Altered fronto‐amygdalar functional connectivity predicts response to cognitive behavioral therapy in pediatric obsessive‐compulsive disorder

Background: Based on findings from adults with obsessive‐compulsive disorder (OCD), this study examined alterations in resting‐state functional connectivity (rs‐fc) between the basolateral amygdala (BLA) and the ventromedial prefrontal cortex (vmPFC) in children and adolescents with OCD. We also assessed whether such BLA‐vmPFC connectivity changed with or predicted response to exposure and response prevention (E/RP), the first‐line treatment for pediatric OCD, given the involvement of these regions in fear processing, regulation, and extinction learning—a probable mechanism of action of E/RP. Methods: Resting state functional magnetic resonance imaging scans were acquired from 25 unmedicated, treatment‐naïve pediatric patients with OCD (12.8 ± 2.9 years) and 23 age‐ and sex‐matched healthy controls (HCs; 11.0 ± 3.3 years). Patients completed a 12–16‐week E/RP intervention for OCD. Participants were rescanned after the 12–16‐week period. ANCOVAs tested group differences in baseline rs‐fc. Cross‐lagged panel models examined relationships between BLA‐vmPFC rs‐fc and OCD symptoms pre‐ and posttreatment. All tests were adjusted for participants' age, sex, and head motion. Results: Right BLA‐vmPFC rs‐fc was significantly reduced (more negative) in patients with OCD relative to HCs at baseline, and increased following treatment. In patients, more positive (less negative) right BLA‐vmPFC rs‐fc pretreatment predicted greater OCD symptoms reduction posttreatment. Changes in BLA‐vmPFC rs‐fc was unassociated with change in OCD symptoms pre‐ to posttreatment. Conclusions: These results provide further evidence of the BLA‐vmPFC pathway as a potential target for novel treatments or prevention strategies aimed at facilitating adaptive learning and fear extinction in children with OCD or subclinical OCD symptoms.

Keywords: biological markers; child and adolescent; cognitive behavioral therapy (CBT); functional magnetic resonance imaging (fMRI); obsessive‐compulsive disorder (OCD)

Obsessive‐compulsive disorder (OCD) is a disabling mental illness characterized by recurrent intrusive thoughts, images, and impulses (obsessions) and repetitive rituals (compulsions) aimed at neutralizing the distress or anxiety generated by the thoughts. Cognitive behavioral therapy (CBT) with exposure and response prevention (E/RP), a first line treatment for both children and adults with OCD (Ost et al., 2015; Rosa‐Alcazar et al., 2015), shows better response than pharmacotherapy in pediatric OCD, yet treatment‐resistant symptoms often remain for many patients following either type of treatment (M. E. Franklin et al., 2011; Pediatric OCD Treatment Study, 2004). Thus, identifying reliable biologically‐based predictors of CBT outcomes is paramount to informing clinical decisions and developing individually‐tailored interventions (i.e., precision medicine). Resting‐state functional magnetic resonance imaging (rs‐fMRI) is increasingly recognized as a promising tool for this endeavor. In the present study, we used longitudinal rs‐fMRI data from children and adolescents with OCD to investigate whether resting‐state functional connectivity (rs‐fc) in a specific neural circuit is altered in pediatric OCD and characterize connectivity‐symptom changes from pre‐ to post‐CBT. Specifically, we focused on rs‐fc between the basolateral amygdala (BLA) and the ventromedial prefrontal cortex (vmPFC), a likely mechanistic component through which E/RP alleviates OCD symptoms.

The prevailing neurobiological model of OCD focuses on disturbances in parallel cortico‐striato‐thalamo‐cortical (CSTC) circuits (Milad & Rauch, 2012). Yet, emerging evidence from fMRI studies pointed to alterations beyond CSTC circuitry (Beucke et al., 2014; Gursel et al., 2018). Notably, alterations in fronto‐amygdalar circuitry that supports fear processing, regulation, and extinction appear to be a key component of OCD pathophysiology (Milad & Rauch, 2012). Fear extinction is central to adaptive learning and fear inhibition, processes known to be disturbed in OCD and anxiety disorders (Graham & Milad, 2011). Animal research examining the neural basis of fear extinction have particularly focused on the amygdala and its associations with the vmPFC (Quirk & Mueller, 2008).

The amygdala is a complex structure involved in processing aversive experience and negative emotional information (Shackman et al., 2017) and comprises at least two cytoarchitectonically distinct subregions: the BLA and the centromedial amygdala (CMA; LeDoux, 2000, 2007; Phelps et al., 2004). Animal and human research point to distinct functional roles of the BLA and CMA. Through its connections with multiple cortical and subcortical brain areas, including prefrontal and sensory cortices, thalamus, hippocampus and striatum, the BLA is primarily implicated in sensory integration and regulation of emotionally significant events, as well as in fear and extinction learning (Janak & Tye, 2015; Orsini & Maren, 2012; Phelps et al., 2004). In contrast, the CMA serves predominantly as amydgalar output via its projections to the brainstem, cerebellum, and sensorimotor area, and is mainly involved in fear expression (LeDoux, 2000, 2007; Roy et al., 2009). Similarly, studies of rs‐fMRI in humans indicate that the BLA is functionally connected with temporal and frontal cortical regions, including the vmPFC, whereas the CMA is primarily connected with the hypothalamus, basal forebrain, striatum, and brainstem (Brown et al., 2014; Roy et al., 2009).

The vmPFC can be described as a major cortico‐limbic hub connecting multiple brain systems involved in an array of cognitive and affective processes and adaptive behavioral responses (Roy et al., 2012). Lesion studies in rodents implicated the vmPFC in fear extinction and highlighted its role in the retention of extinction learning and in the inhibition of conditioned response (Milad & Quirk, 2002). Importantly, the vmPFC appears to regulate BLA activity and support the inhibition of fear (Orsini & Maren, 2012; Senn et al., 2014). Thus, increased interactions between the vmPFC and BLA may mediate successful inhibition of fear responses and facilitate the acquisition and retention of fear extinction.

Fear circuitry (i.e., the amygdala–vmPFC pathway) has previously been implicated in OCD pathophysiology (see Milad & Rauch, 2012 for a review). Notably, the functioning of this circuit may be a predictor of CBT outcome in adults with OCD. e.g., rs‐fMRI findings suggest that lower degree centrality (a graph theory metric of functional connectivity) in the right BLA was associated with diminished response to CBT in a small sample of patients with OCD (Gottlich et al., 2015). Findings from a larger study showed that decreased BLA‐vmPFC rs‐fc predicted better response to CBT in previously medicated adults with OCD who had failed to respond to a course of a selective serotonin reuptake inhibitor (SSRI) treatment (Fullana et al., 2017). Thus, evidence points to the relevance of the BLA‐vmPFC pathway as a potential marker of OCD pathophysiology and treatment outcome. However, such questions remain untested in pediatric OCD.

Relationships between brain function and symptom change with treatment in adults with OCD have typically been assessed using univariate analyses such as correlations, univariate linear regressions, or logistic regression (Bernstein et al., 2018; Feusner et al., 2015; Fullana et al., 2017; Gottlich et al., 2015; Moody et al., 2017; Pagliaccio, Cha, et al., 2019; Pagliaccio, Middleton, et al., 2019). Such analytical approaches do not permit testing the directionality of relationships between brain measures (i.e., connectivity or activity) and symptoms, thereby precluding inferences about causality (Cole & Maxwell, 2003). In contrast, multivariate approaches, such as structural equation modeling with cross‐lagged panel models (CLPMs), can statistically disentangle whether pretreatment rs‐fc predicts CBT‐related changes in OCD symptoms, whether pretreatment symptom severity predicts changes in rs‐fc, or both. CLPMs allow for simultaneously testing of associations between variables over time while controlling for their cross‐sectional association within time points and for the stability of each individual variable across time points. Testing for cross‐lagged associations thus provides a conservative estimate regarding the proportion of change in one variable (e.g., symptoms) uniquely resulting from the other variable (e.g., baseline rs‐fc), thereby suggesting the directionality of relationships (Finkel, 1995).

In the present study, we used a seed‐based approach to examine rs‐fc between the vmPFC and the BLA in a sample of treatment‐naïve children and adolescents with OCD and healthy control (HC) participants. We also used CLPMs to examine relationships between rs‐fc and OCD symptoms pre‐ and post‐CBT, allowing us to identify and disentangle directional effects between these variables. We hypothesized that BLA‐vmPFC rs‐fc would be altered in children and adolescents with OCD relative to HC, and change following CBT. We also hypothesized that BLA‐vmPFC rs‐fc would predict CBT response and that the magnitude of change in connectivity would associate with the magnitude of symptom change pre‐ to post‐CBT. Given the scarcity of prior rs‐fMRI studies in pediatric OCD and heterogeneity in the methods and samples used in previous adult OCD studies, we did not formulate specific hypotheses regarding the direction of effects (i.e., decreased or increased rs‐fc). Finally, we explored whether detected effects were specific to the BLA‐vmPFC pathway or extended to another fronto‐amygdalar pathway (i.e., CMA‐vmPFC).

Participants were 55 unmedicated and treatment‐naïve children and adolescents (age 7–18 years) at baseline, including 28 patients with a primary diagnosis of OCD and 27 matched HCs (Figure 1; see Supporting Infrmation for a detailed description of sample selection). All participants provided informed assent and their caregiver provided informed consent. This clinical trial was approved by the Institutional Review Board at the New York State Psychiatric Institute (NYSPI). Structural and functional imaging data from some of the same participants included in the present study have been reported in two previous publications (Cyr et al., 2020; Pagliaccio, Cha, et al., 2019; Pagliaccio, Middleton, et al., 2019).

During the intake assessment, a structured diagnostic interview, the Anxiety Disorders Interview Schedule for Children (ADIS) adapted for DSM‐IV, child and parent versions (Albano & Siverman, 1996), as well as the Wechsler Abbreviated Scale of Intelligence (WASI; Wechsler, 1999) were administered to all participants. Participants with potential OCD were also administered the Children's Yale‐Brown Obsessive Compulsive Scale (CY‐BOCS; Scahill et al., 1997) to assess OCD symptom severity, as well as the Clinical Global Impressions (CGI) scale (Guy, 1976) to assess global clinical impression of OCD severity (i.e., patient's global functioning in relation to OCD symptom) and change over time. The CY‐BOCS and CGI were also administered at baseline, midpoint, and end of treatment (see below) for patients.

Patients with OCD were included if they met diagnostic criteria for OCD and had clinically significant symptoms (CY‐BOCS ≥ 16) at baseline (M. Franklin et al., 2003; Freeman et al., 2009). Other anxiety disorders were only permitted in the OCD group if OCD was the primary diagnosis. Participants with a history of neurological disorders, past seizures, head trauma with loss of consciousness, WASI IQ < 80, pervasive developmental disorder, tic disorder, or current psychiatric diagnoses (other than OCD and anxiety in the patients) were excluded. HC were excluded if they had any lifetime psychiatric diagnosis. Participants from both groups were scanned at baseline and again after 12–16 weeks. Scanning and treatment initiation in patients were scheduled within one month from baseline assessment.

Following baseline assessment and scan, patients with OCD underwent a course of manualized treatment of CBT with E/RP adapted for pediatric OCD (March & Mulle, 1998) delivered by a licensed clinical psychologist or advanced supervised graduate student in clinical psychology at the NYSPI. CBT treatment consisted of 12–16‐h‐long sessions. For exceptional cases not showing clinical improvement after six CBT treatment sessions (as determined by the therapist and study team psychiatrist based on CY‐BOCS and CGI scores), complementary pharmacological treatment (SSRI) was offered as part of our treatment protocol. See Supporting Information for details.

High‐resolution MRI scans were conducted at the NYSPI using sequences adapted from the Human Connectome Project (HCP) for a GE Signa 3T MR750 scanner. Each participant completed T1 and T2 weighted structural images and functional multi‐band EPI T2*‐weighted images of rs‐fMRI (repetition time = 850 ms, multiband factor = 6, 2 mm isotropic voxels, 7 min 42 s). Participants were instructed to stay still and keep their eyes open while viewing a fixation cross. The HCP preprocessing pipelines (Glasser et al., 2013) v.3.4 were used to preprocess all imaging data. See Supporting Information for details.

Following preprocessing, mean rs‐fMRI BOLD time series was extracted from five regions of interests (ROIs): left and right BLA and CMA, and bilateral vmPFC. Given the substantial number of functional neuroimaging studies reporting lateralized amygdala activation (Baas et al., 2004), separate BLA and CMA ROIs were defined for each hemisphere using the CTI168 Reinforcement Learning Atlas (Pauli et al., 2018). Consistent with prior research most consistently reporting bilateral vmPFC function (e.g., see Hiser & Koenigs, 2018 for a review), a single vmPFC ROI was derived from the FSL Harvard‐Oxford atlas maximum likelihood cortical atlas (http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/Atlases) with a 70% threshold. Rs‐fc strength between ROIs were computed using Fisher Z‐Transformation of the pairwise Pearson correlation coefficients.

Group differences in participant demographics were assessed at baseline and follow‐up using independent t‐tests for continuous variables and χ2 tests for categorical variables. In the patients with OCD, a paired t‐test assessed if CY‐BOCS scores declined pre‐ to posttreatment.

The significance threshold for all a priori tests was set to p < .025 (Bonferroni corrected for the two seeds tested, i.e., left and right BLA). To test for baseline group differences in BLA‐vmPFC rs‐fc, analyses of co‐variances (ANCOVAs) were conducted, adjusting for age, sex, and head motion (mean FD). For connections with significant group differences, within‐group repeated measures ANCOVAs were conducted to examine whether altered rs‐fc in OCD significantly changed after CBT, also adjusting for age, sex, and motion at each time point. Group‐by‐time interaction were not assessed given that follow‐up data were only available from 15 HCs.

CLPMs were constructed to test for relationships between OCD symptom severity (CY‐BOCS total scores) and rs‐fc pre‐/post‐CBT in the patients with OCD. These models provided estimates of the extent to which a variable pre‐CBT (e.g., rs‐fc) predicts another variable post‐CBT (e.g., symptom severity), over and above the variance attributable to the second variable pre‐CBT, and while controlling for covariance among these variables pre‐CBT (Finkel, 1995). Thus, baseline variables are tested as predictors of the residual or change in each variable from pre‐ to post‐CBT. Additionally, these models provide estimates of the association between the residuals of all post‐CBT variables; of interest here, was whether change in rs‐fc associates with change in OCD symptom severity pre‐ to post‐CBT.

The initial model, depicted in Figure 2, included all covariates and all variables at each time point were covaried to adjust for shared variance. Nonsignificant paths (ps > .1) were removed one at a time within each model, and a chi‐square difference test examined whether each removal significantly reduced model fit (see Supporting Information for order of removal and fit indices for the initial and final models). All stability paths were retained, even if nonsignificant, to adjust for baseline levels of each variable.

To ensure that baseline group differences in BLA‐vmPFC rs‐fc were not driven by the presence of comorbid anxiety in the OCD group, we computed Cohen's d values to compare the magnitude of the difference in rs‐fc across the following sub‐groups: (1) N = 23 HC versus all N = 25 OCD, (2) N = 23 HC versus N = 17 OCD with comorbid anxiety, (3) N = 23 HC versus N = 8 OCD without comorbid anxiety, and (4) N = 17 OCD with versus N = 8 without comorbid anxiety. Additionally, we conducted ancillary CLPMs adjusting for comorbid anxiety at each time point (i.e., using the presence/absence of anxiety as a binary time‐varying covariate). Finally, CLPMs were conducted after excluding two patients who began taking SSRIs mid‐treatment.

The same analyses described in Section 2.6 were used with left and right CMA‐vmPFC rs‐fc. The significance threshold for these exploratory analyses was set to p < .025 (Bonferroni correction p < .05/2 tests).

The final analytic sample consisted of 25 OCD and 23 HC participants at baseline and 21 OCD and 15 HC participants at follow‐up (Figure 1; Table 1). No significant group differences in age, sex, IQ, or head motion were detected (ps > .05; Table 1). Patients who completed treatment received an average of 14 CBT sessions. In patients, CY‐BOCS scores declined significantly following treatment (t(20)=6.230, p < .001), with an average decrease of 9.3 ± 6.9 points (percent change=38.6 ± 16.4%). 71% (n = 15) patients responded with >25% reduction in CY‐BOCS and 76% (n = 16) patients had a posttreatment CY‐BOCS score≤12.

1 TableParticipant demographics and clinical characteristics at baseline and follow‐up

1 Note: No significant group differences were detected at either time point (all ps > .05).

2 Abbreviations: CGI, Clinical Global Impression scale; CY‐BOCS, Children's Yale‐Brown Obsessive Compulsive Scale; HC, healthy controls; OCD, obsessive‐compulsive disorder; WASI‐II, Wechsler Abbreviated Scale of Intelligence—Second Edition.

Descriptive statistics and bivariate correlations between outcome variables are presented in Tables S1A and S1B for HC and OCD participants, respectively. Baseline right BLA‐vmPFC rs‐fc differed across groups (F(1,43)=5.822, p = .020), with reduced or more negative rs‐fc in the OCD group (M = −0.04, SD = 0.11) compared to more positive or near zero rs‐fc in the HC group (M = 0.03, SD = 0.09). Right BLA‐vmPFC rs‐fc in patients became significantly more positive following CBT (F(1,16)=6.872, p = .019). No significant effect of age, sex, and scanner motion on rs‐fc were detected (ps > .1). CLPMs revealed that greater (less reduced/negative) right BLA‐vmPFC rs‐fc at baseline predicted greater reduction in OCD symptoms following CBT (r = −.330, p = .018; Table 2; Figure 3). Change in rs‐fc did not associate with change in OCD symptoms pre‐ to post‐CBT, as indicated by the absence of significant relationship between rs‐fc and OCD symptoms post‐CBT after adjusting for the effects of pre‐CBT variables (i.e., their residuals were uncorrelated). See Supporting Information for details about pruning of noncontributory paths and goodness of fit.

2 TableFinal cross‐lagged panel results relating obsessive‐compulsive disorder (OCD) symptoms and resting‐state functional connectivity (rs‐fc) between right basolateral amygdala (rBLA) and ventromedial prefrontal cortex (vmPFC) before and after cognitive behavioral therapy (CBT) in the OCD group (n = 25)

3 Notes: OCD symptoms are based on scores from the Children's Yale‐Brown Obsessive Compulsive Scale. Rs‐fc represents Fisher's r‐to‐z scores. Based on a Bonferroni correction for the 2 models tested (i.e., left and right BLA‐vmPFC), ps < .025 were considered significant and corrected. The association between right BLA‐vmPFC rs‐fc pre‐CBT and OCD symptoms post‐CBT remained significant when the nonsignificant covariance paths within time point were retained in the model.

Cohen's d values for the baseline mean differences in right BLA‐vmPFC rs‐fc across all sub‐groups are reported in Table S2. Similar medium effect sizes were detected for the differences between HC participants and all OCD sub‐groups (i.e., with and/or without comorbid anxiety) and a small effect size was detected for the mean difference between OCD patients with versus without comorbid anxiety. Thus, comorbid anxiety was unlikely to contribute to our main finding of rs‐fc differences between OCD and HC participants. In ancillary CLPMs controlling for comorbid anxiety, the association between baseline right BLA‐vmPFC rs‐fc and OCD symptom change post‐CBT remained significant (r = −.291, p = .022) and comorbid anxiety at either time point did not associate with any other variables (Table S3). Similarly, the association between baseline right BLA‐vmPFC rs‐fc and symptom change post‐CBT remained significant after excluding two patients who began receiving augmentation treatment with an SSRI medication mid‐treatment (r = −.422, p = .007).

CMA‐vmPFC rs‐fc did not significantly differ across groups at baseline (all ps > .1) and did not change with CBT in patients. CLPM revealed that baseline OCD symptoms severity in patients predicted change in left CMA‐vmPFC rs‐fc pre‐ to post‐CBT (r = −.473, p = .016; Table S4). Baseline CMA‐vmPFC rs‐fc did not significantly predict change in symptoms severity from pre‐ to post‐CBT (ps > .1).

This study is first to show altered BLA‐vmPFC rs‐fc in pediatric OCD that predicted response to and changed with CBT. Compared to HCs who exhibited no or positive right BLA‐vmPFC rs‐fc, patients had reduced or negative rs‐fc. Less altered (i.e., more HC‐like) rs‐fc in the patients predicted better response to CBT and rs‐fc became more positive following treatment. These findings from pediatric OCD patients partially replicate and extend prior findings from adults with OCD (Fullana et al., 2017; Gottlich et al., 2015), providing further evidence of the potential of BLA‐vmPFC pathway as a promising target for novel treatments or prevention strategies aimed at facilitating adaptive learning and fear extinction in children with OCD or at risk for developing the disorder (e.g., with subclinical OC symptoms).

Our finding that BLA‐vmPFC rs‐fc predicts CBT outcome in pediatric OCD both aligns with and differs from previous findings from a large sample of adults with OCD (Fullana et al., 2017). Specifically, we detected decreased or negative baseline BLA‐vmPFC rs‐fc in patients compared to no or positive rs‐fc in HCs, whereas the prior adult study found decreased or less positive rs‐fc in OCD compared to more positive rs‐fc in HCs. Further, decreased/negative rs‐fc predicted worse CBT outcome, whereas decreased/less positive rs‐fc predicted better outcome in adults (Fullana et al., 2017). However, our sample consisted of treatment‐naïve, unmedicated children and adolescents with no comorbidites other than anxiety, whereas the prior study consisted of previously medicated adults who had failed to respond to SSRI treatment. Notably, developmental shifts in rs‐fc between the amygdala and prefrontal cortex, with positive versus negative rs‐fc in children and adults, respectively (Gee et al., 2013), may partly account for the discrepant findings between studies. In addition, our sample included patients with comorbid anxiety as long as OCD was primary, whereas several patients in the adult sample had a comorbid nonpsychotic mood (16.4%), anxiety (12.2%), or tic (9.6%) disorders. Thus, differences in sample characteristics across the two studies may account for the slight differences between findings. Nevertheless, both sets of findings suggest that altered BLA‐vmPFC predicts response to CBT in OCD.

The amygdala is involved in the mediation of fear and anxiety. Although many studies have reported unilateral activation of the amygdala and several hypotheses about putative functional differences between left and right amygdala have been proposed, findings are mixed and no clear consensus has emerged (Baas et al., 2004; Wager et al., 2003). Disregarding hemispheric specificity, the BLA (through its connection with vmPFC) is thought to be involved in fear processing and regulation, as well as extinction learning (Orsini & Maren, 2012; Senn et al., 2014). The latter might be a central mechanism of E/RP, the core component of CBT for OCD and treatment response is thought to be mainly accomplished via the successful reduction in fear response through this learning process (Myers & Davis, 2002). Thus, reduced BLA‐vmPFC rs‐fc in patients with OCD may reflect dysfunctional fear regulation and/or fear extinction learning processes that likely influence response to E/RP. Conversely, patients with less reduced rs‐fc may benefit from superior capacity to learn fear extinction and thus respond better to E/RP. Nonetheless, although we detected both significant reduction in symptoms and change in rs‐fc with CBT, our CLPMs suggest that these changes were unrelated to one another (i.e., no significant association between the residuals of the two variables post‐CBT). One possible explanation is that different components of CBT (e.g., psychoeducation, cognitive strategies, and reductions in parental reassurance/accommodation) impact on rs‐fc and symptoms independently and/or that a third variable mediates the association between rs‐fc and symptoms. One such variable may be the extent to which fear extinction was learned in patients. We did not measure extinction learning in this study but should in future studies that examine the specific mechanisms of neural and symptoms changes induced by E/RP.

Prior rs‐fMRI studies of OCD commonly employed a seed‐based, hypothesis‐driven approach focusing on CSTC circuits (Bernstein et al., 2016; Harrison et al., 2009; Posner et al., 2014; Sakai et al., 2011). The present study, along with others either focused on different circuits/networks—for example, default‐mode network (Beucke et al., 2014), BLA‐vmPFC (Fullana et al., 2017)—or employing whole‐brain, data‐driven approaches (Armstrong et al., 2016; Cyr et al., 2020; Feusner et al., 2015; Gottlich et al., 2015; Gruner et al., 2014; Moody et al., 2017; Shin et al., 2014; Tian et al., 2016) points to OCD‐related disturbances in brain circuits/networks that extend beyond the classical CSTC circuits. Few studies have examined rs‐fc in relation to treatment in pediatric OCD, but one small pilot study of youth with OCD (Bernstein et al., 2018) used a seed‐based approach to show increased connectivity in CSTC and salience networks with SSRI treatment that associated with changes in OCD symptoms. Using a whole‐brain, data‐driven approach with the same sample of youth with OCD as used in the present study, we previously identified altered rs‐fc between canonical large‐scale brain networks regions that also predicted response to CBT (Cyr et al., 2020). Together with these prior studies, our present findings point to the relevance of examining such circuits/networks in relation to OCD pathophysiology and treatment.

As in our prior study (Cyr et al., 2020), we carefully controlled for potential confounding effects, including age, sex, motion, and comorbid anxiety and mid‐treatment SSRI augmentation in the OCD group, rendering our analyses quite conservative. CLPMs adjusted for the independent effects of these potential confounds as well as baseline levels of each predicted outcome variable (i.e., rs‐fc and symptoms) in addition to the covariance between variables at each time point. These integrated longitudinal models thus allowed for disentangling the connectivity‐symptom relationships pre‐/post‐CBT, thereby permitting inference about the directional influence of BLA‐vmPFC rs‐fc on CBT‐related change in OCD symptoms, as well as the association between CBT‐related changes in rs‐fc and symptoms. Finally, all participants were medication‐free and treatment‐naïve at baseline, received protocol‐driven CBT with excellent response, and potential effects of mid treatment SSRI initiation in two patients were ruled out in ancillary analyses.

Despite these notable strengths, this study is limited by a modest sample size and thus insufficient power to detect effects of different OCD symptoms dimensions (Abramowitz et al., 2010), comorbidities, and other potentially relevant functional connections. Another limitation is the inclusion of OCD participants with comorbid anxiety disorders secondary to OCD. However, exploratory analyses revealed that this comorbidity did not likely contribute to our findings. In addition, because we did not have a waitlist or placebo control group, we cannot ascertain that the effects detected pre‐/post‐CBT were specifically due to CBT per se. Finally, although our two‐time‐point study design allowed for examining the directionality of the relationships between symptoms and rs‐fc pre‐/post‐CBT, a longitudinal design with more time points would further permit examination over longer term, specifically in relation to post‐CBT maintenance, remission, and relapse.

Notwithstanding these limitations, this is the first fMRI study to examine fronto‐amygdalar rs‐fc and relationships with symptom change pre‐/post‐CBT in pediatric OCD. Our sophisticated modeling of longitudinal data permitted disentangling the directionality of the associations between rs‐fc and OCD symptoms in relation to CBT response, which could not be achieved with univariate regression analyses. Our findings point to an altered BLA‐vmPFC pathway in pediatric OCD that may play a role in patients' ability for fear regulation and extinction learning, thereby enhancing their capacity to benefit from CBT with E/RP. Future longitudinal studies of large samples should replicate and extend these findings that may ultimately lead to treatment targets for future circuit‐based interventions aimed at improving fear extinction (and adaptive learning in general) in youth with OCD or subclinical OC symptoms.

This study was supported by a grant from the National Institute of Mental Health (R21MH101441; MPIs: Marsh and Rynn) and by a fellowship from the Fonds de Recherche du Quebec, Sante (Fellow: Cyr).

Dr. Rynn has received honorarium as a DSMB member from Allergan Inc. The other authors report no competing interests.

The peer review history for this article is available at https://publons.com/publon/10.1002/da.23187

The data that support the findings of this study are available from the corresponding author upon reasonable request.

GRAPH: Supporting information.