*2.6. Statistical Analysis*

Statistical analyses was performed using custom-written MATLAB scripts (V2017B; MathWorks) to test the e ffect of PSA agents and other variables on electrophysiological measures. For all tests, the threshold for statistical significance was set to α = 0.05. Patients were divided into 7 groups according to their sedative administration (no PSA (control group)), PSA discontinued (discontinued before MER), DEX, REMI, CLONI, DEX and REMI, CLONI and REMI (Table 1)).

As a first step of the statistical analysis, one-way ANOVAs were conducted for each electrophysiological measure to test for di fferences between groups. This was followed by multiple two-sample *t*-tests comparing data from each PSA group with data from the no PSA group. *P*-values were corrected for multiple comparison using the Benjamini & Hochberg method for control of the false discovery rate (FDR) [19].

For further insight, linear regression analysis was conducted using the applied PSA drugs, and clinical and demographic variables as predictors to test their e ffect on firing rate, CV and MUA. Additional linear regression analysis with a random e ffect grouped by patient ID were conducted, considering the clustered nature of the data.

Finally, we tested the e ffect of the PSA dose. We focused on DEX since the dose of CLONI was not consistent, which made impossible to run a dose-dependent analysis, while the e ffect of REMI was small. We conducted a correlation analysis including data from patients who received either DEX alone, or a combination of DEX and REMI. A standard linear model (no random e ffects) was constructed.
