Executive functioning (EF) deficits, including inhibitory control, are prominent features of major depressive disorder (MDD), persisting even in remission and not significantly improving with antidepressants. These deficits interfere with emotion regulation, a key skill in cognitive behavioral therapy (CBT). Inhibitory control is crucial for emotion regulation, as it involves inhibiting negative information processing. The Flanker task, measuring selective attention (a component of inhibitory control), is often used to assess this. While the Flanker task literature in MDD is mixed regarding group differences compared to healthy controls, more nuanced measures like post-error adjustments and sequential dependencies (Rabbitt, Laming, and Gratton effects) might be more sensitive. Neuroimaging studies show altered activity in inhibitory control regions (e.g., dorsal anterior cingulate cortex (dACC)) in MDD, often involving compensatory hyperactivity. Importantly, regions involved in inhibitory control overlap with those in emotion regulation (e.g., anterior insula (AI), temporoparietal junction (TPJ)). Resting-state functional connectivity (rsFC) provides a measure of intrinsic connections between these regions, minimizing task performance confounds. Previous studies have linked ACC activity to treatment response, but fewer have examined the broader inhibitory control-emotion regulation network. This study hypothesized that baseline inhibitory control and rsFC within this network would predict better response to internet-based CBT (iCBT), particularly compared to a monitored attention control (MAC) condition. iCBT offers scalability advantages over traditional CBT.

The existing literature shows a mixed relationship between executive functioning (EF) and treatment response in MDD. Some studies suggest that poor baseline EF predicts worse treatment response to psychotherapy or medication, while others find no significant relationship. Regarding inhibitory control specifically, the Flanker task is frequently used, with results showing mixed findings regarding group differences between MDD patients and healthy controls. However, more nuanced analyses, such as post-error adjustments (Rabbitt and Laming effects) and sequential dependencies (Gratton effect), may reveal subtle deficits. Neuroimaging studies provide stronger evidence, indicating hypoactivity or compensatory hyperactivity in inhibitory control brain regions in MDD. These regions often overlap with areas involved in emotion regulation. Resting-state functional connectivity (rsFC) has emerged as a promising measure to explore the role of these regions in depression and treatment response, showing relationships between rsFC and inhibitory control performance. Prior studies have identified ACC activity as a predictor of treatment response, but fewer studies have examined the broader network encompassing inhibitory control and emotion regulation regions. This study aims to address this gap.

This randomized controlled trial (RCT; NCT01598922) included 266 participants initially, with 60 (30 iCBT, 30 MAC) completing the study after excluding those who didn't complete treatment or had inconsistent psychiatric history. Participants had a primary diagnosis of current MDD and PHQ-9 scores between 10 and 23. No participants were taking psychotropic medications. The study involved an initial screening visit (baseline assessment: symptoms, neuroimaging, cognition), a 10-week intervention (iCBT or MAC), and a second in-person assessment. The iCBT involved six online lessons and weekly homework, while the MAC group only completed questionnaires. The PHQ-9 and HAMD assessed depression symptoms. Participants completed a modified Eriksen Flanker task outside the scanner, measuring accuracy and response time (RT), including interference effects, Gratton effects, and post-error adjustments. Resting-state fMRI (rs-fMRI) data were acquired using a Siemens Tim Trio 3.0 Tesla scanner. Preprocessing was done using SPM8 and CONN toolbox. Four regions of interest (ROIs) – dACC, bilateral AI, and right TPJ – were derived from a meta-analysis by Langner et al. (2018). Pearson correlations between time courses of these ROIs were computed, yielding rsFC values. Missing data were imputed using missForest. Elastic net regression was used to predict post-treatment PHQ-9 scores from baseline variables (demographic, PHQ-9, Flanker performance, rsFC). A three-stage approach was used to build the predictive model. Stage 1 included demographic, treatment group, and Flanker variables; Stage 2 used demographic, treatment group, and rsFC variables; Stage 3 combined variables from stages 1 and 2. Ten thousand replicates of elastic net regression were conducted, with variables retained if their non-zero coefficient proportion was ≥0.75. Essential predictors were those with absolute coefficients exceeding the average absolute coefficient. A linear regression model was fit to essential predictors to explore the directionality of effects. Supplemental analyses included examining all variables in a single step, controlling for additional clinical variables, analyzing a restricted sample with no missing data, and using the HAMD or response status as outcome measures.

The iCBT group showed lower baseline right AI-left AI rsFC and lower post-treatment PHQ-9 scores than the MAC group. Flanker interference effects were significant for accuracy and RT across both groups. The Gratton effect was significant for accuracy but not RT, and the post-error effect was significant for RT but not accuracy. Elastic net regression revealed that treatment group, gender, Flanker RT, right AI-TPJ rsFC, and left AI-right AI rsFC were essential predictors of post-treatment PHQ-9 scores. Interactions between treatment group and baseline PHQ-9 scores, age, gender, Flanker RT, Gratton accuracy, post-error accuracy, right AI-TPJ rsFC, and left AI-right AI rsFC were also essential predictors. The Stage 3 model (MSE = 0.89) significantly outperformed the null model (MSE = 0.99), Stage 1 (MSE = 0.94), and Stage 2 (MSE = 0.99) models. Linear regression analysis showed higher Flanker RT (worse inhibitory control) was associated with higher post-treatment depression, an effect stronger in iCBT than MAC. Higher post-treatment depression was associated with weaker right AI-left AI rsFC and stronger right AI-TPJ rsFC, effects stronger in iCBT than MAC. In the iCBT group, higher post-treatment depression was associated with lower Gratton accuracy and higher post-error accuracy. Supplemental analyses generally supported these findings, although using the HAMD instead of the PHQ-9 yielded some differences.

This study provides evidence that baseline inhibitory control predicts treatment response in MDD within an iCBT context. Better baseline inhibitory control (indicated by faster Flanker RT) and stronger sequential dependency effects (Gratton accuracy) were associated with greater symptom improvement, particularly in the iCBT group. Surprisingly, worse post-error accuracy adjustments predicted better outcomes in the iCBT group, potentially indicating iCBT's effectiveness for individuals with hypersensitivity to errors. Regarding rsFC, stronger right AI-left AI rsFC and weaker right AI-TPJ rsFC predicted better outcomes, again stronger in the iCBT group. The lack of dACC involvement might reflect the focus on AI's role in salience detection and network switching. The AI's involvement in error monitoring could explain its predictive role. This study's findings align with the literature linking better baseline EF to more favorable treatment outcomes. This study uniquely demonstrates EF as a prescriptive indicator of iCBT response. Limitations include the unexpected higher baseline rsFC in the iCBT group and the potential influence of emotion regulation. A larger sample is needed to replicate the interaction effects. The generalizability to other treatment modalities and timepoints needs further investigation, as does the reliance on self-report measures. The exclusive use of the Flanker task may also limit generalizability to other inhibitory control measures.

This study identified baseline inhibitory control and rsFC as predictors of iCBT response in MDD. Faster Flanker RT, stronger Gratton accuracy, weaker right AI-TPJ rsFC, and stronger right AI-left AI rsFC predicted better outcomes, particularly within the iCBT group. These findings contribute to the growing body of evidence supporting the importance of inhibitory control in psychotherapy outcomes. Future research should focus on larger samples, longitudinal studies, and exploration of other treatment modalities to confirm and extend these findings, ultimately aiming to personalize treatment selection for individuals with MDD.

This study has several limitations. The unexpected difference in baseline right AI-left AI rsFC between groups requires further investigation. The study did not directly assess emotion regulation, so the interplay between inhibitory control and emotion regulation remains to be clarified. The sample size limits the power to detect smaller effects, especially interaction effects. Longitudinal follow-up is needed to assess the durability of the treatment response. The study relied on self-report measures of depression symptoms. The generalizability of the findings to other treatments and populations warrants further study. The exclusive focus on the Flanker task needs to be expanded in future research.