In Reply.

Abstract

We thank Numan et al. for their interest and comments on our recent publication in Anesthesiology.1 The authors raise two concerns related to the methodology of our study. First, they point out that coherence, which was used to calculate functional cortical connectivity, may be affected by volume conduction. Spectral coherence is a widely used parameter-free feature in electroencephalogram studies and in brain functional connectivity analysis, and it measures the stability and reliability of phase differences between two sequences. We agree that in the analysis of a small number of electroencephalogram time samples, coherence estimates may detect spurious synchrony because of volume conduction or noise at the single subject level. However, in our study, the effect of volume conduction and noise was reduced for a number of reasons. First, we used the Welch modified periodogram averaging method to estimate the coherence from relatively long intervals of electroencephalogram recordings (88 s), followed by averaging across each frequency band. This procedure translates to an increased reliability and stability in coherence estimates, which reduces the effect of noise and random synchrony. More importantly, our analysis and conclusions were not based on an absolute value of localized coherence for each subject; rather, we used a baseline-corrected crossover design and performed a groupwise comparison between the two conditions (placebo vs. remifentanil) with subjectwise averaging. If we assume that for each subject, the effect of volume conduction holds between two consecutive experimental conditions, doing an analysis of differences between conditions would eliminate the effect of volume conduction. In our study, this approach resulted in a number of differences between baseline and remifentanil treatment, whereas there were no differences in the placebo arm. Taken together, we consider our findings to reflect significant changes in functional cortical connectivity specific to remifentanil administration, whereas the effect of spurious electroencephalogram fluctuations and volume conduction were cancelled out.Besides coherence, a number of measures of functional connectivity have been proposed, including the phase lag index as suggested by Numan and coworkers.2,3 We agree that each of these measures have their own properties and therefore provide complimentary information on the brain networks and could be considered in future studies.4Second, Numan et al. suggest normalizing the network measures obtained from graph theoretical analysis (using the corresponding graph measures on a simulated random matched graph with the same number of nodes and edges as the original graph). We disagree with this suggestion because the functional connectivity graph we constructed using coherence was a fully connected undirected graph. Therefore, when performing a comparison of differences between two conditions (based on the same graph structure), such normalization will have no effect on the final result and conclusions.The authors declare no competing interests.