Table 3. Mediation analysis on persecutory belief conviction using mixed effects models.

	2 weeks (mid-treatment)	4 weeks (end of treatment)
Defence behaviours
Total effect	–1·61 (–12·66 to 9·45); p=0·78	–0·07 (–11·03 to 10·89); p=0·99
Direct effect	–0·66 (–11·04 to 9·72); p=0·90	2·27 (–8·10 to 12·65); p=0·67
Indirect effect	0·51 (–2·06 to 3·08); p=0·70	–0·52 (–3·09 to 2·05); p=0·69
Path A (group allocation to mechanism)	0·70 (–2·80 to 4·20); p=0·70	–0·71 (–4·21 to 2·79); p=0·69
Path B (mechanism to treatment outcome)	0·73 (0·42 to 1·04); p<0·001	0·73 (0·42 to 1·04); p<0·001
Proportion mediated*	0·32	7·51
Safety beliefs
Total effect	2·07 (–8·60 to 12·74); p=0·70	3·59 (–7·00 to 14·17); p=0·51
Direct effect	2·68 (–7·52 to 12·89); p=0·61	3·06 (–7·12 to 13·24); p=0·56
Indirect effect	–0·46 (–2·62 to 1·71); p=0·68	1·16 (–1·06 to 3·37); p=0·31
Path A (group allocation to mechanism)	2·32 (–8·66 to 13·30); p=0·68	–5·89 (–16·83 to 5·05); p=0·29
Path B (mechanism to treatment outcome)	–0·20 (–0·29 to –0·11); p<0·001	–0·20 (–0·29 to –0·11); p<0·001
Proportion mediated*	0·22	0·32

Data are estimate (95% CI), unless otherwise indicated.

^*Indirect/total effect.