96 Chapter 5 completed the post-exposure assessment, which included the SPS and the post-exposure AAT, 30 minutes after the second session. One month later, participants completed an online follow-up assessment which included the SPS. The original parent trial was registered in the Dutch trial register (https://www.trialregister.nl/trial/6238) and at EudraCT (2014–004475–23). Data analytic strategy The research questions, hypotheses and data analytic procedures were pre-registered at Open Science Framework (OSF): see https://osf.io/3cxsv. At two minor points the analyses we performed deviated from the pre-registration (https://osf.io/3cxsv), which are discussed in Chapter 5 – Appendix 1. For the parent trial a sample size of 52 participants was deemed necessary to detect group differences with at least a moderate effect size and a power of 80%. Consistent with pre-registration, we ran preparatory analyses to further specify our AAT predictors in our analyses. Specifically, we first tested if participants showed an avoidance bias toward facial expressions and if this bias was affected by picture type with a Repeated Measures ANOVA with factor picture type (happy, angry, neutral, control) and response direction (push, pull) on the AAT reaction times. Based on the results of this analysis (e.g., all facial stimuli showed faster push than pull RTs (see Chapter 5 – Appendix 1 for details), we decided to analyze AAT reaction times for social stimuli (i.e., all facial expressions) versus non-social stimuli (i.e., the checkerboards). To test the moderator hypothesis, we conducted mixed model analyses for the first (enhanced, with testosterone (T) or placebo (P)) and second (unenhanced) session separately. More specifically, to determine whether avoidance tendencies toward facial expressions moderated (testosterone enhanced) exposure efficacy in terms of self-reported fear (SUDs) during the exposure sessions, AAT combined effect score, group (T/P) and time (start, 2 min, 4 min, 6 min, 8 min, end) were included as predictors. Because we found that SUD scores did not follow a linear pattern (Hutschemaekers et al., 2021), we included linear and quadratic time terms. Participant was included as random intercept. Initial SUD scores were included as a fixed factor to control for variance in fear levels unrelated to time or group. To test the relation between social avoidance tendencies and symptom severity, we modeled SPS scores, with AAT combined effect score, group (T/P) and time (pre/post/FU) as predictors and participant as the random intercept. To test the effects of treatment condition on pre- to post changes in avoidance tendencies toward facial expressions, we modelled AAT RTs with Time (pre-post exposure), Group (T/P), Response direction (Push/Pull) and Picture type (Social/Non-social) as predictors. Participant and Stimulus model (e.g., the model presented on the stimulus) were included as random intercepts. Response direction and time (and their interaction)
RkJQdWJsaXNoZXIy MTk4NDMw