Improved prediction of awakening or nonawakening from severe anoxic coma using tree-based classification analysis.

OBJECTIVE: To evaluate the utility of sensory and event-related evoked potentials for the prediction of awakening/nonawakening in severe anoxic coma and to design a decision tree helping decision for any patient in this condition.

DESIGN: Prospective cohort study.

SETTING: Clinical neurophysiology unit and intensive care unit of a French university hospital.

PATIENTS: Sixty-two consecutive severe comatose patients after out-of-hospital cardiac arrest.

INTERVENTIONS: None.

MEASUREMENTS AND MAIN RESULTS: We gathered clinical variables and recorded the somatosensory, auditory, and cognitive evoked potentials within an average period of 8 days after cardiac arrest. The patients were followed for 12 months and classified as awake or nonawake (permanent vegetative state or death). The statistical study included measurements of specificity, sensitivity, and positive and negative predictive value for each clinical and electrophysiologic variable recorded at the early stage of coma. Furthermore, a tree-based classification analysis was performed.All patients in whom somatosensory evoked potentials or middle-latency auditory evoked potentials were abolished did not awaken (100% specificity). All patients in whom mismatch negativity (MMN) was present awakened (100% specificity). MMN was superior to somatosensory evoked potentials for the prediction of awakening and had the best specificity and positive predictive value for awakening. On the decision tree, the awakening/nonawakening explicative variables were, by order of importance, MMN, pupillary reactivity, and somatosensory evoked potentials.

CONCLUSIONS: There is a need to predict early and accurately awakening or nonawakening in postanoxic comas. Using sensory and cognitive evoked potentials to assess the functional condition of the brain, a prognostic tree for the prediction of awakening/nonawakening in severe anoxic coma has been designed. It is applicable to any patient in this condition and offers the possibility to predict with very high probability awakening when MMN, the earliest component of event-related potentials, is present and nonawakening when MMN and pupillary light reflex are absent or cortical components of somatosensory evoked potentials are abolished.