Ty Smith was my graduate advisor (UCSD 1985, Systems Science), and I've used both BIS and SEDLine in research studies, including one I'm working on writing today.

The first problem is defining what you mean by "conditions are the same", since you need the output of the device to be changing to develop a probability density function. The assumption is that some sort of open loop infusion will do this, but it is subject to modeling errors. See the discussion in the appendix of the referenced paper. The problem is that while you may think you are at a steady state because you set the TCI to 3 µg/ml, you are slowly increasing or decreasing over the length of your EEG epoch.

The second problem is that these devices aren't measuring a fundamental property of the brain, they are simply combining multiple characteristics (burst suppression ratio, spectral edge, ...) that in the original population correlated with a clinical observation. So just because you discover that the mean effect site concentration associated with a BIS of 50 is 3.0 µg/ml in 50 year olds and 6.0 in 6 month olds, this doesn't mean that 6 month olds require twice as much anesthesia for other clinical endpoints.

The third problem is that these devices don't track induction and emergence very well. You need to stabilize in an anesthetic state for a little while before the numbers mean much. This requires a very slow induction, which you can do in most adults, but is more problematic in infants.

Having said all of this, if you have reliable drug levels for reliable EEG endpoints, you can compare probability density functions if you use appropriate statistical methods. Remember, the distributions will be skewed, and truncated at 0 and 100, so they may need to be fit to beta distributions. I had to find a Wharton School PhD mathematician to help me with that one! Good luck.

Article Safety and Efficacy of Drug-Induced Sleep Endoscopy Using a ...

Dear Jeff, thanks so much for your explanation and apologies for my late reply.

My study is focused on comparing BIS and Entropy concurrently applied in paediatric patients and we are trying to understand which one is more reliable at different ages. We used anaesthetic gas and we modelled the effect site concentration on the end-tidal and monitors values.

We used PK statistics to directly compare the two monitors' predictive performance to detect the anaesthetic concentrations within each age group separately, but we wondered if and how would be also possible to compare estimated PK values between different age groups( e.i infants vs children). 

Regard the 'same condition' I refer to what Smith's article says:' ..when comparing indicators,..,it is necessary to gather data using the same stimulus procedure and over the same distribution of anaesthetic depths.  A good way to ensure appropriate conditions for comparing the performance of two anaesthetic depth indicators is to measure the indicator values simultaneously for the same subjects-hence the importance of carry out paired data comparisons.'

Other authors have used PK values to compare monitors' performance from different data sets (with different distribution and sample sizes), but I am not sure whether this is statistically appropriate.

There is a group at University of British Columbia that is working on a device that will record EEG so it can be replayed into different monitors. Mark Ansermino would be a good contact.

https://www.researchgate.net/profile/Mark_Ansermino