Predicting relapse to alcohol and drug abuse via quantitative electroencephalography

Abstract

A sensitive and specific screening test that would identify the subset of substance-abusing patients at highest risk for relapse would constitute an important advance for treatment planning. This study examined the relative value of quantitative electroencephalography as a rapid, inexpensive, and noninvasive measure of relapse potential. The subjects were 107 substance-dependent patients enrolled in residential treatment programs. All were unmedicated and free of the complicating effects of major medical and neurological disorders. Structured clinical interview data and a 5-minute recording of the resting, eyes-closed electroencephalogram were obtained after patients had verifiably maintained abstinence for 1-5 months. Patients were then monitored for relapse or successful abstinence by research staff for an ensuing 6-month period. ANCOVAs of EEG power spectral density within pre-defined frequency bands revealed an enhanced amount of high frequency (19.5-39.8 Hz) beta activity among the 48 patients who later relapsed compared to both 59 patients who maintained abstinence and 22 additional subjects with no history of substance dependence. Importantly, in subsequent logistic regression analyses, fast beta power was found to be superior to severity of illness, depression level, and childhood conduct problems in predicting relapse. With fast beta power as the sole predictor, the sensitivity, specificity, and positive and negative predictive value parameters for discriminating outcomes were 0.61, 0.85, 0.75, and 0.74, respectively. Additional ANCOVAs revealed that the EEG difference between relapse-prone and abstinence-prone groups was related to the interaction of two premorbid factors, viz., childhood Conduct Disorder and paternal alcoholism. The enhancement of fast beta electroencephalographic activity in patients who will later relapse most likely originates from a premorbid and subtle dysfunction involving frontal brain regions.