Abstract
Background
Delusions, false beliefs held in the face of disconfirming evidence, are a prevalent and highly distressing feature of psychotic disorders. The neurobiology of delusions remains unknown but recent evidence suggests a role for abnormal prediction error neural signaling. Prediction error is neurocognitive process in which the brain signals the need to update beliefs when presented with information that disconfirms expectations. Task based neuroimaging studies have identified delusional beliefs correlate with altered activation in frontal and subcortical brain regions during prediction error, though such work is limited in scope. In a large sample of transdiagnostic psychotic patients we modeled the resting state effective connectivity of the delusion-associated predication error (D-PE) circuit.
Methods
Resting state fMRI was obtained from 289 psychotic subjects (schizophrenia, schizoaffective disorder, bipolar disorder with psychotic features) and 219 healthy controls, recruited as part of the multisite Bipolar & Schizophrenia Network on Intermediate Phenotypes (BSNIP1) study. Neuroimaging data were processed using CONN software with strict quality control criteria. Five D-PE regions of interest (ROIs) were created based on peak coordinates from published task-based prediction error fMRI studies: right Dorsolateral Prefrontal Cortex [r DLPFC], r Ventrolateral Prefrontal Cortex [r VLPFC], r Caudate, l Caudate and l Midbrain. In each subject the first eigenvariate was extracted from the rs-fMRI timeseries of each D-PE ROI. Spectral Dynamic Causal Modeling (spDCM) was performed on a fully connected model of the 5 ROIs. Parameters for the full model were fit using Parameter Empirical Bayes (PEB) and then passed to the group level where they were reduced using Bayesian Model Averaging (BMA). The association of effective connectivity with current delusional severity was tested using PEB-BMA controlling for antipsychotic medication, sex, age and scanner site. Significant effective connectivity was identified as parameters with free energy evidence greater than 95% probability. Additionally, we assessed the effective connectivity differences of this circuit between psychotic probands and healthy controls.
Results
Greater delusional severity was significantly associated with inhibition of the r Caudate by the r VLPFC, excitation of the r DLPFC by the l Caudate, and decreased self-inhibition of the r VLPFC and r DLPFC. Effective connectivity of the D-PE network in psychotic probands compared to healthy controls was associated with inhibition of the r Caudate by the r VLPFC, the r DLPFC by the l Midbrain, the l Midbrain by the r Caudate, and decreased self-inhibition of the r Caudate, r VLPFC, and r DLPFC.
Discussion
We found that resting state effective connectivity of the prediction error circuit is disrupted in psychotic subjects experiencing delusions. Specifically, delusion severity was associated with both increased bottom-up and decreased top-down frontostriatal connectivity along with greater disinhibition of the r VLPFC and r DLPFC. These effective connectivity results provide novel insight into the causal paths which may underlie delusion neural circuitry. This provides further evidence that dysconnectivity of prediction error system is a biomarker of delusions in psychosis. Furthermore, these transdiagnostic results implicate frontostriatal dysconnectivity as common neuropathology in delusions.
Regression dynamic causal modeling for resting-state fMRI
