Specification dependence of EEG theta/alpha ratio associations with cognitive impairment: a multiverse analysis

Objective: The theta/alpha ratio, an index of electroencephalographic slowing, has been proposed as a marker of cognitive impairment, but estimates may depend on preprocessing, spectral, spatial, and statistical decisions. We used multiverse analysis to evaluate its robustness across plausible pipelines. Methods: Two resting-state EEG datasets were analyzed. Dataset 1 included 49 controls and 100 participants with Parkinson's disease spanning normal cognition to dementia. Dataset 2 included 29 controls, 36 patients with Alzheimer's disease, and 23 with frontotemporal dementia. Eight decisions were varied: filtering, artifact correction, independent component analysis (ICA), referencing, spatial summary, alpha-band definition, spectral method, and covariate adjustment, yielding 1,280 specifications (universes). Results: Findings varied across the unrestricted multiverse. In Dataset 1, all universes in the exploratory family combining ICA with absolute fast Fourier transform (FFT) or Welch power yielded p < 0.05 for comparisons of controls and cognitively normal Parkinson's disease with Parkinson's disease dementia and produced significant associations with Montreal Cognitive Assessment scores in more than 95% of universes. In Dataset 2, absolute FFT or Welch power supported control versus Alzheimer's disease differences in more than 95% of universes, while ICA with absolute power yielded significant Mini-Mental State Examination associations in all specifications. Higher theta/alpha ratios were consistently associated with greater cognitive impairment. Conclusions: The theta/alpha ratio is a promising but specification-dependent marker of cognitive impairment, contingent particularly on ICA and spectral power quantification. Significance: Identifying the analytic conditions under which theta/alpha findings remain stable supports standardized, reproducible evaluation of quantitative EEG markers of cognitive impairment.