interpersonal dependency, to such a model. These authors found a significant relationship between sociotropy and both negative affect and ideal body internalization. Although these structural models have been tested with college women and adolescent girls (Duemm, et al., 2003; Stice, Shaw, & Nemeroff, 1998), no studies appear to have tested similar models in a clinical sample. * Corresponding author. Ottawa Hosptial — General Campus, 501 Smyth Road, Room 7300, Ottawa, Ontario, Canada, K1H 8L6. Tel.: +1 613 737 8035; fax: +1 613 737 8895. E-mail addresses: [email protected] (G.A. Tasca), [email protected] (J. Kowal), [email protected] (L. Balfour), [email protected] (K. Ritchie), [email protected] (B. Virley), [email protected] (H. Bissada). 1 Mailing address: 430 Gilmour Street, Suite 302, Ottawa, ON, Canada, K2P 0R8, Tel.: +1 613 798 3014. 1471-0153/$ - see front matter D 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.eatbeh.2005.09.004
G.A. Tasca et al. / Eating Behaviors 7 (2006) 252–257
253
The goal of this study was to assess whether attachment insecurity was related to negative affect and restrained eating in a structural model among women with a diagnosed eating disorder. Attachment insecurity (Bowlby, 1988) (i.e., attachment anxiety and attachment avoidance) was hypothesized to play a role in the development and maintenance of eating disorder symptoms, although the nature of this role remains largely unknown due to a paucity of research (Ward, Ramsey, & Treasure, 2000). In one study in this area, Ward, Ramsey, Turnbull, Benedettini, and Treasure (2000) found that individuals with eating disorders scored higher than a comparison group on measures of attachment anxiety and attachment avoidance, but no differences were found among individuals with different eating disorder diagnoses. For the current investigation, it was hypothesized that attachment insecurity would influence the development of eating disorder symptoms through negative affect. That is, attachment insecurity influences negative affect which, in turn, influences eating disordered behavior. Research has supported the relationship between negative affect and eating disorder symptoms (e.g., Stice, 2001). Avoidant and anxious attachment styles have been associated with binge eating under stress (Brennan & Shaver, 1995), and binge eating has been conceptualized as a maladaptive means of coping with stress (Wonderlich, Mitchell, Peterson, & Crow, 2001). Further, attachment avoidance was associated with dietary restraint and a lack of interoceptive awareness, both of which were associated with anorexic symptoms (Brennan & Shaver, 1995). A second path by which attachment insecurity could lead to eating disorder symptoms (e.g., dietary restraint) is through body dissatisfaction. Duemm et al. (2003) found a relationship between interpersonal dependence and ideal body internalization, which is known to be correlated with body dissatisfaction. Hence, in the current study, it was predicted that attachment insecurity, which encompasses aspects of interpersonal dependence, would be related to body dissatisfaction. A number of other directional relationships (i.e., paths) were hypothesized based on previous research with non-clinical samples (e.g., Duemm, et al., 2003; Stice, 2001). The hypothesized model tested in the present study is presented in Fig. 1. 2. Method 2.1. Participants Participants were 268 women seeking treatment at a center for eating disorders of a general hospital. Of these participants: 30 were diagnosed with Anorexia Nervosa–Restricting subtype (ANR) (mean age = 26.4 F 10.4; mean BMI = 15.3 F 1.7); 43 were diagnosed with Anorexia Nervosa–Binging or Purging subtype (ANB) (mean age = 29.8 F 10.8; mean BMI = 16.3 F 1.6); 57 were diagnosed with Bulimia Nervosa–Purging subtype (BN) (mean age = 29.0 F 8.0; mean BMI = 25.6 F 6.0); 115 were diagnosed with Binge Eating Disorder (BED) (mean age = 42.3 F 10.8; mean BMI = 40.5 F 8.7); and 23 were diagnosed with an Eating Disorder Not Otherwise Specified (EDNOS) (mean age = 26.2 F 8.6; mean BMI = 22.4 F 6.0). Most women (96%) were Caucasian, and the median family income was 40 to 49 thousand Canadian dollars. 2.2. Procedure Participants were referred by a physician for an assessment of an eating disorder and for possible treatment. Eating disorder diagnoses were based on a semi-structured interview conducted by a psychologist or psychiatrist using DSMIV (APA, 1994) criteria. To test for reliability, 10% of all assessment reports with the diagnosis removed were reviewed by an independent psychologist. Agreement between the original diagnosis and the independent review was high (Cohen’s n = .76). In all cases, the original diagnosis was retained. All participants completed questionnaires used in this study and provided informed consent prior to beginning treatment. The research was approved by the institution’s research ethics board. 2.3. Factors, variables and measures The Attachment Insecurity factor was comprised of the Anxious and Avoidant attachment scales from the Attachment Styles Questionnaire (Feeney, Noller, & Hanrahan, 1994). The Avoidant Attachment scale has 16 items (possible range 16–96), and the Anxious Attachment scale has 13 items (possible range 13–78) (J. Feeney,
254
G.A. Tasca et al. / Eating Behaviors 7 (2006) 252–257
Avoidant 0.76
Anxious
Depression
0.82
0.81
.74***
Attachment insecurity
Authority 0.70
Current 0.87
** * .59
0.77
0.11
Perceived pressure
.76***
BMI
.18*
Restrained eating -.99***
.23 **
.67
***
0.74
Drive for thinness
.34***
0.68
Restraint
Body dissatisfaction
0.98
BMI
0.88
Negative affect
.17 *
Family
Ineffectiveness
0.67
Dissatisfaction
0.86
Appearance
0.82
Weight
0.34
Fig. 1. Measurement model and structural model with standardized path coefficients. Notes. N=268. Measurement model: pb.001 for all parameters; v 2 (df=50)=125.22, pb.001, GFI=.93, CFI=.96, RMSEA=.08 (90%CI=.06,.09). Structural model: *** pb.001, ** p=.003, * p=.01; v 2 (df=55)=150.87, pb.001, GFI=.92, CFI=.95, RMSEA=.08 (90%CI=.07,.10). Elipses represent latent variables or factors, rectangles represent indicator or measured variables, small circles represent error terms of indicators (shown only for Body Dissatisfaction and Drive for Thinness), double headed arrows represent correlated error terms. Standardized path coefficients (single headed arrows between elipses) appear above arrows and indicate direction and strength of relationships between factors.
November 3, 2002, personal communication). In the current sample, coefficient alpha for the Avoidant Attachment scale was .76, and coefficient alpha for the Anxious Attachment scale was .85. Perceived social pressure to diet was assessed by three scales derived from the Diagnostic Survey of Eating Disorders–Revised (Johnson & Connors, 1987). Items were scored on a six-point Likert scale (range: 1 = never to 6 = always) and mean item scores were calculated. The items asked: bRate the degree to which any of the following people encouraged you to dietQ. The Pressure to Diet by Family of Origin scale had four items for the patient’s mother, father, sister, and brother. The Pressure to Diet by Current Relationships scale consisted of two items for the patient’s spouse/boyfriend/girlfriend and children. The Pressure to Diet by Authority scale consisted of two items for the patient’s physician and teacher/coach. Because of the small number of items in these scales, mean inter item correlation were assessed as the appropriate measure of internal consistency (Clark & Watson, 1995). Mean inter item correlations of the three scales were .49, .22, and .25, respectively, demonstrating adequate internal consistency. Body Mass Index (BMI; kg/m2) was a single indicator variable based on patient self report at the time of assessment. The correlation between self-reported and actual body weight was very high for adolescents (r = .97) (Attie & Brooks-Gunn, 1989), and for obese adults with BED (r = .99) (Masheb & Grilo, 2001). The Body Dissatisfaction factor was made up of three scales. The Body Dissatisfaction scale was used from the Eating Disorders Inventory (EDI; Garner & Olmsted, 1984). Coefficient alpha for this scale was .95. The Body Esteem – Appearance (BEA) and the Body Esteem – Weight (BEW) scales were taken from the Body Esteem Scale for Adolescents and Adults (Mendelson, Mendelson, & White, 2001). For the current sample, coefficient alphas were .87 and .85, respectively. Scores on BEA and BEW were reversed so that high scores on the Body Dissatisfaction Factor represented greater body dissatisfaction.
G.A. Tasca et al. / Eating Behaviors 7 (2006) 252–257
255
The Dietary Restraint factor consisted of two scales. The Drive for Thinness scale was taken from the EDI (Garner & Olmsted, 1984). Coefficient alpha was .95. The Restraint scale from the Eating Disorders Examination Questionnaire (Fairburn & Beglin, 1994) was also used. Coefficient alpha for this scale was .71. One of the two scales comprising the Negative Affect factor was the Center for Epidemiological StudiesDepression scale (Radloff, 1977). Coefficient alpha for the current sample was .76. The Ineffectiveness scale from the EDI (Garner & Olmsted, 1984) was also used. Coefficient alpha was .84. 2.4. Overview of structural equation modeling analyses Structural equation modeling with EQS (Bentler, 1995) was used to test the hypothesized model (Fig. 1). The data were analyzed in two steps. First, a confirmatory factor analysis (CFA) was conducted to evaluate the measurement model. This allowed for an assessment of the relationships (loadings) between indicators (observed variables) and factors (latent variables). Because the BMI factor had a single indicator, the error term was set to one minus the reliability, and this was multiplied by the indicator variance (Bollen, 1989). Because the Body Dissatisfaction and the Restrained Eating factors each had one indicator taken from the EDI, the error variances of these two scales were correlated (see Fig. 1). Second, the hypothesized directional relationships among latent variables were tested (Fig. 1). The exogenous (independent) variables were Attachment Insecurity and BMI. Endogenous (dependent) variables were Perceived Pressure to Diet, Body Dissatisfaction, Restrained Eating, and Negative Affect. A maximum likelihood method of estimation was used for all analyses. Several goodness-of-fit indices were calculated. For the comparative fit index (CFI) and the goodness-of-fit index (GFI), values greater than .90 are indicative of a well fitting model (Bentler, 1992; Bentler & Bonnett, 1980). For the root-mean-square error of approximation (RMSEA), values less than .08 represent an acceptable fit, values between .08 and .10 indicate a moderate fit, and values above .10 indicate a poor fit (Browne & Cudeck, 1993). 3. Results The following variables had significantly skewed distributions: Pressure to Diet by Family of Origin, Pressure to Diet by Current Relationships, Pressure to Diet by Authority, Body Dissatisfaction, BEA, BEW, and Drive for Thinness (all zs z 3.3). These variables were transformed using logarithmic or square root transformations, resulting in normally distributed data. The original sample (N = 271) had three multivariate outliers that were removed from the Table 1 Correlations among indicator variables, including their univariate statistics for the total eating disordered sample (N = 268) Indicator Variable 1. Attachment avoidant 2. Attachment anxiety 3. Pressure — family 4. Pressure — authority 5. Pressure — current 6. Body mass index 7. Body dissatisfaction 8. Body esteem appear 9. Body esteem weight 10. Drive for thinness 11. Dietary restraint 12. Depression 13. Ineffectiveness Mean Standard deviation Skewness Kurtosis
1
2
.62 .20 .14 .25 .27 .23 .31 .18 .31 .25 .59 .58 59.95 11.58 0.20 0.24
.20 .17 .16 .25 .31 .39 .22 .40 .29 .59 .66 54.92 10.15 0.42 0.07
3
4
5
.61 .65 .58 .17 .17 .26 .18 .29 .17 .06 2.26 1.29 0.18 1.40
.60 .48 .14 .16 .27 .18 .27 .21 .08 1.81 0.98 0.53 1.17
.67 .27 .29 .42 .13 .35 .18 .03 2.23 1.07 0.02 1.23
6
7
.29 .28 .41 .36 .54 .28 .17 29.14 12.44 0.59 0.60
.59 .59 .39 .05 .29 .39 21.11 7.02 0.65 0.84
8
9
.72 .32 .01 .38 .52 2.26 0.59 0.71 0.30
.22 .11 .24 .34 2.25 0.63 0.21 0.61
10
.52 .42 .43 13.54 5.79 0.64 0.64
11
.33 .25 4.38 1.78 0.32 0.98
12
13
.71 30.80 12.57 0.14 0.85
12.30 7.98 0.29 0.85
Item means are reported for pressure — family, pressure — authority, pressure — current, body esteem — appear, body esteem — weight, and dietary restraint. Total score means are reported for all other scales. Data for pressure — family, pressure — authority, pressure — current, body dissatisfaction, body esteem — appearance, and body esteem — weight were transformed due to univariate non-normality. The means and standard deviations are for non-transformed data. All other values are based on the transformed data.
256
G.A. Tasca et al. / Eating Behaviors 7 (2006) 252–257
subsequent analyses. This resulted in the final sample of 268. Pearson correlations among indicator variables and univariate statistics are presented in Table 1. 3.1. Confirmatory factor analysis and structural model Results of the CFA demonstrated significant factor loadings of all indicator variables onto their respective latent variables ( ps b .001) (Fig. 1). Fit indices indicated that the measurement model provided an adequate fit of the data, CFI = .96, GFI = .93, and RMSEA= .08 (90% CI = .06 to .09). Results of the structural model also appear in Fig. 1. In terms of significant direct effects, attachment insecurity was positively related to body dissatisfaction and to negative affect ( ps b .001). BMI was positively associated with perceived pressure and body dissatisfaction, but negatively related to restrained eating ( ps b .001). Perceived pressure to diet was positively associated with body dissatisfaction ( p = .003), but not to restrained eating ( p = .313). Body dissatisfaction was positively related to both restrained eating ( p b .001) and negative affect ( p = .01). Finally, restrained eating was positively related to negative affect ( p = .01). The hypothesized model adequately fit the data, CFI = .95, GFI = .92, and RMSEA= .08 (90% CI = .07 to .10). The structural model accounted for 58% of variance in perceived pressure, .64% of variance in body dissatisfaction, 66% of variance in restrained eating, and 87% of variance in negative affect. 4. Discussion Results indicated that attachment insecurity was related to body dissatisfaction and negative affect for women with a diagnosed eating disorder. These results are consistent with attachment theory as well as research demonstrating an association between attachment insecurity and negative affect (Cole-Detke & Kobak, 1996; Kobak & Sceery, 1988). The current study extended attachment research to an eating disordered sample. Specifically, these results suggested that attachment insecurity may be a vulnerability factor for the development of an eating disorder, and attachment insecurity may lead to eating disorder symptoms through its relationship with body dissatisfaction and negative affect. Consistent with previous studies (Stice et al., 1998; Stice, 2001), body dissatisfaction was directly related to negative affect, however the relationship observed in the current study was modest. The nature of the relationship among body dissatisfaction, restrained eating and negative affect suggested that body dissatisfaction might be indirectly associated with negative affect through its influence on restrained eating. It is likely that the addition of attachment insecurity as an independent variable accounted for much of the variance in the negative affect factor. Thus, attachment insecurity may have reduced the relationship between body dissatisfaction and negative affect. It is possible that the development of negative affect among women with a diagnosed eating disorder may, in part, have its origin in attachment insecurity, which is likely rooted in early negative family interactions (Ward, et al., 2000). Although a number of therapeutic interventions have targeted body dissatisfaction in the treatment of individuals with eating disorders (Fairburn, Marcus, & Wilson, 1993), the present findings suggest that attachment insecurity and problems in interpersonal relationships could be targeted to address negative affect associated with eating disorders. This is consistent with interpersonal psychotherapies in the treatment of individuals with eating disorders (Tasca et al., in press; Wilfley et al., 1993). Results of this investigation may provide a model for the processes of change among eating disordered individuals who benefit from interpersonally based therapies. This study had several limitations. First, the eating disordered sample was heterogeneous in terms of diagnoses, which likely increased error variance in the model. The tested model also assumed that underlying relationships between insecure attachment and other latent variables were similar across diagnostic categories, but this may not be the case. The dual pathway model (Stice et al., 1998) has been typically discussed in terms of bulimic symptoms, but not all patients in this study had these symptoms. Future studies could focus on more homogeneous groups of individuals with an eating disorder and could include an eating disorders symptom factor. Second, there was a high correlation between anxious and avoidant attachment scales and a single attachment insecurity factor emerged. Attachment research has demonstrated differences between anxious and avoidant attachment styles in terms of coping and negative affect (Wei, Heppner, & Mallinckrodt, 2003). Anxious and avoidant attachment styles may play differing and complementary roles in the development of eating disorder symptoms, and this could be evaluated in future studies. Third, the current study was cross-sectional in nature. Longitudinal studies examining the consistency of attachment styles over time, as well as the role played by attachment insecurity in the development of eating disorder symptoms are needed.
G.A. Tasca et al. / Eating Behaviors 7 (2006) 252–257
257
Acknowledgements Thanks go to Ann Barber, Dr. Alan MacDonald, and Dr. Amy Silverman for their help with this manuscript. This study was partially supported by a grant from the Ontario Mental Health Foundation. References American Psychiatric Association. (1994). Diagnostic and statistical manual of mental disorders (4th ed.). Washington7 Author. Attie, I., & Brooks-Gunn, J. (1989). Development of eating problems in adolescent girls: A longitudinal study. Developmental Psychology, 25, 70 – 79. Bentler, P. M. (1992). On the fit models to covariance methodology in the Bulletin. Psychological Bulletin, 112, 400 – 404. Bentler, P. M. (1995). Structural equation program manual. Encino, CA7 Multivariate Software. Bentler, P. M., & Bonnett, D. G. (1980). Significance tests and goodness of fit in the analysis of covariance structures. Social and Behavioral Sciences, 19, 16. Bollen, K. A. (1989). Structural equations with latent variables. New York7 Wiley and Sons. Bowlby, J. (1988). A secure base. New York7 Basic Books. Brennan, K. A., & Shaver, P. R. (1995). Dimensions of adult attachment, affect regulation, and romantic relationship functioning. Personality and Social Psychology Bulletin, 21, 267 – 283. Browne, M. W., & Cudeck, R. (1993). Alternative ways of assessing model fit. In K. A. Bollen, & J. S. Long (Eds.), Testing structural equation models (pp. 445 – 455). Newbury Pk, C.A7 Sage. Clark, L. A., & Watson, D. (1995). Constructing validity: Basic issues in objective scale development. Psychological Assessment, 7, 309 – 319. Cole-Detke, H., & Kobak, R. (1996). Attachment processes in eating disorders and depression. Journal of Consulting and Clinical Psychology, 64, 282. Duemm, I., Adams, G. R., & Keating, L. (2003). The addition of sociotropy to the dual pathway model of bulimia. Canadian Journal of Behavioral Sciences, 35, 281 – 291. Fairburn, C. G., & Beglin, S. J. (1994). Assessment of eating disorders: Interview or self-report questionnaire? International Journal of Eating Disorders, 16, 363 – 370. Fairburn, C. G., Marcus, M. D., & Wilson, G. T. (1993). Cognitive-behavior therapy for binge eating and bulimia nervosa: A comprehensive treatment manual. In C. G. Fairburn, & G. T. Wilson (Eds.), Binge eating: Nature, assessment and treatment (pp. 361 – 404). NY7 Guilford. Feeney, J. A., Noller, P., & Hanrahan, M. (1994). Assessing adult attachment. In M. B. Sperling, & W. H. Berman (Eds.), Attachment in adults (pp. 128 – 151). New York7 Guilford. Garner, D. M., & Olmsted, M. P. (1984). The eating disorders inventory. Odessa, FL7 Psychological Assessment Resources. Johnson, C., & Connors, M. E. (1987). The etiology and treatment of bulimia nervosa: A biopsychosocial perspective. New York7 Basic. Kobak, R. R., & Sceery, A. (1988). Attachment in late adolescence: Working models, affect regulation, and representations of self and others. Child Development, 59, 135 – 146. Masheb, R. M., & Grilo, C. M. (2001). Accuracy of self reported weight in patients with Binge Eating Disorder. International Journal of Eating Disorders, 29, 29 – 36. Mendelson, B. K., Mendelson, M. J., & White, D. R. (2001). Body-Esteem Scale for adolescents and adults. Journal of Personality Assessment, 76, 90 – 106. Radloff, L. S. (1977). The CES-D scale: A self-report depression scale for research in the general population. Applied Psychological Measurement, 1, 385 – 401. Stice, E. (2001). A prospective test of the dual-pathway model of bulimic pathology: Mediating effects of dieting and negative affect. Journal of Abnormal Psychology, 110, 124 – 135. Stice, E., Shaw, H., & Nemeroff, C. (1998). Dual pathway model of bulimia nervosa: Longitudinal support for dietary restraint and affect-regulation mechanisms. Journal of Social and Clinical Psychology, 17, 129 – 149. Tasca, G. A., Ritchie, K., Conrad, G., Balfour, L., Gayton, J., Daigle, V. & Bissada, H. (in press). Attachment scales predict outcome in a randomized controlled trial of two group therapies for Binge Eating Disorder: An aptitude by treatment interaction. Psychotherapy Research. Ward, A., Ramsay, R., & Treasure, J. (2000). Attachment research in eating disorders. British Journal of Medical Psychology, 73, 35 – 51. Ward, A., Ramsey, R., Turnbull, S., Benedettini, M., & Treasure, J. (2000). Attachment patterns in eating disorders: The past in the present. International Journal of Eating Disorders, 28, 370 – 376. Wei, M., Heppner, P. P., & Mallinckrodt, B. (2003). Perceived coping as a mediator between attachment and psychological distress: A structural equation modeling approach. Journal of Counseling Psychology, 50, 438 – 447. Wilfley, D. E., Agras, W. S., Telch, C. F., Rossiter, E., Schneider, J. A., Cole, A. G., Sifford, K., & Raeburn, S. (1993). Group cognitive-behavioral therapy and group interpersonal therapy for the nonpurging bulimic individual: A controlled comparison. Journal of Consulting and Clinical Psychology, 61, 296 – 305. Wonderlich, S. A., Mitchell, J. E., Peterson, C. B., & Crow, S. (2001). Integrative cognitive therapy for bulimic behavior. In R. H. Striegel-Moore, & L. Smolak (Eds.), Eating Disorders: Innovative directions in research and practice (pp. 173 – 195). Washington7 American Psychological Association.