to regulate blood glucose. Adaptation to IDDM isoften more difficult during adolescence when fam-ily communication and conflict resolution tend todeteriorate. The treatment burden pervades dailylife, complicating other challenges of adolescence,and the regimen often becomes the focus of parent-adolescent conflict.

Family conflict has been associated with adoles-cents’ treatment adherence and diabetic control

Journal of Pediatric Psychology, Vol. 25, No. 1, 2000, pp. 23–33

Randomized, Controlled Trial of BehaviorTherapy for Families of Adolescents With Insulin-Dependent Diabetes Mellitus

Tim Wysocki,1,2 PhD, Michael A. Harris,3 PhD, Peggy Greco,1 PhD, Jeanne Bubb,4 MSW,Caroline Elder Danda,3 AB, Linda M. Harvey,2 MS, Kelly McDonell,1 BA, AlexandraTaylor,1 MA, and Neil H. White,3,4 MD, CDE1Nemours Children’s Clinic, 2University of Florida, 3Washington University School of Medicine, and4St. Louis Children’s Hospital

Objective: To describe the short-term results of a controlled trial of Behavioral Family Systems Therapy

(BFST) for families of adolescents with diabetes.

Methods: We randomized 119 families of adolescents with diabetes to 3 months’ treatment with either

BFST, an education and support Group (ES), or current therapy (CT). Family relationships, psychological ad-

justment to diabetes, treatment adherence and diabetic control were assessed at baseline, after 3 months

of treatment (reported here), and 6 and 12 months later.

Results: Compared with CT and ES, BFST yielded more improvement in parent-adolescent relations and

reduced diabetes-specific conflict. Effects on psychological adjustment to diabetes and diabetic control

were less robust and depended on the adolescent’s age and gender. There were no effects on treatment

adherence.

Conclusions: BFST yielded some improvement in parent-adolescent relationships; its effects on diabetes out-

comes depended on the adolescent’s age and gender. Factors mediating the effectiveness of BFST must be

clarified.

Key words: behavior therapy; family therapy; adolescents; diabetes.

Treatment of insulin-dependent diabetes mellitus(IDDM) is designed to maintain near-normal bloodglucose levels (DCCT Research Group, 1994). Theregimen includes several daily insulin injections,self-monitoring of blood glucose, a prescribed mealplan, regular exercise, and problem-solving tactics

q 2000 Society of Pediatric Psychology

All correspondence should be sent to Tim Wysocki, Nemours Children’sClinic, 807 Nira Street, Jacksonville, Florida 32207. E-mail: twysocki@nemours.org.

in cross-sectional studies (Anderson, Miller, Aus-lander, & Santiago, 1981; Marteau, Bloch, & Baum,1987). Because similar associations have been foundin longitudinal studies, we may infer that familyconflict may be related causally to poor diabetesoutcomes (Gustafsson, Cederblad, Ludvigsson, &Lundin, 1987; Hauser et al., 1990). Other studiespoint to parent-adolescent conflict specifically as acorrelate of poor diabetes outcomes (Bobrow,AvRuskin, & Siller, 1985; Miller-Johnson et al.,1994; Wysocki, 1993). The association betweenparent-adolescent relationships and family conflictmay be bi-directional, but it is plausible that a treat-ment targeting family communication and conflictresolution could improve adaptation to IDDM,treatment adherence, and diabetic control. A fewstudies support the effectiveness of family therapywith this population, but none was a well-controlled trial of treatments that target parent-adolescent communication (Ryden et al., 1994; Sny-der, 1987).

Robin and Foster’s (1989) behavioral family sys-tems model suggests promising research directions,portraying parent-adolescent conflict as a productof the clash between the adolescent’s need for au-tonomy and parental needs to maintain stability.They argue that parent-adolescent conflict is modu-lated by four factors: family problem-solving skills;family communication; the degree to which familymembers hold extreme beliefs about one another;and the extent of family structural or systemicanomalies. They have validated several assessmenttools based on the model (Prinz, 1977; Robin,Koepke, & Moye, 1990) and behavioral family sys-tems therapy (BFST), an intervention targeting theircentral constructs. The model is supported by stud-ies confirming the benefits of BFST and similar ther-apies (e.g. Barkley, Guevremont, Anastopoulos, &Fletcher, 1992; Foster, Prinz, & O’Leary, 1983; Guer-ney, Coufal, & Vogelsong, 1981). Since BFST reducesparent-adolescent conflict, it could help families ofadolescents to cope with IDDM by improving theircommunication and conflict resolution skills. Suchimprovements could enhance parental social sup-ports for diabetes self-care, reduce overall familystress and clarify responsibility for diabetes tasks(Wysocki et al., 1997). This paper compares theshort-term outcomes of three treatments: currentmedical therapy alone or augmented by ten sessionsof participation in either BFST or an educationalsupport group.

Method

Participants

The sampling plan was designed to recruit familieswho were appropriate candidates for BFST. Theenrollment criteria were designed to ensure thatparent-adolescent conflict in each family might beexpected to impede management of diabetes. Be-cause we considered severe psychopathology to bea contraindication for BFST, families with recenttreatment for certain psychiatric diagnoses were ex-cluded. Enrollment was limited to adolescents withadequately stable family structure to enable com-pletion of the various study requirements.

Adolescents with IDDM and their parents wererecruited in St. Louis, Missouri, or Jacksonville, Flor-ida. Recruitment included an initial confirmationof eligibility based on demographic factors fol-lowed by a screening process to ensure that enrolledfamilies had at least moderate levels of parent-adolescent conflict. Initially, 380 families were con-tacted about the study and to verify that the adoles-cent met these criteria: age between 12 years (an ageat which parent-adolescent conflict often increases)and 16.75 years (to ensure that adolescents lived athome during the 15-month study); IDDM for atleast 1 year; no other major chronic diseases; nomental retardation; no incarceration, foster care, orresidential psychiatric treatment; and absence of di-agnoses of psychosis, major depression, or sub-stance abuse disorder in parents or adolescentsduring the prior 6 months. Families were not askedto report on psychiatric diagnoses prior to thatpoint in time. Parents or step-parents living withthe patient were required to participate in the studyand other adult caregivers were allowed to partici-pate. Of the 380 families contacted, 28 did not meetall of the demographic enrollment criteria. The 174families who denied interest in the study cited timeconstraints (41%), travel distance (17%), minimalparent-adolescent conflict (33%), and other factors(9%) as reasons for not participating.

Eligible, interested families (n 5 178) thensigned an approved informed consent form andcompleted two screening tools: the Conflict Behav-ior Questionnaire (CBQ; Prinz, Foster, Kent, &O’Leary, 1979) and the Diabetes Responsibility andConflict Scale (DRC; Rubin, Young-Hyman, &Peyrot, 1989). This was done to limit study enroll-ment to families reporting parent-adolescent con-

24 Wysocki et al.

needed for measuring socioeconomic status (SES)with the Hollingshead Four-Factor Index of SocialStatus (Hollingshead, 1975). Tanner stage informa-tion was retrieved from each adolescent’s medicalrecord. Demographic data were updated at follow-up evaluations.

General Parent-Adolescent Relationships. The Par-ent-Adolescent Relationship Questionnaire (PARQ;Robin et al.,1990) assesses the primary constructs inthe behavioral family systems model. It yields 16subscales that load on three factors: Overt Conflict/Skill Deficits; Extreme Beliefs; and Family Structure,with higher scores indicative of worse family rela-tions. There are separate forms for adolescents (314items) and parents (280 items), and the normativegroup included 314 adolescents and 427 parents.Internal consistency based on the present sampleranged from .73 to .89 for the three scales and didnot differ among adolescents, mothers, and fathers.

The Issues Checklist (IC) obtains ratings of thefrequency and intensity of recent conflicts around44 issues (Prinz et al.,1979). It yields scores for thenumber of conflict items endorsed and for conflictfrequency and intensity. Higher scores indicatemore parent-adolescent conflict. Internal consis-tency based on the present sample was .74 for ado-lescents, .72 for mothers, and .79 for fathers.

Montemayor and Hanson’s (1985) telephone re-call interview was used to collect participants’ de-scriptions of conflict situations that occurred in theprior day. The topic, participants, intensity, dura-tion, and manner of conflict resolution (negotia-tion, withdrawal, and authoritarian parental action)were recorded. Higher scores are less favorable. Thisstudy yielded significant correlations between par-ents and youths for frequency (r 5 .68), intensity(r 5 .57), and duration (r 5 .53) of conflict events.

IDDM-Specific Psychological Adjustment. The TeenAdjustment to Diabetes Scale (TADS) is a 21-itemLikert-type scale with parallel parent and adolescentforms that measures adolescents’ behavioral, af-fective, and attitudinal adjustment to IDDM (Wy-socki, 1993). Higher scores indicate more favorableadjustment to IDDM. Internal consistency, calcu-lated from data obtained from the present sample,was .81 for adolescents, .87 for mothers, and .88for fathers.

The DRC (Rubin et al.,1989) assesses parent-child conflict over 15 IDDM tasks. Higher scores in-dicate more conflict about the diabetes regimen. In-ternal consistency based on the present sample was

flict at levels that could impede family managementof IDDM. In consultation with the authors of thesetools, we identified cutoff scores that were expectedto exclude 60% of families (CBQ . 5; DRC . 24).Only families in which at least one family memberobtained scores above these cutoffs on one or bothscales were eligible to enroll. Of 132 families ex-ceeding this criterion, 119 (90%) enrolled in thestudy. The CBQ was used only for pre-enrollmentscreening purposes, while the DRC was also treatedas an outcome measure. Participants included 119adolescents, 117 female caregivers, and 82 malecaregivers. The 46 families excluded by the screen-ing procedure did not differ demographically fromthose who enrolled. The apparent enrollment rateof 31% (119 of 380 contacted) is artifically low asthe denominator includes families who were ineli-gible demographically (n 5 28), who failed the con-flict screening criterion (n 5 46), or who reportedminimal parent-adolescent conflict as a reason forrefusing to participate (n 5 58). With these familieseliminated from consideration, the enrollment rateis 52% (119 of 228 families enrolled). No patientsreceived mental health services from any of the re-searchers other than those received in this project.

Measures

Participants completed a baseline evaluation andfollow-up evaluations scheduled at posttreatment(3 months), and at 6 and 12 months after the con-clusion of treatment. This article reports only re-sults of the baseline and 3-month (immediateposttreatment) evaluations. Each evaluation in-cluded collection of interview, questionnaire, andbiochemical data; order of administration of instru-ments was counterbalanced among families. A re-search assistant administered questionnaires atevaluation sessions; the research assistant com-pleted telephone interviews during the 2 weeks pre-ceding each of the four evaluations. A detailedprocedural manual promoted equivalence of meth-ods across the two sites. Measures were chosen toprovide varied perspectives of the family processestargeted by BFST and of the general and diabetes-specific outcomes expected to be affected bychanges in those factors. The following are the spe-cific measures used.

Demographic Factors. Parents reported the pa-tient’s age, gender, race, duration of IDDM, familycomposition, family size, and the information

Behavioral Family Therapy in IDDM 25

.92 for adolescents, .86 for mothers, and .89 for fa-thers.

IDDM Treatment Adherence. Parents and teenswere interviewed separately during three 20–30 mintelephone interviews over 2 weeks to elicit their re-call of IDDM self-care during the prior day. The 24-Hour Recall Interview (Johnson, 1995) yields reli-able and valid scores for five adherence factors: DietComposition, Diet Amount, Insulin, Testing andEating Frequency, and Exercise. Higher scores indi-cate worse adherence. Each interview began withassessment of IDDM treatment adherence and thenof teen-parent conflict using the Montemayor andHanson (1985) method.

The 14-item Self-Care Inventory (SCI) validatedby Greco et al. (1990) was used to sample adherenceover a longer interval than is captured by the recallinterviews. Higher scores indicate better treatmentadherence. Internal consistency based on the pres-ent sample was .76 for adolescents, .81 for mothers,and .82 for fathers.

Health Status. At each evaluation, a 3 cc venousblood sample was collected from each patient forglycated hemoglobin (GHb) assays to index recentdiabetic control. A regression equation, based onconcurrent measurements on 56 split samples, wasused to enable treatment of all results as if they hadbeen obtained from one laboratory (i.e., GHbSt. Louis 5

1.007[GHbJacksonville] 2 .032). The normal range forthe assay is about 6%–8% and higher values indi-cate poorer metabolic control.

Parents reported hospitalizations, emergencyroom visits, and contacts with other mental healthprofessionals at the 3-month evaluation. These re-ports were verified by chart review or contact withthe pertinent health professionals when possible.The study did not include collection of pre-enrollment measures of these variables.

Procedure

After the baseline evaluation, the research assistantat the opposing center randomly assigned each fam-ily to one of the three conditions described below.Randomization was stratified by the adolescent’sgender and treatment center so that each center en-rolled a similar number of boys and girls into thethree groups.

Current Therapy (CT). Patients in the CT group(and in the other groups) continued in standardtherapy for IDDM directed by their physicians, in-cluding examination by a physician and GHb assay

three or more times annually; two or more daily in-jections of mixed intermediate and short-actinginsulins; home blood glucose monitoring and re-cording of test results; IDDM self-managementtraining; a prescribed diet; physical exercise; andannual evaluation for long-term diabetic complica-tions.

Education and Support (ES). In the first 12 weeksof the study, ES families attended 10 group meetingsemphasizing diabetes education and social support.The meetings were designed to emulate a commonmental health service for families of chronically illadolescents and to serve as a “best alternative ther-apy” comparison. A master’s level social workerwith extensive diabetes experience and a master’slevel health educator served as group facilitators.Panels of two to five families began and completeda 10-session series together, attended by the parentsand adolescent with diabetes. Session content wasorganized around the chapters of the American Dia-betes Association’s Diabetes Support Groups for YoungAdults: A Facilitator’s Manual (1990). The same mate-rials and session outlines were used at both sites,and the two facilitators conferred weekly by tele-phone to ensure consistency of the intervention.Family communication and conflict resolutionskills were excluded from session content. Each ses-sion included a 45-min educational presentation bya diabetes professional on one of the 10 topics, fol-lowed by 45 minutes of family interaction aboutthat topic led by the facilitator.

Behavioral Family Systems Therapy (BFST). Adoles-cents and parents in this group received 10 sessionsof Robin and Foster’s (1989) BFST. Sessions wereconducted by one of two licensed psychologistswho each received about 150 hours of training andsupervised BFST experience and were certified asproficient by Dr. Robin. Extensive efforts ensuredthat each psychologist’s technical proficiency wasmaintained throughout the study; every BFST ses-sion was audiotaped and rated by either Dr. Robinor one of the project psychologists, and feedbackfrom these ratings was provided in weekly confer-ence calls. These ratings verified excellent treatmentfidelity throughout the study. Neither psychologistdemonstrated any consistent or significant depar-ture from prescribed therapy content or delivery. Adetailed therapy manual supplemented the guide-lines offered by Robin and Foster (1989) and in-cluded session outlines, educational handouts, andhomework assignments used at both sites. BFSTconsisted of four therapy components matched to

26 Wysocki et al.

Data Reduction

To reduce the number of statistical comparisons,clarify data presentation, and decrease measure-ment error, we calculated family composite scoresby summing and averaging the scores of individualfamily members (e.g. Hanson, Henggeler, & Bur-ghen, 1987). This was justified conceptually sinceall family members reported on the same family be-haviors and, in each case, there were significantpositive correlations (range .45 to .83) between fam-ily members’ scores. This reduced the number ofunivariate tests from 45 to 21, reducing both therisk of Type I error and variability in some measures.Our conclusions did not differ when we analyzedindividual family members.

Results

Sampling and Randomization

The sampling plan was designed to enroll familieswith parent-adolescent relationship difficulties thatwere severe enough to impede family managementof diabetes. With the assay used for this study, aGHb level of 10% was considered indicative of gooddiabetic control. This criterion was exceeded by73% of the enrolled adolescents, indicating thatmost were in poor or fair diabetic control. Meanscores for normative nondistressed families were ex-ceeded by a substantial percentage of enrolled fami-lies on study measures for which these data wereavailable. The percentage of families in which atleast one family member’s baseline scores exceededthe normative mean by one standard deviation ormore were CBQ: 74%; DRC: 64%; PARQ Overt Con-flict/Skill Deficits: 27%; PARQ Extreme Beliefs: 21%;PARQ Family Structure: 29%; and IC Number ofItems Endorsed: 28%. For those measures withoutsuch a normative comparison group, 32% of thesample had SCI scores below 42, indicative of aver-age adherence below 50% for each of the 14 diabe-tes management tasks, whereas on the TADS, 29%of the sample had scores below 63, indicative ofpoor emotional or social adjustment to 21 diabetes-related challenges. A total of 31 families (26%) didnot meet any of these criteria, and these familieswere distributed equally among the three groups.Taken as a whole, these data suggest that the distri-butions of scores for the study sample were shiftedin the direction of more frequent and severe parent-

families’ treatment needs as identified by the proj-ect psychologists based on study data and familyinteraction in sessions: Problem-solving training pro-vided families with a behavioral contracting ap-proach to conflict resolution with training inproblem definition, generation of alternative solu-tions, group decision making, planning, implemen-tation and monitoring of the selected solution, andrenegotiation or refinement of ineffective solutions.Communication skills training included instructions,feedback, modeling, and rehearsal targeting com-mon parent-adolescent communication problems.Cognitive restructuring was used to identify andchange family members’ exaggerated beliefs, atti-tudes, and attributions that may have impededeffective parent-adolescent communication andconflict resolution. Functional and structural familytherapy interventions targeted anomalous familysystemic characteristics (e.g., weak parental coali-tions; cross-generational coalitions) that may haveimpeded effective problem solving and communi-cation.

Families received an individualized BFST treat-ment plan designed by the three project psycholo-gists in accord with the results of baselineassessments and observation of family interactions.Sessions consisted of family problem-solving discus-sions and focused on IDDM-specific or general con-flictual issues as appropriate for each family. Thepsychologist used standard behavior therapy tech-niques of instructions, feedback, modeling, and re-hearsal. Behavioral homework was assigned at eachsession and reviewed at the next session. Familieswere asked to practice the targeted skills at homeand to apply them to new problems.

Participation Incentives and InterventionAdherence

To promote adherence to the study requirements,we paid families $100 ($50 each for the parents andadolescent) upon completing each evaluation. TheES and BFST families could earn another $100 ifthey completed all 10 treatment sessions. The 3-month follow-ups were completed by 115 families(96%). All 10 treatment sessions were completed by87% of BFST families and 91% of ES families. Psy-chological services outside of the study were re-ceived by five CT families (22 sessions), three ESfamilies (21 sessions), and no BFST families. Therewere no psychiatric admissions.

Behavioral Family Therapy in IDDM 27

adolescent conflict and poorer adaptation to diabe-tes, confirming that a clinically appropriate sampleof families was enrolled.

Table I describes the three groups at baselinewith respect to the adolescents’ age, duration ofIDDM, gender, race, GHb level, Tanner stage, familysize and composition, and parental socioeconomicstatus. Despite careful randomization, the threetreatment groups differed at baseline on severaldemographic dimensions. The BFST group includedsignificantly fewer intact families (Kruskal-WallisH 5 7.05; p , .03) and more single-parent families(Kruskal-Wallis H 5 7.27; p , .03) than did theother two groups. The divorce rate for the CT groupwas significantly lower than that for either the ESor BFST groups (Kruskal-Wallis H 5 5.47; p , .05).

Table II shows that these demographic differ-ences were accompanied by baseline differences inseveral measures, indicating greater conflict andpoorer adaptation to IDDM among BFST families.Analyses of variance (ANOVA) with treatmentgroup (degrees of freedom 5 2, 116) as the between-subjects factor were conducted for family compositebaseline scores on the PARQ, DRC, IC, SCI, TADS,and Recall Interviews and for GHb values. A sig-nificant main effect for groups, in each case indica-tive of less favorable status for the BFST groupcompared with one or both of the other two groups,was obtained on the following measures: PARQ Skill

Deficits/Overt Conflict scale (F 5 4.43; p , .02), ICIntensity scale (F 5 3.19; p , .05), DRC (F 5 3.61;p , .03), TADS (F 5 3.08; p , .05), SCI (F 5 3.29; p ,

.05), Recall Interview Testing/Eating Frequency (F 5

4.03; p , .03) and Diet Amount (F 5 3.71; p , .03)factors, and Recall Interview scores for duration ofconflict events (F 5 3.15; p , .05). Subsequent anal-yses were designed to compensate for these pre-treatment group differences as described below.

Statistical Analysis Strategy

Initial analyses consisted of repeated measures anal-yses of variance (MANOVA) and analyses of covari-ance (ANCOVA) using the baseline values of theoutcome measures as the covariates and with group,adolescent age, and gender as between-subjects fac-tors. The MANOVA revealed no significant group 3

time interactions for any measure. The ANCOVAyielded no significant main effects at the 3-monthfollow-up when baseline values of the outcomemeasures served as covariates.

Pretreatment inequality of the groups may im-pede discrimination of true treatment effects fromthose due to regression toward the mean. With suchbaseline differences, interpretation of statisticalanalyses may be impeded by strong correlations be-tween the baseline value of a variable and the mag-nitude of change in that variable (Fleiss, 1986). This

28 Wysocki et al.

Table I. Characteristics of Study Participants at Baseline

CT BFST ES

Age (mean yrs 6 1 SD) 14.3 6 1.4 14.5 6 1.2 14.1 6 1.4

Duration of IDDM (mean yrs 6 1 SD) 5.2 6 3.8 5.4 6 3.8 4.5 6 3.7

Hollingshead index raw score (mean 6 1 SD) 43.9 6 12.9 41.3 6 11.8 44.3 6 11.1

Family size (mean # persons 6 1 SD) 4.2 6 1.5 4.2 6 1.8 4.2 6 1.4

Glycated hemoglobin (mean % 6 1 SD) 11.8 6 3.1 11.9 6 3.3 11.8 6 2.9

Gender

Male 20 (49%) 15 (39%) 15 (38%)

Female 21 (51%) 23 (61%) 25 (62%)

Race

Caucasian 32 (78%) 29 (79%) 32 (80%)

African American 9 (22%) 9 (21%) 7 (17%)

Hispanic 0 (0%) 0 (0%) 1 (3%)

Tanner stage

Prepubertal (stage I) 0 (0%) 1 (3%) 2 (5%)

Midpubertal (stages II-IV) 21 (51%) 17 (45%) 23 (58%)

Pubertal (stage V) 20 (46%) 20 (52%) 15 (37%)

Family composition

Living with both biological parents 23 (56%) 15 (39%) 27 (68%)

Living with one biological parent 14 (34%) 17 (45%) 5 (12%)

Living with one biological and one step-parent 3 (7%) 5 (13%) 7 (17%)

Other 1 (3%) 1 (3%) 1 (3%)

Conflict and Skill Deficits scale, F(2, 103) 5 2.98,p 5 .050, and the Extreme Beliefs scale, F(2, 103) 5

5.45, p 5 .006, but no significant difference betweengroups on the Family Structure scale. Post-hoc anal-yses using the Scheffe test showed that the BFSTgroup improved more on the Overt Conflict/SkillDeficits scale than the CT, but not the ES, group.On the Extreme Beliefs scale, the BFST group hadsignificantly greater improvement than either theCT or ES groups. Neither the group 3 age, group 3

gender, nor group 3 age 3 gender interaction ef-fects were significant for any of the three PARQ fac-tor scores.

The ANCOVA revealed a significant main effectfor groups on change in the IC scores for numberof items endorsed, F(2, 103) 5 4.75, p 5.011, andconflict intensity, F(2, 103 5 3.99, p 5 .022, but noeffect on conflict frequency, possibly due to exces-sive variability in that measure. Post-hoc analysesconfirmed greater reduction in number of items andconflict intensity for BFST families than for eitherCT or ES families. No age or gender interaction ef-fects were significant for any IC score. No group orinteraction effects were significant for family con-flict reported during recall interviews.

problem can be countered by treating baselinescores as covariates (Llabre, Spitzer, Saab, Ironson, &Schneiderman, 1991). Hence, ANCOVAs for post-treatment change in each primary outcome mea-sure were completed, using the baseline value of theoutcome measure as a covariate. This approach con-trolled statistically for the baseline differencesbetween groups, reducing the influence of corre-lations between baseline status and change scores.Because inspection of the data showed differencesas a function of the age and gender of the adoles-cent, all analyses treated the youth’s age group(older [. 14.3 yrs] versus younger [, 14.3 yrs] basedon a median split in order to equate sample size)and gender as additional between-subjects factors.Table II shows the mean (6 1 SD) family compositescores on each measure at baseline and 3-monthfollow-up.

Measures of General Parent-AdolescentRelationships

The ANCOVA revealed a significant main effect forgroups on mean change from baseline to posttreat-ment in family composite scores on the PARQ Overt

Behavioral Family Therapy in IDDM 29

Table II. Family Composite Scores and GHb Values (Mean 6 1 SD) for each Group at Baseline and 3-Month Follow-Up

Baseline Posttreatment

Measures CT ES BFST CT ES BFST

n 41 40 38 41 39 35

Parent-Adolescent Relationship Questionnairea

Overt Conflict/Skill Deficitsb 51.2 6 3.9 52.8 6 5.4 53.3 6 5.7 51.0 6 5.4 51.4 6 5.6 50.2 6 6.7

Extreme Beliefs 49.6 6 3.4 51.2 6 5.1 51.1 6 4.4 49.4 6 3.9 50.1 6 6.3 46.9 6 5.3

Family Structure 51.7 6 6.6 52.3 6 6.4 51.7 6 5.6 50.8 6 6.4 51.3 6 7.0 49.8 6 6.4

Issues Checklista

No. of items endorsed 15.4 6 4.5 16.9 6 6.0 17.4 6 6.8 13.9 6 5.1 14.0 6 5.0 12.9 6 5.8

Total frequency of conflict 58.7 6 42.3 70.8 6 47.7 94.0 6 133.1 53.2 6 42.9 54.9 6 45.5 42.2 6 38.3

Total intensity of conflictb 31.0 6 13.1 36.5 6 13.9 40.8 6 20.2 29.0 6 16.9 31.9 6 16.3 26.7 6 16.2

Recall Interview conflict scoresa

Frequency 2.1 6 1.9 2.1 6 1.3 2.3 6 1.3 1.7 6 5.9 1.9 6 4.4 2.1 6 4.1

Intensity 1.7 6 1.3 1.7 6 0.7 1.9 6 1.2 1.8 6 3.2 1.3 6 3.1 1.2 6 2.3

Durationb 8.5 6 9.1 11.0 6 19.3 10.7 6 15.5 7.3 6 12.8 7.1 6 22.3 4.6 6 12.2

Diabetes Responsibility and Conflict Scalea,b 28.6 6 8.3 29.5 6 8.1 32.5 6 9.4 25.5 6 6.5 26.2 6 7.0 24.8 6 7.6

Teen Adjustment to Diabetes Scaleb 72.8 6 10.5 77.0 6 10.2 78.2 6 9.7 77.3 6 9.6 77.0 6 10.7 73.6 6 11.3

Recall interview adherence factorsa

Insulin 2.11 6 .39 .09 6 .51 .02 6 .49 2.07 6 .77 2.01 6 .89 .09 6 .79

Testing/eating frequencyb 2.17 6 .78 2.31 6 .58 .52 6 .75 2.01 6 .96 2.27 6 .83 .32 6 .93

Diet composition 2.14 6 .37 .10 6 .89 .04 6 .46 2.12 6 .76 2.11 6 .76 .26 6 .81

Diet amountb 2.22 6 .91 2.09 6 .97 .32 6 .87 .23 6 .94 2.03 6 .74 2.21 6 .58

Exercise .15 6 .83 .12 6 .78 2.29 6 .58 2.05 6 .45 .02 6 .51 .03 6 .82

Self-Care Inventoryb 51.1 6 6.6 49.4 6 7.7 46.7 6 9.3 49.7 6 6.8 49.5 6 7.6 47.5 6 8.7

Glycated hemoglobina (%) 11.8 6 3.1 11.8 6 2.9 11.9 6 3.3 11.7 6 3.2 11.6 6 2.5 12.3 6 2.9

aHigher scores are less favorable for these measures. For all others, lower scores are less favorable.bA significant ANOVA main effect for groups was obtained at baseline.

IDDM-Specific Psychological Adjustment

The ANCOVA revealed a significant main effect forgroups on change in DRC family composite scoresfavoring the BFST group, F(2, 103) 5 3.08, p 5 .049,indicative of decreased IDDM-specific conflict.There were no significant age or gender interac-tions.

The ANCOVA analysis of change in family com-posite scores on the TADS revealed no significantmain effects for groups, but significant group 3

gender, F(2, 103) 5 3.35, p 5 .039, and group 3 age3 gender, F(2, 103) 5 3.18, p 5 .046, interactioneffects were obtained, as shown in Table III. Olderboys showed improved adjustment to IDDM (e.g.,higher TADS scores) after treatment with BFST andworse adjustment following treatment with ES,whereas older girls demonstrated the opposite treat-ment effects.

IDDM Treatment Adherence

No significant main or interaction effects were ob-tained for either the SCI family composite score orthe five factor scores obtained with the 24-Hour Re-call Interview.

Health Status Measures

The ANCOVA revealed no significant main effectsfor groups on baseline to 3-month change in GHblevels. However, significant group 3 age, F(2, 103) 5

3.34, p 5 .041, and group 3 gender 3 age, F(2,103) 5 3.72, p 5 .028, interaction effects were ob-

tained. Older adolescents in the BFST group demon-strated a mean increase in GHb of 1.51%, whereasyounger adolescents displayed a .89% decrease, in-dicative of improved metabolic control. The sig-nificant group 3 age 3 gender interaction effect ispresented in Table III, which shows that, amongBFST participants, older girls demonstrated a 2.19%increase in GHb, while substantial decreases oc-curred for younger girls (21.40%) and moderate de-creases were found for both younger (2.60%) andolder (2.54%) boys. A variety of analyses designedto explore pretreatment differences between oldergirls and other participants failed to reveal anymeaningful differences that might have mediatedthese significant group 3 age 3 gender interactions.

There were no significant effects on the low fre-quencies of hospital admissions (2) or emergencyroom visits (5) reported at the 3-month follow-upevaluation.

Discussion

This article compares the short-term benefits ofBFST compared with continued current medicaltherapy or participation in a diabetes support groupwith a large, clinically relevant sample of families ofadolescents with diabetes. The average GHb level ofthe study patients indicated very poor diabetic con-trol, and the sample had generally unfavorable sta-tus on a variety of measures of parent-adolescentrelationships and adaptation to diabetes. This studyadvances the methodology of previous trials of psy-chological treatments for families of youths with

30 Wysocki et al.

Table III. Illustration of Significant Group by Age and Group by Age by Gender Interaction Effects on Baseline to PostTreatment Changein Glycated Hemoglobin (GHb) Assays and Family Composite Scores on the Teen Adjustment to Diabetes Scale (TADS)

Group Age group Gender GHb changea SD TADS changeb SD

CT Younger M 1.23 .34 3.21 1.73

F 2.53 .56 21.36 .61

Older M .06 .63 21.14 2.37

F 2.63 1.19 22.83 2.51

ES Younger M 2.69 .65 .33 2.43

F .86 .37 .11 1.82

Older M 2.05 .39 24.13 2.65

F 2.10 .57 4.79 2.08

BFST Younger M 2.60 .68 21.22 1.96

F 21.40 1.12 .73 2.13

Older M 2.54 1.16 6.03 1.91

F 2.19 .77 23.13 1.88

aLower scores are favorable.bHigher scores are favorable.

plications Trial (DCCT Research Group, 1994)showed that long-term maintenance of near-normal blood glucose levels reduces the onset andprogression of diabetic complications by 50%-75%.Those results were achieved with intensified use ofavailable medical tools, suggesting that translationof the DCCT findings into clinical practice may de-pend on the validation of interventions for promot-ing adaptation to this more demanding treatment.

This study suggests some promise for BFST inthis regard, and the results suggest avenues forfurther research to increase the impact of BFST ondiabetes outcomes. Targeting families of youngeradolescents with BFST to prevent, rather than rem-edy, family conflict around IDDM may be more ef-fective. Others have also reported greater effective-ness of behavioral interventions among younger, ormore recently diagnosed, children with IDDM (De-lamater et al., 1990; Kaplan, Chadwick, & Schim-mel, 1985). Further, the realization of benefits inimportant diabetes outcomes may require BFST ses-sions targeted specifically at each family’s uniquebarriers to adequate treatment adherence and dia-betic control. Clearly, giving families general skillsthat improve parent-adolescent relationships doesnot guarantee that those skills will be applied to en-hance family coping with diabetes. Other possibili-ties for improving the diabetes-specific impact ofBFST might include integrating it with other effec-tive intervention strategies such as multifamily sup-port groups (Satin, LaGreca, Zigo, & Skyler, 1989),training in use of blood glucose data for diabetesproblem solving (Anderson, Wolf, Burkhart, Cor-nell, & Bacon, 1989; Delamater et al., 1990), em-ploying a longer duration of intervention (Dela-mater et al., 1990), and implementing regularlyoccurring “booster” sessions (Foster et al., 1983).Our findings provide reason for optimism that fur-ther research on BFST can yield a disseminable andbroadly applicable intervention that can improvefamily adaptation to IDDM.

Acknowledgments

This work was supported by grant 1-RO1-DK43802“Behavior Therapy for Families of Diabetic Adoles-cents” awarded by the National Institutes of Health(National Institute of Diabetes, Digestive and Kid-ney Diseases) to the first author and by the Pediatricand General Clinical Research Centers of Washing-ton University (RR06021 and RR00036). We thank

IDDM (Rubin & Peyrot, 1992) by using multiplewell-validated outcome measures and two appro-priate comparison groups. In addition to these im-portant features, the study goals went beyond thoseof conventional treatment outcome studies by at-tempting to show that change in a clinically relevantprocess (parent-adolescent relationships) yieldedchanges in disease-related functioning and healthstatus.

Despite careful randomization, the three groupsdiffered at baseline along many clinically meaning-ful dimensions, impeding the confirmation of cleartreatment effects. Conventional MANOVA andANCOVA methods did not confirm treatment bene-fits for BFST; ANCOVA treating the baseline valuesof the outcome measures as covariates enabled sta-tistical control of these pretreatment differences.These analyses revealed significant between-groupeffects favoring BFST in terms of changes in severalfamily composite scores on the PARQ, IC, and DRC,suggesting that BFST yielded some improvementsin parent-adolescent relationships. Change indiabetes-specific outcomes such as GHb and TADSscores was less robust and depended on the age andgender of the adolescent. Older adolescent girlstended to deteriorate along these dimensions afterBFST, whereas boys and younger girls derived bene-fits. There were no main or interaction effects ob-tained on measures of treatment adherence orparent-adolescent conflict derived from recall inter-views.

Given the analytic problems inherent in evalua-tion of change scores, particularly with substantialgroup differences at baseline, the results of theANCOVA analyses should be interpreted very cau-tiously. Although treatment of the baseline out-come measures enabled some degree of statisticalcontrol over these complications, the groups mayhave differed qualitatively despite this statisticalmanipulation. This report offers only suggestive ev-idence in support of the use of BFST with this popu-lation. The failure of the present randomizationillustrates the importance of stratification based onkey outcome variables in order to increase the prob-ability of pretreatment equivalence of the groups.

This study expands the research literature onpsychological interventions for this population.Previous studies supporting the efficacy of familytherapy for youths with IDDM (Ryden et al., 1994;Snyder, 1987) lacked the large sample size, compari-son groups, and multimodal assessment methodsused in this study. The Diabetes Control and Com-

Behavioral Family Therapy in IDDM 31

the following physicians and their respective clinicstaffs for their assistance in recruiting families forthis study: Thomas Aceto, George Bright, Domi-nique Darmaun, Myrto Frangos, John Galgani, Lu-igi Garibaldi, Santosh Gupta, Morey Haymond,Nelly Mauras, Robert Miller, and Patricia Wolff. Wealso thank Arthur L. Robin, PhD, and DianaGuthrie, PhD, for their consultation in interventiondesign and evaluation. A detailed BFST Treatment

References

American Diabetes Association. (1990). Diabetes supportgroups for young adults: A facilitator’s manual. Alexan-dria, VA: American Diabetes Association.

Anderson, B. J., Miller, B., Auslander, W. F., & Santiago,J. V. (1981). Family characteristics of diabetic adoles-cents: Relationships to metabolic control. DiabetesCare, 4, 586–594.

Anderson, B. J., Wolf, F. M., Burkhart, M. T., Cornell,R. G., & Bacon, G. E. (1989). Effects of a peer groupintervention on metabolic control of adolescents withIDDM: Randomized outpatient study. Diabetes Care,12, 184–188.

Barkley, R. A., Guevremont, D.C., Anastopoulos, A. D., &Fletcher, K. E. (1992). A comparison of three familytherapy programs for treating family conflicts in ado-lescents with attention deficit hyperactivity disorder.Journal of Consulting and Clinical Psychology, 60,450–462.

Bobrow, E. S., AvRuskin, T. W., & Siller, J. (1985). Mother-daughter interactions and adherence to IDDM regi-mens. Diabetes Care, 8, 146–151.

DCCT Research Group. (1994). Effect of intensive treat-ment on the development and progression of longterm complications in adolescents with insulin-dependent diabetes mellitus: Diabetes Control andComplications Trial. Journal of Pediatrics, 125,177–188.

Delamater, A. M., Bubb, J., Davis, S. G., Smith, J. A.,Schmidt, L., & White, N. H. (1990). Randomized pro-spective study of self-management training withnewly diagnosed diabetic children. Diabetes Care, 13,492–498.

Fleiss, J. L. (1986). The design and analysis of clinical experi-ments. New York: Wiley.

Foster, S. L., Prinz, R. J., & O’Leary, K. D. (1983). Impact ofproblem solving communication training and gener-alization procedures on family conflict. Child and Fam-ily Behavior Therapy, 5, 1–23.

Greco, P., La Greca, A. M., Auslander, W. F., Spetter, D.,Skyler, J. S., Fisher, E., & Santiago, J. V. (1990). As-sessing adherence in IDDM: A comparison of twomethods. Diabetes, 40(suppl. 2), 108A (abstract).

Guerney, B. G., Coufal, J., & Vogelsong, E. L. (1981). Rela-

Implementation Manual can be obtained by send-ing a $20 check or money order payable to theNemours Children’s Clinic to Tim Wysocki, PhD,Nemours Children’s Clinic, 807 Nira Street, Jack-sonville, FL 32207.

Received September 2, 1997; revisions received February24, 1998, July 23, 1998, and December 13, 1998; ac-cepted December 29, 1998

tionship enhancement versus a traditional approachto therapeutic/preventive/enrichment parent-adoles-cent programs. Journal of Consulting and Clinical Psy-chology, 49, 927–939.

Gustafsson, P., Cederblad, M., Ludvigsson, J., & Lundin, B.(1987). Family interaction and metabolic balance injuvenile diabetes mellitus: A prospective study. Diabe-tes Research and Clinical Practice, 4, 7–14.

Hanson, C. L., Henggeler, S. W., & Burghen, G. A. (1987).Social competence and parental support as mediatorsof the link between stress and metabolic controlamong adolescents with insulin-dependent diabetesmellitus. Journal of Consulting and Clinical Psychology,55, 529–533.

Hauser, S. T., Jacobson, A. M., Lavori, P., Wolfsdorf, J. I.,Herskowitz, R. D., Milley, J. E., & Wertlieb, D. (1990).Adherence among children and adolescents withinsulin-dependent diabetes mellitus over a four yearfollow-up. II: Immediate and long term linkages withthe family milieu. Journal of Pediatric Psychology, 15,527–542.

Hollingshead, A. B. (1975). Four factor index of social status.Unpublished manuscript, Yale University, New Ha-ven, CT.

Johnson, S. B. (1995). Managing insulin-dependent diabe-tes mellitus in adolescence: A developmental perspec-tive. In Wallander, J. L. and Siegel, L. J. (Eds.), Advancesin pediatric psychology: Adolescent health problems: Be-havioral perspectives (pp. 265–288). New York: Guil-ford Press.

Kaplan, R. M., Chadwick, M. W., & Schimmel, L. E.(1985). Social learning intervention to improve meta-bolic control in type I diabetes mellitus: Pilot experi-ment. Diabetes Care, 8, 152–155.

Llabre, M. M., Spitzer, S. B., Saab, P. G., Ironson, G. H., &Schneiderman, N. (1991). The reliability and specific-ity of delta versus residualized change as measures ofcardiovascular reactivity to behavioral challenges. Psy-chophysiology, 28, 701–711.

Marteau, T. M., Bloch, S., & Baum, J. D. (1987). Family lifeand diabetic control. Journal of Child Psychology andPsychiatry, 28, 823–833.

Miller-Johnson, S., Emery, R. E., Marvin, R. S., Clarke,

32 Wysocki et al.

Parent-child responsibility and conflict in diabetescare. Diabetes, 38(suppl. 2), 28A (abstract).

Ryden, O., Nevander, L., Johnsson, P., Hansson, K., Kron-vall, P., Sjoblad, S., & Westbom, L. (1994). Family ther-apy in poorly controlled juvenile IDDM: Effects ondiabetic control, self-evaluation and behaviouralsymptoms. Acta Paediatrica, 83, 285–291.

Satin, W., La Greca, A. M., Zigo, S., & Skyler, J. S. (1989).Diabetes in adolescence: Effects of multifamily groupintervention and parent simulation of diabetes. Jour-nal of Pediatric Psychology, 14, 259–276.

Snyder, J. (1987). Behavioral analysis and treatment ofpoor diabetic self-care and antisocial behavior: Asingle-subject experimental study. Behavior Therapy,18, 251–263.

Wysocki, T. (1993). Associations among parent-adolescentrelationships, metabolic control and adjustment to di-abetes in adolescents. Journal of Pediatric Psychology,18, 443–454.

Wysocki, T., Harris, M. A., Greco, P., Harvey, L. M., McDo-nell, K., Elder, C. L., Bubb, J., & White, N. H. (1997).Social validity of support group and behavior therapyinterventions for families of adolescents with insulin-dependent diabetes mellitus. Journal of Pediatric Psy-chology, 22(5), 635–649.

W. L., Lovinger, R., & Martin, M. (1994). Parent-childrelationships and the management of insulin-dependent diabetes mellitus. Journal of Consulting andClinical Psychology, 62, 603–610.

Montemayor, R., & Hanson, E. (1985). A naturalistic viewof conflict between adolescents and their parents andsiblings. Journal of Early Adolescence, 5, 23–30.

Prinz, R. J. (1977). The assessment of parent-adolescent rela-tions: Discriminating distressed and non-distressed dyads.Doctoral dissertation, State University of New York,Stony Brook, NY.

Prinz, R. J., Foster, S. L., Kent, R. N., & O’Leary, K. D.(1979). Multivariate assessment of conflict in dis-tressed and non-distressed parent-adolescent dyads.Journal of Applied Behavior Analysis, 12, 691–700.

Robin, A. L., & Foster, S. L. (1989). Negotiating parent-adolescent conflict: A behavioral family systems approach.New York: Guilford Press.

Robin, A. L., Koepke, T., & Moye, A. (1990). Multidimen-sional assessment of parent-adolescent relations. Psy-chological Assessment,10, 451–459.

Rubin, R. R., & Peyrot, M. (1992). Psychosocial problemsand interventions in diabetes. Diabetes Care, 15, 1640–1647.

Rubin, R. R., Young-Hyman, D., & Peyrot, M. (1989).

Behavioral Family Therapy in IDDM 33