Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2024 Feb 1.
Published in final edited form as: Behav Res Ther. 2023 Jan 10;161:104251. doi: 10.1016/j.brat.2022.104251

Mechanisms of Cognitive Behavioral Therapy vs. Supportive Psychotherapy in Body Dysmorphic Disorder: An Exploratory Mediation Analysis

Jennifer L Greenberg a, Katharine A Phillips b,c, Susanne S Hoeppner a, Nicholas C Jacobson a,d, Angela Fang a,e, Sabine Wilhelm a
PMCID: PMC9892287  NIHMSID: NIHMS1865145  PMID: 36640457

Abstract

Body dysmorphic disorder (BDD) is common, severe, and often chronic. Cognitive behavioral therapy (CBT) is the first-line psychosocial treatment for BDD, with well-established efficacy. However, some patients do not improve with CBT, and little is known about how CBT confers its effects. Neurocognitive processes have been implicated in the etiology and maintenance of BDD and are targeted by CBT-BDD treatment components. Yet, the malleability of these factors in BDD, and their potential role in mediating symptom improvement, are not well understood. Understanding how treatment works could help optimize treatment outcomes. In this secondary data analysis of a randomized clinical trial of CBT vs. supportive psychotherapy (SPT) in BDD (n=120), we examined whether treatment-related changes in detail processing (Rey-Osterrieth Complex Figure test), maladaptive appearance beliefs (Appearance Schemas Inventory-Revised), and emotion recognition (Emotion Recognition Task) mediated treatment outcome. All constructs improved over time and were associated with symptom improvement. CBT was associated with greater improvements in maladaptive beliefs than SPT. None of the variables examined mediated symptom improvement. Findings suggest that with successful treatment, individuals with BDD demonstrate reduced neurocognitive deficits (detail processing, emotion recognition, maladaptive beliefs) and that CBT is more likely than SPT to improve maladaptive appearance beliefs. More work is needed to understand mechanisms of change and thus maximize treatment outcomes.

Keywords: body dysmorphic disorder, cognitive behavioral therapy, mechanisms, mediators, neurocognitive, information processing

Body dysmorphic disorder (BDD) is a common, chronic, and often severe disorder characterized by distressing or impairing preoccupation with one or more perceived appearance flaws and time-consuming rituals intended to check, fix, hide, or obtain reassurance about perceived flaws (APA, 2013). BDD currently affects 1.7-2.9% of the general population and is associated with substantial psychosocial impairment and morbidity (Buhlman et al., 2010; Gunstad & Phillips, 2003; Koran et al., 2008; Phillips et al., 2005; Schieber et al., 2015). Many trials have demonstrated the efficacy of cognitive behavioral therapy (CBT) for BDD, with response rates ranging from 48% to 84.6% (Harrison et al., 2016; Wilhelm et al., 2019). Despite the robust treatment effects of CBT for BDD, some patients (15-52%) do not respond or remit with treatment (Flygare et al., 2020; Weingarden et al., 2021), and very little is known about the processes underlying symptom improvement. Supportive psychotherapy (SPT) is the most commonly received therapy for BDD (Phillips et al., 2013); yet only two studies have examined its efficacy (Enander et al., 2016; Wilhelm et al., 2019, based on the same study sample as presented herein), and only one study has examined mechanisms underlying SPT for BDD (Bernstein et al., 2021, also based on the same sample examined in the current report). To optimize BDD treatment outcomes, more work is needed to understand how treatment confers its effects.

Theoretically derived mechanisms of change come from the CBT model of BDD (Wilhelm et al., 2013). The CBT-BDD model posits that individuals with BDD overfocus on (selective attention to detail) and overestimate the importance of (maladaptive beliefs) nonexistent or slight perceived imperfections in their appearance, which leads to emotional distress and maladaptive coping strategies, including appearance-related compulsions (e.g., excessive grooming) and social avoidance. These maladaptive coping behaviors in turn reinforce selective attention, negative beliefs, and distress (Wilhelm et al., 2013). Data from neurocognitive research in BDD corroborate clinically observed abnormalities in attention, emotion processing, and belief patterns that are thought to underlie core symptoms of the disorder, consistent with this model (Buhlmann & Hartmann, 2017).

Studies using the Rey-Osterrieth Complex Figure test (ROCF; Osterrieth, 1944) suggest that BDD patients overfocus on details of visual stimuli in lieu of seeing “the big picture” (Deckersbach et al., 2000; Greenberg et al., 2018; Yousefi et al., 2020). While one small study failed to find deficits in BDD on the ROCF (Hanes, 1998), visuospatial processing deficits on this test have been reported regarding both accuracy (Deckersbach et al., 2000; Yousefi et al, 2020) and organization (Deckersbach et al., 2000; Greenberg et al., 2018). Moreover, Deckersbach et al. (2000) found that deficits in visuospatial memory were mediated by impaired organizational strategy, with excessive attention to small details rather than larger “holistic” configural elements of the figure. Selective attention to detail and holistic processing deficits in BDD have also been observed in studies using eye tracking, fMRI, and computerized tests of facial processing (Beilharz et al., 2017; Feusner, Moller et al., 2010; Feusner, Moody et al., 2010; Greenberg et al., 2014; Grocholewski et al., 2012; Jefferies et al., 2012; Toh et al., 2017).

In addition to overfocusing on small details of appearance, individuals with BDD exaggerate the importance, meaning, and influence of appearance, often conflating attractiveness with self-worth (Buhlmann et al., 2008; Buhlmann et al., 2009; Buhlmann et al., 2011; Grocholewski et al., 2012; Hrabosky et al., 2009). Indeed, even when compared to those with other body image disorders (i.e., bulimia nervosa, anorexia nervosa), individuals with BDD report greater appearance overvaluation and appearance-managing investment (Hrabosky et al., 2009). Moreover, some studies of obsessive compulsive disorder (OCD), a disorder closely related to BDD (Simberlund & Hollander, 2017), suggest that a reduction in maladaptive disorder-related beliefs mediates CBT response (Wilhelm et al., 2015; Woody et al., 2011).

Maladaptive beliefs about the importance and influence of appearance are frequently accompanied by referential thinking in BDD. About three-quarters of individuals with BDD have experienced ideas/delusions of reference, believing others take special notice of (e.g., stare, laugh, mock) the perceived flaw (Phillips et al., 1993). Consistent with referential thinking, studies using the Emotion Recognition Task (ERT; Beilharz et al., 2007; Buhlmann et al., 2004; Buhlmann et al., 2006; Buhlmann et al., 2011; Buhlmann et al., 2013; Grace et al., 2019) have shown that persons with BDD, compared to those with OCD or healthy controls, are less accurate in identifying emotion, particularly negative emotion (e.g., anger, fear, sadness). In addition, individuals with BDD are more likely to misinterpret other people’s neutral facial emotional expressions as threatening or angry. Interestingly, this emotion recognition impairment is present only in scenarios where individuals believe the emotions to be personally relevant (self-referent) and not when they believe that the emotions are about someone else (other-referent; Beilharz et al., 2007; Buhlmann et al., 2004; Buhlmann et al., 2006; Grace et al., 2019). Furthermore, BDD is characterized by high levels of rejection sensitivity (Kelly et al., 2014). Perceiving others as rejecting, threatening, or angry likely reinforces individuals’ maladaptive beliefs about their perceived unattractiveness and its importance (Buhlmann et al., 2004; Buhlmann et al., 2013).

CBT-BDD specifically addresses the above-noted components of the CBT-BDD model, for example via cognitive restructuring and other cognitive strategies, perceptual/attention retraining, and exposure and response prevention. Thus, these hypothesized pathways should change in response to CBT-BDD. To our knowledge, only two studies have examined mechanisms of CBT for BDD (Bernstein et al., 2021; Fang et al., 2020). Fang and colleagues (2020) examined the cognitive and behavioral facets of this model in patients (n=45) receiving CBT for BDD and found that within-subject changes in maladaptive beliefs about appearance and associated excessive behaviors (checking, excessive grooming, avoidance) mediated symptom improvement (Fang et al., 2020). More recently, using network intervention analyses in the same sample presented herein, Bernstein et al. (2021) found that CBT and SPT differentially affected change. Specifically, CBT was associated with greater efforts to disengage from unhelpful thoughts and resist BDD rituals. Moreover, the timing of these symptom changes aligned with the CBT-BDD treatment protocol, such that cognitive effects emerged first (paralleling cognitive restructuring in earlier sessions), followed by behavioral changes (consistent with an emphasis on exposure and ritual prevention in later sessions). SPT was associated with better BDD-related insight earlier in treatment than CBT, although greater improvements in insight began emerging later in treatment and during follow-up among participants in the CBT group compared to the SPT group.

Given the role of detail processing, maladaptive beliefs, and emotion recognition deficits in the development and maintenance of the disorder, it is possible that CBT changes specific BDD symptoms via improvement in these processes. Yet, very little is known about whether or how detail processing, maladaptive beliefs, and emotion recognition deficits change in response to treatment, or their potential role in mediating symptom improvement. Understanding the role of neurocognitive change in BDD is important to understanding the malleability of these processes. Moreover, if changes in detail processing, maladaptive beliefs, and emotion recognition deficits occur early in treatment, prior to BDD symptom change, this may help explain how CBT for BDD works, and thus, facilitate efforts to optimize treatment. For example, clinicians may use strategies that target these symptoms (e.g., cognitive restructuring, perceptual/attention retraining) earlier in treatment, or new treatments could be developed to specifically target these symptoms.

In this secondary analysis using data from a randomized clinical trial of CBT vs SPT for adults with BDD (Wilhelm et al., 2019), we examined treatment-related changes in detail processing (Rey-Osterrieth Complex Figure test [ROCF; Osterrieth, 1944]), emotion recognition deficits (Emotion Recognition Task [ERT; Buhlmann et al., 2006]) and maladaptive appearance beliefs (Appearance Schemas Inventory-Revised [ASI-R; Cash et al., 2004]) and explored whether changes in these processes mediated response to treatment. These mediation analyses were pre-specified as exploratory follow-up analyses in the original grant proposal, as well as in the protocols approved by the institutional review board at each study site. To our knowledge, no prior study has examined these constructs as therapy outcomes or as potential mediators of treatment outcome in BDD. Comparing CBT to another active psychotherapy enables us to identify unique effects of CBT rather than change due to time or common factors of psychotherapy. Specifically, we hypothesized that 24 weeks of CBT-BDD, as compared to SPT, would lead to (1) greater reductions in maladaptive appearance beliefs (ASI-R), (2) greater reductions in detail processing in (a) copy and (b) delayed recall conditions of a neurocognitive detail processing task (ROCF), and (3) greater accuracy in emotion recognition (ERT), and that changes in these constructs would partially mediate treatment response.

Methods

Participants

Participants were 120 adults recruited at two sites, Massachusetts General Hospital (MGH)/Harvard Medical School and Rhode Island Hospital (RIH)/Brown University, from 11/24/2011 to 7/7/2016. Recruitment methods and demographic characteristics are described in detail in the primary outcome paper (Wilhelm et al., 2019). Eligible participants had a primary DSM-IV diagnosis of BDD and a score ≥ 24 on the Yale-Brown Obsessive-Compulsive Scale Modified for BDD (BDD-YBOCS), reflecting at least moderate symptom severity. Exclusion criteria were: intellectual impairment (i.e., estimated intelligence quotient < 80); current diagnosis of mania, psychosis, borderline personality disorder, or substance use disorder (except nicotine use disorder); current suicidality or psychopathology requiring a higher level of care; concurrent therapy or a history of ≥10 sessions of CBT for BDD. Concurrent pharmacotherapy was allowed, provided the dosage had been stable for at least 2 months before study baseline and would not be changed during the study period.

Procedure

The investigation was carried out in accordance with the latest version of the Declaration of Helsinki. Both study sites’ institutional review boards approved the study, and all participants provided written informed consent.

Treatment

Participants were randomly assigned to receive CBT for BDD (CBT-BDD; Wilhelm et al., 2013) or SPT. Both treatments consisted of 22 individual 60-minute sessions administered over 24 weeks, with the last two sessions spaced 2 weeks apart. Two booster sessions were offered at 1- and 3-month follow-up to participants in either condition who achieved at least 30% symptom reduction at post-treatment on the BDD-YBOCS. Manualized therapy for both treatment arms was provided by trained masters or doctoral-level clinicians, and adherence and treatment fidelity were assessed throughout the trial.

CBT-BDD (Wilhelm et al., 2013) is a neurobiologically informed modular, skills-based treatment designed to specifically address BDD symptoms. Treatment consists of psychoeducation, values and goal setting, cognitive restructuring, exposure with ritual prevention, mindfulness and attentional/mirror retraining, and relapse prevention. Optional modules target symptoms that some, but not all, patients experience (e.g., skin picking, surgery seeking). SPT is a non-specific, nondirective treatment intended to support patients’ self-esteem, expression of emotions, and use of adaptive coping skills more broadly. SPT treatment followed the Pinsker manual (Pinsker, 1997), which was enhanced with BDD-specific psychoeducation.

Measures

Clinician-administered assessment were conducted by masters- or doctoral-level independent evaluators (IEs) who were blinded to treatment received. The Structured Clinical Interview for DSM-IV Axis I and Axis II Disorders (First et al., 2002; First et al., 1997) were administered at baseline to determine inclusion/exclusion diagnoses and to characterize the sample. The BDD-YBOCS (Phillips et al., 1997; Phillips et al., 2014), a 12-item semi-structured clinician-administered measure of BDD symptom severity, was the primary outcome measure. Total scale scores range from 0 to 48, with higher scores indicating more severe BDD symptoms. The internal consistency of the BDD-YBOCS in this sample ranged from α=.68 at baseline to α=.94 at week 24. We defined treatment response as a ≥30% reduction in BDD-YBOCS from pre- to post-treatment (Phillips et al., 1997; Phillips et al., 2014) and “remission” as a post-treatment BDD-YBOCS score of ≤16 (Fernandez de la Cruz et al., 2019), which combines partial and full remission. The Appearance Schemas Inventory-Revised (ASI-R; Cash et al., 2004) is a 20-item self-report measure that assesses beliefs about the importance, meaning, and influence of appearance in one’s life. The ASI-R evaluates body image investment by assessing both the importance an individual places on his or her physical appearance for their definition of self-worth (12 items; e.g., “When I meet people for the first time, I wonder what they think about how I look.”) as well as an individual’s efforts to look or feel attractive (8 items, e.g., “I try to be as physically attractive as I can be.”). In this analysis, we used the ASI-R composite mean score, which has a score range from 1.0 to 5.0, where higher scores indicate higher body image investment. In our sample, the internal consistency of ASI-R composite scores ranged from α=.83 at baseline to α=.91 at week 24. The Rey-Osterrieth Complex Figure test (ROCF; Osterrieth, 1944) evaluates visuospatial ability, executive functioning, and memory. Participants are shown and asked to reproduce a complex geometric figure (copy condition) and then by memory, immediately (immediate recall condition) and after a 20-30 minute delay (delayed recall condition). Accuracy (participants’ ability to construct and recall visual details of the figure), and organization (participants’ ability to construct basic configural elements of the figure as unfragmented units) were scored using the Denman (1984)] and Savage et al. (1999) systems, respectively. In this analysis, we focused on ROCF organization scores (range 0 – 6; higher scores indicating greater organization) in the copy and delayed recall conditions, as these deficits in BDD have been previously demonstrated (Deckersbach et al., 2000; Greenberg et al., 2018). The Emotion Recognition Task (ERT; Buhlmann et al., 2006), presents participants with 24 facial photographs of different emotions (e.g., anger, neutral; Ekman & Friesen, 1975, 1976)). Each emotion is presented 6 times, and each face is followed by a self-referent scenario (“Imagine this bank teller is looking in your direction”) or other-referent scenario (“Imagine this bank teller is looking in your friend’s direction”). Participants are asked to identify the emotion for each scenario (accuracy score; range 0-24), then rate whether the person’s expression was caused by the participant (self-attributional) or another factor (other-attributional). In the current analysis, we used two scores: ERT accuracy (range 0-24), where higher scores indicate better emotion recognition accuracy, and ERT self-attribution score (range 0-24), with higher scores indicating greater levels of self-attribution; both scores were taken from the self-referent (SR) scenarios. All potential mediators used in this secondary analysis were measured at baseline (week 0), mid-treatment (week 12), and end-of-treatment (week 24). The outcome measure (i.e., BDD-YBOCS) was assessed monthly during the treatment phase, but only the data from weeks 0, 12, and 24 were included in this analysis.

Statistical Analysis

Basic demographics and baseline values of BDD symptom severity and each hypothesized mechanistic construct were summarized by site, and site differences were tested using two-sample t-tests for continuous variables and either chi-square tests or Fisher’s exact tests for categorical outcomes. We also used Pearson correlations to assess baseline correlations between the outcome (BDD-YBOCS total score) and all hypothesized mechanisms. Missing data for the hypothesized mediators and the outcome ranged from 0.0% (BDD-YBOCS, ROCF organization copy) to 5.0% (ERT self-attribution SR) at baseline, 25.8% (BDD-YBOCS, ASI-R) to 34.2% (ERT accuracy SR) at week 12, and 23.3% (BDD-YBOCS) to 28.3% (ROCF organization copy and delay) at week 24 (see Supplementary Table S1). Statistical significance was evaluated at a 2-tailed p<.05 and all analyses were performed using SAS System for Windows, version 9.4 (IBM).

To estimate the total effect of treatment and sites on these variables as secondary outcomes, we used hierarchical mixed models with repeated measures. The models used site, treatment, the site by treatment interaction, as well as time, time-by-treatment, time-by-site, and time-by-treatment-by-site interactions as fixed effects. For the ERT models, we also adjusted for the order effect (self-referent vs. other-referent) as a fixed effect. Repeated measures were modeled as random effects with an unstructured covariance matrix, and degrees of freedom were estimated using the Kenward-Roger method. We used specific linear contrasts to estimate pre-to-post (i.e., baseline to week 24) treatment effects overall, by site, and the treatment-by-site difference. Pre-to-post effect estimates are presented as model-adjusted estimates with 95% confidence intervals.

To estimate the hypothesized longitudinally mediated effects, we used the SAS macro “THREEWAVEMED” for computing causal mediated effects in three-wave longitudinal models (Valente et al., 2018). For data sets with missing data, Valente and colleagues (2018) recommended the path analysis method implemented in SAS PROC CALIS with METHOD=FIML for full-information maximum likelihood, which performs well with missing outcomes in longitudinal data (Ji et al., 2018). We estimated single-mediator models for each of the 5 hypothesized mediators: (1) ASI-R composite mean scores, (2) ROCF organization copy scores, (3) ROCF organization delay scores, (4) ERT accuracy in the self-referent condition, and (5) ERT self-attribution in the self-referent condition. The path model (shown in Figure 1) estimated the effect of the randomized treatment on the mediator at weeks 12 (am2x) and 24 (am3x) and on the outcome at weeks 12 (c’y2x) and 24 (c’y3x), the effects of the mediator on subsequent observations of the outcome (by2ml and by3m2), the effects of the outcome on subsequent observations of the mediator (fm2y1 and fm3y1), and autoregressive effects of successive observations of the mediator (M1->M2, M2->M3) and outcome (Y1->Y2, Y2->Y3). We specified site as a baseline covariate, which was assumed to be correlated with the baseline measures of the mediator and the outcome (i.e., M1 and Y1) and uncorrelated with X. The mediator model further assumed that the baseline covariate affected only the wave 2 (i.e., week 12) mediator and outcome variables (i.e., M2 and Y2) and not the wave 3 (i.e., week 24) mediator and outcome variables (i.e., M3 and Y3; Valente et al., 2018). All equations defining these paths are estimated simultaneously in the path model, so that the effects in each equation were adjusted for the other effects in the model. In this framework, two components of the mediated effect were estimated: mediated effect 1, which estimated the effect of the treatment (X) on the mediator at week 12 (M2) and its concurrent impact on the outcome at week 12 (Y2) (i.e., pathway X on Y2 through M2, estimated as the product of am2x * by2m2 in Figure 1); and mediated effect 2, which estimated the longitudinal effect of treatment on the mediator at week 12 on the outcome subsequently measured at the post-treatment (week 24) (i.e., pathway X on Y3 through M2, estimated as the product am2x * by3m2 in Figure 1; Valente, 2018). For the path analysis method used, the macro computed an effect size for each mediated effect by dividing the respective mediated effect by the standard deviation of the outcome (i.e., for mediated effect 1, the effect size is (Mediated effect 1/standard deviation of Y2) and for the mediated effect 2 the effect size is Mediated effect 2/standard deviation of Y3).

Figure 1.

Figure 1.

Open in a new tab

Diagram of the three-wave causal longitudinal single mediator model with randomized intervention, X (i.e., CBT vs. SPT), pretest measures of the mediator (M1; i.e., ASI-R, ROCF, or ERT outcomes) and outcome (Y1; i.e., BDD-YBOCS), treatment intermediate measures of mediator and outcome (M2 and Y2), and end-of-treatment measures of the mediator and outcome (M3 and Y3).

In addition, we ran two types of sensitivity analyses: (1) running all mediation models with completer-only data; and (2) running all mediation models in multi-group path models by site (see Supplementary Materials). Model fit characteristics were assessed using the following statistics: chi-square tests of model fit, where significance indicates poor fit; root mean square error of approximation (RMSEA), where values less than 0.10 indicate a fair fit (MacCallum et al., 1996) and a cut-off value less than .06 indicates a good fit (Hu and Bentler, 1999); the Bentler Comparative Fit Index and the Bentler-Bonett Non-normed Index, where values greater than .95 indicate good fit (Hu & Bentler, 1999); and mediator and outcome r2 as measures of how much of their variation is explained by other variables in the model.

In follow-up analyses, we also examined the hypothesized mediator variables as predictors of treatment response or BDD remission in two separate multiple logistic regression models for participants with complete data (n=88). Continuous predictors were mean-centered and standardized before adding them to the multiple logistic regression models, and each model also included treatment, site, and a treatment-by-site interaction as predictors. Neither model included interaction terms between predictors and treatment (i.e., moderator effects). We excluded ROCF organization delay scores from the multiple logistic regression models due to its strong correlation with ROCF organization copy scores at baseline (Supplementary Table S2a).

Power considerations

Exploratory analyses of the mediation of treatment effects via maladaptive appearance beliefs, detail processing, and emotion recognition changes were prespecified in the grant proposal and study protocol of this study. However, the trial was not powered to detect specific mediation effects; instead, the clinical trial was powered to detect a main treatment effect on efficacy outcomes at the end of treatment. A widely cited guideline for researchers to determine the sample size necessary to conduct mediational studies (Fritz & Mackinnon, 2007) suggests that with the available sample size of n=120 in our study sample, we would likely have 80% power to detect mediated effects via the joint significance test if at least one of the mediation paths is large (coefficient > .59) and the other path is at least medium-to-small (coefficient >.26).

Results

Participants were predominantly white, non-Hispanic, and female, with a mean (SD) age of 34.0 (13.1) years. The demographic and clinical characteristics of participants did not differ significantly between sites (Table 1). Raw means of the outcome and the hypothesized mediators by treatment group over time are available in the supplementary materials in Table S2b.

Table 1.

Baseline characteristics of 120 participants with BDD at baseline enrolled at Massachusetts General Hospital (MGH) or Rhode Island Hospital (RIH).

MGH (n=65) RIH (n=55) Site diff.
Variable M/% (SD/N) M/% (SD/N) p
Demographics
 Age, mean (SD), y 32.2 (12.8) 36.0 (13.1) 0.11
 Male, % (N) 29 (19) 16 (9) 0.10
 Race, % (N)
  White 85 (55) 89 (49) 0.47
  Black 3 (2) 0 (0) 0.50
  Asian/Pacific Islander 9 (6) 2 (1) 0.12
  Other 3 (2) 9 (5) 0.24
 Ethnicity, % (N) 6 (4) 7 (4) 1.00
 Education, % (N)
  High school graduate or less 9 (6) 20 (11) 0.09
  Technical school/some college/college degree 63 (41) 53 (29) 0.25
  Graduate or professional school 28 (18) 27 (15) 0.96
 Single, never married, % (N) 66 (43) 55 (30) 0.19
 Unemployed, % (N) 15 (10) 24 (13) 0.25
Psychiatric Characteristics
 Any comorbid Axis I diagnosis, % (N) 77 (50) 75 (41) 0.76
 Current psychiatric medication, % (N) 38 (25) 55 (30) 0.08
 BDD duration, years M (SD) 16.6 (14.8) 19.5 (14.5) 0.28
 BDD-YBOCS total score, M (SD) 32.6 (5.0) 31.0 (4.5) 0.07
Potential mediators
 ASI-R composite mean, M (SD) 4.3 (0.5) 4.2 (0.5) 0.50
 ROCF organization, copy, M (SD) 4.0 (2.0) 4.1 (1.8) 0.62
 ROCF organization, delay, M (SD) 4.3 (1.7) 3.7 (2.1) 0.08
 ERT accuracy, SR, M (SD) 17.9 (2.9) 18.1 (3.3) 0.75
 ERT self-attribution, SR, M (SD) 11.7 (59) 11.1 (7.0) 0.59
Open in a new tab

Note: BDD = body dysmorphic disorder; BDD-YBOCS = Yale-Brown Obsessive-Compulsive Scale Modified for Body Dysmorphic Disorder; ASI-R = Appearance Schema Inventory - Revised; ROCF = Rey-Osterrieth Complex Figure test; ERT = Emotion Recognition Task; SR = self-referent scenarios.

Hypothesized mediators as secondary treatment outcomes

When examining the hypothesized mediators as secondary outcomes, only ASI-R composite scores (i.e., beliefs about the importance, meaning, and influence of appearance in one’s life) differed by treatment group. This treatment difference in ASI-R composite mean scores was not yet detectable at mid-treatment (week 12, p=.36), but ASI-R composite mean scores were lower in the CBT group compared to the SPT group by post-treatment (p=.001; Table 2 and Figure 2). We were unable to detect a significant site difference in the treatment effect (p=.49). We failed to detect any significant treatment differences in ROCF organization scores (copy or delayed recall), ERT accuracy in the self-referent condition, or ERT self-attribution in the self-referent condition at either mid-treatment or posttreatment (Table 2).

Table 2.

Treatment effects of CBT vs. SPT in participants with BDD on hypothesized mediators as secondary treatment outcomes.

Potential mediators Treatment (Tx) difference overall
 Tx difference at MGH
 Tx difference at RIH
 Tx by site differences
Est.  95% CI Pr>F Est. 95% CI Pr>F Est. 95% CI Pr>F Est. 95% CI Pr> F
Week 12
  ASI-R composite mean −0.09 [−0.30, 0.11] 0.36 −0.10 [−0.37, 0.18] 0.48 −0.09 [−0.39, 0.21] 0.55 −0.01 [−0.42, 0.40] 0.97
  ROCF org., copy −0.19 [−0.92, 0.54] 0.61 0.08 [−0.89, 1.05] 0.87 −0.46 [−1.54, 0.63] 0.41 0.54 [−0.92, 1.99] 0.47
  ROCF org., delay −0.72 [−1.44, 0.01] 0.052 −0.79 [−1.76, 0.17] 0.11 −0.64 [−1.71, 0.44] 0.24 −0.16 [−1.60, 1.29] 0.83
  ERT accuracy, SR 0.46 [−0.81, 1.73] 0.47 1.14 [−0.56, 2.84] 0.18 −0.22 [−2.09, 1.64] 0.81 1.37 [−1.15, 3.88] 0.28
  ERT self-attribution, SR 0.27 [−2.74, 3.27] 0.86 1.64 [−2.38, 5.65] 0.42 −1.10 [−5.55, 3.35] 0.62 2.73 [−3.25, 8.71] 0.37
Week 24
  ASI-R composite mean −0.47 [−0.75, −0.20] 0.001 −0.38 [−0.74, −0.01] 0.043 −0.57 [−0.98, −0.16] 0.007 0.19 [−0.36, 0.74] 0.49
  ROCF org., copy −0.40 [−1.07, 0.28] 0.25 0.04 [−0.86, 0.94] 0.93 −0.83 [−1.85, 0.19] 0.11 0.87 [−0.49, 2.23] 0.21
  ROCF org., delay −0.45 [−1.06, 0.15] 0.14 −0.43 [−1.23, 0.37] 0.29 −0.48 [−1.39, 0.43] 0.30 0.05 [−1.17, 1.26] 0.94
  ERT accuracy, SR −0.42 [−1.46, 0.63] 0.43 0.38 [−1.01, 1.76] 0.59 −1.21 [−2.77, 0.35] 0.13 1.59 [−0.49, 3.66] 0.13
  ERT self-attribution, SR −1.63 [−4.37, 1.10] 0.24 0.03 [−3.59, 3.65] 0.99 −3.29 [−7.38, 0.80] 0.11 3.32 [−2.14, 8.77] 0.23
Open in a new tab

Note: Participants: MGH: n=65, RIH: n=55. Estimates of treatment differences are based on specific contrasts contrasting the change in scores from baseline to end-of-treatment between both treatment conditions. MGH = Massachusetts General Hospital; RIH = Rhode Island Hospital; ASI-R = Appearance Schema Inventory - Revised; ROCF org. = Rey-Osterrieth Complex Figure test organization score; ERT = Emotion Recognition Task; SR = self-referent scenarios.

Figure 2.

Figure 2.

Open in a new tab

Changes in hypothesized mediators over time by treatment group and site. When analyzed as secondary outcomes, the hypothesized mediators all showed significant improvements over time overall (i.e., decreases in ASI-R composite scores and ERT self-attribution in the self-referent condition, increases in ROCF organization and ERT accuracy in the self-referent condition). Only ASI-R composite scores showed greater improvement with CBT compared to SPT (top left panel).

Causal mediated effects in three-wave longitudinal models

We were unable to detect any significant mediation effects (i.e., combinations of significant a- and b-paths) of any of the hypothesized mediators on either the BDD-YBOCS scores at week 12 (mediated effect 1) or at week 24 (mediated effect 2) in any of the three-wave longitudinal mediation models (Table 3). Effect sizes for all modeled mediation pathways (i.e., combinations of a- and b-paths) were very small (< 0.10 for positive and >−0.10 for negative effects). Examining the path models more closely, the lack of significant mediation pathways was due to non-significant effects of CBT compared to SPT on the hypothesized mediators at week 12 (i.e., non-significant a-paths, shown as parameters am2x in Figure 1; Table 4). Two mediation models, the ASI-R composite mean model and the ERT self-attribution SR model, indicated significant pathways from the hypothesized mediators at week 12 to BDD-YBOCS scores at week 12 and week 24 (i.e., b-paths). In both models, the concurrent effect estimate of the week 12 mediator on the week 12 outcome was positive (b-paths shown as parameter by2m2 in Figure 1; Table 4), indicating that higher ASI-R composite scores and higher ERT self-attribution in the self-referent condition were associated with concurrent higher BDD-YBOCS scores. Conversely, the effect estimates of the two significant week 12 hypothesized mediators on week 24 BDD-YBOCS scores were negative (b-paths shown as parameter by3m2 in Figure 1; Table 4), indicating that higher ASI-R composite scores and higher ERT self-attribution SR at mid-treatment were associated with subsequently lower BDD-YBOCS scores at post-treatment. However, due to the non-significant a-paths leading to these hypothesized mediators noted above, neither of these relationships mediated the treatment effect on BDD-YBOCS scores. The model fit was good for the mediation model that used ASI-R composite scores as a mediator (RMSEA=0.046, 95%CI: [0.000, 0.090]; Bentler-Bonett Non-Normed Index=0.974), marginal in the model that used ERT self-attribution SR as a mediator (fair fit based on RMSEA=0.066, 95%CI: [0.037, 0.106]; poor fit based on Bentler-Bonett Non-Normed Index=0.907), and relatively poor in all other models (Table 5).

Table 3.

Estimates of the intermediate (mediated effect 1) and longitudinal (mediated effect 2) causal mediation effects of CBT vs. SPT treatment on BDD symptom severity among 120 participants with BDD at baseline.

Mediated Effect 1 Mediated Effect 2
Mediator Estimate Effect size z-value p-value Estimate Effect size z-value p-value
ASI-R composite mean −0.85 −0.10 −0.77 0.779 0.36 0.04 0.76 0.224
ROCF organization copy 0.00 0.00 0.02 0.493 0.00 0.00 −0.02 0.508
ROCF organization delay 0.01 0.00 0.02 0.494 0.59 0.06 1.16 0.123
ERT accuracy, SR 0.22 0.03 0.59 0.279 −0.19 −0.02 −0.57 0.716
ERT self-attribution, SR −0.03 0.00 −0.03 0.513 0.01 0.00 0.03 0.487
Open in a new tab

Note: ASI-R = Appearance Schema Inventory - Revised; ROCF = Rey-Osterrieth Complex Figure test; ERT = Emotion Recognition Task; SR = self-referent scenarios.

Table 4.

Key paths in the causal mediation three-wave longitudinal models.

Mediator and Path Parameter Estimate 95% Confidence Interval t Value Pr > |t| Std. Estimate
Mediator: ASI-R composite mean scores
  Treatment ===> Mediator week 12 am2x −0.06 [−0.21, 0.09] −0.77 0.4392 −0.06
  Treatment ===> Mediator week 24 am3x −0.34 [−0.55, −0.13] −3.20 0.0014 −0.24
  Mediator week 12 ===> BDD-YBOCS week 12 by2m2 14.51 [11.01, 18.01] 8.12 <.0001 0.86
  Mediator week 12 ===> BDD-YBOCS week 24 by3m2 −6.25 [−9.37, −3.12] −3.92 <.0001 −0.32
  Treatment ===> BDD-YBOCS week 12 cy2x −2.95 [−5.49, −0.42] −2.28 0.0225 −0.17
  Treatment ===> BDD-YBOCS week 24 cy3x −0.38 [−2.81, 2.05] −0.31 0.7602 −0.02
Mediator: ROCF organization copy scores
  Treatment ===> Mediator week 12 am2x 0.01 [−0.62, 0.63] 0.02 0.9843 0.00
  Treatment ===> Mediator week 24 am3x −0.23 [−0.76, 0.31] −0.84 0.4032 −0.07
  Mediator week 12 ===> BDD-YBOCS week 12 by2m2 0.02 [−1.11, 1.16] 0.04 0.9668 0.00
  Mediator week 12 ===> BDD-YBOCS week 24 by3m2 −0.25 [−1.53, 1.02] −0.39 0.6948 −0.04
  Treatment ===> BDD-YBOCS week 12 cy2x −3.78 [−7.10, −0.46] −2.23 0.0257 −0.22
  Treatment ===> BDD-YBOCS week 24 cy3x −3.30 [−6.18, −0.43] −2.25 0.0243 −0.17
Mediator: ROCF organization delay scores
  Treatment ===> Mediator week 12 am2x −0.59 [−1.18, 0.00] −1.95 0.0512 −0.17
  Treatment ===> Mediator week 24 am3x 0.00 [−0.42, 0.42] −0.01 0.9912 0.00
  Mediator week 12 ===> BDD-YBOCS week 12 by2m2 −0.01 [−1.21, 1.20] −0.02 0.9875 0.00
  Mediator week 12 ===> BDD-YBOCS week 24 by3m2 −1.01 [−2.37, 0.36] −1.45 0.1481 −0.18
  Treatment ===> BDD-YBOCS week 12 cy2x −3.86 [−7.26, −0.46] −2.23 0.0261 −0.22
  Treatment ===> BDD-YBOCS week 24 cy3x −3.58 [−6.46, −0.69] −2.43 0.0151 −0.18
Mediator: ERT accuracy, self-referent
  Treatment ===> Mediator week 12 am2x 0.93 [−0.13, 1.99] 1.72 0.0859 0.16
  Treatment ===> Mediator week 24 am3x −0.59 [−1.46, 0.28] −1.33 0.1829 −0.12
  Mediator week 12 ===> BDD-YBOCS week 12 by2m2 0.23 [−0.50, 0.97] 0.62 0.5330 0.08
  Mediator week 12 ===> BDD-YBOCS week 24 by3m2 −0.20 [−0.84, 0.45] −0.60 0.5458 −0.06
  Treatment ===> BDD-YBOCS week 12 cy2x −4.13 [−7.54, −0.72] −2.37 0.0177 −0.23
  Treatment ===> BDD-YBOCS week 24 cy3x −2.87 [−5.70, −0.05] −2.00 0.0458 −0.14
Mediator: ERT self-attribution, self-referent
  Treatment ===> Mediator week 12 am2x −0.04 [−2.61, 2.52] −0.03 0.9743 0.00
  Treatment ===> Mediator week 24 am3x −1.75 [−4.01, 0.52] −1.51 0.1311 −0.13
  Mediator week 12 ===> BDD-YBOCS week 12 by2m2 0.78 [0.53, 1.03] 6.07 <.0001 0.60
  Mediator week 12 ===> BDD-YBOCS week 24 by3m2 −0.29 [−0.58, −0.01] −2.00 0.0458 −0.20
  Treatment ===> BDD-YBOCS week 12 cy2x −3.72 [−6.58, −0.85] −2.54 0.0111 −0.21
  Treatment ===> BDD-YBOCS week 24 cy3x −2.43 [−5.13, 0.27] −1.76 0.0776 −0.12
Open in a new tab

Note: ASI-R = Appearance Schema Inventory - Revised; ROCF = Rey-Osterrieth Complex Figure test; ERT = Emotion Recognition Task; BDD-YBOCS = Yale-Brown Obsessive-Compulsive Scale Modified for Body Dysmorphic Disorder. Paths that are shown in bold are pathways that are part of the mediated effects, while non-bolded pathways show other effects of treatment on either the potential mediator at post-treatment or on the outcome at either mid-treatment or post-treatment.

Table 5.

Fit indices and variance explained for the structural equation path models used to estimate causal mediation effects in longitudinal three-wave models.

ASI-R composite mean ROCF organization, copy ROCF organization, delay ERT accuracy, SR ERT self-attribution, SR
Chi-Square 7.501 27.513 12.389 21.872 9.155
Chi-Square DF 6 6 6 6 6
Pr > Chi-Square 0.277 0.000 0.054 0.001 0.165
RMSEA Estimate 0.046 0.173 0.094 0.149 0.066
RMSEA Lower 50% Confidence Limit 0.000 0.150 0.070 0.126 0.037
RMSEA Upper 50% Confidence Limit 0.090 0.203 0.129 0.180 0.106
Bentler Comparative Fit Index 0.994 0.864 0.963 0.895 0.980
Bentler-Bonett Non-normed Index 0.974 0.364 0.829 0.508 0.907
Mediator, week 12 (r2) 0.521 0.233 0.373 0.332 0.266
BDD-YBOCS total scores, week 24 (r2) 0.703 0.540 0.551 0.579 0.608
Open in a new tab

Note: ASI-R = Appearance Schema Inventory - Revised; ROCF = Rey-Osterrieth Complex Figure test; ERT = Emotion Recognition Task; SR = self-referent scenarios; DF = degrees of freedom; RMSEA = Root Mean Square Error of Approximation; BDD-YBOCS = Yale-Brown Obsessive-Compulsive Scale Modified for Body Dysmorphic Disorder.

Path estimates in the mediation models that used complete data only (i.e., no incidental missing data at baseline, no missing longitudinal outcomes) were remarkably similar to the intent-to-treat models that used full information maximum likelihood methods to account for missing data. The path estimates in the completer-only models were similar in magnitude and direction, and no path estimate changed in significance (Supplemental Table S4). In the multi-group mediation analyses by site, we were unable to detect any site-specific treatment effects on the mediators at week 12. The site-specific direct treatment effect estimates of CBT vs SPT on BDD-YBOCS scores (paths Cy2x and Cy3x, Figure 1) indicated that the overall treatment effects observed in the model of the combined sample were largely site-specific to RIH (Supplemental Table S6). A similar site-specificity was also observed for the pathway from treatment to mediators at week 24 (path am3x) in the models testing ASI-R composite mean scores or ERT accuracy SR as mediators, but not in the models for ROCF organization (copy or delay conditions) or ERT self-attribution SR. In the ASI-R model, CBT treatment was associated with lower ASI-R composite mean scores at week 24, but this effect was stronger at RIH (path am3x in the ASI-R model, estimate [95% CI]: −0.42 [−0.71, −0.12], p=0.0059) than at MGH (−0.26 [−0.54, 0.02], p=0.0670). In the ERT self-attribution SR model, CBT treatment was associated with lower ERT accuracy SR at week 24 than SPT, but this effect was stronger at RIH (path am3x in the ERT accuracy SR model, estimate [95% CI]: −1.13 [−2.21, −0.04], p=0.0415) than at MGH (−0.16 [−1.42, 1.09], p=0.7972). The site-specific mediation model estimates need to be interpreted with caution due to the low sample size and relatively poor model fit (Supplemental Table S7). However, the site-specificity of the treatment effect of CBT vs. SPT aligns well with the observed site difference in the treatment effect reported in the study’s main outcome paper (Wilhelm et al., 2019).

Hypothesized mediators as predictors of treatment response and remission

Of the 88 participants who completed the post-treatment assessment and who had complete baseline data, 71.6% (n=63) participants were classified as treatment responders and 64.8% (n=57) were classified as in remission from BDD symptoms at the post-treatment assessment (week 24). After accounting for treatment and site effects, as well as a potential treatment by site interaction, none of the hypothesized mediators assessed at baseline were significantly associated with treatment response (OR [95%CI]; ASI-R: 0.80 [0.44, 1.44]; ROCF organization copy: 1.36 [0.82, 2.24]; ERT accuracy SR: 0.90 [0.52, 1.56]; ERT self-attribution SR: 1.21 [0.69, 2.13]; all p>.20; c-statistic=.72). Similarly, none of the baseline values of hypothesized mediators assessed at baseline were significantly associated with remission from BDD (OR [95%CI]; ASI-R: 0.79 [0.46, 1.35]; ROCF organization copy: 1.29 [0.79, 2.10]; ERT accuracy SR: 0.65 [0.37, 1.14]; ERT self-attribution SR: 1.51 [0.87, 2.60]; all p>.10; c-statistic=.75).

Discussion

In this secondary data analysis of a randomized clinical trial of therapist-delivered in-person CBT vs. SPT for BDD (n=120), we examined treatment-related changes in detail processing, emotion recognition, and maladaptive appearance beliefs, and the potential role of these constructs as mediators of treatment response. To our knowledge, this is the first study to examine these constructs as therapy outcomes or mediators of treatment outcome in BDD. All constructs improved over time. Consistent with our hypothesis, CBT was associated with a greater reduction in maladaptive appearance beliefs (ASI-R) compared to SPT. However, contrary to our hypothesis, changes in detail processing and emotion recognition were not significantly greater for CBT than for SPT, and none of the constructs examined mediated treatment response.

It is encouraging that those who improved with treatment for BDD demonstrated more holistic (ROCF) and emotionally accurate (ERT) processing, regardless of which treatment they received. This suggests that these neurocognitive deficits in BDD are malleable and supports earlier findings that emotion recognition deficits in BDD can be modified with training (Buhlmann et al., 2011). It is possible that common factors (e.g., making a patient feel supported and understood) enabled patients in both treatments to develop more holistic, adaptive beliefs. Although SPT does not involve explicit training in cognitive, attentional, or behavioral skills, repeated encounters with a non-judgmental, supportive therapist, irrespective of theoretical orientation, could mimic the effect of certain CBT skills by providing a broader focus in sessions beyond appearance concerns (e.g., self-esteem, general life issues) and by providing corrective social feedback that contradicts patients’ feared beliefs of rejection. Of note, treatment fidelity in the study was high, indicating no cross-contamination between conditions (Wilhelm et al., 2019). Overall improvements over time in detail processing, emotion recognition, and maladaptive appearance beliefs might be a combined effect of the benefits of common factors of therapy and, particularly for the ROCF, practice effects. However, results for the ASI-R also suggest distinct therapeutic effects of CBT.

Learning to disengage from and reevaluate unhelpful beliefs is a core component of CBT-BDD, which is addressed via specific cognitive strategies (mindfulness, cognitive restructuring, advanced strategies to target core beliefs). Thus, it is unsurprising that changes in maladaptive appearance beliefs over time were greater for CBT than for SPT. However, contrary to prior findings in BDD (Bernstein et al., 2021; Fang et al., 2020) and OCD (Wilhelm et al., 2015), changes in maladaptive beliefs did not mediate the treatment effect on symptom improvement in the current study. Using the same sample as the current study, Bernstein et al. (2021) showed CBT was directly associated with more adaptive efforts to disengage from unhelpful thoughts and to resist BDD rituals than SPT. Notably, Bernstein et al. (2021) examined longitudinal shifts that occurred during treatment in different symptom components of the primary outcome (BDD-YBOCS). The BDD-YBOCS assesses BDD symptoms (e g., maladaptive appearance beliefs and behaviors) specifically targeted by CBT, and thus could inflate the positive effects of CBT. The current study, on the other hand, examined neurocognitive processes (detail processing, maladaptive beliefs, and emotion recognition), which are distinct from the outcome (i.e., BDD-YBOCS), as mechanistic mediators.

None of the constructs examined predicted treatment response. Individuals with greater deficits in detail processing, emotion recognition, and maladaptive appearance beliefs were as likely to benefit from treatment as those with less severe pathology. Changes in most of these constructs also occurred later in the treatment. Thus, clinicians should not be discouraged by initial neurocognitive impairment, nor lack of early improvement in these processes. This aligns with a recent finding, based on a separate analysis of early responders to CBT in a sample that combined participants from the study described herein as well as participants from two smaller CBT trials, that minimal reduction in BDD symptom severity early in CBT is not indicative of eventual non-response (Greenberg et al., 2022).

The lack of significant mediators of the treatment effect in the current study could be explained by a few factors. First, we were likely underpowered for mediation. The primary trial (Wilhelm et al., 2019) was not powered for this exploratory aim, and our sample size was small for mediation modeling, making it likely we would detect only relatively strong effects (Fritz & Mackinnon, 2007).

Second, baseline impairments in ROCF and ERT accuracy in the current sample were less severe than those reported in prior studies (Buhlmann et al., 2006; Deckersbach et al., 2000; Greenberg et al., 2018). Mixed findings regarding neurocognitive deficits in BDD likely stem in part from methodological differences (e.g., sample size, clinical characteristics) and may also reflect the heterogeneous nature of neurocognitive functioning in BDD, which can range from intact to markedly impaired (Malcolm et al., 2021). Anecdotally, participants in the current study often reported that the ERT task was repetitive and long, so it is possible some participants may not have fully attended to the task, thus compromising the quality of the data. Future research could use a wider battery of neuropsychological tests, focusing on tests that can be administered quickly, easily, and repeatedly, without practice effects. Selection of tasks should carefully balance potential benefit with potential participant burden.

Third, although trials of CBT for BDD have demonstrated continued, linear progress over the full course of treatment, up to 50% of response may be achieved by 12 weeks (Wilhelm et al., 2014). Thus, a week 12 timepoint may be too late to adequately capture mediation. Given that neurocognitive changes in the current study occurred after the mid-point, future studies should measure potential mediators of treatment earlier in the treatment (e.g., Week 4) before we would expect to see significant symptom change (Greenberg et al., 2022).

Fourth, in the current study, we hypothesized that changes in maladaptive beliefs about appearance and deficits in detail and emotion processing would lead to changes in BDD symptom severity. However, clinical observations suggest that the converse could also be true. It may be the case that BDD patients are initially too rigid or too ill to consider reevaluating their beliefs or shifting their attentional focus until their BDD symptoms have improved and they are less distressed, later in the treatment, which might then bring about changes in detail processing, emotion recognition, and maladaptive beliefs. Indeed, in the current study, participants who were still worse off in terms of ASI-R and ERT self-attribution at mid-treatment (week 12) had significantly lower BDD-YBOCS scores at post-treatment, suggesting that these symptoms changed later in treatment and were associated with concurrent improvements in BDD symptoms. These changes also correspond with the timing of the CBT-BDD protocol, in which mindfulness/attention retraining, which likely leads to a reduction in detail processing on the ROCF, is often not introduced until later in treatment (e.g., week 10 or later). Similarly, automatic thoughts (e.g., “the cashier just stared at my pimple”) are addressed early in treatment, whereas intermediary rules and assumptions (“I must be attractive to be happy”) and deeper-level core beliefs (“I’m worthless”) are typically not addressed until after the mid-treatment (week 12) timepoint, when patients are usually better poised to reevaluate deeply held beliefs around the importance and influence of appearance. It is also possible that changes in BDD behaviors contribute to subsequent neurocognitive changes. Changes in behavior, rather than cognitive change, might also explain the more robust effects of CBT vs SPT. In a recent network analysis examining mechanisms of CBT vs SPT for BDD in the same study sample used herein, Bernstein and colleagues (2021) found that the observed treatment differences in favor of CBT were largely due to the unique impact of CBT vs. SPT in reducing rituals. More work is needed to understand the direct and downstream contributions of different components of our current treatment protocol.

Lastly, the lack of significant mediation in our study does not imply that the mechanistic variables we examined are not important or relevant. As described above, there are a number of reasons why the treatment effect on week 12 mediators was largely not detectable in our study. Still, all the mediators (detail processing, maladaptive appearance beliefs, and emotion recognition) improved over time and were associated with reduction in BDD symptom severity.

Limitations of the current study include a predominantly female, well-educated sample, which may not be fully representative of all individuals with BDD, and use of an enhanced version of SPT, which likely bolstered the effects of SPT in this study compared to SPT more typically received in the community. As previously noted, our sample size was low for estimating complex structural equation models, especially given the site differences in the treatment effect reported in the study’s main outcome paper (Wilhelm et al., 2019). Thus, the chance of type II error is relatively high. Also as noted, baseline impairments in detail processing (ROCF organization) and emotion recognition (ERT accuracy) were less severe than in other BDD samples described in the literature (Buhlmann et al., 2006; Deckersbach et al., 2000; Greenberg et al., 2018). With regard to model fit for the mediation models, Bentler’s NFI and NNFI, as well as RMSEA are all sensitive to sample size and underestimate model fit for small samples (Bentler, 1990; Kenny et al., 2015).

In conclusion, successful treatment of BDD was associated with reduced impairments in detail processing, emotion recognition, and maladaptive appearance beliefs, regardless of whether CBT or SPT was received. However, CBT was more effective than SPT in reducing maladaptive appearance beliefs. Taken together with results from the main outcome paper that CBT was consistently and highly effective in reducing BDD symptom severity (Wilhelm et al., 2019), our results support the importance of targeting maladaptive beliefs around the importance and influence of attractiveness in the treatment of BDD (Hrabosky et al., 2009; Wilhelm et al., 2019). The variables examined in this study do not likely capture the full range or heterogeneity of neurocognitive processes in BDD. Thus, more work is needed to understand how these different treatments work. Future research should identify reliable, easily measured neurocognitive targets of treatment and look for other early indicators of treatment response that could explain symptom improvement, for example self-esteem (Buhlmann et al., 2009), self-efficacy (Bandura et al., 1977; Bouchard et al., 2007), or cognitive flexibility (e.g., set shifting) (Greenberg et al., 2018). Understanding how treatment works could help optimize treatment response, as clinicians could emphasize components directly contributing to symptom change earlier, or more intensely, in the treatment, and potentially develop new interventions to target these mechanisms more potently and efficiently.

Supplementary Material

1

Highlights.

  • Examined mediators of CBT vs supportive psychotherapy for body dysmorphic disorder

  • Both treatments improved neurocognitive functioning and maladaptive beliefs

  • CBT was more effective than SPT in improving maladaptive appearance beliefs

  • There were no significant mediators of symptom improvement

Role of the funding source

This research was supported by a National Institute of Mental Health Collaborative R01 grant to Dr. Wilhelm (R01MH091078) and Dr. Phillips (R01 MH091023). The sponsor had no role in study design; in the collection, analysis and interpretation of data; in the writing of the report; or in the decision to submit the article for publication.

Disclosures

Jennifer L. Greenberg: Dr. Greenberg has received salary support from Koa Health and is a presenter for the Massachusetts General Hospital Psychiatry Academy in educational programs supported through independent medical education grants from pharmaceutical companies.

Katharine A. Phillips: Dr. Phillips receives book royalties from Oxford University Press, International Creative Management, Inc., American Psychiatric Association Publishing, and Guilford Press. She receives writing royalties from UpToDate/Wolter’s Kluwer and has received writing honoraria from the Merck Manual and from Simple and Practical Medical Education. She has received honoraria for scale use from Nview Health and OCD scales. She has also received speaking honoraria from Informa Exhibitions and from academic institutions and professional organizations.

Susanne S. Hoeppner: None

Nicholas C. Jacobson: None

Angela Fang: None

Sabine Wilhelm: Dr. Wilhelm is a presenter for the Massachusetts General Hospital Psychiatry Academy in educational programs supported through independent medical education grants from pharmaceutical companies; she has received royalties from Elsevier Publications, Guilford Publications, New Harbinger Publications, Springer, and Oxford University Press. Dr. Wilhelm has also received speaking honoraria from various academic institutions and foundations, including the International Obsessive Compulsive Disorder Foundation, Tourette Association of America, and Brattleboro Retreat. In addition, she received payment from the Association for Behavioral and Cognitive Therapies for her role as Associate Editor for the Behavior Therapy journal, as well as from John Wiley & Sons, Inc. for her role as Associate Editor on the journal Depression & Anxiety. Dr. Wilhelm has also received honoraria from One-Mind for her role in PsyberGuide Scientific Advisory Board. Dr. Wilhelm has received research and salary support from Novartis and Koa Health.

Footnotes

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

CRediT Author Statement

Jennifer Greenberg: Conceptualization, Methodology, Investigation, Writing—Original Draft, Writing—Review & Editing, Project administration

Katharine A. Phillips: Conceptualization, Methodology, Investigation, Writing—Review & Editing, Funding Acquisition, Resources

Susanne S. Hoeppner: Conceptualization, Methodology, Data curation, Formal analysis, Writing—Review & Editing

Nicholas C. Jacobson: Formal analysis, Writing—Review & Editing

Angela Fang: Writing—Original Draft, Writing—Review & Editing

Sabine Wilhelm: Conceptualization, Methodology, Investigation, Writing—Review & Editing, Funding Acquisition, Resources

References

  1. American Psychiatric Association. (2013). Diagnostic and statistical manual of mental health disorders (5th ed.). [Google Scholar]
  2. Bandura A, Adams NE, & Beyer J (1977). Cognitive processes mediating behavioral change. Journal of Personality and Social Psychology, 35(3), 125–139. 10.1037//0022-3514.35.3.125 [DOI] [PubMed] [Google Scholar]
  3. Beilharz F, Castle DJ, Grace S, & Rossell SL (2017). A systematic review of visual processing and associated treatments in body dysmorphic disorder. Acta Psychiatrica Scandinavica, 136(1), 16–36. 10.1111/acps.12705 [DOI] [PubMed] [Google Scholar]
  4. Bentler PM (1990). Comparative fit indexes in structural models. Psychological Bulletin, 107(2), 238–246. 10.1037/0033-2909.107.2.238 [DOI] [PubMed] [Google Scholar]
  5. Bernstein E, Phillips K, Greenberg J, Curtiss J, Hoeppner S, & Wilhelm S (2021). Mechanisms of cognitive-behavioral therapy effects on symptoms of body dysmorphic disorder: A network intervention analysis. Psychological Medicine, 1–9. doi: 10.1017/S0033291721004451 [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Bouchard S, Gauthier J, Nouwen A, Ivers H, Valliéres A, Simard S, & Fournier T (2007). Temporal relationship between dysfunctional beliefs, self-efficacy and panic apprehension in the treatment of panic disorder with agoraphobia. Journal of Behavior Therapy and Experimental Psychiatry, 38(3), 275–292. 10.1016/j.jbtep.2006.08.002 [DOI] [PubMed] [Google Scholar]
  7. Buhlmann U, Etcoff NL, & Wilhelm S (2006). Emotion recognition bias for contempt and anger in body dysmorphic disorder. Journal of Psychiatric Research, 40(2), 105–111. 10.1016/j.jpsychires.2005.03.006 [DOI] [PubMed] [Google Scholar]
  8. Buhlmann U, Etcoff NL, & Wilhelm S (2008). Facial attractiveness ratings and perfectionism in body dysmorphic disorder and obsessive-compulsive disorder. Journal of Anxiety Disorders, 22(3), 540–547. 10.1016/j.janxdis.2007.05.004 [DOI] [PubMed] [Google Scholar]
  9. Buhlmann U, Glaesmer H, Mewes R, Fama JM, Wilhelm S, Brähler E, & Rief W (2010). Updates on the prevalence of body dysmorphic disorder: a population-based survey. Psychiatry Research, 178(1), 171–175. 10.1016/j.psychres.2009.05.002 [DOI] [PubMed] [Google Scholar]
  10. Buhlmann U, Gleiss MJ, Rupf L, Zschenderlein K, & Kathmann N (2011). Modifying emotion recognition deficits in body dysmorphic disorder: an experimental investigation. Depression and Anxiety, 28(10), 924–931. 10.1002/da.20887 [DOI] [PubMed] [Google Scholar]
  11. Buhlmann U, & Hartmann A (2017). Cognitive and emotional processing in body dysmorphic disorder. In: Body Dysmorphic Disorder: Advances in Research and Clinical Practice. Phillips KA, editor. Oxford University Press. 10.1093/med/9780190254131.003.0022 [DOI] [Google Scholar]
  12. Buhlmann U, McNally RJ, Etcoff NL, Tuschen-Caffier B, & Wilhelm S (2004). Emotion recognition deficits in body dysmorphic disorder. Journal of Psychiatric Research, 38(2), 201–206. 10.1016/s0022-3956(03)00107-9 [DOI] [PubMed] [Google Scholar]
  13. Buhlmann U, Teachman BA, Naumann E, Fehlinger T, & Rief W (2009). The meaning of beauty: implicit and explicit self-esteem and attractiveness beliefs in body dysmorphic disorder. Journal of Anxiety Disorders, 23(5), 694–702. 10.1016/j.janxdis.2009.02.008 [DOI] [PubMed] [Google Scholar]
  14. Buhlmann U, Winter A, & Kathmann N (2013). Emotion recognition in body dysmorphic disorder: application of the Reading the Mind in the Eyes Task. Body Image, 10(2), 247–250. 10.1016/j.bodyim.2012.12.001 [DOI] [PubMed] [Google Scholar]
  15. Cash TF, Melnyk SE, & Hrabosky JI (2004). The assessment of body image investment: an extensive revision of the appearance schemas inventory. The International Journal of Eating Disorders, 35(3), 305–316. 10.1002/eat.10264 [DOI] [PubMed] [Google Scholar]
  16. Deckersbach T, Savage CR, Henin A, Mataix-Cols D, Otto MW, Wilhelm S, Rauch SL, Baer L, & Jenike MA (2000). Reliability and validity of a scoring system for measuring organizational approach in the Complex Figure Test. Journal of Clinical and Experimental Neuropsychology, 22(5), 640–648. 10.1076/1380-3395(200010)22:5;1-9;FT640 [DOI] [PubMed] [Google Scholar]
  17. Deckersbach T, Savage CR, Phillips KA, Wilhelm S, Buhlmann U, Rauch SL, Baer L, & Jenike MA (2000). Characteristics of memory dysfunction in body dysmorphic disorder. Journal of the International Neuropsychological Society, 6(6), 673–681. 10.1017/s1355617700666055 [DOI] [PubMed] [Google Scholar]
  18. Denman SB (1984). Denman Neuropsychology Memory Scale, S.B. Denman, Charleston, SC [Google Scholar]
  19. Ekman P, & Friesen W (1975). Unmasking the face: A guide to recognizing emotions from facial cues. Prentice-Hall. [Google Scholar]
  20. Ekman P, & Friesen W (1976). Pictures of facial affect. Consulting Psychologists Press. [Google Scholar]
  21. Enander J, Andersson E, Mataix-Cols D, Lichtenstein L, Alström K, Andersson G, Ljótsson B, & Rück C (2016). Therapist guided internet based cognitive behavioural therapy for body dysmorphic disorder: single blind randomised controlled trial. BMJ, 352, i241. 10.1136/bmj.i241 [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Fang A, Steketee G, Keshaviah A, Didie E, Phillips KA, & Wilhelm S (2020). Mechanisms of change in cognitive behavioral therapy for body dysmorphic disorder. Cognitive Therapy and Research, 44(3), 596–610. 10.1007/s10608-020-10080-w [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Feusner JD, Moller H, Altstein L, Sugar C, Bookheimer S, Yoon J, & Hembacher E (2010). Inverted face processing in body dysmorphic disorder. Journal of Psychiatric Research, 44(15), 1088–1094. 10.1016/j.jpsychires.2010.03.015 [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Feusner JD, Moody T, Hembacher E, Townsend J, McKinley M, Moller H, & Bookheimer S (2010). Abnormalities of visual processing and frontostriatal systems in body dysmorphic disorder. Archives of General Psychiatry, 67(2), 197–205. 10.1001/archgenpsychiatry.2009.190 [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. First MB, Gibbon M, Spitzer RL, Williams JBW, & Benjamin LS (1997). Structured clinical interview for DSM-IV Axis II personality disorders (SCID-II). American Psychiatric Press, Inc. [Google Scholar]
  26. First MB, Spitzer RL, Gibbon M and Williams J (2002) Structured clinical interview for DSM-IV-TR Axis I disorders, research version. Biometrics Research. [Google Scholar]
  27. Flygare O, Enander J, Andersson E, Ljótsson B, Ivanov VZ, Mataix-Cols D, & Rück C (2020). Predictors of remission from body dysmorphic disorder after internet-delivered cognitive behavior therapy: a machine learning approach. BMC Psychiatry, 20(1), 247. 10.1186/s12888-020-02655-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Fritz MS, & Mackinnon DP (2007). Required Sample Size to Detect the Mediated Effect. Psychological Science, 18(3), 233–239. 10.1111/j.1467-9280.2007.01882.x [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Grace SA, Toh WL, Buchanan B, Castle DJ, & Rossell SL (2019). Impaired recognition of negative facial emotions in body dysmorphic disorder. Journal of the International Neuropsychological Society, JINS 25(08), 884–889. 10.1017/S1355617719000419 [DOI] [PubMed] [Google Scholar]
  30. Greenberg JL, Reuman L, Hartmann AS, Kasarskis I, & Wilhelm S (2014). Visual hot spots: an eye tracking study of attention bias in body dysmorphic disorder. Journal of Psychiatric Research, 57, 125–132. 10.1016/j.jpsychires.2014.06.015 [DOI] [PubMed] [Google Scholar]
  31. Greenberg JL, Weingarden H, Reuman L, Abrams D, Mothi SS, & Wilhelm S (2018). Set shifting and visuospatial organization deficits in body dysmorphic disorder. Psychiatry Research, 260, 182–186. 10.1016/j.psychres.2017.11.062 [DOI] [PubMed] [Google Scholar]
  32. Greenberg JL, Jacobson NC, Hoeppner SS, Bernstein EE, Snorrason I, Schwartzberg A, Steketee G, Phillips KA, & Wilhelm S (2022). Early response to cognitive behavioral therapy for body dysmorphic disorder as a predictor of outcomes. Journal of Psychiatric Research, 152, 7–13. 10.1016/j.jpsychires.2022.06.001. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Grocholewski A, Kliem S, & Heinrichs N (2012). Selective attention to imagined facial ugliness is specific to body dysmorphic disorder. Body Image, 9(2), 261–269. 10.1016/j.bodyim.2012.01.002 [DOI] [PubMed] [Google Scholar]
  34. Gunstad J, & Phillips KA (2003). Axis I comorbidity in body dysmorphic disorder. Comprehensive Psychiatry, 44(4), 270–276. 10.1016/S0010-440X(03)00088-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Hanes KR (1998). Neuropsychological performance in body dysmorphic disorder. Journal of the International Neuropsychological Society, 4(2), 167–171. 10.1017/s1355617798001672 [DOI] [PubMed] [Google Scholar]
  36. Harrison A, Fernández de la Cruz L, Enander J, Radua J, & Mataix-Cols D (2016). Cognitive-behavioral therapy for body dysmorphic disorder: A systematic review and meta-analysis of randomized controlled trials. Clinical Psychology Review, 48, 43–51. 10.1016/j.cpr.2016.05.007 [DOI] [PubMed] [Google Scholar]
  37. Hrabosky JI, Cash TF, Veale D, Neziroglu F, Soil EA, Gamer DM, Strachan-Kinser M, Bakke B, Clauss LJ, & Phillips KA (2009). Multidimensional body image comparisons among patients with eating disorders, body dysmorphic disorder, and clinical controls: a multisite study. Body Image, 6(3), 155–163. 10.1016/j.bodyim.2009.03.001 [DOI] [PubMed] [Google Scholar]
  38. Hu LT, & Bentler PM (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling: A Multidisciplinary Journal, 6, 1–55. 10.1080/10705519909540118 [DOI] [Google Scholar]
  39. Jefferies K, Laws KR, & Fineberg NA (2012). Superior face recognition in Body Dysmorphic Disorder. Journal of Obsessive-Compulsive and Related Disorders, 1(3), 175–179. 10.1016/j.jocrd.2012.03.002 [DOI] [Google Scholar]
  40. Ji L, Chow S-M, Schermerhorn AC, Jacobson NC, & Cummings EM (2018). Handling missing data in the modeling of intensive longitudinal data. Structural Equation Modeling. 10.1080/10705511.2017.1417046 [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Kelly MM, Didie ER, & Phillips KA (2014). Personal and appearance-based rejection sensitivity in body dysmorphic disorder. Body Image, 11(3), 260–265. 10.1016/j.bodyim.2014.03.004 [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Kenny DA, Kaniskan B, & McCoach DB (2015). The performance of RMSEA in models with small degrees of freedom. Sociological Methods & Research, 44(3), 486–507. 10.1177/0049124114543236 [DOI] [Google Scholar]
  43. Kerwin L, Hovav S, Hellemann G, & Feusner JD (2014). Impairment in local and global processing and set-shifting in body dysmorphic disorder. Journal of Psychiatric Research, 57, 41–50. 10.1016/j.jpsychires.2014.06.003 [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Koran LM, Abujaoude E, Large MD, & Serpe RT (2008). The prevalence of body dysmorphic disorder in the United States adult population. CNS Spectrums, 13(4), 316–322. 10.1017/s1092852900016436 [DOI] [PubMed] [Google Scholar]
  45. MacCallum RC, Browne MW, and Sugawara H,M (1996). Power analysis and determination of sample size for covariance structure modeling. Psychological Methods, 1(2), 130–149. [Google Scholar]
  46. Malcolm A, Pikoos T, Castle DJ, & Rossell SL (2021). Cross-diagnostic cognitive heterogeneity in body dysmorphic disorder and obsessive-compulsive disorder. Journal of Behavior Therapy and Experimental Psychiatry, 73, 101674. 10.1016/j.jbtep.2021.101674 [DOI] [PubMed] [Google Scholar]
  47. Osterrieth PA (1944). Le test de copie d’une figure complex : Contribution a l’etude de la perception et de la memoire [The Complex Figure Test: Contribution to the study of perception and memory]. Archives de Psychologie, 28, 1021–1034. [Google Scholar]
  48. Phillips KA, Coles ME, Menard W, Yen S, Fay C, & Weisberg RB (2005). Suicidal ideation and suicide attempts in body dysmorphic disorder. The Journal of Clinical Psychiatry, 66(6), 717–725. 10.4088/jcp.v66n0607 [DOI] [PubMed] [Google Scholar]
  49. Phillips KA, Hart AS, & Menard W (2014). Psychometric evaluation of the Yale–Brown Obsessive-Compulsive Scale Modified for Body Dysmorphic Disorder (BDD-YBOCS). Journal of Obsessive-Compulsive and Related Disorders, 3, 205–208. 10.1016/j.jocrd.2014.04.004 [DOI] [Google Scholar]
  50. Phillips KA, Hollander E, Rasmussen SA, & Aronowitz BR (1997). A severity rating scale for body dysmorphic disorder: Development, reliability, and validity of a modified version of the Yale-Brown Obsessive Compulsive Scale. Psychopharmacology Bulletin, 33, 17–22. 725. [PubMed] [Google Scholar]
  51. Phillips KA, McElroy SL, Keck PE Jr, Pope HG Jr, & Hudson JI (1993). Body dysmorphic disorder: 30 cases of imagined ugliness. The American Journal of Psychiatry, 150(2), 302–308. 10.1176/ajp.150.2.302 [DOI] [PubMed] [Google Scholar]
  52. Pinsker H (1997). A Primer of Supportive Psychotherapy. Hillsdale, NJ: The Analytic Press. [Google Scholar]
  53. Savage CR, Baer L, Keuthen NJ, Brown HD, Rauch SL, & Jenike MA (1999). Organizational strategies mediate nonverbal memory impairment in obsessive–compulsive disorder. Biological Psychiatry, 45(7), 905–916. 10.1016/s0006-3223(98)00278-9 [DOI] [PubMed] [Google Scholar]
  54. Schieber K, Kollei I, de Zwaan M, & Martin A (2015). Classification of body dysmorphic disorder - what is the advantage of the new DSM-5 criteria? Journal of Psychosomatic Research, 78(3), 223–227. 10.1016/j.jpsychores.2015.01.002 [DOI] [PubMed] [Google Scholar]
  55. Simberlund J, & Hollander E (2017). The relationship of body dysmorphic disorder to obsessive-compulsive disorder and the concept of the obsessive-compulsive spectrum. In Body Dysmorphic Disorder: Advances in Research and Clinical Practice. Phillips KA, editor. Oxford University Press. [Google Scholar]
  56. Toh WL, Castle DJ, & Rossell SL (2017). Face and object perception in body dysmorphic disorder versus obsessive-compulsive disorder: The Mooney Faces Task. Journal of the International Neuropsychological Society, JINS, 23(6), 471–480. 10.1017/S1355617717000327 [DOI] [PubMed] [Google Scholar]
  57. Valente MJ, & MacKinnon DP (2018). SAS(®) Macros for Computing Causal Mediated Effects in Two- and Three-Wave Longitudinal Models. SAS global forum, 2018, 2499. https://pubmed.ncbi.nlm.nih.gov/30221259 [PMC free article] [PubMed] [Google Scholar]
  58. Weingarden H, Hoeppner SS, Snorrason I, Greenberg JL, Phillips KA, & Wilhelm S (2021). Rates of remission, sustained remission, and recurrence in a randomized controlled trial of cognitive behavioral therapy versus supportive psychotherapy for body dysmorphic disorder. Depression and Anxiety, 38, 700–707. 10.1002/da.23148 [DOI] [PMC free article] [PubMed] [Google Scholar]
  59. Wilhelm S, Berman NC, Keshaviah A, Schwartz RA, & Steketee G (2015). Mechanisms of change in cognitive therapy for obsessive compulsive disorder: role of maladaptive beliefs and schemas. Behaviour Research and Therapy, 65, 5–10. 10.1016/j.brat.2014.12.006 [DOI] [PMC free article] [PubMed] [Google Scholar]
  60. Wilhelm S, Phillips KA, Didie E, Buhlmann U, Greenberg JL, Fama JM, Keshaviah A, & Steketee G (2014). Modular cognitive-behavioral therapy for body dysmorphic disorder: a randomized controlled trial. Behavior Therapy, 45(3), 314–327. 10.1016/j.beth.2013.12.007 [DOI] [PMC free article] [PubMed] [Google Scholar]
  61. Wilhelm S, Phillips KA, Greenberg JL, O’Keefe SM, Hoeppner SS, Keshaviah A, …, Schoenfeld DA, 2019. Efficacy and posttreatment effects of therapist-delivered cognitive behavioral therapy vs supportive psychotherapy for adults with body dysmorphic disorder: A randomized clinical trial. JAMA Psychiatry, 76, 363–373. 10.1001/jamapsychiatry.2018.4156. [DOI] [PMC free article] [PubMed] [Google Scholar]
  62. Wilhelm S, Phillips K, & Steketee GA (2013). Cognitive Behavioral Treatment Manual for Body Dysmorphic Disorder. Guilford Press. [Google Scholar]
  63. Woody S,R, Whittal ML, & McLean PD (2011). Mechanisms of symptom reduction in treatment for obsessions. Journal of Consulting and Clinical Psychology, 79, 653–664. 10.1037/a0024827 [DOI] [PubMed] [Google Scholar]
  64. Yousefi A, Rossell SL, Fakour Y, Ashayeri H, Naimijoo P, Fathi H, Toh WL, & Asgharnejad-Farid A (2020). Examining memory performance in body dysmorphic disorder (BDD): A comparison study with obsessive compulsive disorder (OCD). Asian Journal of Psychiatry, 53, 102–110. 10.1016/j.ajp.2020.102110 [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

1
NIHMS1865145-supplement-1.docx (74.3KB, docx)

RESOURCES