Predictors of Outcome for Two Treatments forBulimia Nervosa: Short and Long Term

Susan J. TurnbullUlrike Schmidt

Nicholas A. TroopJane TillerGill Todd

Janet L. Treasure

(Accepted 19 September 1995)

Objective: This study examined pretreatment variables to predict outcome in two treatmentsfor bulimia nervosa. Method: Patients were offered either 16 weeks of cognitive-behavioraltherapy (CBT) or a self-treatment manual followed by up to 8 weeks of CBT (sequentialgroup). Using complete data, stepwise regression analyses were performed. Results: It wasfound that a longer duration of illness and lower binge frequency predicted a better outcomeboth at the end of treatment (p < .001) and at 18 months of follow-up (p < .005). In thesequential group, lower pretreatment binge frequency predicted better outcome at the end oftreatment (p < .05) and at 18 months of follow-up (p < .05). In the CBT group, longer durationof illness predicted better outcome at the end of treatment (p < .02). Discussion: It is con-cluded that (1) those with more frequent binging may require a more intense intervention and(2) those who have been ill longer may be more motivated to respond to treatment. © 1997by John Wiley & Sons, Inc.

Over the past decade much effort has been put into the development of effective thera-peutic techniques to treat patients with bulimia nervosa. The factors which influence howwell a patient will respond to a treatment have not, however, been consistently studied.With the development of a range of therapeutic options, from self-care (Schmidt, Tiller, &Treasure, 1993; Treasure et al., 1994) to guided self-care (Cooper, Coker, & Fleming, 1994)

Susan J. Turnbull, M.A.(hons.), is Clinical Research Assistant; Nicholas A. Troop, B.Sc.(hons.), is Doctoral Student(funded by the Medical Research Council); Jane Tiller, M.B.Ch.B., M.R.C.Psych., is Senior Registrar; Gill Todd,B.A.(hons.), R.M.N., is Clinical Nurse Leader; Janet L. Treasure, M.D., Ph.D., M.R.C.Psych., is Consultant, EatingDisorder Research Group, Institute of Psychiatry, London. Ulrike Schmidt, M.Phil., M.R.C.Psych., is Consultant,Psychiatry Department, St. Mary’s Hospital, London. Address reprint requests to Dr. Treasure at Eating DisorderResearch Group, Institute of Psychiatry, De Crespigny Park, Denmark Hill, London SE5 8AF, United Kingdom(e-mail, SPJEJJT@iop.bpmf.ac.uk).

International Journal of Eating Disorders, Vol. 21, No. 1, 17–22 (1997)© 1997 by John Wiley & Sons, Inc. CCC 0276–3478/97/010017–06

Prod. #1147

to long-term psychotherapy, it is important to be able to rationally predict which level ofintervention will be appropriate for an individual.

Several studies have looked at the possible predictors of outcome in the treatment ofbulimia nervosa. Unfortunately, these studies are often difficult to compare as they haveused different treatment approaches, different outcome measures, and various methods ofstatistical analysis. The range of possible predictors that has been generated from thesestudies is large and care must be taken when interpreting findings.

Factors that have been found to be associated with predicting a poor outcome rangefrom discord in the family environment (Blouin et al.,) and the absence of a family historyof alcohol abuse (Collings & King, 1994) to the presence of depression (Davis, Olmsted, &Rockert, 1992), low self-esteem (Fairburn, Peveler, Jones, Hope, & Doll, 1993; Baell &Wertheim, 1992; Fairburn, Kirk, O’Connor, Anastasiades, & Cooper, 1987), and a higherseverity of bulimic symptoms (Baell & Wertheim, 1992; Garner et al., 1990; Keller, Herzog,Lavori, Bradburn, & Mahoney, 1991). Few of these findings have been replicated andwhere they have been the findings have been controversial, that is, in other studies norelation has been found between outcome and bulimic symptoms (Blouin et al., 1994;Fairburn, Peveler, et al., 1993; Davis et al., 1992; Fairburn et al., 1987) or self-esteem (Daviset al., 1992; Garner et al., 1990).

This study seeks to find the predictors of long- and short-term outcome in a group ofpatients who were treated with either 16 sessions of cognitive-behavioral therapy (CBT)or a self-help manual followed by up to 8 weeks of CBT. In keeping with many of thestudies which have looked at possible predictors of outcome in therapy, a global measureof symptom severity will be used as the outcome variable both at end of treatment and at18 month follow-up.

METHOD

Subjects

One hundred and ten patients presenting at a tertiary referral center with ICD-10bulimia nervosa or atypical bulimia nervosa were randomly assigned to one of twotreatment conditions: a sequential treatment group (8 weeks with a self-care manualfollowed by up to eight sessions of therapy) or 16 sessions of CBT. There was a significantimprovement of all bulimic symptoms in both treatments with no significant differencebetween the two groups on any of the measures at the end of treatment and at 18 monthsfollow-up (Treasure et al., 1996).

Assessment Procedures

Expert ratings of bulimic behaviors were made using a rating schedule used in a recentmulticenter study (British Bulimia Group and Duphar Laboratories, obtainable on requestfrom authors). The Structured Clinical Interview for DSM-III-R (Spitzer, Williams, Gib-bon, & First, 1989) was used to measure additional psychiatric morbidity. Family envi-ronment was assessed using the parental care in childhood interview (Bifulco, Brown, &Harris, 1987) and a family history of psychiatric disorder was also taken.

Follow-Up Procedure

Eighteen months (range 14–26 months) after treatment, the 110 patients originally as-signed to treatment were written to so as to assess their current status. After receipt of a

18 Turnbull et al.

questionnaire the patients were assessed using the expert rating of bulimic symptoms bya phone interview.

Statistical Analysis

On the basis of previous research, possible pretreatment predictor variables were cho-sen to be entered into forward stepwise regressions. These included the following vari-ables: duration of eating disorder, age of onset of eating disorder, a history of anorexianervosa, frequency of bingeing, severity of attitude disturbance (weight and shape dis-satisfaction), diagnosis of DSM-III-R (3rd Rev. ed. of the Diagnostic and statistical manual ofmental disorders [American Psychiatric Association, 1987]) depression at time of assess-ment, discord in the family home, a high degree of parental control in home, and a familyhistory of alcohol abuse as reported by the patient. Unfortunately, there was not sufficientdata on pretreatment self-esteem to include this as an independent variable.

Those who had a full course of treatment and/or were available for follow-up werecompared with those who had dropped out of the study on the above predictor variables.The posttreatment outcome variable used as a dependent value was the overall severityof eating disorder psychopathology assessed by summing the frequency of bingeing,vomiting, abuse of laxatives or diuretics, and intense exercising. Forward stepwise mul-tiple regressions were performed on end-of-treatment data and 18-month follow-up data.All data were analyzed using SPSS version 6.0 for Windows.

RESULTS

Patient Sample for Outcome

Data from 86 patients (78%) were obtained at the end of treatment and for 64 patients(58%) at 18-month follow-up. There were no significant differences between those whoremained in treatment at the end of treatment and/or were available for follow-up andthose who had dropped out on any of the predictor variables chosen for regressionanalysis. Further details of this patient group are presented elsewhere (Treasure et al.,1994, 1996).

Treatment Differences

There were no differences in the outcome between the two treatment groups (Treasureet al., 1996). The median of the expert rating scale for severity of eating disorder psycho-pathology at the beginning of treatment was 6 (25th and 75th percentiles: 4 and 9, n = 110),at the end of treatment it was 2 (25th and 75th percentiles: 0 and 4.25; n = 86), and atfollow-up the median was 1 (25th and 75th percentiles: 0 and 4; n = 50).

Predictors of Outcome

Stepwise linear regressions were performed using the above predictor variables toestablish any significant associations with overall eating disorder psychopathology at theend of treatment and again at 18-month follow-up. At the end of treatment, duration ofillness before treatment and binge frequency were significantly associated with outcome(p < .001). These accounted for 24% of the variance (adjusted R2 = 0.243). The relationship

Predictors of Outcome 19

was in the direction of a long duration of illness and a lower severity of bingeing at thestart of treatment, indicating a better outcome in terms of less global symptoms (see Table1 for full details of model).

The CBT and manual groups were then looked at separately in regression analyses.Duration of illness was found to be a significant predictor of outcome in the CBT group(n = 23, t = −2.69, p < .02), explaining 22% of the variance. Binge frequency was the onlypredictor for the manual group (n = 25, t = 2.10, p < .05), explaining 12% of the variance.However, duration of illness appeared as a trend to predict outcome in this model (t =−1.94, p < .07).

At the 18-month follow-up, frequency of bingeing at the beginning of treatment andduration of illness before treatment were again the variables in the model that weresignificantly associated with outcome (p < .005). This accounted for 28% of the variance(adjusted R2 = 0.276). The relationship was in the direction of a lower frequency ofbingeing and a longer duration of illness, indicating a better outcome (see Table 2 for fulldetails of model).

Looking at the CBT and manual groups separately found no indicators for the CBTgroup. Pretreatment binge frequency was the only predictive variable for the manualgroup (n = 14, t = 2.20, p < .05). However, it must be remembered that the numbers in theequation are very small.

DISCUSSION

It was found that a longer duration of the eating disorder and a lower frequency ofbinging at the beginning of treatment were predictive of a better outcome at the end oftreatment and were still predictive at the 18-month follow-up. No other variables in theequation approached significance.

It is important to acknowledge the limitations of this study. The most important prob-lem was that the proportion of patients available for the 18-month follow-up was low.This is frequently found in research on large inner city samples and in young women whoare at the most mobile period in their lives. The majority of follow-up interviews weredone by telephone. Others have found that this is a reliable method of assessment (e.g.,

Table 1. Regression of eating disorder psychopathology at end oftreatment on duration

Source of VariationDegrees ofFreedom

Sum ofSquares

MeanSquares F p

Regression on durationand binge frequency 2 100.42 50.21 8.56 .0007

Residual 45 263.90 5.86Total 47 364.32Residual SD = 2.42

Variable Coefficient bStandard Error

SE (b) t p

Duration −0.186 0.057 −3.280 .002Binge frequency 0.501 0.233 2.150 .037(Constant) 2.751 0.881

20 Turnbull et al.

Paulsen, Crowe, Noyes, & Pfohl, 1988) and in the pilot phase of this project we foundgood agreement between the phone assessments made by the research assistant andface-to-face interviews made by a clinician.

The finding that binge frequency at the start of treatment can be predictive of outcomesupports previous findings (Baell & Wertheim, 1992; Garner et al., 1990). When the twotreatment groups (CBT alone and manual followed by brief CBT) were considered sepa-rately, binge frequency was predictive only in the manual group at the end of treatmentand at the 18-month follow-up. This finding can also be seen to have clinical validity: Ifthe patient presents with an extremely chaotic binging pattern, then a minimal interven-tion such as a self-care book is less likely to be acceptable or useful. It may be impossiblefor those who are markedly symptomatic to make progress even on a weekly psycho-therapeutic approach without support. It has been suggested that such patients requiretwo weekly sessions (Fairburn, Marcus, & Wilson, 1993) or it is possible that such patientswould benefit from an antidepressant to give short-term remission from some symptomsto supplement psychotherapy.

On the other hand, the finding of the predictive value of the amount of binging at thestart of treatment when using an outcome measure of global symptoms (which includesbinge frequency) may be masking progress in some patients. It is possible that there couldhave been a relative improvement in those with a higher pretreatment binge frequency.

A longer duration of illness accounted for some of the variance in predicting a betteroutcome. This variable has not been found to be significant in previous studies looking atpredictors of outcome, however, our sample included many who had had a long durationof illness. We found previously that using the manual, that is, compliance with treatment,was positively related to duration of illness (Troop et al., in press) and that the overallcompliance with the manual was predictive of a reduction of symptoms at that stage oftherapy. Troop et al. suggest that this may be due to patients with a longer illness beingfurther along in Prochaska and DiClemente’s (1986) ‘‘stages of change’’ model and there-fore more ready to use treatment. It is possible that there may be some interaction effectin that those with a longer duration and a lower binge frequency might have started tomake progress before commencement of treatment, that is, it could be that some of thesepatients may already be in the ‘‘action’’ stage of change and hence already making stepstoward recovery.

Table 2. Regression of eating disorder psychopathology at 18-monthfollow-up on binge frequency and duration

Source of VariationDegrees ofFreedom

Sum ofSquares

MeanSquares F p

Regression on durationand binge frequency 2 85.95 42.98 6.71 .004

Residual 28 179.40 6.41Total 30 265.35Residual SD = 2.53

Variable Coefficient bStandard Error

SE (b) t p

Duration −0.148 0.064 −2.30 .029Binge frequency 0.769 0.273 2.82 .009(Constant) 2.0930 1.0941

Predictors of Outcome 21

The best test of significance is replication. The present paper supports the finding thathigher pretreatment binge frequency predicts a poorer outcome in terms of posttreatmentbulimic symptoms. It also adds another variable to consider, that of duration of illness.However, if it is correct that those who have been ill longest may be more ready to change,then a measure of the stage of change may be a more useful addition to the pretreatmentvariables to consider in future studies.

REFERENCES

American Psychiatric Association. (1987). Diagnostic and statistical manual of mental disorders (3rd Rev. ed.).Washington, DC: Author.

Baell, W. K., & Wertheim, E. H. (1992). Predictors of outcome in the treatment of bulimia nervosa. British Journalof Clinical Psychology, 31, 330–332.

Bifulco, A., Brown, G. W., & Harris, T. O. (1987). Childhood loss of parent, lack of adequate parental care andadult depression: A replication. Journal of Affective Disorders, 12, 115–128.

Blouin, J. H., Carter, J., Blouin, A. G., Tener, L., Schnare-Hayes, K., Zuro, C., Barlow, J., & Perez, E. (1994).Prognostic indicators in bulimia nervosa treated with cognitive-behavioural group therapy. InternationalJournal of Eating Disorders, 15, 113–123.

Collings, S., & King, M. (1994). Ten-year follow-up of 50 patients with bulimia nervosa. British Journal ofPsychiatry, 164, 80–87.

Cooper, P. J., Coker, S., & Fleming, C. (1994). Self help for bulimia nervosa: A preliminary report. InternationalJournal of Eating Disorders, 16, 401–404.

Davis, R., Olmsted, M. P., & Rockert, W. (1992). Brief group psychoeducation for bulimia nervosa. II: Predictionof clinical outcome. International Journal of Eating Disorders, 11, 205–211.

Fairburn, C. G., Kirk, J., O’Connor, M., Anastasiades, P., & Cooper, P. J. (1987). Prognostic factors in bulimianervosa. British Journal of Clinical Psychology, 26, 223–224.

Fairburn, C. G., Marcus, M. D., & Wilson, G. T. (1993). Cognitive-behavioural therapy for binge eating andbulimia nervosa: A comprehensive treatment manual. In C. G. Fairburn & G. T. Wilson (Eds.), Binge eating:Nature, assessment, and treatment (pp. 361–404). New York: Guildford Press.

Fairburn, C. G., Peveler, R. C., Jones, R., Hope, R. A., & Doll, H. A. (1993). Predictors of 12-month outcome inbulimia nervosa and the influence of attitudes to shape and weight. Journal of Consulting and Clinical Psy-chology, 61, 696–698.

Garner, D. M., Olmsted, M. P., Davis, R., Rockert, W., Goldbloom, D., & Eagle, M. (1990). The associationbetween bulimic symptoms and reported psychopathology. International Journal of Eating Disorders, 9, 1–15.

Keller, M. B., Herzog, D. B., Lavori, P. W., Bradburn, I. S., & Mahoney, E. M. (1991). The naturalistic history ofbulimia nervosa: Extraordinarily high rates of chronicity, relapse, recurrence, and psychosocial morbidity.International Journal of Eating Disorders, 12, 1–9.

Paulsen, A. S., Crowe, R. R., Noyes, R., & Pfohl, B. (1988). Reliability of the telephone interview in diagnosinganxiety disorders. Archives of General Psychiatry, 45, 62–63.

Prochaska, J. O., & DiClemente, C. C. (1986). Toward a comprehensive model of change. In W. R. Miller & N.Heather (Eds.), Treating addictive behaviours: Processes of change (pp. 3–27). New York: Plenum Press.

Schmidt, U. L., Tiller, J. M., & Treasure, J. H. (1993). Self-treatment of bulimia nervosa: A pilot study. InternationalJournal of Eating Disorders, 13, 273–277.

Spitzer, R. L., Williams, J. B. W., Gibbon, M., & First, M. B. (1989). Structured Clinical Interview for DSM-III-R:Patient Edition (SCID-P, 9/1/89 Version). New York: Biometrics Research Department, New York State Psy-chiatric Institute.

Treasure, J. L., Schmidt, U. H., Troop, N. A., Tiller, J. M., Todd, G., Keilen, M., & Dodge, E. (1994). First step inmanaging bulimia nervosa: Controlled trial of a therapeutic manual. British Medical Journal, 308, 686–689.

Treasure, J. T., Schmidt, U., Troop, N. A., Tiller, J. M., Todd, G., & Turnbull, S. J. (1996). A randomised controlledtrial of sequential treatment for bulimia nervosa incorporating a self-care manual: Outcome at the end oftreatment and at eighteen month follow-up. British Journal of Psychiatry. 168, 94–98.

Troop, N. A., Schmidt, U. H., Tiller, J. M., Todd, G., Keilen, M., & Treasure, J. L. (in press). Compliance with aself-care manual for bulimia nervosa: Predictors and outcome. British Journal of Clinical Psychology.

22 Turnbull et al.