Extended Data Fig. 1: Change in Montgomery-Åsberg Depression Rating Scale (MADRS) ratings over time.

Estimated mean MADRS rating scores from baseline to week 10 in the modified intention-to-treat analysis sample (n = 173) in active tDCS and sham tDCS treatment arms. Error bars represent ± 1 standard error (SE). MADRS scores range from 0 to 60 with higher values indicating more severe depressive symptoms. A significant improvement was observed in the change in MADRS ratings from baseline to week 10 in the active tDCS treatment arm, MADRS change 11.31 ± 8.81(standard deviation (SD)) (mean week 10 MADRS 12.46 ± 1.09 (SE)) as compared to sham tDCS treatment arm, MADRS change 7.74 ± 8.47 (SD) (mean week 10 MADRS 15.30 ± 1.07 (SE)) (95% CI 1.1 to 6.1, p = 0.006). The difference in change scores was also significant at week 4 (95% CI 1.2 to 5.5, p = 0.003) and week 7 (95% CI 1.1 to 5.8, p = 0.005) with a greater score decrease in the active treatment arm. Fully Conditional Specification (FCS) approach was used to produce 20 multiply imputed completed data sets. The FCS approach accommodates nonmonotonicity in the pattern of missing data and requires regression models to be specified for each variable with missing values needing imputation. All models included age, sex, in psychotherapy at baseline, use of any antidepressants at baseline and treatment group. The resulting completed datasets were combined using Rubin’s Rules. ** = p < 0.01.