An insecure attachment style (AS), described as being highly anxious and/or avoidant, is often assumed to be stable over time, yet some studies show that AS can change. To the extent that AS may be malleable over shorter time periods, it potentially impacts key therapy processes and outcomes. In the present study, we first investigated the stability of AS in patients with panic disorder (N = 49) treated with short-term cognitive behavioral therapy (CBT) including follow-up. Second, we tested whether time-specific change of AS predicted subsequent symptom severity, interpersonal distress (ID), and alliance, or vice versa. Third, we investigated if anxious attachment and ID average levels impact the alliance-outcome relation. Analyses were conducted at within-and between-patient levels with 5 measurements (baseline, intermediate, end, and follow-up after 4 and 8 months) over the course of CBT (12 sessions/8 weeks, 2 booster sessions) using linear mixedeffects models. A strong decrease in anxious attachment was found that was stable until 8 months after therapy. At the within-patient level, a reciprocal relationship of reduction in anxious attachment with less symptom severity, stronger alliance, and a prediction of less ID was found. ID decrease preceded less avoidance. At the between-patient level, anxious attachment and ID moderated the relation of alliance change and subsequent depressive symptoms. The prolonged improvement in interpersonal patterns suggests that short-term CBT positively modifies attachment working models in panic disorder patients. Effects on therapy process indicate that anxiously attached and interpersonally distressed patients seem to benefit more from alliance increase. 206 Psychotherapy