11
BEHAVIOR II-I~RAPY 21, 99-109, 1990 Developing Computer-Assisted Therapy for the Treatment of Obesity W. STEWART AGRAS, C. BARR TAYLOR, DANIEL E. FELDMAN, MARTHA LOSCH Stanford Universzty KENT F. BURNETT Umversity o f Wisconsin-Madison Ninety mild to moderately overweight women participated in a study examining the potential utility of a hand-held computer in the treatment of overweight. The study compared the use of a computer with one introductory session; computer therapy with four additional group support sessions; and a therapist conducted weight loss pro- gram, in a randomized design. The weight losses in all three groups were modest being approximately half of that expected. No differences in weight losses were found be- tween groups to one-year follow-up, nor were differences found on process measures, suggesting that ambulatory computer treatment is as effective, in te~ms of education and outcome, as therapist conducted treatment. The computer without group support was significantly more cost-effective at immediate outcome than the other two groups. There was some indication, however, that group support may enhance long-term main- tenance of weight loss, thus the cost advantage for the computer alone group declined over the follow-up period. Further developmental work is needed to ascertain whether robust weight losses can be achieved with a hand-heid computer or whether this method has inherent limitations. The coincidental development of effective structured therapies and the com- puter may signal an opportunity to provide psychological therapies for indi- This research was supported by Grant #AM37150--01 from the National Institute of Arthritis, Diabetes, and Digestive and Kidney Diseases. We thank Helena C. Kraemer, Ph.D. for providing statistical expertise to this study. Enquiries regarding the availability of the computer used in this project should be directed to Behavioral Science Products, 451 Chaucer St., Palo Alto, CA 94301. Reprint requests should be directed to W. Stewart Agras, Behavioral Medicine Program TD 205, Stanford University School of Medicine, Stanford, CA 94305. 99 0005-7894/90/0099 010951.00/0 Copyright 1990 by Association for Advancementof B,'havior Therapy All rights of r©t, roduction in any form res,aved.

Developing computer-assisted therapy for the treatment of obesity

Embed Size (px)

Citation preview

Page 1: Developing computer-assisted therapy for the treatment of obesity

BEHAVIOR II-I~RAPY 21, 99-109, 1990

Developing Computer-Assisted Therapy for the Treatment of Obesity

W . STEWART AGRAS,

C . BARR TAYLOR,

DANIEL E. FELDMAN,

MARTHA LOSCH

Stanford Universzty

KENT F. BURNETT

Umversity of Wisconsin-Madison

Ninety mild to moderately overweight women participated in a study examining the potential utility of a hand-held computer in the treatment of overweight. The study compared the use of a computer with one introductory session; computer therapy with four additional group support sessions; and a therapist conducted weight loss pro- gram, in a randomized design. The weight losses in all three groups were modest being approximately half of that expected. No differences in weight losses were found be- tween groups to one-year follow-up, nor were differences found on process measures, suggesting that ambulatory computer treatment is as effective, in te~ ms of education and outcome, as therapist conducted treatment. The computer without group support was significantly more cost-effective at immediate outcome than the other two groups. There was some indication, however, that group support may enhance long-term main- tenance of weight loss, thus the cost advantage for the computer alone group declined over the follow-up period. Further developmental work is needed to ascertain whether robust weight losses can be achieved with a hand-heid computer or whether this method has inherent limitations.

The coincidental development of effective structured therapies and the com- puter may signal an opportunity to provide psychological therapies for indi-

This research was supported by Grant #AM37150--01 from the National Institute of Arthritis, Diabetes, and Digestive and Kidney Diseases. We thank Helena C. Kraemer, Ph.D. for providing statistical expertise to this study. Enquiries regarding the availability of the computer used in this project should be directed to Behavioral Science Products, 451 Chaucer St., Palo Alto, CA 94301. Reprint requests should be directed to W. Stewart Agras, Behavioral Medicine Program TD 205, Stanford University School of Medicine, Stanford, CA 94305.

99 0005-7894/90/0099 010951.00/0 Copyright 1990 by Association for Advancement of B,'havior Therapy

All rights of r©t, roduction in any form res,aved.

Page 2: Developing computer-assisted therapy for the treatment of obesity

100 AGRAS ET AL.

viduals at reduced cost (Agras, 1987a). Well defined therapy packages, partic- ularly those for which manuals have been developed, can be programmed for use on a computer in an interactive mode. Reducing the cost of psychological treatment is important from several viewpoints. First, as Ghosh and Greist (1988) pointed out in their recent review of computer therapy, treating everyone with mild to moderate depression in the United States using individual cognitive-behavioral therapy would cost $15 billion each year, a task well be- yond the resources of current health services. Second, there are not enough therapists available to perform such treatment even in the USA. Thus, it is imperative that less costly, yet effective, treatment programs are developed so that those who cannot now afford it, will have access to treatment.

Although still in its infancy, computer therapy using desktop microcom- puters, has been shown to be effective in the treatment of mild depression (Selmi, 1983), agoraphobia (Ghosh, Marks, & Carr, 1984; Ghosh & Marks, 1987), and weight loss (Foree-Gavert & Gavert, 1980). In each of these controlled studies, computer therapy proved as effective as therapist-administered ver- sions of similar treatments, suggesting that computer therapy may in some circumstances be as effective as therapist provided treatment.

The advent of ever more powerful and low priced hand-held computers may signal another opportunity, allowing easy access to treatment for patients in their own environment. There are, of course, trade-offs in such a development. The memory available in a microcomputer much exceeds that of a hand-held computer, thus limiting the sophistication of the therapeutic program in the latter. In our previous work (Burnett et al., 1985) we demonstrated that a hand- held computer, specifically designed and built for the study, was more effec- tive in producing weight loss than was a condition using the same procedures delivered in paper and pencil form. Mean weight loss was 3.7 kg for the computer-treated subjects and 1.5 kg for the control subjects with eight weeks of treatment. Since few computers were available, the group size was small; consisting of six matched subjects in each group; hence the generalizability of these results is limited. Moreover, the control condition was a weak form of treatment, since therapist time, which was minimal in the computer condi- tion, was equated across the two groups. Thus, the computer condition was not tested against a standard intervention.

A further question regarding computer therapy is how much therapist time is necessary to produce optimal results. Obviously some therapist time is es- sential, if only to explain the philosophy behind treatment and the method of operation of the computer, yet the more therapist time needed, the less cost-effective the program. The development of commercially available, pro- grammable hand-held computers since our original study (Burnett et ai., 1985) has now made a larger scale study feasible. Thus, in this paper we present the results of a randomiTed trial of weight loss therapy in which commercially available hand-held computers, used in conjunction with two levels of ther- apist support, were compared with standard behavior therapy administered in a group format.

Page 3: Developing computer-assisted therapy for the treatment of obesity

C O M P U T E R - A S S I S T E D 1 I . .~RAPY 101

METHOD Subjects

Ninety overweight women were admitted to the study. Since we consider computer therapy to be most suitable for the relatively mildly overweight person at this stage of development, women whose Body Mass Index (BMI) defined as Wt (kg)/Ht (m) 2 was between 25 and 35 were selected for study. Potential subjects were recruited by means of newspaper advertisements for a no-cost weight control study. Subjects replying to the advertisements were sent a brief questionnaire and their responses were screened by study personnel. If they met the initial screen, ie. were female, were over 18 years of age, and were within the BMI cut off points, they were scheduled for an assessment interview at the Stanford Eating Disorders Clinic. These interviews were carried out by physicians or by PhD psychologists. During the interview, the study was ex- plained to each potential subject and a consent form was signed. A semi- structured interview was used to delineate the subject's overweight problem together with a brief review of psychiatric history and current psychopathology. Subjects meeting DSM-III-R criteria for bulimia nervosa (past or present), current major depression, alcohol or drug dependence, or psychosis of any kind were excluded from the study. Subjects with a physical disability preventing them from following the exercise program used in the treatment regimen were also excluded. Following a measurement of weight and height to confirm that subjects met the entry criterion for BMI, potential participants over the age of 45 years were asked to obtain a physical examination from their own physi- cian. When the physician's permission to participate was received, subjects were entered into the study. All participants were asked to deposit $75, of which $25 was returned for completion of each of the three posttreatment assess- ments. Unreturned deposits were mailed to a charity chosen by the participant.

The mean age of participants was 45.2 years (SD = 12.4). Of the subjects 15% were single, 64% were married, and 21% were divorced or separated. The mean weight of the participants pre-treatment was 78.0 kg (SD = 30.2), mean BMI was 29.7 (SD = 4.3), and mean percentage overweight based on the Met- ropolitan Tables (1983) was 125.3 (SD = 21.1).

Research design The design chosen compares a novel treatment administered in two forms

with a standard treatment (behavior therapy) that has been shown in many studies to be effective (Agras, 1987b). The 90 subjects entering the study were randomly allocated, by means of a computer generated table of random num- bers, to one of three conditions: Computer therapy with one introductory ses- sion; Computer therapy with one introductory session and four follow-up group sessions; Behavior therapy conducted in ten sessions over a twelve week period.

Assessment The primary outcome variables were, change in weight over the time of

Page 4: Developing computer-assisted therapy for the treatment of obesity

102 AGRAS ET AL.

the study, and the cost of delivering each treatment. Weights were measured pre-treatment, post treatment (12 weeks), and at six months and one year follow-up, by a research technician using a balance scale which was periodi- cally calibrated. To estimate cost efficacy, the expenses associated with weight loss treatment in the Stanford Eating Disorders Clinic (therapists salary (Ph.D. psychologist) plus overhead costs which include benefits, rent, light and heat- ing, and materials and staff support) were calculated for each of the three programs. Only the costs of therapy, i.e. the total time spent treating patients, not those associated with the research aims of the study e.g., recruitment, screening, data collection, collation, and analysis, were included in the esti- mation of therapy costs. For the computer program, the cost of the com- puter was pro-rated for each use. The computers cost $300 each and it was estimated that a useful life would be 12 uses of twelve weeks each. We believe this to be a conservative estimate since in three years of use only one com- puter was broken when a participant dropped it and the remainder are still in use. Only the cost of the hand-held computer for therapy purposes was included in the analysis, not the cost of the computer used for analysis of down-loaded data. The costs were totaled for each of the three groups and the weight loss (kg) for each $100 cost was then calculated for each individual.

In addition, the following secondary measures were obtained: 1. Adherence to the program was assessed by the frequency o f computer

use as compared with the program goal for such use. In both computer groups participants were asked to use the computer to monitor their caloric intake and exercise levels at least four times each week. At month- ly intervals participants in each o f the groups met briefly with a research assistant who downloaded the hand held computer to a larger computer for analysis. The actual interaction time was approximately five-minutes and no therapeutic issues were discussed. In the behavior therapy group exactly the same goals were set for self-monitoring. Participants in the latter group were asked to bring in self-monitoring records of food con- sumption and exercise each week, thus adherence to the goal could be assessed.

2. Stanford Eating Disorders Questionnaire (Agras, 1987b): This question- naire which focuses on eating behavior both past and present and al- lows determination of binge eating status was administered at pretreat- ment. The subscale from this questionnaire applicable to binge eating was administered at each assessment period.

3. Eating Patterns Questionnaire: This questionnaire originally developed by Wollersheim (1970) and modified by Stalonas and Kirschenbaum (1985) consists of 72 items yielding 12 factors including: eating frequency, food quantity, food assertiveness, self-monitoring, food knowledge, eat- ing topography, eating in response to inappropriate stimuli, food storage, rewards, eating areas, eating as a pure experience, and cognitive ecolo- gy. This questionnaire was administered pretreatment, post treatment, and at 12 month follow-up.

4. Beck Depression Inventory: This widely used and well validated mea- sure of depression is a self-report inventory consisting of 21 items each

Page 5: Developing computer-assisted therapy for the treatment of obesity

COMPUTER=A,.~$IS 1 I~D I HI~I~Y 103

comprising four alternative statements scored on a 0-3 basis for sever- ity (Beck, Ward, Mendelson, Erbaugh, & Mock, 1961).

5. Treatment Acceptability Questionnaire: This questionnaire adapted for both the computer and behavior therapy groups evaluated the ease of use and acceptability of treatment and was administered at post treat- ment.

6. Computer literacy questionnaire: This questionnaire was designed to elicit the individuals past and present use of computers and was ad- ministered pretreatment, post treatment, and at 12 month follow-up.

Treatment conditions 1. Computer therapy with one introductory session. The computer used

in this study was the Casio PB-700 with 16K random access memory and a battery life of 2.5 months. A back up battery is included that will preserve data for over one year in the event of main battery failure. The unit weighs 11 ounces and measures 7.5" x 8" x 3.5". The computer has a four line liq- uid crystal display with twenty characters per line upon which text and graphics can be shown. Information concerning the availability of the hand-held com- puter and program may be obtained from Behavioral Science Products at the address shown in the footnote.

The program began each day with a request that the subject set goals in terms of caloric intake and exercise. Participants were warned in the instruc- tion manual accompanying the computer not to eat less than 1200 calories each day and the warning was repeated by the computer if the daily total in- take entered was less than 1200 calories. Participants entered caloric informa- tion separately for snacks and meals during the day using a book of caloric values. At any point during the day the computer could be used to total the caloric and exercise values. Meals could be planned on the computer using a planning function. Other options included a trainer to promote slow eating, and random messages aimed at motivating the participant and reminding them of distorted cognitions regarding eating. There were also reinforcing statements contingent upon progress. Longer term feedback was afforded by three graphs that plotted daily caloric intake, amount of exercise, and weight, over a 14- day period with a moving window. By studying the relationship between these graphs over time the participant was encouraged to problem solve, to set new goals, and to change behavior appropriately in response to progress.

Participants in this condition attended one 90 minute group session in which the philosophy of the weight loss program was briefly presented, the opera- tion of the computer explained to them, practice in using it afforded, and a telephone number given in case the computer broke down or they could not remember how to proceed. Participants were also given a manual of opera- tion for computer and the weight loss program together with a book of ca- loric values.

2. Computer therapy with group support. In this condition, in addition to the introductory session described above, participants met with a therapist in a group format at weeks 2, 4, 6 and 8. In these sessions problems concerning the use of the computer were addressed and any problems regarding weight

Page 6: Developing computer-assisted therapy for the treatment of obesity

I 0 4 AGRAS ET AL.

loss discussed. Brief education on the principles of weight loss, including di- etary and exercise information were presented and discussed by the group.

3. Behavior Therapy. In this condition a skilled weight loss therapist met with the group for ten sessions over the 12-week treatment period. The ther- apist, a Ph.D. psychologist with five years experience in leading weight loss groups, followed a treatment manual used in the Stanford Eating Disorders Clinic and also provided handouts as used in the clinic. Participants were asked to self-monitor food intake and the circumstances surrounding such intake and activity levels during this treatment. The program incorporated educa- tion concerning the problem of overweight and its treatment, principles of behavior change and application of these principles to enhance self-control of eating behavior and to increase exercise. The calorie and exercise goals for the behavior therapy program were identical to those for the computer pro- gram, namely to gradually reduce caloric intake so that weight loss occurred, but not below 1200 cal/day, and to exercise from 20-45 minutes three or more times per week. The mean weight loss associated with this course of treatment in routine clinical use at the Stanford Eating Disorder Clinic was 5.0 kg over a 12 week course of treatment in patients averaging 135070 of ideal body weight.

Statistical Analysis. The primary outcome measures for this study, based on the research questions being addressed, were weight loss and the cost-efficacy of the three conditions. A repeated measures analysis of co-variance, using the baseline weight as the co-variate, was used to determine whether there was a statistically significant difference between the three groups. A Scheffe test was then used to determine where the significant differences lay between pairs of groups. Secondary measures included the Beck Depression Inventory, and the subscales of the Eating Patterns Inventory. Other measures, such as the computer literacy questionnaire, were used for descriptive purposes.

RESUU S Weight loss and cost efficacy. The weights for all three experimental groups

over time are displayed in Table 1. The weight loss from pre to post-treatment for the three groups combined was 4.82 pounds (2.19 kg), indicating a significant change (t (88) = 6.93, p < .001). There were no significant differences between the groups at any point from post treatment to one year follow-up, suggesting equivalent effects of the three different therapy conditions. There was a significant difference between the groups in the cost-efficacy of treatment at immediate outcome (df = 2,86 F = 7.72, p < .01), with the computer alone group being significantly superior to both behavior therapy (p < .05) and the computer supported condition (p < .05).

Program Adherence. The adherence results over the twelve weeks of treat- ment are shown in Figure 1. As can be seen, all three groups began with en- thusiasm, using the computer or self-monitoring more frequently than the program goals. The two computer groups demonstrated similar declining ad- herence patteias until the lOth week of treatment, when the group using the computer without support showed a sharp decline in adherence. The behavior

Page 7: Developing computer-assisted therapy for the treatment of obesity

COMPUTER-ASSISTED 1 HI~KAPY 105

TABLE 1 WEIGHT, CHANGES IN WEIGHT (kg) AND COST EFFECTIVENESS (kg lost/S100) FROM

PRETt~ATMEHT TO 1 Y E ~ FOR THE TI-mEE GROUPS.

Weight (change in weight)

Pretreatment Post t reatment 6 mon th 12 mon th

Computer alone 76.9 74.6 (2.3) 75.5 (1.4) 76.6 (0.3) n = 3 0 n = 3 0 n = 2 9

Computer plus group support 78.7 76.1 (2.6) 76.6 (2.1) 76.8 (1.9) n = 3 0 n = 2 9 n = 2 9

Behavior therapy 78.5 76.7 (1.8) 75.8 (2.7) 77.5 (1.0) n = 30 n = 30 n = 30

Computer alone

Computer plus group support

Behavior therapy

Pretreatment

Cost effectiveness (kg/$100)

Post t reatment 6 mon th 12 mon th

4.12 3.02 0.52

1.57 1.29 1.1

0.68 1.11 0.43

therapy group demonstrated a similar pattern to the two computer groups until the sixth week of treatment when they showed a rapid decline in the use of self-monitoring which corrected slightly until the eleventh week of treatment before falling off to equal the computer alone group. It is notable that the computer treatment plus group support participants were adhering at 70% in the final week of treatment, while the other two groups were at 29% adher- ence in the same week. This difference, however, was not statistically significant.

Secondary measures. No differences were found between groups on any of the factors of the Eating Patterns Questionnaire. Significant within group changes were found for all groups at both 12 weeks and 12 months, with the exception of food knowledge which showed no change at either assessment, and cognitive ecology, eating topography, and self-monitoring, which showed no changes at 12 weeks, but significant changes at 12 months (all significant p values exceeded .001).

The Beck Depression Inventory decreased for all groups at both 12 weeks and 12 months, with no significant differences between groups. The decrease was highly significant for all groups, the mean at baseline being 7.4, at 12 weeks 5.5, and at 12 months 3.3. The number of individuals who engaged in binge eating defined by DSM-III criteria declined from 9.0% of the group at base- line to 3.5% of the group at 12 months with no significant differences between groups. At the 12 month follow-up 17% of the participants were satisfied with their weight as compared with 4% at pre-treatment, again there were no significant differences between groups.

Computer literacy and confidence. Only the two computer groups partici-

Page 8: Developing computer-assisted therapy for the treatment of obesity

106 AGRAS ET AL.

,.J ,< O

UJ 0 z uJ re uJ -l- a ,< I -

uJ 0 iX: uJ n

150

100

50

O COMPUTER A L O N E O-O, A . . • COMPUTER & O ' ~ : ' ~ • . . . • . L ~ GROUP SUPPORT

-".. .~--0,:" - . . BEHAVIOR THERAPY

" " .0 . . . ; , . . ,•. -'=',%..= Q.... •

O......Q , • . , . •

0 I I I I I I 2 4 6 8 10 12

WEEKS FIG. 1. Percentage adherence for the use of the computer or self-monitoring in each of the

three groups during each week of treatment.

pated in this measurement. Of the group, 69% had experience with computers at work before joining the weight loss program, almost all o f whom used a computer on a daily basis. Forty six percent had access to a home computer, al though only 33°70 used such a computer at least once a week. Thus, overall, the group was reasonably well acquainted with computers. There were no changes in computer usage either at home or at work as a result o f participa- tion in the program.

Treatment confidence. The three groups were equally confident in the treat- ment modality to which they were assigned, felt that the treatment made sense, and that they would recommend it to a friend. All mean scores were above 7.5 on a 0-10 rating scale, where 0 represented no confidence in the treatment.

DISCUSSION The development of the hand-held computer as a therapeutic device is still

in its infancy, thus we present these data more as a test o f a potential ther-

Page 9: Developing computer-assisted therapy for the treatment of obesity

COMPUTER-ASSISTED i i-I~RAPY 107

apeutic model than as a definitive therapeutic application. It was for this reason that only mildly to moderately overweight individuals were admitted to the study. The main results seem quite clear, no differences were found in weight loss between the three groups, suggesting that computer therapy, with or without group support, was as effective as behavior therapy conducted by a skilled clinician. Cost efficacy was significantly superior for the computer alone group at immediate outcome, 17°70 of the cost of therapist conducted behavior therapy and 38o/0 of the cost of group supported computer therapy. This cost advantage was no longer significant at the one year follow-up, at which time the group supported condition was one half the cost of the computer alone and behavior therapy conditions at one year follow-up. In addition, adher- ence during treatment appeared superior for the group supported computer condition adhering at 70°/0 of goal v e r s u s 29~0 for the other two groups after ten weeks, while adherence with the computer alone was equal to that of be- havior therapy over the same time period. Thus, it seems that the computer combined with group support may be superior to the use of the computer alone in terms of maintenance of weight loss.

The weight losses recorded in this study were smaller than the changes that we might have expected to see in a similar group conducted in our clinic. There are several possible reasons for the low apparent weight losses. First, the follow- up rates for the three groups were very high, 100~0 posttreatment, 98.8°/0 at six months follow-up, and 97.7~0 at one year follow-up. The usual follow-up percentages for weight loss studies are much lower, averaging 75°70 (Wing & Jeffery, 1984). Since those who do poorly tend to drop out of studies, then studies with high follow-up rates will inevitably show lower average weight losses. Second, we chose to target a group of women who were mildly to moder- ately overweight, thus, we would expect weight losses to be less in such a popu- lation. Third, while the 1200 calorie diet recommended to all participants may be sufficient to achieve the weight loss goal, self-reported intakes of this level may be too high to produce the desired weight losses because of under reporting of actual caloric consumption. It is possible that in studies achieving higher weight losses, subjects have a caloric intake below 1200 calories by self-report, a behavior minimized by the computer delivered warnings generated by reports under 1200 calories a day. Fourth, the treatment period was only 12 weeks in length. A longer intervention period would be expected to produce more substantial weight losses (Brownell & Wadden, 1986; Perri, Nezu, Patti, & McCann, in press).

One might have expected that the behavior therapy group would have an advantage in terms of providing information regarding diet, exercise, seif- monitoring, stimulus control procedures, and negative cognitions. There were, however, no differences between the three conditions concerning these aspects of the process of change as measured by the Eating Patterns Questionnaire, suggesting that the computer taught these more complex aspects of behavior change to the participants as effectively as did the therapists. Depression scores, although not high to begin with, demonstrated significant decreases over the entire follow-up period. Thus, there was no tendency for the group as a whole to become more depressed as weight was regained.

Page 10: Developing computer-assisted therapy for the treatment of obesity

108 AGRAS ET P~L.

Overall, the results of this feasibility study suggest that commercially avail- able hand-held computers can be programmed to provide therapeutic care al- though the results obtained in this study in all three treatment groups were very modest. Although the computer groups were as effective as the behavior therapy condition in this study, it can be argued that the effectiveness o f com- puter therapy is similar to that achieved with bibliotherapy (Hagen, 1974; Stunkard, 1975). This study therefore raises some questions and directions for future research in applications of computer therapy. Foremost among these questions is whether the weight-loss achieved with computer-assisted treat- ment can be brought to a level equal to that of the average short-term be- havior therapy treatment, i.e. 5.0 kg. In this study we achieved only half of this goal. Superior hand-held computers having more memory and greater ease of use are now available at the same cost of the computer used in this study. It is possible that with more memory available a more sophisticated interactive program might achieve better results. This question is eminently testable in the next phase o f research.

The second question involves the issue of maintenance of treatment effects. Methods to enhance maintenance will need to be specifically developed for computer therapy. There is some suggestion from the results of this study that maintenance may be improved by increased therapist contact. There is, o f course, a trade offhere in that cost-efficacy declines as the amount of therapist contact increases. Alternative approaches to improving maintenance without major cost increases include: allowing participants to keep the computer longer; the use of low cost telephone contacts; and the use of interactive feedback between the hand-held computer and more powerful computers located in a clinic setting which are able to provide more detailed and specific advice to participants.

There are two ways to look at these results. First, despite the low weight losses for all three groups, computer-assisted therapy would appear to be as effective as a therapist-led behavior therapy group. We would not wish to gener- alize this statement beyond the particular mild to moderately overweight sub- jects who participated in this study and we would note that computer therapy alone had a very limited long-term efficacy. Second, given the modest results obtained, there is clearly a long way to go before we can be satisfied with our efforts to translate behavior therapy principles to computer format, particu- larly hand-held computers with their inherent limitations dictated by size. Thus, we present this as the second step in a program of research aimed at enhancing the efficacy of hand-held computer treatment, a larger scale test following the relative success of a smaller scale initial controlled study (Burnett et al., 1985). Clearly more research is needed to ascertain the feasibility of this potentially interesting form of treatment.

REFERENCES Agras, W. S. (1987a). So where do we go from here? Behavior Therapy, 18, 264-273. Agras, W. S. (1987b). Eating Dtsorders: Management o f obesity, buhmia, and anorexia nervosa.

Elmsford, N.Y. Pergamon Press.

Page 11: Developing computer-assisted therapy for the treatment of obesity

COMPUTER-ASSISTED i HhRAPY 109

Beck, A. T., Ward, C. H., Mendelson, M., Erbaugh, G., & Mock, G. (1961). An inventory for measuring depression. Archives of General Psychiatry, 20, 561-571.

Brownell, K. D., & Wadden, T. A. 0986). Behavior therapy for obesity: Modem approaches and better results. In: Brownell, K. D., & F o r ~ , J. P. (Eds). Handbook of eating disorderg" Phys- iology, psychology, and treatment of obesity, anorexi(~ and bulimia. New York, Basic Books.

Bumett, K. E, Taylor, C. B., & Agras, W. S. (1985). Ambulatory computer therapy for obesity: A new frontier for behavior therapy. Journal of Consulting and Clinical Psychology, 53, 698-703.

Foree-Gavert, S., & Gavert, L. (1980). Obesity: Behavior therapy with computer-feedback versus traditional starvation treatment. Scandinavian Journal of Behavior Therapy, 9, 1-14.

Ghosh, A., & Greist, J. H. (1988). Computer treatment in psychiatry. Psychiatric Annals, 18, 246-250.

Ghosh, A., & Marks, 1. M. (1987). Self-treatment of agoraphobia by exposur~ Behavior Therapy, 18, 3-15.

Ghosh, A., Marks, I. M., & Carr, A. C. (1984). Controlled study of self-exposure treatment for phobics: Preliminary communication. Journal of the Royal Society of Medicin~ 77, 483-487.

Hagen, R. (1974). Group therapy versus bibliotherapy in weight reduction. Behavior Therapy, 5, 222-234.

Metropolitan Height and Weight Tables. (1983). Metropolitan Life Insurance Company. Society of Actuaries and Association of Life Insurance Medical Directors, New York.

Perri, M. G., Nezu, A. M., Patti, E. T., & McCann, K. L. (In press). Effect of length of treatment on weight loss. Journal of Consulting and Clinical Psychology.

Selmi, P. M. (1983). Computer-assisted cognitive behavior therapy in the treatment of depres- sion. D~sertation Abstract, Illinois Institute of Technology.

Stalonas, P. M., & Kirschenbaum, D. S. (1985). Behavioral treatments for obesity: Eating habits revisited. Behavzor Therapy, 16, 1-14.

Stnnkard, A. J. (1975). From explanation to action in psychosomatic medicine: The case of obesity. Psychosomatic Medicine, 37, 195-236.

Wing, R. R., & Jeffery, R. W. (1984). Sample size in clinical outcome research: The case of be- havioral weight control. Behawor Therapy, IS, 550-556.

Wollersheim, J. P. (1970). The effectiveness of group therapy based upon learning principles in the treatment of overweight women. Journal of Abnormal Psychology, 76, 462-474.

RECeiVED: August 22, 1989 FINAL ACCEPTANCE: October 2, 1989