Using Micro-Cognition Biomarkers of Neurosystem Dysfunction to Define ADHD Subtypes: A Scalable Digital Path to Diagnosis Based on Brain Function

BackgroundDiagnostic categories in psychiatry are based on symptoms and include individuals with different underlying pathology. This within-diagnosis heterogeneity confounds new treatment development and treatment selection for individual patients. A research priority is to discover biomarkers that define groups of patients with similar neuropathology. Digital neurocognitive therapy (DNT) and assessment can provide micro-cognition biomarkers of unprecedented precision. The brain is hierarchically organized from single cells to neurosystems that integrate action of millions of neurons across the brain necessary for cognition and emotion. Micro-cognition biomarkers identify dysfunction at the level of neurosystems that produce clinical illness. We used micro-cognition biomarkers to identify subgroups of children diagnosed with ADHD but with different neuropathology. MethodsK-means clustering was applied to 69 children 6-9 years old with ADHD using performance variables from a Go/NoGo test and the results analyzed against 58 typically developing (TD) children. Neurosystem dysfunction in each group was further characterized by micro-cognition biomarkers extracted from thousands of responses by each child during DNT. FindingsFour highly reproducible clusters were identified that differed on emblematic features of ADHD. Cluster 4 showed two core ADHD features, poor response inhibition and inconsistent attention. Cluster 3 showed only poor response inhibition and the other two showed neither. Cluster 2 showed faster and more consistent responses, higher detection of simple targets and better working memory than TD children but showed the most marked performance decrements when required to track multiple targets or ignore distractors. Cluster 1 showed much greater ability recognizing members of abstract categories rather than natural categories that children learn through physical interaction with the environment while Cluster 4 was the opposite. InterpretationDNT provides data-rich fine-grained, low-cost, noninvasive, and scalable micro-cognition biomarkers that characterize subgroups of patients with the same symptom-based psychiatric diagnosis but differing neuropathology.