Martijn van Teffelen

Interpretation bias modification for hostility 143 6 expectations, results revealed that CBM-I decreased general psychiatric symptoms to a lesser extent than AC. Also, findings showed only a significant non-differential decrease over time for self-reported state and trait hostility measures. Moreover, explorative findings showed no difference between conditions in working alliance for people who went into psychotherapy after our interventions. Interestingly, across conditions reductions in aggressive behavior (VDT) were related to increased working alliance, whereas changes in other outcome variables were not significantly related to working alliance. GENERAL DISCUSSION The present work investigated the effects of a novel CBM-I intervention for hostility. We first tested its feasibility in a single-blind randomized sham-controlled feasibility trial using a mixed clinical-community male sample ( N = 20, Study 1) and then tested its efficacy in a double-blind, randomized sham-controlled clinical trial in a mix-gender sample of people with clinical levels of hostility ( N = 135, Study 2). Overall, we found that eight sessions of CBM-I across four weeks increased benign bias. Study 1 indicated moderate to good feasibility but showed no significant effects on hostile bias. In Study 2, CBM-I increased benign bias and reduced hostile bias and aggressive behavior with small effect sizes. However, we observed no differential changes in self-reported hostility measures. Surprisingly, results indicated that general psychiatric symptoms decreased to a lesser extent in the CBM-I condition, compared to AC. Results furthermore showed that reductions in aggressive behavior were related to an increased quality of working alliance in subsequent psychotherapy following both interventions, but we did not observe a differential impact of CBM-I on working alliance in subsequent psychotherapy; however, this analysis was very underpowered. The strength of therapy effects on benign and hostile bias in the present study (i.e., d = 0.29 and d = 0.26 for benign bias, and d = -0.17 and -0.33 for hostile bias), are smaller than those in a number of previous studies. One study compared one CBM-I session to a sham condition in students and found medium effect sizes ( d = 0.44 for benign bias and d = 0.66 for hostile bias) (Hawkins & Cougle, 2013). Two other studies compared an eight-session CBM-I training to a sham condition and demonstrated medium to large effect sizes ( d = 1.17 for benign bias and d = 0.65 for hostile bias) in non-treatment seeking people with alcohol use disorder (Cougle et al., 2017) and major depressive disorder ( d = 1.06 for benign bias and d = 0.64 for hostile bias) (Smith et al., 2018). A fourth study compared eight sessions CBM-I to a waitlist condition in aggressive boys and found large effect sizes ( d = 1.40 for benign bias and d = 2.21 for hostile bias) (Vassilopoulos et al., 2014). A fifth study compared one session of CBM-I benign training to a negative training condition and revealed a large effect size (approximate d = 0.85 for benign bias) (AlMoghrabi et al., 2018). However, the effect size on aggression in the current