Groups had statistically equivalent BDI-II scores at baseline, both when examining all participants (t(36) = 0.50, ns) and participants with post-program data, (t(23) = 0.16, ns; Table 1). A one-way analysis of covariance (ANCOVA) with group as the independent variable conducted on post-program BDI-II scores while controlling for baseline BDI-II scores revealed a highly significant effect for group, with MBCT + TAU (M = 10.60, SD = 8.12) exhibiting lower BDI-II scores post-program than controls (M = 25.46, SD = 14.91; F(1,22) = 22.51, p < 0.001, η
= 0.505, 90% CI (0.23, 0.65); Fig. 3). BDI-II scores from week 6 were used for four participants who discontinued before post-testing. This effect remained highly significant after excluding these participants F(1,18) = 29.14, p < 0.001, η
= 0.618, 90% CI (0.33, 0.74).
Groups had statistically equivalent baseline HAM-D-28 scores, both when examining all participants t(35) = 0.47, ns (Table 1) and when examining only participants with post-program data (t(23) = 0.71, ns). ANCOVA conducted on post-program HAM-D-28 scores while controlling for baseline scores revealed a significant effect for group, with MBCT + TAU (M = 12.63, SD = 8.76) exhibiting lower HAM-D-28 scores post-program than wait-list + TAU (M = 20.44, SD = 6.35), F(1,22) = 4.77, p < 0.05, η
= 0.178, 90% CI (0.005, 0.39; Fig. 3).
Groups were equivalent at baseline in reaction time (RT; t(38) = 0.36, ns) and error rates (t(38) = 1.00, ns) when including all participants and when including only participants with post-program data (maximal t(25) 1.38, ns). An ANCOVA with group as the independent variable conducted on post-program RT while controlling for baseline RT revealed a significant effect for group, with the MBCT + TAU group (M = 44 ms, SD = 77 ms) exhibiting an increased CRS effect compared to the control group (M = −41 ms, SD = 109 ms; F(1,24) = 6.13, p = 0.02, η
= 0.20, 90% CI (0.02, 0.40); Fig. 4). A similar ANCOVA conducted on post-program error rates while controlling for baseline error rates did not reach significance (F(1,24) = 1.57, ns).
The relationship between CRS and depressive symptoms
Multiple regression was used to predict change in BDI-II based on change in CRS error rates and RT. A significant regression model was found (F(1,19) = 6.91, p = 0.02) with change in error rates associated with change in depressive symptoms. The multiple correlation coefficient was 0.52, indicating that overall improvement in irrelevant mental-set suppression explained 26.7% of the variance in BDI-II scores (β = 0.52, p = 0.02; R
2 = 0.267). A similar regression was calculated with group as an additional factor. A significant model was found (F(2,18) = 19.20, p < 0.001), with a multiple correlation coefficient of 0.82 (R
2 = 0.64). Group (β = 0.68, p < 0.001) was a significant factor and error rates just attained significance (β = 0.30, p = 0.05).