The Impact of a Podcast-based Body Image Intervention Among Millennial Women

Overview

The goal of this three-arm randomized controlled trial is to evaluate the acceptability and efficacy of two podcast-based interventions -- a podcast with an accompanying journal and the same podcast without a journal -- which aim to improve body confidence among millennial-aged women (29-44 years of age) in the United Kingdom. A total of 1800 women in the UK (29-44 years), who are podcast listeners, will be recruited by a research agency. Participants will be randomised to one of three groups: body confidence podcast episode with journal; body confidence podcast episode alone; a non-body confidence-related podcast episode from the same series as the intervention podcast. The research aims are to determine the following: 1. Are the two versions of the intervention (i.e., podcast with and without a journal) effective in improving appearance esteem (primary outcome), internalisation of appearance ideals, self-objectification, and weight esteem among UK-based women (i) 1 day post-intervention and (ii) 1 week post-intervention, compared to a control podcast? 2. Are the two interventions (i.e., podcast with and without a journal) effective in generating immediate change in state body satisfaction and state mood, compared to a control podcast? 3. Are the two versions of the intervention acceptable to women?

Background: Around 25% of women aged 29-54 report body dissatisfaction, and 50% have moderate shape and weight concerns (Becker et al., 2019). Millennial women, including those from diverse groups, could benefit from targeted body image interventions. However, existing interventions typically focus on younger women. Interventions: The Dove Self-Esteem Project has developed a three-part series of podcasts aimed at fostering body confidence in millennial-aged women. Each podcast episode comes with an accompanying journal covering evidence-based activities to improve body confidence. The podcasts were developed by Dove's Marketing and Communications team with academics from the Centre for Appearance Research (CAR) providing expert advice on content. CAR academics and EVERYBODY Consulting provided the content for the journals. This trial evaluates the impact of one of these podcasts. At just over halfway through the podcast, the host describes a cognitive dissonance-based exercise for listeners to try at home. During this segment, the host asks listeners to: 1. identify the costs (e.g., energy, time, societal) of trying to adhere to appearance pressures; 2. Identify the downsides of pursuing beauty ideals; and 3. Identify something in their routine that makes them feel good that isn't a consequence of adhering to appearance ideals. Listeners are asked to identify how they can amplify that in their own life. As well, they are asked to identify something they do as a consequence of adhering to appearance ideals that feels draining or not helpful. They are asked to consider doing less of that activity. The podcast episode is supplemented with evidence-based journal exercises, which reinforce and build upon the exercise described in the podcast, are based on cognitive dissonance. Research questions: The primary research question (PRQ) and its corresponding hypothesis (H) are as follows: PRQ1: Does listening to the podcast with a journal (PRQ1a) or without a journal (PRQ1b) increase appearance esteem 1 day later, relative to a control podcast? H-PRQ1: The researchers anticipate that women randomised to listen to the podcast with a journal (H-PRQ1a) and without a journal (H-PRQ1b) will experience improvements in appearance esteem 1 day later, relative to a control podcast. The secondary research questions (SRQ) and the corresponding hypotheses (H) are as follows: SRQ1: Does listening to the podcast with a journal (SRQ1a) or without a journal (SRQ1b) decrease internalisation, decrease self-objectification, and/or increase weight esteem 1 day later, relative to a control podcast? H-SRQ1: The researchers anticipate that women randomised to listen to the podcast with a journal (H-SRQ1a) and without a journal (H-SRQ1b) will experience decreased internalisation, decreased self-objectification, and/or increased weight-esteem 1 day later, relative to a control podcast. SRQ2. Does listening to the podcast with a journal (SRQ2a) or without a journal (SRQ2b) increase appearance esteem, decrease internalisation, decrease self-objectification, and/or increase weight esteem 7 days later, relative to a control podcast? H-SRQ2: The researchers anticipate that women randomised to listen to the podcast with a journal (H-SRQ2a) and without a journal (H-SRQ2b) will experience decreased internalisation, decreased self-objectification, and/or increased weight-esteem 7 days later, relative to a control podcast. SRQ3. Does listening to the podcast with a journal (SRQ3a) or without a journal (SRQ3b) lead an immediate increase in body satisfaction and/or mood, relative to a control podcast? H-SRQ3: The researchers anticipate that women randomised to listen to the podcast with a journal (H-SRQ3a) and without a journal (H-SRQ3b) will experience an immediate increase in body satisfaction and/or mood, relative to a control podcast. SRQ4. Does the impact of the Dove podcast depend on the extent women engaged in the content? H-SRQ4: The researchers anticipate that women who demonstrate high engagement in the intervention will experience greater benefits than those who demonstrate lower engagement. SRQ5. Does the impact of the Dove podcast depend on baseline levels of appearance esteem? H-SRQ5: The researchers anticipate that women with lower levels of appearance esteem will experience the greatest impacts. SRQ6. Is the intervention more impactful for older millennials compared to younger millennials? H-SRQ6: The researchers anticipate that older millennials will experience greater positive impact than younger millennials. Design: The podcast episode is developed as a standalone episode, and thus it could hold more potential as a scalable intervention beyond the podcast and journal combined. As such, we are seeking to understand the unique impact both intervention versions may have via a three-arm randomised controlled trial (RCT) delivered online. The RCT aims to assess the effectiveness of both interventions by comparing them to an active control condition. Trait-based measures will be asked at baseline, 1 day post-intervention with a 24-hour window, and 7 days later with a two-day (48-hour) window. State measures will also be assessed. For all arms, both state outcomes will be measured immediately before listening to their randomised podcast. For Arm A, they will be measured again immediately after completing the journal. For Arms B and C, they will measured again immediately after listening to their assigned podcast. Intervention acceptability and engagement will also be measured immediately post-intervention. (See Outcome Measures for further details.) Sample size: Sample size depends on (i) required statistical power (assumed minimum 1 - beta = 0.8, 80%), (ii) statistical significance (alpha = 0.05, 5%), (iii) extent of correlation in appearance esteem (BESAA-A) between T1 and T2, (iv) mean difference between arms, and (v) extent of missing data. The degree of correlation between T1 and T2 will be based on r = 0.7 (i.e. 50% of the variation in the T2 measure not accounted for by the measure at T1). On this basis, in two-sided testing, a minimum sample size of N = 393 per arm would have 80% power to detect a small standardised effect (Cohen's d = 0.2, one-fifth of a standard deviation) between intervention podcast and control podcast but with N = 477 per arm needed for either podcast intervention arm demonstrating an effect compared with control whilst simultaneously maintaining overall family-wise error at alpha = 0.05. If N = 1800 are recruited (n = 600 per arm) then 20% loss between T1 and T2 (N = 480 per arm) would retain sufficient power for study hypotheses. Data loss of 35% would retain power for similar styled analyses at T3. Participants and procedure: UWE Bristol is collaborating with a research agency (C\&R) experienced with UK-based recruitment. C\&R will be employed by our funding partners, Unilever/Dove. The research agency will recruit 1800 UK-based women. The agency will use their existing research panel membership to recruit participants along with their trusted and vetted agency partners experienced in recruiting national samples. Recruitment information will be sent to a sample that closely reflects UK demographics on the dimensions of age, ethnicity, and socio-economic status. Potential participants will be contacted via email and/or text, depending on how they have opted to receive study information. Participants will complete the research using their own digital device with internet access. The research agency will send participants reminders as per the following: * Survey 1: Follow-up reminder 24 hours later, then 12 hours later; * Survey 2: 12-hour pre-reminder before survey 2 is sent. Then, a 24-hour follow-up reminder, followed by one last reminder 12 hours later; * Survey 3: 24-hour pre-reminder before survey 3 is sent. Then, a 24-hour follow-up reminder, followed by one last reminder 12 hours later. Block randomisation will be used to randomise participants to one of the three arms (blocks of 3; 6 block sequences). The research agency will execute block randomisation using the statistical design software built into Sawtooth Software. Analyses Primary Endpoint Analysis: Analyses will proceed with an intention-to-treat analysis set meaning any participant who provides some outcome data at T2 or T3 will be included in the analyses and analysed according to randomised arm. Standardised effect sizes will be reported. The primary analysis will be a repeated measures analysis of covariance (ANCOVA) with the appearance subscale of the Body Esteem Scale for Adolescents and Adults (BESAA-A) at T1 as the covariate, randomised arm as a three level between subjects factor and BESAA-A at T2 as the outcome. The ANCOVA model will be extended to include a randomised arm by baseline covariate interaction term to model whether differences in outcome differ between groups according to BESAA-A at T1 and to investigate the ANCOVA parallel lines assumption. If the interaction term is not considered necessary, then this term will be removed from the ANCOVA model for parsimony. If the interaction term is deemed necessary, then it will be retained to avoid biased estimates of effects. The ANCOVA model will be used to conduct pairwise comparisons between randomised arms reporting adjusted 95% confidence intervals for mean differences. For the interaction model, Johnson-Neyman ANCOVA will be used to identify regions of difference between randomised arms. These analyses will also answer secondary research question 5. The above analyses are ITT analyses and are used to evaluate effectiveness. The modified ITT analyses are sensitivity analyses. The per protocol analysis set is used to examine efficacy. Secondary Analyses: Secondary research question 1 will use the same analysis strategy as for the primary analysis. Similarly, secondary research question 2 will use the same ANCOVA modelling strategy using the appropriate data at times T1 and T3. Single item VAS measures taken immediately before intervention and immediately after intervention will be assessed using the same ANCOVA approach +/- interaction term as deemed necessary. Secondary research question 4 will evaluate effect of engagement. To better understand the effects of the intervention(s), the levels of engagement will be assessed by demographics and randomised arm using two-way between-subjects ANOVA. Effect of engagement on each secondary outcome will be assessed using ANCOVA. The functional form of the ANCOVA model will be as per the main analyses but will additionally include Engagement as a main effect and an engagement by randomised arm interaction term. For secondary research question 6 (Is the intervention more impactful for older millennials compared to younger millennials?) the ANCOVA model will be extended to include a main effect for Age and an Age by Randomised Arm interaction effect. Statistical significance of the interaction term and associated effect size will be used to assess modification of impact by Age. These analyses will be done for appearance esteem, internalisation of appearance ideals, self-objectification, and weight esteem at T2 and T3. Underpinning statistical assumptions will be checked in analyses to ensure robust statistical conclusions are obtained. If assumptions are broken to the extent of placing statistical conclusions in doubt then remedial action will be taken as deemed necessary whilst adhering to the broad principles outlined. Missing Data: Moderate levels of missing data are anticipated at T2. The extent of missing data at T3 will most likely be higher. Similar styled research indicates that missing data will be either Missing Completely at Random (MCAR) or Missing at Random (MAR) as opposed to Missing Not at Random (MNAR). If \< 50% of participants have missing data at T2 then Multiple Imputation will be used as a sensitivity analysis to assess robustness of any statistically significant results. If \> 50% of data is missing, then results will be reported descriptively. Tertiary Analyses: The Primary and Secondary analyses are concerned with changes in mean values. It is of potential value to the field of appearance research to better understand if other distributional changes are present. In these regards, the ANCOVA model (a form of linear regression) will be re-cast using quantile regression (Quantiles 0.1, 0.25, 0.5, 0.75, 0.9) to determine if the lower 10%, lower 25%, median, upper 25%, and upper 10% of the outcome distributions of BESAA-A are intervention dependent. If these analyses show results of interest to the wider field of body image research, then they will be reported in a wider dissemination. If no such interesting findings emerge then this will be stated in the dissemination. Exploratory Analyses: All other analyses including any subset analysis or subgroup analysis would be considered to exploratory hypothesis generating analyses. Pilot study: An internal pilot with 180 participants (10% of the total main sample) will assess and enhance the main trial's methodology based on progression criteria (i.e., participant recruitment; participant retention rate; data quality; intervention acceptability; journal completion; and technical issues), permitting assessment of any issues with the intervention and ensure smooth processes are in place among all stakeholders (UWE researchers, research agency, participants, funders). A stop/go protocol will be used as follows: The research team will PROCEED to the main trial if: * 100% of participants (i.e., 180 participants) were recruited in the time allocated * at least 80% of participants completed all three surveys * at least 95% correctly answered at least 2 attention checks at each time point * intervention acceptability did not significantly differ across conditions * at least 80% of participants in Arm A provided at least one comprehendible response to journal questions * less than 5% of participants experienced technical issues The research team will REVIEW (pause, consult with research agency, and correct) research processes if: * between 50-99% were recruited in the time allocated * between 61-79% completed all three survey timepoints * between 80-94% correctly answered at least 2 attention checks correctly at each time point * intervention acceptability was significantly different across some (but not all) acceptability items * between 51-79% of participants in Arm A provided at least one comprehendible response to journal questions * between 5% and 14% experienced technical issues The research team will consider NOT proceeding to the main trial or making significant changes if: * equal to or less than 50% were recruited in the time allocated * equal to or less than 60% completed all three survey timepoints * equal to or less than 79% correctly answered 2 attention checks at each time point * Intervention acceptability was significantly different across all items * equal to or less than 50% of participants in Arm A provided at least one comprehendible response to journal questions * 15% or more participants experienced technical issues