107 Social Avoidance and Testosterone Enhanced Exposure Efficacy in Women with Social Anxiety Disorder from one another (all p > .605). Moreover, there was a significant interaction between response direction and picture type, F(3, 159) = 10.32, p = .015. Therefore, the effect of response direction was analyzed at each type of stimulus (angry, happy, neutral, control). P-values were adjusted using the Bonferroni multiple testing correction method. The effect of response direction was significant for the angry (p = .04), neutral (p < .001) and control stimuli (p = .008), but not for happy stimuli (p = .102). However, the direction of the effect of control stimuli was opposite to the direction of other stimuli, indicating that participants show a significant tendency to faster push angry and neutral faces away compared to pulling them, (e.g., an avoidance tendency). The same pattern is found for happy faces, although not significant. While for checkerboards participants show a significant tendency to pulling checkerboards faster than pushing them away (e.g., an approach tendency). Based on the results of this analysis we computed a combined AAT effect score for all facial expressions (happy, angry, neutral): ([Push RTs – Pull RTs of all facial expressions combined]/3 – [Control Push – Control Pull]). The second preparatory analyses showed that social anxiety symptoms at baseline were not related to avoidance tendencies toward facial expressions at baseline r = .08, p = .556. Details results inclusion of baseline testosterone Social anxiety symptom analysis Social avoidance tendencies did not affect social anxiety symptoms over time, pre-post Estimate = - 0.001(.02), 95% CI [-.05, .04]; t(92) = -.025, p =.980, pre-FU: Estimate = -.039(.02), 95% CI [-.003, .08]; t(92) = 1.76, p = .083 nor an interaction effect of AAT effect score x Time x Group, pre-post: Estimate = .002(.02), 95% CI [-.04, .04]; t(92) = -.108, p =.914, pre-FU: Estimate = .004(.02), 95% CI [-.04, .04]; t(98) = -.018, p =.985. The effects of exposure and testosterone on social avoidance tendencies4 There was a main effect of time Estimate = 13.24(3.33), 95% CI [6.82, 20.29]; t(36) = 3.97, p <.001. Reaction times on the AAT become less from pre-exposure to after exposure. A marginal significant interaction effect was found for picture type and response direction Estimate = 6.68(3.74), 95% CI [1.57, 15.33]; t(6) = 2.32, p = .058. Showing the same pattern 4 These are the results of the same model as reported in the paper without baseline testosterone. Inclusion of baseline testosterone resulted in model converging errors. Estimates of simpler random effects structures with baseline testosterone included were highly similar to the current model. Therefore, we decided to report the results of this model to stay in line with main analyses without baseline testosterone. 5
RkJQdWJsaXNoZXIy MTk4NDMw