101 Social Avoidance and Testosterone Enhanced Exposure Efficacy in Women with Social Anxiety Disorder Avoidance tendencies over time3 Results revealed a main effect of Time, Estimate = 13.19(3.25), 95% CI [6.10, 19.65]; t(37) = 4.06, p <.001, suggesting that reaction times on the AAT reduced from pre- to post-exposure. A significant interaction effect for picture type (social, non-social) and response direction (push, pull), Estimate = 8.84(3.66), 95% CI [1.30, 15.81]; t(6) = 2.42, p = .050, showed that participants were faster in pushing compared to pulling social faces (e.g. an avoidance bias), whereas they were faster in pulling compared to pushing non-social stimuli (e.g., an approach bias). This effect was present for both groups and did not change over time (pre- to post treatment). All other main effects and interactions were not significant (see Figure 5.2). Table 5.2 Multiple linear regression predicting exposure success (SUD reductions), without and with baseline testosterone included. Exposure success SUDs session 1 SUDs session 2 Estimate (95% CI) P-value Estimate (95% CI) P-value Model step 1 AAT effect score .004(-.08, .08) .911 .05(-.03, .14) .223 AAT effect score * time (linear) .21(-.20, .65) .346 .21(-.17, .58) .278 AAT effect score * time (quadratic) -.01(-.42, .43) .963 -.08(-.43, .28) .683 AAT effect score * group .05(-.02, .12) .205 .08(.001, .17) .065 AAT effect score * group * time (l) -.44(-.89, .02) .051 -.08(-.44, .31) .680 AAT effect score * group * time(q) -.35(-.81, .07) .099 -.02(-.39, .30) .924 Model step 2 (with baseline testosterone) AAT effect score .01(-.06, .08) .862 .06(-.02, .14) .153 AAT effect score * time (linear) .19(-.28, .63) .399 .18(-.19, .55) .338 AAT effect score * time (quadratic) -.06(-.05, .03) .794 -.10(-.45, .26) .582 AAT effect score * group .05(-.02, .12) .135 .09(.003, .18) .041 AAT effect score * group * time (l) -.41(-.87, .001) .067 -.07(-.44, .31) .727 AAT effect score * group * time(q) -.34(-.74, .04) .108 -.01(-.36, .34) .937 Note. there was no correlation between baseline testosterone and AAT effect scores prior to exposure: r = .029, p = .837 or post exposure: r = -.079, p = .580. 3 The residuals of this model showed that the assumption of normality was violated. Therefore, a log transformation was performed. This improved the distribution of the residuals and yielded similar results compared to the model without this transformation. To improve interpretation of the estimates we reported the results of the non-transformed data. Moreover, mixed model analyses are fairly robust against violations of normality (Knief & Forstmeier, 2021; Schielzeth et al., 2020). 5
RkJQdWJsaXNoZXIy MTk4NDMw