Who Benefits from a Digital Intervention for Appearance Concerns? A Randomized Controlled Trial of a Large Mixed-Gender Sample

Abstract

Objective:

Individuals with excessive appearance concerns commonly engage in appearance-related safety behaviors (ARSBs), or behaviors aimed at managing or avoiding negative evaluation of one’s physical appearance. Targeting these behaviors can lead to reductions in appearance concerns and related symptoms. However, prior studies have only examined these effects among women scoring above a specific cutoff, excluding men and individuals with lower levels of these concerns. The present study further evaluated the efficacy of this brief ARSB intervention and examined who benefits from this intervention.

Method:

Adults (N=400) interested in treatment for appearance concerns were recruited across the United States and randomized to a 1-month smartphone-based intervention targeting ARSBs or a four session relaxation video control.

Results:

Baseline appearance concerns moderated condition effects such that, among those with elevated concerns, ARSB fading led to lower appearance concerns than control at post, while conditions did not differ among those with low concerns. Probing the moderated effect using the Johnson-Neyman technique indicated ARSB fading led to significantly lower appearance concerns at post among those scoring in the upper two-thirds in appearance concerns (SMD=0.39). ARSB fading also led to lower body dissatisfaction (SMD=0.15–0.19) at post and follow-up, and among those high in appearance concerns, ARSB fading led to lower bulimic symptoms at post (SMD=0.34). No demographic factor moderated effects of treatment, indicating comparable efficacy between men and women.

Conclusions:

These findings further support the efficacy of ARSB fading and provide novel evidence for individuals who are likely to benefit from this intervention.

Concerns with one’s physical appearance are a common experience (Goodwin et al., 2020), but excessive appearance concerns are prominent in several psychological conditions. Appearance concerns are a central feature of body dysmorphic disorder (BDD), with affected individuals placing excessive importance on appearance and experiencing negative consequences related to how they believe others perceive them (American Psychiatric Association [APA], 2022; Veale, 2002). Likewise, those with eating disorders (EDs) often display cognitive distortions involving the overvaluation of body shape and weight (Linardon et al., 2018). Appearance-related concerns are also prevalent among individuals with social anxiety disorder (SAD; Moscovitch et al., 2013). As in BDD and EDs, appearance concerns in SAD may contribute to maladaptive behaviors, including social avoidance, which serve to maintain the disorder (Moscovitch et al., 2009). In fact, a recent study found that appearance concerns may maintain social anxiety symptoms and change in these concerns were indirectly related to changes in social anxiety (Patel, Zech, & Cougle, 2025). Another study found that individuals high in appearance concerns are significantly more likely to meet criteria for BDD, EDs, and SAD compared to those low in appearance concerns (Patel, Moore, & Cougle, 2025). Taken together, these findings suggest that appearance concerns function as a transdiagnostic construct and targeting them may promote improvement across appearance-related psychopathology (i.e., BDD, EDs, SAD).

A behavioral feature that is common among individuals with elevated appearance concerns is appearance-related safety behavior (ARSB). ARSBs are behavioral strategies that are used by individual anxious about their appearance to reduce or avoid this fear or any perceived negative consequences of negative evaluation of appearance (Patel et al., 2022). Example behaviors include social comparison, excessive grooming, mirror gazing, body checking, seeking reassurance from others, and avoidance. While ARSBs may reduce anxiety in the short term (Patel & Cougle, 2024a), they can serve to actually maintain or increase appearance concerns over time (Legenbauer et al., 2017; Stentz et al., 2022; Summers & Cougle, 2018). ARSBs may also lead an individual to falsely attribute safety to engaging in these behaviors. For example, a person may think that they are accepted by a friend due to repeated body checking.

Importantly, ARSBs have been found to be strongly related to appearance concerns, BDD, ED, and SAD symptoms (Patel et al., 2022). Additionally, several studies have demonstrated that these behaviors may have a causal role in appearance-related psychopathology. For example, one study found that instructions to increase ARSBs led to greater subsequent BDD symptoms, SAD symptoms, and body dissatisfaction compared to two other active control conditions (Summers & Cougle, 2018). Further, two studies, using clinical and non-clinical samples, showed that instructions to reduce these behaviors led to appearance concerns and related psychopathology compared to control conditions (Patel & Cougle 2024a; Wilver et al., 2020). These behaviors have also been found to predict symptoms of BDD following treatment (Patel et al., 2023), suggesting they can maintain symptoms if not properly addressed. These studies provide compelling evidence suggesting that targeting ARSBs will lead to changes across multiple symptom domains.

Notably, recent studies have examined the efficacy of directly targeting these behaviors as a treatment strategy among women. These studies included rationale for reducing ARSB use and instructed participants to reduce them over the course of one month using daily reminders (i.e., ARSB fading) to do so (Patel & Cougle, 2024b, 2025a, 2025b). Notably, these studies have found that participants instructed to reduce ARSBs experience reductions in appearance concerns, ED symptoms, SAD symptoms, and depression. Further, a recent randomized clinical trial (RCT) evaluated the efficacy of ARSB fading compared to an active control condition (reducing unhealthy behaviors; Patel & Cougle, 2024b). This study recruited a large sample of women and found that compared to this control condition, individuals in the ARSB fading condition reported greater reductions in appearance concerns, ED symptoms, social anxiety, and appearance importance.

While ARSB fading shows promise, more evidence is needed in determining who may actually benefit from this treatment. Importantly, most of the work on ARSB fading has focused primarily on women (Patel & Cougle, 2024b, 2025a; Patel et al., 2024; Stentz et al., 2022; Summers et al., 2018; Wilver et al., 2020), despite research demonstrating that men similarly experience appearance concerns (Patel, Robison, & Cougle, 2025). Additionally, a recent study found that ARSBs function similarly across gender and demonstrated that ARSB fading may be efficacious in men (Patel & Cougle, 2025b). However, no RCT targeting ARSBs has included men, thus far. Another limitation of existing research is that these studies have used a cutoff representing the upper quartile of appearance concerns (i.e., score of 53 on Social Appearance Anxiety Scale), drawn from one sample of data (Patel, Robison, & Cougle, 2025). In the previously mentioned RCT (Patel & Cougle, 2024b), 328 individuals were excluded from participation for not meeting this cutoff, despite the absence of evidence indicating whether they would benefit from this treatment. While it is possible that floor effects may reduce the observed benefits of treatment, it is possible that another cutoff may be allow us to better determine who is likely to benefit. Taken together, further work is necessary to understand who may actually benefit from this treatment approach.

Present Study

The goal of the present study was to evaluate the efficacy and determine who benefits from one-month ARSB fading compared to a relaxation video (RV) control group (Cougle et al. 2017; Zech et al., 2024). In deciding a comparator condition for the present study, we aimed to have an active control condition that did not directly or indirectly target ARSBs. Previous RCTs of SB fading used control conditions with similar behavior change instructions but focused on unhealthy behaviors (Patel & Cougle, 2024b, Zech et al., 2025). Importantly, both of these studies found that despite targeting different behaviors, this control condition may have indirectly led to reductions in safety behaviors. Given our primary aims were to more clearly determine the efficacy of this strategy as well as examine who benefits from ARSB fading, we chose a control group that was less likely to lead to reductions in ARSBs. Comparing this ARSB treatment to different control conditions also allowed us to evaluate the replicability of treatment effects, potentially giving us greater confidence in this intervention approach.

Accordingly, the present study had three primary aims: (1) to evaluate the efficacy of one-month ARSB fading relative to an active relaxation-based control, (2) to empirically derive a Social Appearance Anxiety Scale cutoff at which ARSB fading produces statistically and clinically meaningful benefits, and (3) to identify clinical and demographic moderators of treatment response. To address these aims, we recruited a large sample of men and women seeking treatment for appearance concerns from across the United States and randomly assigned them to either ARSB or RV control. One of the primary aims of our study was to evaluate the point on the Social Appearance Anxiety Scale (Hart et al., 2008) where both the within- and between-group effect of treatment was statistically significant to derive an empirical cutoff to help guide the prescription of this approach for individuals with appearance concerns. Additionally, we aimed to evaluate clinical (i.e., appearance concerns, depression, social anxiety, and ED symptoms) and demographic (i.e., gender, age, race, sexual orientation) moderators of treatment. Given we recruited across the full spectrum of appearance concerns for the first time, we were unsure about treatment effects, broadly. However, given the evidence supporting ARSB fading (Patel & Cougle, 2024b, 2025a, 2025b), we hypothesized that this condition would lead to lower primary (i.e., appearance concerns, social anxiety, ED symptoms) and secondary symptom outcomes (i.e., depression) compared to control. Importantly, we expected that some of these effects would be qualified by a moderation effect on one of our tested moderators, with more severe presentation pre-treatment being related to larger treatment effects. Taken together, this study intended to clarify the boundary conditions of ARSB fading, inform personalized treatment selection, and further explore the efficacy of ARSB fading as a treatment approach.

Methods

Participants

Study participants included 400 individuals recruited across the United States through online advertisements (i.e., Meta ads) and ResearchMatch from April 2025 to July 2025. Advertisements were focused on highlighting a novel digital treatment for appearance concerns that is self-administered. Inclusion criteria included being aged 18–65 years and having access to a cellular phone to receive text messages. Counter to prior work (Patel & Cougle, 2024b, 2025a, 2025b), we did not recruit participants based on elevations on the Social Appearance Anxiety Scale (i.e., the measure was not used as a screener). All individuals desiring treatment for appearance concerns were welcome to participate in the study. Individuals were excluded if they were currently receiving psychotherapy of any kind, had changes in their psychotropic medications in the past month, or failed an attention check in baseline data collection. Participants who completed the posttreatment survey were entered into a raffle to win one of ten $20 Amazon gift cards, and individuals who completed the follow-up survey were given another entry into the raffle. The participant study flow can be found in Figure 1. Notably, of the 400 individuals who were enrolled, 296 individuals completed posttreatment, and 262 individuals completed one-month follow-up. All participants provided informed consent, and study procedures were approved by the Institutional Review Board.

Figure 1.

Consolidated Standards of Reporting Trials Participants Flow Through Stages of the Study.

Measures

Social Appearance Anxiety Scale (Hart et al., 2008).

The SAAS is a 16-item measure of appearance concerns. Items are rated on a 5-point Likert scale ranging from 1 (not at all) to 5 (Extremely) indicating agreement with each statement (e.g., “I feel nervous when having my picture taken.”). Total scores are the sum of each item with scores ranging from 16 to 80. In the present study, the SAAS demonstrated excellent internal consistency at baseline (α = .95), posttreatment (α = .95), and follow-up (α = .96). The SAAS was used as the primary outcome measure of treatment.

Appearance Behaviors Checklist (Adapted from Summers & Cougle, 2018).

The ABC is a measure of checklists used in a prior experimental study using the same paradigm (Summers & Cougle, 2018; Patel & Cougle, 2024b). Notably, we added an additional item to what was used in prior studies focusing on social media use. The 13-item checklist was delivered everyday via text messages to participants during the treatment phase (28 consecutive days). A past week version of the checklist was completed at the three timepoints. The daily version used a sliding scale (0 to 5 hours) indicating how much time individuals engaged in the behavior (duration; “Closely examine your facial features”) or a 0–10+ scale indicating how many times a behavior was used in the past day (frequency; “Check your appearance in reflective surfaces throughout the day”). For the past week version, items were scored using a 0 (Never) - 10 (All the time) scale to assess frequency in the past week. The past week version of the checklist demonstrated good to excellent internal consistency at all three timepoints (baseline α = .88, post α = .90, follow-up α = .91).

Center of Epidemiologic Studies Depression Scale (CESD; Radloff, 1977).

The CESD is a 20-item measure of depression severity. Items assess the frequency of depressive symptoms in the past week on a 0 (rarely or none of the time) to 4 (most or all of the time) scale. The CESD demonstrated good internal consistency at baseline (α = .87), posttreatment (α = .88), and at follow-up (α = .89).

Social Phobia Inventory (SPIN; Connor et al., 2000).

The SPIN is a 17-item self-report measure of social anxiety in the past week. Items assessed how much a statement applies to a participant and were rated on a scale from 0 (not at all) to 4 (extremely). The SPIN demonstrated excellent internal consistency at all three timepoints (baseline α = .93, post α = .92, follow-up α = .93) in the present study.

The Eating Disorder Inventory (EDI; Garner et al., 1983).

The EDI is a 64-item self-report questionnaire of eating disorder pathology. Respondents are asked to rate a series of items on a 6-point scale (“never” to “always”) to determine the frequency of each symptom. In the present study, we used three subscales of the EDI: drive for thinness, bulimic symptoms, and body dissatisfaction. The drive for thinness scale demonstrated good to excellent internal consistency at all three timepoints (baseline α = .89, post α = .89, follow-up α = .90). The bulimic symptoms subscale demonstrated good to excellent internal consistency at all three timepoints (baseline α = .89, post α = .90, follow-up α = .91). The body dissatisfaction subscale demonstrated good internal consistency at all three timepoints (baseline α = .86, post α = .88, follow-up α = .88).

Credibility and Expectancy Questionnaire (CEQ; Devilly & Borkovec, 2000).

The CEQ is a measure of an individual’s perceptions of the credibility and expected efficacy of a treatment completed after receiving treatment rationale. Three items, rated a scale of 0 to 9, assess the participants’ perceived logic and utility of the treatment and how confident a participant feels they would recommend the treatment to a friend. These three items are combined to form a subscale measuring credibility. One additional item is also included assessing the extent individuals believe they will see improvement in their appearance concerns rated on a 0%−100% scale, measuring expectancy. The credibility scale demonstrated good internal consistency (α = .80).

Interventions

Appearance Related Safety Behavior Fading.

The intervention followed the same procedures as a previous RCT examining ARSB fading (see Patel & Cougle, 2024b). Participants were given information on ARSBs and how research indicates that they are strongly related to appearance anxiety. The rationale given can be found in Supplemental Table 1. Participants were instructed to reduce or eliminate the listed ARSBs for the 28 days. Participants were then sent a daily text message at 8:00AM EST that contained a short message (i.e., “Hi! Here is today’s checklist:”) and a link to a survey. Participants completed the daily version of the ABC logging ARSB use in the past day. Afterwards, one of five short messages providing encouragement and additional rationale (e.g., “Please remember to avoid using your checklist behaviors. As you do, you will become less anxious about your appearance!”) was shown to foster engagement and provide a reminder to continue reducing these behaviors.

Relaxation Video Control.

Participants in this condition completed four computerized sessions consisting of practicing relaxation techniques. In these sessions, the participants were invited to close their eyes and focus their attention on the present moment while listening to relaxing music or ambient noises. The rationale given to participants can be found in Supplemental Table 1. Versions of this approach have been used in previous studies and were perceived as credible in these studies (Cougle et al., 2017; Zech et al., 2024). Each video was 15 minutes long and was delivered on a weekly basis over the course of 28 days. We chose to match this condition with ARSB fading based on duration (i.e., total time spent) rather than dose frequency. This choice was made for two reasons. First, we believed it would be more feasible as well as more therapeutic to fully practice relaxation in 15 minutes rather than one or two minutes; this duration is also consistent with the recommended time devoted to relaxation exercises in treatments such as applied relaxation (Öst, 1987). Second, the two treatments would have been mismatched if we had individuals practice relaxation daily for 15 minutes.

Procedures

Potential participants were directed to a Qualtrics survey to determine eligibility criteria. After providing consent, participants completed the screening questionnaire. Participants were informed of the general nature of both treatments, but specific details about the differences between the treatments was withheld. Eligible individuals were directed to complete the baseline questionnaire. At the end of this survey, participants were enrolled in EZText, a text messaging marketing platform, that was used to send daily treatment messages and weekly videos. The next day, participants were sent a link to a survey that included four validity check questions (e.g., “I understand that this a treatment study for appearance-related anxiety and concerns.”), treatment rationale, and treatment instructions. Participants who passed the validity checks were randomized to treatment via Qualtrics such that the two conditions would be equally distributed among participants in a random manner. Please see Figure 1 for the CONSORT flow diagram. The participants were then informed that the treatment would begin the next day. At the end of the one-month treatment period, participants were sent a posttreatment survey via email. One month later, the follow-up survey was sent. Participants were provided a debriefing document that fully explained the nature of the two treatments and the purpose of withholding this information at the end of the study.

Power Analysis

A priori power analyses were conducted using G*Power to detect a small effect (f² = .02) for linear multiple regression (Faul et al., 2009). To detect this effect with α = .05 and power = .80, and three predictors (condition, baseline symptoms, and the interaction of the two) power analyses revealed that a total sample size of 395 participants would be sufficient. This size effect was chosen given prior research on ARSB fading demonstrating small-to-medium effects on appearance concerns (Patel & Cougle, 2024b). We aimed to recruit 400 individuals to be sufficiently powered to examine this moderation effect (appearance concerns × condition).

Data Analysis

All analyses were conducted in R version 4.4.2 (R Core Team, 2024). Baseline differences in demographic variables between groups were tested using one-way ANOVA and Chi-square tests of differences. Data were also screened for issues or normality by assessing for values outside of skewness and kurtosis range of −2 to 2.

To evaluate treatment effects, we use latent difference score (LDS) modeling, an extension of growth curve modeling, via structural equation modeling (Mun et al., 2009). This approach enables the simultaneous estimation of treatment effects across multiple timepoints within a single model. Compared to traditional approaches such as ANOVA, LDS modeling is advantageous because it captures within- and between-group change from baseline to posttreatment and follow-up within one framework, reducing the number of analyses. Further, LDS models account for baseline variability within the models, similar to adjusting for baseline in a regression-based model. LDS modeling also estimates missing data through robust maximum likelihood, allows calculation of effect sizes using all available data, offers greater power, is robust to nonnormality, and estimates model fit not available in traditional methods (Muthén & Curran, 1997). Given missing data are estimated in these models, all analyses in the present study employed an intent-to-treat approach whereby every participant’s data were used for analyses, and any missing data were estimated with maximum likelihood.

Model fit was evaluated using multiple criteria, including the Yuan-Bentler (Y-B) χ² test (Yuan & Bentler, 1998), comparative fit index (CFI), Tucker-Lewis index (TLI), root mean square error of approximation (RMSEA), and standardized root mean square residual (SRMR; Bentler, 1990). Effect sizes were expressed as standardized mean differences (SMDs), calculated by taking the estimated parameter and dividing by the baseline standard deviation (Feingold, 2009). The size of effects are consistent with Cohen’s d guidelines (small=0.20, medium=0.50, large=0.80).

The intercept was centered at baseline, allowing for the estimation of two change parameters: Change 1 (baseline to posttreatment) and Change 2 (baseline to one-month follow-up). These parameters reflect differences in estimated means between baseline and the respective follow-up timepoints. To test treatment effects, the intercept, Change 1, and Change 2 were regressed on condition. We also calculated a reliable change index (RCI; Jacobson & Truax, 1992) for SAAS scores in each condition to contextualize the number of participants in each group who achieved clinically significant improvements in appearance concerns.

To examine our main aim of testing moderation, we use a regression-based approach with structural equation modeling, where the outcome at post or follow-up is entered as a dependent variable and the baseline symptoms, condition, and the interaction of the two were entered as independent variables. Given our main aim was to identify the exact cutoff value for where we see an effect of condition, we used a Johnson-Neyman approach (Bauer & Curran, 2005; Lin, 2020), where we extracted the parameter estimates and the covariance matrix from the model to identify the exact point where which the effect of condition was significant (See supplementary materials for code). The Johnson-Neyman approach is similar to the popular simple slope approach, but rather than picking arbitrary points to examine the moderation effect, the Johnson-Neyman approach plots the effects at every point of a moderator with accompanying confidence intervals. This approach identifies where the effect is significant (confidence intervals not containing zero) and allows for the examination for where this effect is most stable (i.e., narrow confidence intervals; Bauer & Curran, 2005). To quantify the effect size across baseline appearance concerns, we extracted the effects of condition from this analysis and divided them by the baseline standard deviation to derive SMDs across the range of baseline appearance concerns (Feingold, 2009). For our other moderation analyses, we used the same regression-based approach, but to probe the interaction we mean centered the independent variables and tested the simple slopes at high (+1SD) and low (−1SD) values of the moderator.

To examine whether baseline appearance concerns moderated the within-group effect of treatment for ARSB fading, we used a latent change score (LCS) approach using SEM (Kievit et al., 2018). This approach allows for the variance in the change to be modeled while also allowing for the inclusion of an auto-regressive parameter. For this analysis, we modeled the change in appearance concerns from baseline to posttreatment and examined whether baseline appearance concerns predicted this change. To identify the size of effects, we used the Johnson-Neyman approach to quantify the effect sizes as SMDs across levels of baseline appearance concerns.

Transparency and Openness

We report how our sample size was determined and all manipulations in the study. All study procedures were approved by the Florida State University Institutional Review Board (IRB). Data are not publicly available due to not receiving IRB approval to share data publicly. Analysis code and output are available by contacting the authors. This study’s design was pre-registered on ClinicalTrials.gov (NCT06896435). However, the analytic plan was not pre-registered due to the authors needing to seek consultation on the best method to achieve study aims. A-priori, we planned to conduct LDS models to test treatment effects, but our probing of the moderation effect required consultation. Notably, we did deviate from our pre-registered protocol, such that we do not report on findings related to a measure of distress tolerance and concerns with mistake making as these were included for preliminary data purposes for an unrelated project. No data from the present study have been previously published.

Results

Baseline Comparisons

Prior to main analyses, we examined baseline differences in demographic variables. As can be seen in Table 1, there were no demographic differences between groups (all ps > .177). No outcome variables violated normality and no outliers were outside of expected ranges (i.e., a value outside the range of the median plus or minus two interquartile ranges).

Table 1.

Participants Characteristics by Condition

Characteristic	M (SD) or n (%)		F or χ2
	ARSB Fading (n = 200)	RV Control (n = 200)
Age	38.84 (13.84)	40.73 (14.11)	1.83
Gender			2.71
Man	28 (14.0%)	29 (14.5%)
Woman	159 (79.5%)	165 (82.5%)
Non-Binary/Trans	13 (6.5%)	6 (3.0%)
Race1			5.99
White	170 (85%)	165 (82.5%)
Black or African American	18 (9.0%)	26 (13.0%)
Asian	17 (8.5%)	14 (7.0%)
Native Hawaiian or Pacific Islander	0 (0.0%)	0 (0.0%)
Native American/Alaskan Native	4 (2.0%)	4 (2.0%)
Other	6 (3.0%)	3 (1.5%)
Ethnicity			0.23
Hispanic or Latino	23 (11.5%)	20 (10.0%)
Not Hispanic or Latino	177 (88.5%)	180 (90.0%)
Sexual Orientation			7.29
Heterosexual/Straight	140 (70.0%)	148 (74.0%)
Gay/Lesbian/Homosexual	9 (4.5%)	14 (7.0%)
Bisexual	31 (15.5%)	26 (13.0%)
Other/Not Sure	20 (10.0%)	12 (6.0%)
Marital Status			2.13
Married	76 (38.0%)	73 (36.5%)
Never Married/Single/Divorced	124 (62.0%)	127 (63.5%)
Education			5.72
Less than High school	0 (0.0%)	1 (0.5%)
High School Graduate	11 (5.5%)	20 (10.0%)
Some College	40 (20.0%)	43 (21.5%)
4 year degree	75 (37.5%)	78 (39.0%)
Graduate degree	74 (37.0%)	58 (29.0%)
Income	7.39 (3.31)	7.26 (3.58)	0.15

Note. ARSB = Appearance-related Safety Behaviors; RV = Relaxing Video; M = Mean; SD = standard deviation.

¹Individuals were allowed to select more than one race.

Treatment Adherence and Perception

Participants in the ARSB condition completed an average of 16.64 (SD = 10.60) daily checklists, and the RV condition on average viewed 2.9 (SD = 1.2) videos. Group differences in credibility and expectancy of the effectiveness of treatment were compared at baseline. The ARSB condition (M = 15.43, SD = 4.15) did not significantly differ from the RV condition (M = 15.64, SD = 5.06) on perceived treatment credibility, F(1, 392) = 0.21, p = .649. Similarly, the ARSB condition (M = 39.44, SD = 21.15) did not significantly differ from the RV condition (M = 40.26, SD = 23.84) on expected levels of improvement in their appearance concerns, F(1, 392) = 0.13, p = .720.

Treatment Retention

Retention was high, with 70% of the ARSB condition and 70% of the RV condition completing posttreatment, and 67.0% of the ARSB condition and 64.0% of the RV condition completing follow-up assessments. These proportions were not significantly different from one another (p’s > .970).

Main Effects of Treatment on Outcomes

Descriptive statistics for all study variables and estimated within-group effect sizes based on the LDS models are presented in Table 2. A full summary of all LDS models for all seven outcomes as well as SMD effect sizes and 95% confidence intervals can be found in Table 3. Briefly, all models fit the data well. However, the ARSB model initially showed acceptable fit, but model fit improved with the inclusion of a regression path of pre-to-post change predicting pre-to-follow-up change recommended by the modification index.

Table 2.

Summary of Descriptive Statistics of Study Variables Across Conditions (N = 400)

Measure	Baseline M (SD)	Posttreatment M (SD)	One-Month Follow-up M (SD)	Within-Group SMD Effect Size1
				Pre-to-Post	Pre-to-Follow-up
ARSB Fade	n = 200	n = 148	n = 134
SAAS	52.29 (14.14)	40.84 (11.10)	40.91 (11.22)	0.80 [0.70, 0.90]	0.78 [0.67, 0.88]
ABC	68.48 (28.62)	40.00 (25.96)	39.36 (25.09)	0.93 [0.76, 1.10]	1.00 [0.86, 1.16]
CESD	12.51 (6.70)	10.04 (6.34)	10.64 (6.93)	0.34 [0.24, 0.44]	0.24 [0.13, 0.36]
SPIN	33.33 (14.94)	27.71 (13.40)	26.95 (13.43)	0.39 [0.27, 0.50]	0.40 [0.29, 0.51]
EDI-BN	16.40 (7.41)	14.74 (6.80)	14.92 (6.59)	0.21 [0.10, 0.33]	0.24 [0.13, 0.35]
EDI-BD	38.39 (9.31)	36.32 (10.34)	35.97 (10.03)	0.28 [0.19, 0.37]	0.28 [0.18, 0.37]
EDI-DT	25.90 (9.18)	24.40 (8.97)	23.94 (9.10)	0.17 [0.06, 0.28]	0.19 [0.08, 0.30]
RV Control	n = 200	n = 148	n = 128
SAAS	51.86 (15.01)	45.04 (15.14)	44.93 (15.30)	0.45 [0.33, 0.56]	0.46 [0.32, 0.60]
ABC	68.86 (30.46)	56.43 (29.31)	58.34 (30.17)	0.40 [0.29, 0.52]	0.38 [0.26, 0.50]
CESD	12.75 (6.78)	11.63 (6.93)	11.90 (7.17)	0.20 [0.08, 0.31]	0.20 [0.07, 0.32]
SPIN	32.91 (14.60)	28.17 (14.50)	28.34 (15.07)	0.35 [0.24, 0.45]	0.34 [0.22, 0.47]
EDI-BN	16.56 (7.15)	16.04 (7.20)	15.46 (7.66)	0.09 [−0.03, 0.21]	0.22 [0.09, 0.34]
EDI-BD	37.52 (9.60)	36.24 (9.32)	36.91 (9.95)	0.13 [0.05, 0.21]	0.09 [−0.01, 0.17]
EDI-DT	25.82 (9.04)	24.80 (8.71)	24.92 (9.37)	0.17 [0.09, 0.26]	0.18 [0.09, 0.28]

Note. M = Mean; SD = Standard Deviation; ARSB = Appearance-related Safety Behaviors; RV = Relaxing Video; SAAS = Social Appearance Anxiety Scale; ABC = Appearance Behaviors Checklist; SPIN = Social Phobia Inventory; CESD = Center of Epidemiologic Studies Depression Scale; EDI = Eating Disorder Inventory; BN = Bulimic Symptoms Subscale; BD = Body Dissatisfaction subscale; DT = Drive for Thinness Subscale

¹All effect sizes include all cases and are derived from the latent difference score models

Table 3.

Model Parameters for Latent Difference Score Models (N = 400)

Fit Statistics	ABC	SAAS	CESD	SPIN	EDI-BN	EDI-BD	EDI-DT
Y-B χ2 (df)	1.42 (1)	3.95 (2)	1.90 (2)	2.32 (2)	0.14 (2)	3.03 (2)	1.15 (2)
RMSEA	.038	.052	<.001	.015	<.001	.045	<.001
CFI	0.99	1.00	1.00	1.00	1.00	0.99	1.00
TLI	0.99	0.98	1.00	0.99	1.00	0.99	1.00
SRMR	.014	.026	.019	.022	.006	.022	.014
Models Parameters	ABC	SAAS	CESD	SPIN	EDI-BN	EDI-BD	EDI-DT
Condition→ Intercept
B	−0.38	0.42	−0.25	0.42	−0.24	0.82	−0.01
SE	2.95	1.45	0.67	1.47	0.74	0.95	0.91
p-value	.899	.288	.716	.776	.745	.384	.994
Condition→ Change 1
B	−13.95	−4.65	−0.95	−0.70	−0.93	−1.41	0.01
SE	3.04	1.15	0.52	1.18	0.62	0.59	0.65
p-value	<.001	<.001	.068	.550	.132	.017	.992
SMD	0.47	0.32	0.14	0.05	0.03	0.15	−0.001
95% CI	[0.27, 0.67]	[0.16, 0.47]	[−0.01, 0.29]	[−0.11, 0.20]	[−0.04, 0.29]	[0.03, 0.27]	[−0.18, 0.17]
Condition→ Change 2
B	−13.09	−4.05	−0.30	−0.93	−0.24	−1.82	−0.10
SE	2.97	1.30	0.59	1.25	0.62	0.63	0.69
p-value	<.001	.002	.617	.457	.699	.004	.880
SMD	0.44	0.28	0.04	0.06	0.03	0.19	−0.01
95% CI	[0.25, 0.64]	[0.10, 0.45]	[−0.13, 0.22]	[−0.10, 0.23]	[−0.13, 0.20]	[0.06, 0.32]	[−0.17, 0.20]

Note. Y-B χ² = Yuan-Bentler scaled Chi square value; SRMR = Standardized root mean square residual; RMSEA = Root mean square error of approximation; CFI = comparative fit index; TLI = Tucker Lewis index. SE = Standard error. Baseline intercept = estimated mean of outcome at baseline. Change 1 = Pre-to-posttreatment; Change 2 = Pretreatment to 1-month follow-up. B = Unstandardized parameter estimate; SE = Standard error; SMD = standardized mean difference effect size CI = confidence interval. SAAS = Social Appearance Anxiety Scale; ABC = Appearance Behaviors Checklist; SPIN = Social Phobia Inventory; CESD = Center of Epidemiologic Studies Depression Scale; EDI = Eating Disorder Inventory; BN = Bulimic Symptoms Subscale; BD = Body Dissatisfaction subscale; DT = Drive for Thinness Subscale

We found that the two groups did not significantly differ in any outcomes at baseline, indicated by the effect of condition on the intercepts (i.e., baseline scores). We found that there was an effect of condition on pre-to-posttreatment (Change 1) ARSBs, appearance concerns, and body dissatisfaction, such that the ARSB fading condition led to lower symptoms of all three constructs at posttreatment compared to RV control. Further, we found there was an effect of condition on pretreatment-to-follow-up (Change 2) ARSBs, appearance concerns, and body dissatisfaction, such that the ARSB fading condition led to lower symptoms of all three constructs at follow-up compared to RV control. There was no effect of condition at either timepoints for depression, social anxiety, bulimic symptoms, or drive for thinness.

Reliable Change

A reliable change criterion of 9.20 was calculated for appearance concerns (following Jacobson & Truax, 1991, using Cronbach’s alpha = .95 and SD = 14.56 for the SAAS at baseline). Based on this criterion, 69% of participants in the ARSB condition displayed reliable change from pre-to-posttreatment, and 51.5% of those in the RV condition displayed reliable change from pre-to-posttreatment; these rates were significantly different from one another, χ² = 12.75, p < .001. Importantly, these rates assume that participants who did not complete the follow-up assessments were nonresponders. Regarding pre-to-follow-up change, 70% of the ARSB condition displayed reliable change, and 60% of the UHB condition displayed reliable change. These rates were significantly different from one another, χ² = 4.82, p = .028.

Moderation Effect of Appearance Concerns

In testing our primary aim, we found that there was a statistically significant interaction between condition and appearance concerns predicting posttreatment appearance concerns (β = −0.34, p = .019). We probed the interaction using a Johnson-Neyman approach to examine at what points of baseline appearance concerns the effect of condition was significant. We found that the exact point estimate of baseline appearance concerns for which there was an effect of condition was above 40.72; there was no effect of condition below this point. The plotted moderation effect can be seen in Figure 2. At the threshold (i.e., 40.71), the effect of condition was small¹ (SMD = 0.17, [0.01, 0.35], and at the high end of the significant condition effect range, the effect of condition was medium (SMD = 0.61, [0.32, 0.90]). At the threshold and above, the average effect size was 0.39 [0.21, 0.58]. At the previously used cutoff on the SAAS (i.e., 53 and above) the between-group effect was 0.46, [0.26, 0.66]). Notably, we found that the confidence intervals for the effect of condition were most stable (i.e., narrow) at 52 on baseline appearance concerns. To quantify the difference in effect sizes, we also examined the simple slopes at high (+1SD) and low (−1SD) baseline appearance concerns. We found that at high appearance concerns, there was a medium effect of condition (b = −6.72, SE = 1.43, p <.001, SMD = 0.46 [0.27, 0.66]), such that ARSB fading led to lower appearance concerns than RV control. There was no effect of condition at low appearance concerns (b = −1.99, SE = 1.43, p =.163, SMD = 0.14 [−0.06, 0.33]). The full range is plotted in Supplemental Figure 1 and all values are listed in Supplemental Table 2.

Figure 2.

Plotted simple slopes of the effect of treatment on appearance concerns based on baseline values of appearance concerns.

We next examined whether baseline appearance concerns moderated the within-group effect of ARSB fading. Within our LCS model, we found that baseline appearance concerns predicted change in appearance concerns (β = −0.47, p <.001) indicating that individuals with higher baseline scores saw greater within-group effects. In examining the magnitude of these effects, we plotted the within-group effect sizes across values of baseline appearance concerns in Figure 3. SMD effect sizes² ranged from −0.18 [−0.34, −0.01] at the floor of the baseline values (i.e., SAAS = 17) to 1.27 [1.09, 1.44] at the ceiling of the scale (i.e., SAAS = 80). At the previously identified threshold and above (i.e., 40.71), the within-group effect was large (SMD = 0.82, [0.72, 0.92]), and this was comparable to the within-group effect (SMD = 0.96, [0.84, 1.08]) at the prior used cutoffs (i.e., 53) and above.

Figure 3.

Plotted Standardized Mean Difference Within-Group Effect Size of ARSB Fading on appearance concerns based on baseline values of appearance concerns. CI = Confidence Intervals

We also found that baseline appearance concerns moderated the effect of condition on ARSBs (β = −0.11, p = .015). Probing the interaction revealed that there was an effect of condition on ARSBs at high (+1SD) appearance concerns such that ARSB fading led to lower ARSBs compared to RV control (b = −21.50, SE = 3.73, p <.001, SMD = 0.95 [0.62, 1.27]). There was also an effect (albeit smaller in size) of condition on ARSBs at low (−1SD) appearance concerns (b = −8.90, SE = 3.69, p =.016, SMD = 0.39 [0.07, 0.71]). This same pattern was found for follow-up ARSBs (β = −0.11, p = .017). There was an effect of condition on follow-up ARSBs at both high (b = −22.52, SE = 3.73, p <.001, SMD = 1.05 [0.71, 1.40]) and low ARSBs (b = −10.30, SE = 3.65, p =.005, SMD = 0.48 [0.15, 0.82]).

We also found that baseline appearance concerns moderated the effect of condition on bulimic symptoms at posttreatment (β = −0.09, p = .033). Probing the interaction revealed that there was an effect of condition on bulimic symptoms at high (+1SD) appearance concerns such that ARSB fading led to lower bulimic symptoms compared to RV control (b = −1.65, SE = 0.82, p =.045, SMD = 0.34 [0.01, 0.68]). There was no effect of condition on bulimic symptoms at low (−1SD) appearance concerns (b = −0.59, SE = 0.82, p =.475, SMD = 0.12 [−0.21, 0.45]). There were no other moderating effects of baseline appearance concerns (p’s > .288).

Exploratory Moderation Analyses

We next conducted exploratory analyses examining whether other baseline symptoms (besides appearance concerns) moderated the effect of condition on the same symptoms at posttreatment and follow-up. We found that baseline ARSBs moderated the effect of condition on ARSBs (β = −0.19, p < .001). Probing the interaction revealed that there was an effect of condition on ARSBs at high (+1SD) appearance concerns such that ARSB fading led to lower ARSBs compared to RV control (b = −26.22, SE = 3.72, p <.001, SMD = 1.17 [0.85, 1.50]). There was no effect of condition on ARSBs at low (−1SD) appearance concerns (b = −5.12, SE = 3.52, p =.145, SMD = 0.23 [−0.08, 0.54]). We also found a significant interaction for follow-up ARSBs (β = −0.13, p = .003). There was a large effect of condition on follow-up ARSBs at high (b = −24.01, SE = 3.73, p <.001, SMD = 1.13 [0.79, 1.47]) and a medium effect at low ARSBs (b = −8.75, SE = 3.68, p =.017, SMD = 0.41 [0.07, 0.75]). We found that no other baseline scores moderated the effect of condition on the respective outcome at posttreatment or follow-up (p’s > .091).

We also examined whether there were any demographic moderators (i.e., gender, race [white vs. not], ethnicity, education, marital status, sexual orientation, and income) of the effect of condition on appearance concerns via a regression based approach in SEM. We found that none of our observed demographic factors moderated the effect of treatment on posttreatment or follow-up appearance concerns (p’s > .289). Notably, the effect sizes for ARSB fading on appearance concerns at post were comparable among men (SMD = 0.31 [−0.07, 0.70]) and women (SMD = 0.30 [0.15, 0.44]). Lastly, we examined whether credibility or expectancy at baseline moderated treatment effects on appearance concerns. We found that neither moderated the effect of treatment on either posttreatment or follow-up appearance concerns (p’s > .329).

Discussion

The goal of the present study was to further evaluate the efficacy of a recently developed digital intervention for appearance concerns, ARSB fading, and address gaps in knowledge about who benefits from this treatment (Patel & Cougle, 2025a, 2024b). While prior studies have used non-empirical cutoffs based off one measure of appearance concerns and excluded men, the present study was able to recruit a large mixed-gender sample to examine whether ARSB fading was efficacious among these individuals. Of note, we found that there was a between-group effect for individuals who scored a 41 and above on the SAAS (Hart et al., 2008), corresponding to a small effect size (i.e., SMD = 0.39). This is comparable to the previous RCT examining ARSB fading in women elevated (i.e., 53 and above) in appearance concerns (SMD = 0.34; Patel & Cougle, 2024b). Our within-group findings showed that individuals at this threshold and above experienced large within-group changes in appearance concerns, consistent with prior research using the conservative cutoff (Patel & Cougle, 2025a, 2025b, 2024b). These findings indicate that the previously used threshold may have excluded individuals who would have otherwise benefited from ARSB fading. However, it is important to note that the control condition used in this previous study also led to reduction in ARSBs, which may have diluted the between-group effect of ARSB fading. Further, we found that the effect of condition was most stable for individuals who scored a 52 at baseline, suggesting the prior cutoff may have been appropriate. We also found that individuals randomized to ARSB fading demonstrated significantly greater reliable change in appearance concerns compared to our control condition, further supporting its efficacy as a treatment approach.

We also found that the effect of ARSB fading was not moderated by gender, indicating that men and women equally benefited from this intervention for appearance concerns. These findings align with recent studies indicating that both appearance concerns and ARSBs are relevant in men (Patel & Cougle, 2025b; Patel, Robison, & Cougle, 2025). Similarly, we found that none of our other demographic factors moderated the effect of treatment. While further work will be necessary to examine these effects, these findings are promising in suggesting that individuals across gender, race, sexuality, income, and education may benefit from a treatment such as ARSB fading for appearance concerns.

With regard to treatment effects of our other outcomes, we found that there was a small effect of condition on body dissatisfaction (SMD = 0.17), such that ARSB fading led to lower body dissatisfaction compared to control at both posttreatment and follow-up. We also found that, among individuals high (+1SD) in baseline appearance concerns, ARSB fading led to lower bulimic symptoms compared to control, corresponding to a small effect (SMD = 0.34). These findings are consistent with a previous RCT examining ARSB fading that found it led to greater reductions in ED symptoms than an active control (SMD = 0.25; Patel & Cougle, 2024b). The effects of ARSB fading on body dissatisfaction are also comparable to effects of in-person dissonance-based interventions for EDs (d =.18; Stice et al., 2019). Future research should continue to evaluate ARSB fading as a treatment and preventative intervention for ED risk factors and symptoms.

Interestingly, we found that there was no effect of condition on depression or social anxiety. This finding was partially consistent with the previous RCT (Patel & Cougle, 2024b), which found no condition differences on depression, though ARSB fading led to lower social anxiety at follow-up. Study design differences may account for these differences in findings, and it is possible that effects on these outcomes may be present only for individuals who meet for appearance-related psychopathology (Patel & Cougle, 2024a). Loneliness may also be a factor influencing these findings, as a recent study found that the effect of ARSB fading on social anxiety was only present among women low in loneliness (Patel, Sala, & Cougle, 2025).

This study has several strengths that are worth highlighting. First, we recruited a large sample of mixed-gender individuals to evaluate the efficacy of ARSB fading. Additionally, we had low attrition within the study, and the recruitment window was relatively short demonstrating that high demand for a treatment for appearance concerns. Second, treatment adherence (measured by checklist completion and videos watched) was high across conditions, especially for a study that did not offer guaranteed compensation. Third, we examined ARSB fading to an active comparator condition, which individuals found equally credible. Fourth, we were able to leverage power statistical techniques (i.e., SEM, LDS, Johnson-Neyman) to answer a complex question demonstrating crucial strategies to conceptualize and answer novel research questions.

Limitations and Future Directions

The present findings should be considered in the context of limitations. While we were able to recruit a large sample, there was still heterogeneity in participants (white women). It is also important to highlight that while men made up a small portion of our sample (i.e., 14.3%), this is consistent with other large scale digital intervention trials (Eberle et al., 2024; Steinman & Teachman, 2014). Future studies may oversample men and people of color to better examine ARSB fading within these populations. Similarly, trans/non-binary folks made up a small portion of the sample (i.e., 4.8%). We are also limited in that we did not assess for diagnosis of appearance-related psychopathology. A crucial next step will be examining the efficacy of this intervention or approach among a clinical sample. Additionally, future research may examine this approach as an adjunctive therapeutic tool. The present study excluded 332 people who were presently in treatment. While the standalone effect of ARSB fading would be difficult to isolate, it is possible that this approach could be helpful as an adjunctive tool to increase efficacy in treatment of disorders such as BDD, EDs, and SAD.

The follow-up period was also brief. Also, while the two interventions were matched for duration, they differed in dose frequency (i.e., 1x/day vs. 1x/week). We also relied solely on daily checklists to support treatment; however, future research could incorporate measures of daily appearance concerns to more precisely capture fluctuations in treatment outcomes. Additionally, future studies could add behavior tracking summaries to better facilitate and motivate the individual to reduce ARSBs. It would also be important to evaluate the efficacy of this treatment protocol using alternative behavioral assessments, such as laboratory-based appearance stressors (Patel & Cougle, 2024b; Summers & Cougle, 2018; Wilver et al., 2020). Future research could also assess motivation to reduce ARSBs and how that may affect study findings. Future work may want to incorporate more extensive psychoeducational components than a single paragraph. Perhaps a video-based, interactive psychoeducation on ARSBs could be efficacious in its own right or improve efficacy of ARSB fading. Lastly, future research may consider discussing behavioral substitution and teach participants to not substitute one ARSB for another.

Conclusion

The findings of the present study extend prior work on ARSB fading by further evaluating its efficacy against a new control condition and examining who actually benefits from the treatment. The findings provide novel evidence that individuals with lower appearance concerns might still benefit from this intervention. Further, the study provides evidence that ARSB fading may be equally efficacious in men as it is in women. The treatment was found to be both accessible and well tolerated, and the protocol can be readily implemented as a standalone intervention or in conjunction with other treatments. Although relatively straightforward and minimal in design, the protocol could be refined in future research by incorporating interactive features, checklist-based tracking and summaries, or delivering the intervention in a downloadable app format. Future studies may also examine this approach across diverse samples, with the inclusion of additional treatment components, extended follow-up periods, and evaluation of its utility as both a therapeutic and preventative intervention.

What is the public health significance of this article?

This study evaluated two one-month digital interventions for appearance concerns, with the goal of examining who actually benefits from a daily self-monitoring intervention to reduce maladaptive behaviors (i.e., appearance-related safety behaviors) such as checking one’s appearance, asking others how they look, and avoiding situations due to their appearance. Compared to a relaxation control, this intervention led to lower appearance concerns among individuals higher in these concerns and showed comparable efficacy among men and women.

Data Transparency

No data from the present study are in press or have been previously published.