Attention-deficit/hyperactivity disorder (ADHD) is associated with an elevated risk of substance use disorders (SUDs), yet most affected individuals do not develop addiction, indicating that vulnerability is selectively distributed rather than intrinsic to the disorder itself. This Review synthesizes current evidence on the neurodevelopmental mechanisms linking ADHD to SUD vulnerability, examines reward-related dysregulation as a unifying mechanistic framework underlying this association, and proposes a subgroup-based precision psychiatry perspective to explain the heterogeneity of addiction risk among individuals with ADHD. Integrating evidence across genetics, molecular neuroscience, neurodevelopment, systems-level circuitry, computational modeling, and environmental modulation, converging findings implicate polygenic variation affecting synaptic and dopaminergic pathways that interacts with atypical maturation of frontostriatal and frontocortical systems supporting reward valuation, inhibitory control, and salience processing. These multilayered perturbations may manifest computationally as dysregulated reinforcement learning, including altered dopamine signaling dynamics and impaired integration of delayed outcomes, biasing decision-making toward immediate reward. At the circuit level, disrupted coordination between mesolimbic reward systems and prefrontal control networks may facilitate a shift from goal-directed behavior to habit-based responding following repeated substance exposure, further shaped by experience-dependent neuroplasticity. Importantly, available evidence suggests that addiction risk is concentrated within distinct ADHD subgroups characterized by reduced responsiveness to anticipated rewards, impaired reinforcement learning, preference for immediate over delayed rewards, impaired inhibitory control, emotional dysregulation, and impulsive decision-making that emerge and evolve across development. Collectively, this synthesis supports a precision psychiatry framework integrating genetic, neuroimaging, computational, and digital phenotyping approaches to improve the identification, prediction, and prevention of SUD vulnerability in ADHD.

Treatment-resistant depression (TRD) remains a highly burdensome clinical condition, and improvement in depressive severity does not necessarily imply parallel modification of maladaptive cognitive-affective processes such as rumination. We compared 6-month trajectories of depressive symptoms and depressive rumination across three active treatment conditions in routine clinical care: intranasal esketamine-based treatment without depression-specific metacognitive training (ESK), depression-specific metacognitive training without esketamine exposure (D-MCT), and combined esketamine plus D-MCT (D-MCT + ESK). Of 65 screened patients, 12 were excluded because of at least one absolute contraindication and 53 were enrolled (ESK n = 15, D-MCT n = 12, D-MCT + ESK n = 26). Seven participants did not complete the 6-month endpoint assessment (six in ESK and one in D-MCT + ESK), yielding a complete-case analysis set of 46 patients (ESK n = 9, D-MCT n = 12, D-MCT + ESK n = 25). Generalized estimating equations showed significant time effects and time-by-group interactions for both MADRS and RRS total scores. Endpoint ANCOVA models adjusting for baseline outcome value showed significant group effects at 6 months for MADRS (F(2,42) = 24.97, p<.001) and RRS (F(2,42) = 5.59, p=.007). Six-month MADRS response among completers was 3/9 (33.3%) in ESK, 4/12 (33.3%) in D-MCT, and 23/25 (92.0%) in D-MCT + ESK; corresponding intention-to-treat estimates using non-responder imputation were 3/15 (20.0%), 4/12 (33.3%), and 23/26 (88.5%). Concomitant treatment profiles showed marked imbalance in previous non-pharmacological treatment exposure, representing an important potential confounder. Findings support a cautious, hypothesis-generating interpretation because allocation was nonrandomized, assessment was open-label, the sample was modest, and baseline rumination was higher in the combined group.