BACKGROUND: The ability of current depth-of-anaesthesia monitors to differentiate subtle changes in the conscious state has not been well characterised. We examine the variability in bispectral index (BIS) scores associated with disconnected conscious and unconscious states as confirmed by a novel serial awakening paradigm. METHODS: Seventy adult participants, given propofol or dexmedetomidine, had a cumulative 1381 electroencephalographic (EEG) recordings across two centres. Participants were awakened periodically, and their recent conscious experience interrogated by structured questioning. BIS were reconstructed from EEG using openibis, and the distribution of BIS scores were compared using linear mixed effects modelling. The predictive capacity of BIS across states of consciousness was also examined. RESULTS: Reconstructed BIS scores correlated significantly with blood concentrations of propofol and dexmedetomidine (all P<
0.001). However, while the average BIS was different between baseline wakefulness (mean BIS=95.1 [standard deviation=3.5])
connected consciousness with drug present (84.0 [10.9])
disconnected consciousness (70.0 [16.9])
and unconsciousness (68.1 [16.1]), the interquartile range of these states (3.6, 15.1, 23.3 and 26.8, respectively) indicated high degrees of overlap and individual variability. Connected consciousness could be differentiated from either disconnected consciousness or unconsciousness with 86% accuracy (i.e. 14% error rate), and disconnected consciousness differentiated from unconsciousness with 74% accuracy. CONCLUSIONS: These results agree with previous studies that BIS scores fail to reliably differentiate between states of consciousness, exacerbated by segregating connected, disconnected, and unconscious states. To develop a method that reliably identifies the conscious state of an individual (not an average), work is needed to establish the causal mechanisms of disconnection and unconsciousness.