Five Structural Biases in Neurofeedback-for-ADHD Research: Why the Evidence Looks Weaker Than It Should

Background. Neurofeedback is the most extensively studied non-pharmacological intervention forattention-deficit/hyperactivity disorder (ADHD). The most recent meta-analysis (Westwood et al.,2025) reports no clinically meaningful group-level advantage over control conditions (SMD = 0.04).Several secondary findings in the same literature, however, do not sit comfortably with a purenull, including delayed separation at follow-up, executive function improvements, dose–responserelationships, and effects that grow rather than fade over time.Methods. This narrative review examines the sham-controlled ADHD neurofeedback literatureand identifies five structural features of existing trials that likely reduce sensitivity to detect aspecific effect: (1) protocol standardization applied to neurobiologically heterogeneous populations;(2) reliance on the theta-beta ratio (TBR) despite its erosion as a reliable ADHD biomarker; (3)training doses below the range associated with clinical effect; (4) sham conditions that are ther-apeutically active rather than inert; and (5) exclusive reliance on symptom questionnaires ratherthan neurophysiological outcomes.Results. Each of these features is empirically documented and each pushes between-group sensitiv-ity in the same direction. An illustrative sensitivity analysis shows that phenotype mismatch alonecan shrink a within-responder effect of d = 0.8 to a group-level d between 0.16 and 0.48, dependingon how many enrolled participants match the targeted phenotype. Multiple double-blind studiesdemonstrate contingency-dependent operant neural responses that are absent under sham condi-tions (He et al., 2020; Guler et al., 2021; Hill, 2026), which fits a specific-mechanism account. Thereview proposes concrete design modifications: QEEG-guided protocol selection, adequate sessioncounts, stratified enrollment, and EEG co-primary endpoints.Conclusions. The arguments presented here do not establish that individualized neurofeedbackis effective for ADHD. They do constrain what the current null findings generalize to: the versionof neurofeedback the trials tested, not the version clinical practice delivers. What is warranted istherefore a sharper research agenda, not a settled verdict in either direction.