DEPRESSION AND ANXIETY 27 : 476–484 (2010)
WAITLIST-CONTROLLED TRIAL OF COGNITIVE BEHAVIOR THERAPY FOR HOARDING DISORDER
Gail Steketee, Ph.D.,1� Randy O. Frost, Ph.D.,2 David F. Tolin, Ph.D.,3 Jessica Rasmussen, M.A.,4
and Timothy A. Brown, Psy.D.4
Objective: This study investigated a multicomponent cognitive behavioral treatment (CBT) for hoarding based on a model proposed by Frost and colleagues and manualized in Steketee and Frost (2007). Method: Participants with clinically significant hoarding were recruited from the community and a university-based anxiety clinic. Of the 46 patients randomly assigned to CBT or waitlist (WL), 40 completed the 12-week assessment and 36 completed 26 sessions. Treatment included education and case formulation, motivational interviewing, skills training for organizing and problem solving, direct exposure to nonacquiring and discarding, and cognitive therapy. Measures included the Saving Inventory-Revised (self-report), Hoarding Rating Scale-Interview, and measures of clinical global improvement. Between group repeated measures analyses using general linear modeling examined the effect of CBT versus WL on hoarding symptoms and moodstate after 12 weeks. Within group analyses examined pre-post effects for all CBT participants combined after 26 sessions. Results: After 12 weeks, CBT participants benefited significantly more than WL patients on hoarding severity and mood with moderate effect sizes. After 26 sessions of CBT, participants showed significant reductions in hoarding symptoms with large effect sizes for most measures. At session 26, 71% of patients were considered improved on therapist clinical global improvement ratings and 81% of patients rated themselves improved; 41% of completers were clinically significantly improved. Conclusions: Multicomponent CBT was effective in treating hoarding. However, treatment refusal and compliance remain a concern, and further research with independent assessors is needed to establish treatment benefits and durability of gains. Depression and Anxiety 27:476–484, 2010. r 2010 Wiley-Liss, Inc.
Key words: compulsive hoarding; controlled trial; cognitive behavior therapy; adult intervention; collecting
Published online 24 March 2010 in Wiley InterScience (www.
Portions of this study were presented at the Association for Behavioral
and Cognitive Therapy, Philadelphia, PA, November 16–18, 2007.
Drs. Steketee, Frost, and Tolin receive royalties from Oxford University
Press (OUP) for publication of a self-help version of the therapy
methods (Tolin, Frost and Steketee, Buried in Treasures). Drs. Steketee and Frost also receive royalties from the therapist guide and client
workbook describing the methods used in this study (Steketee and
Frost, Compulsive Hoarding and Acquiring), and are under contract with Harcourt Press to publish Stuff: Compulsive Hoarding and the Meaning of Things for a public audience. Dr. Tolin received funding from the following sources: Schering-Plough, Indevus Pharmaceuticals, and Eli
Received for publication 7 November 2009; Revised 12 January 2010;
Accepted 13 January 2010
The authors disclose the following financial relationship within the past 3 years: Contract grant sponsor: National Institute of Mental Health; Contract grant number: R21 MH68539.
�Correspondence to: Gail Steketee, Boston University School of Social Work, 264 Bay State Rd., Boston, MA 02215-1409.
1Boston University School of Social Work, Boston,
Massachusetts 2Department of Psychology, Smith College, Northampton,
Massachusetts 3Institute of Living, Anxiety Disorders Center and Yale
University School of Medicine, Department of Psychiatry,
New Haven, Connecticut 4Department of Psychology, Boston University, Boston,
rr 2010 Wiley-Liss, Inc.
Features of hoarding include difficulty parting with personal possessions, even those of apparently useless or limited value, resulting in the accumulation of large amounts of clutter in the living areas of the home and often other personal and/or work environments.
Excessive acquiring through buying or collecting free items is also evident in most cases. These symptoms impair functioning and/or pose significant health and safety risks, as well as distress to those who hoard and/ or those living with or near them.[3–6]
Although hoarding has traditionally been considered a subtype of OCD, increasing evidence points to substantial differences in clinical and biological fea- tures.[7–10] Hoarding is also included in DSM-IV-TR as a symptom of obsessive–compulsive personality disorder, but does not appear strongly associated with other features of this condition (see review by Pertusa and colleagues, in press). Epidemiological findings indicate that clinically significant hoarding occurs in 2–5% of the population, making it a strikingly common problem.[11–13]
Retrospective treatment studies have recruited OCD patients with hoarding symptoms rather than people with hoarding as a primary problem. Most large-scale pharmacological studies have found that hoarding symptoms predict poor outcomes following SRI treat- ment, and another study reported nonsignificant trends for hoarding to predict worse outcome (e.g.,). A prospective study by Saxena et al. reported no difference in response to paroxetine among hoarding and nonhoarding OCD patients; however, both groups showed only modest improvement (approximately 25%) on standard measures of OCD symptoms.
Findings from retrospective studies of behavioral treatments for OCD patients with hoarding symptoms have followed the trend of hoarding, predicting worse outcomes. This was evident for a computer-based behavioral therapy and for therapist-administered exposure and response prevention (ERP), a cognitive- behavioral therapy (CBT) method developed for OCD that utilizes prolonged exposure to increasingly feared obsessive situations and gradual blocking of rituals associated with these obsessions, to achieve habituation of fear and reduction or elimination of rituals. Abramowitz and colleagues reported that only 31% of hoarders exhibited a clinically significant response compared to 46–76% of patients with nonhoarding OCD, a relatively poor response for hoarding to this typically effective ERP method. Unfortunately, these studies suffer from sampling and measurement problems, with most recruiting hoarding patients from OCD clinics and utilizing the 2-item Symptom Checklist of the Yale–Brown Obsessive Compulsive Scale to identify hoarding. Retrospective studies that have combined serotonergic medication with behavior therapy for OCD also reported disappointing outcomes for hoarding compared to nonhoarding OCD patients.[19–21] Descriptive case
reports of hoarding patients receiving behavior therapy have reported generally negative treatment outcomes accompanied by poor insight, treatment refusal, and lack of cooperation.[22–27]
Over the past decade, a cognitive-behavioral model of compulsive hoarding has emerged[1,7,28] that posits that the excessive acquisition, difficulty discarding, and clutter that comprise hoarding stem from information processing deficits, problematic beliefs and behaviors, and emotional distress and avoidance. Research findings support many aspects of the model, including problems with focusing and sustaining attention,[29,30] categoriz- ing possessions, and decision making [Maltby et al., submitted] as well as problematic beliefs about posses- sions.[32–34] The model proposes that strong negative emotional reactions to possessions (e.g., anxiety, grief, guilt) lead to avoidance of discarding and organizing, whereas strong positive emotions (pleasure, joy) re- inforce acquiring and saving possessions.
In a recent open trial, Tolin and co-workers tested a multicomponent CBT intervention for hoarding based on this model. Treatment included office and home visits with motivational interviewing to address low insight and limited motivation, decision-making training to improve cognitive processing, exposure to reduce negative emotions associated with discarding and resisting acquiring, and cognitive restructuring to alter beliefs. This treatment resulted in reductions of the major manifestations of hoarding (clutter, difficulty discarding, acquiring) in 14 adults, of whom 10 completed 26 sessions of treatment for more than 7–12 months. Significant decreases from pre to posttreatment were evident on standardized measures of hoarding symptoms, and at posttreatment, half the sample (n 5 5) were rated much or very much improved on clinical global improvement ratings.
Following upon this pilot work and minor modifica- tions to the treatment protocol, this study tested these CBT methods in a waitlist-controlled trial conducted at two sites. CBT was hypothesized to lead to greater improvement in hoarding symptoms than a wait period of 12 weeks for patients seeking treatment for hoarding. In addition, 26 sessions of this treatment was expected to produce significant and substantial benefits on hoarding symptoms.
DESIGN AND PARTICIPANTS
Participants were recruited from anxiety clinics at Boston University and The Institute of Living from community service providers and from local and national media presentations. No formal advertising was employed and participants were not recruited through OCD specialty clinics. Inclusion criteria were adults Z18 whose most severe problem was hoarding; symptoms of clutter and difficulty discarding were at least moderately severe (Z4 on 0–8 ratings of the Hoarding Rating Scale (see below). Patients were excluded if they reported current psychotic symptoms, bipolar disorder, serious cognitive impairment that would interfere with
477Research Article: Waitlist-Controlled Trial of Cognitive Behavior Therapy
Depression and Anxiety
accurate reporting, substance use disorder within the past 6 months, were receiving concurrent psychotherapy, or had received psychiatric medications within the past month. We also excluded participants who hoarded animals, as treatment was not designed to address this problem. The need for monthly home visits required that we include participants who lived within 45 min of the clinic.
Figure 1 illustrates participant flow through the study. After a detailed description of the study to 46 patients who met criteria and provided written informed consent, 23 were randomly assigned to immediate CBT (26 sessions) and 23 to a 12-week waitlist (WL) before receiving 26 sessions of CBT. Two participants assigned to CBTrefused treatment and another two discontinued before session 10, leaving 19 completers in the immediate CBT group. Of those assigned to WL, two discontinued (animal hoarding was discovered at one home visit and one did not complete assessments), leaving 21 waitlist completers who were invited to continue into CBT. One refused CBT and two discontinued before week 10. Taken together, of 44 patients offered CBT immediately (23) or following WL (21), three declined and four discontinued (o10 sessions), representing 16% of the sample; the reasons given were scheduling conflicts, long travel times to therapy sessions, and feeling too overwhelmed to engage in therapy.
The sample included 75% women and was 87% Caucasian (minority representation included one Latino, five Black, one Asian); mean age was 54. Thirty-six percent had never married, 40% were married/living with partner, and 24% were divorced/separated or widowed. Most (66%) worked full- or part-time or were students and 32% were unemployed. In this generally well-educated sample, 40% reported having graduate education, 19% completed college, 26% had some college education, and 11% completed high school.
The Anxiety Disorders Interview Schedule for DSM-IV (ADIS-IV) was used to determine diagnosis for anxiety, mood, somatoform, substance use disorders, and to screen for the presence of other conditions (e.g., psychosis). The ADIS-IV-L has produced good- to-excellent reliability estimates for the majority of anxiety and mood
disorders, including k5 .85 for principal OCD diagnoses. Cross- site and rater reliability was accomplished via extensive ongoing training by ADIS developers and by a research team consensus regarding principal and secondary diagnoses for all participants.
The Saving Inventory-Revised (SI-R) is a 23-item self-report questionnaire scored from 0 (no problem) to 4 (very severe, extreme); 3 factor analytically defined subscales are difficulty discarding, excessive clutter, and compulsive acquisition. The SI-R showed good internal consistency, test–retest reliability, known groups validity, and concurrent and divergent validity in clinical and nonclinical samples. Cronbach’s a for baseline SI-R in this sample was .88. Mean scores of 50 and above are typical of hoarding samples with clinically significant symptoms, whereas means for nonhoarding clinical samples and community samples score are in the 22–24 range.
Hoarding Rating Scale-Interview (HRS-I; 41) is a brief 5-item semi- structured interview completed by the assessor to rate clutter,
difficulty discarding, acquisition, distress, and impairment on scales
from 0 to 8. All 5 items were summed to yield a total score ranging
from 0 to 40. The HRS-I has shown high internal consistency and
reliability across time, context, and raters, as well as known groups
and construct validity. Internal consistency for the sample was .79.
A total cutoff score of 14 and above distinguished hoarding from
OCD patients; mean scores of 20 and above are typical for clinical
hoarding samples, and mean scores in the 3–4 range were typical of
nonhoarding clinical and nonclinical samples.
The NIMH Clinician Global Impression—Improvement ratings (CGI) were made by therapists at the end of the waitlist period, session 12, and session 26 and ranged from 1 (very much improved) to 7 (very much worse). Patient CGI-I ratings were collected at CBT sessions 12 and 26 (self-ratings were not collected at postwaitlist).
The Beck Depression Inventory (BDI-II), a 21-item self-report inventory of depression, shows good internal consistency and reasonable construct validity. Cronbach’s a in this sample was .94.
The Beck Anxiety Inventory (BAI) is a 21-item self-reported rating of anxiety with high internal consistency and satisfactory test–retest reliability. Internal consistency was .95 in this sample.
73 Met Criteria
23 Assigned to CBT 23 Assigned to WL
45 Did Not Meet Criteria
27 Declined to Participate
21 Completed 12 wks of WL
1 refused CBT, 2 discontinued
18 Completed 26 wks CBT
2 refused CBT, 2 discontinued
19 Completed 26 wks CBT
19 Completed 12 wks of CBT
Figure 1. Flow chart for hoarding treatment study enrollment.
478 Steketee et al.
Depression and Anxiety
This study was approved by the Institutional Review Boards at Boston University, Smith College, and Hartford Hospital. After initial telephone screening, potential participants met with a trained assessor who provided detailed information about the study, obtained informed consent, reviewed inclusion and exclusion criteria and severity of hoarding, and other psychiatric conditions. Consenting participants with hoarding as their primary problem, who met other criteria, were randomly assigned to begin CBT immediately or after a 12-week wait period. Assessments were completed before treatment, after the wait period, after session 12, and after 26 sessions of CBT by patients and by clinicians who provided treatment and were not blind to the waitlist versus CBT condition.
Therapists were advanced psychology graduate students and masters-level social workers who received extensive training in CBT for hoarding, watched motivational interviewing training videotapes, received weekly supervision from RF or GS who listened to audio recordings sessions and provided ongoing feedback. Formal fidelity ratings were not computed, but two pairs of graduate psychology students, who assisted in developing adherence ratings, reported good adherence for six patients treated by three different therapists. Treatment included 26 individual sessions scheduled approximately weekly in the office (1 hr each), with every fourth session (2 hr) conducted in the home or at a local acquiring location. The average therapy duration was 44.8 weeks (range 28–77) due mainly to patient scheduling and motivational problems.
Treatment began with assessment and case formulation, develop- ing an idiographic CBT model, and treatment planning. Motivational interviewing strategies were applied whenever therapists detected ambivalence (e.g., homework compliance, attendance problems). Therapists applied interventions skills training for organizing, decision making, and problem solving; exposure to nonacquiring and discarding; and cognitive therapy for problematic hoarding- relevant beliefs. Techniques were applied flexibly based on the initial treatment plan and progress in various areas. Homework tasks decided at each session required patients to practice therapy methods several times per week. The last two sessions on relapse prevention addressed future progress and managing stressors without reverting to hoarding behaviors. In addition, eight participants (four at each site) with very severe clutter or physical limitations received one or two ‘‘marathon’’ sessions of 3–6 hr each in which the therapist and up to four other research team members traveled to the patient’s home to help with sorting, organizing, and discarding. All such intensive sessions occurred late in therapy after patients had learned skills and could provide direction to team members on what to discard. All decisions were made according to the patients’ rules.
This hoarding-focused CBT intervention differs from standard ERP for OCD, by including specialized aspects of the hoarding problem in the patient case formulation (e.g., familial and biological vulnerability factors, cognitive processing problems, positive and negative emotions that reinforce saving and acquiring behaviors), as well as nonstandard treatment components, such as motivational interviewing and cognitive skills training in decision making and problem solving. Other therapy elements—gradual exposure to discarding and nonacquiring situations (exposure), encouraging voluntary restriction of acquiring (response prevention), and Beckian cognitive therapy methods—are similar to those used in ERP and more recent CBT for OCD.
Data were analyzed using SPSS v. 15 with SPSS Missing Value Analysis v. 7.5 (SPSS, 1997). Whenever possible, missing values were
first replaced with actual data collected from an assessment close in time to the missing value. Accordingly, scores from the first home visit replaced missing baseline data for 3 participants (4 measures), session 16 data were substituted for missing session 12 data for 3 participants (5 measures), and session 8 data substituted for session 12 for 1 participant (1 measure). Remaining missing values were imputed using the expectation-maximization method, which uses a conditional expectation of the missing data given the observed values and current estimates of the parameters, then computes maximum likelihood estimates of the parameters. The extent and effect of missing data are noted below where appropriate. In reporting treatment outcomes, we have included effect sizes calculated via partial Z2 (Z2P) as well as the conversion to Cohen’s d as per Cohen.
Chi square and t-test (two-tailed) analyses first examined compar- ability of groups on baseline demographic and mood state measures for WL and CBT participants and for CBT completers versus discontinuers. Repeated measures general linear modeling (GLM) analyses tested the hypothesis that 12 sessions of CBT would outperform waitlist using a 2 (site: Boston, Hartford)� 2 (group: WL, CBT)� 2 (time: baseline, 12 weeks) repeated measures mixed factor GLM. Similar 2 (site)� 3 (time: pre, mid, post) repeated measures mixed GLM analyses tested the hypothesis that CBT participants would show significant improvement in hoarding symptoms from pretreatment to mid (12 sessions) and posttreatment (26 sessions) for the 41 patients who began CBT. Outcome analyses were performed on both the intention-to-treat and completer samples, but because findings were very similar, only the former are reported here.
RESULTS BASELINE COMPARISONS
Baseline comparisons of CBT versus WL partici- pants (Table 1) indicated no significant differences for gender, age, ethnicity/race, marital status, employment, anxiety (BAI), or depression (BDI, all Ps4.05). Comparisons of treatment completers (n 5 36) to noncompleters (n 5 11) at baseline on demographic variables and hoarding severity also indicated no significant differences (all Ps4.05) in age, gender, race, marital status, or employment, or mood state (BDI, BAI).
CBT VERSUS WL AT WEEK 12
Intent-to-treat analyses included 46 patients (23 WL and 23 CBT).1 Mean scores for outcome variables are given in Table 2. SI-R subscale scores (Clutter, Difficulty Discarding, Acquisition)2 were examined first. As ex- pected, a significant main effect of time (F1,44 5 15.603, P 5.001, Z2P 5 0.271 [Cohen’s d 5 1.219]) was
1Examination of missing data on the main outcome variables at 12 weeks showed that 7 (15.2%) participants were missing SI-R and HRS-I data, and 11 (23.9%) were missing CGI data. Little’s Missing Completely at Random (MCAR) test was not significant [w2(5) 5 2.607, P 5.760], suggesting randomness in the missing data. 2In these analyses, SI-R subscale scores were summed and divided by the number of items per scale to yield average subscale scores. However, Table 2 provides the summed score for each subscale to facilitate comparison with earlier published findings.
479Research Article: Waitlist-Controlled Trial of Cognitive Behavior Therapy
Depression and Anxiety
modified by a significant group� time interaction (F1,44 5 10.749, P 5.002, Z2P 5 0.204 [Cohen’s d 5 1.012]). Because none of the interaction effects invol- ving time and subscale were significant (all Ps4.05), follow-up tests were conducted using the SI-R total score. Again, a significant main effect of time emerged (F1,44 5 15.007, P 5.001, Z2P 5 0.263 [Cohen’s d 5 1.195]) as well as the predicted group� time interaction (F1,44 5 10.719, P 5.002, Z2P 5 0.203 [Cohen’s d 5 1.009]). No site by time interactions were significant (Ps4.05). As Table 2 indicates, the WL group showed minimal reduction in SI-R scores, t(22) 5 0.50, P4.05, whereas the CBT group showed a nearly 10-point (15%) average reduction in total SI-R scores, t(22) 5 4.54, Po.001.
Results from analyses of HRS-I scores showed a significant main effect of time (F1,42 5 60.037, Po.001, Z2P 5 0.588 [Cohen’s d 5 2.389]), modified by a sig- nificant group� time interaction (F1,42 5 8.446, P 5.006, Z2P 5 0.167 [Cohen’s d 5 0.896]). Although both the CBT and WL groups improved (ts 5 6.0 and 4.5, respectively, Pso.05), CBT led to more improvement than did WL (Table 2). However, site interactions were significant for this variable, including time� site (F1,42 5 5.829, P 5.020, Z2P 5 0.122 [Cohen’s d 5 0.764]) and time� site� group (F1,42 5 5.612, P 5.023, Z2P 5 0.118 [Cohen’s d 5 0.732]). Post hoc analyses indicated that Boston
TABLE 2. Means (standard deviations) and effect sizes from general linear modeling (GLM) analyses comparing treated versus waitlisted hoarding participants after 12 weeks
Measure N Baseline SD Wk 12 SD Partial eta squared Cohen’s d conversion
SI-R Total CBT 23 63.42 14.95 53.73 18.47 .223 1.071 WL 23 62.39 10.60 61.21 13.05
Clutter CBT 23 27.37 7.09 24.16 9.42 .135 0.790 WL 23 27.17 4.07 27.17 4.89
Discarding CBT 23 20.03 4.40 16.93 5.57 .090 0.629 WL 23 20.22 3.38 19.67 4.03
Acquiring CBT 23 16.10 6.45 12.67 6.96 .120 0.739 WL 23 15.55 6.43 14.37 6.73
HRS CBT 23 29.88 4.47 21.68 6.97 .111 0.707 WL 23 29.48 5.48 25.83 4.40
BDI-II CBT 14 11.25 10.53 11.59 11.69 .017 0.263 WL 18 19.18 10.51 17.84 10.71
BAI CBT 14 10.49 11.89 9.66 7.51 .001 0.063 WL 20 9.38 6.72 9.00 7.92
CGI therapist CBT 23 – – 2.72 0.74 .403 1.643 WL 23 – – 3.84 0.62
CGI, Clinical Global Improvement; HRS, Hoarding Rating Scale; SI-R, Saving Inventory Revised.
TABLE 1. Demographic information and baseline symptom severity (frequency/mean, P value) for hoarding participants (intention-to-treat) assigned to waitlist (WL) versus cognitive behavioral treatment (CBT) conditions
WL participants CBT participants
Variable N 5 23 N 5 23 P value
Age 53.3 (8.0) 54.5 (8.2) .60 % female 69.6 82.6 .30 % Caucasian 91.3 82.6 .38 % married or
cohabiting 69.6 47.8 .13
% employeda 56.5 73.9 .20 % with BA or
more 54.5 65.2 .47
% with MDD 47.8 39.1 .55 % with GAD 34.8 34.8 1.0 % with social
phobia 21.7 26.1 .73
SI-R total 62.4 (10.6) 63.5 (15.6) .78 HRS 29.5 (5.5) 29.9 (4.5) .79 BDI 19.7 (11.6) 13.5 (11.4) .08 BAI 11.0 (10.8) 11.3 (12.1) .93
aEmployed full-time or part-time or student. BA, Bachelors of Arts degree; BAI, Beck Anxiety Inventory; BDI, Beck Depression Inventory; GAD, generalized anxiety disorder; HRS, Hoarding Rating Scale; MDD, major depressive disorder; SI-R, Saving Inventory-Revised.
480 Steketee et al.
Depression and Anxiety
patients treated with CBT showed substantial reduc- tion in hoarding symptoms, whereas Hartford CBT patients and WL participants from both sites improved only modestly.
Analyses of session 12 outcomes for depressed (BDI-II) and anxious (BAI) mood indicated no significant effect of time or group� time interaction (all Ps4.479). As evident in Table 2, the samples for these analyses were small (Ns 5 14 for CBT and 18–20 for WL), standard deviations were large, and severity levels were low-to-moderate. Univariate ANOVAs used to examine therapist CGI-I ratings at session 12 showed a main effect of group (F1,42 5 28.341, P 5.001, Z2P 5 .403) in which CBT participants were rated as showing more average improvement than WL participants (Table 2). No site effect or site� group interaction emerged (Ps4.05). When therapist CGI-I scores at 12 weeks were recoded into ‘‘very much’’ or ‘‘much improved’’ (scores of 1 or 2) versus ‘‘not improved’’ (score of 3–7),3 none (0%) of the WL patients and 10 (43.5%) of the CBT patients were rated improved, a significant difference according to Fisher’s Exact Test (P 5.001).
PRE-MID-POST-CHANGES IN THE COMBINED CBT SAMPLE
Intent-to-treat analyses examined all participants assigned to immediate and delayed CBT (N 5 41). Mean scores for pretest, session 12, session 26, and effect sizes (partial Z2 (Z2P) from GLM analyses are reported in Table 3. Missing data were addressed, as described earlier.4 SI-R scores were analyzed using a 3
(time: pretreatment, session 12, session 26)� 3 (scale: Clutter, Difficulty Discarding, Acquisition)� 2 (site: Bos- ton, Hartford) repeated measures GLM with time and scale as repeated measures. Results showed a significant main effects of time (F2,38 5 33.265, Po.001, Z2P 5 0.460) and of subscale (F2,78 5 86.506, Po.001, Z2P 5 0.689). The time by site interactions were not significant (time� site, F2,38 5 1.682, P 5.20; time� site� subscale F4,36 5 1.940, P 5.125), indicating that o