Local and global effects of sedation in resting-state fMRI: a randomized, placebo-controlled comparison between etifoxine and alprazolam

Abstract

TSPO ligands are promising alternatives to benzodiazepines in the treatment of anxiety, as they display less pronounced side effects such as sedation, cognitive impairment, tolerance development and abuse potential. In a randomized double-blind repeated-measures study we compare a benzodiazepine (alprazolam) to a TSPO ligand (etifoxine) by assessing side effects and acquiring resting-state fMRI data from 34 healthy participants after 5 days of taking alprazolam, etifoxine or a placebo. To study the effects of the pharmacological interventions in fMRI in detail and across different scales, we combine in our study complementary analysis strategies related to whole-brain functional network connectivity, local connectivity analysis expressed in regional homogeneity, fluctuations in low-frequency BOLD amplitudes and coherency of independent resting-state networks. Participants reported considerable adverse effects such as fatigue, sleepiness and concentration impairments, related to the administration of alprazolam compared to placebo. In resting-state fMRI we found a significant decrease in functional connection density, network efficiency and a decrease in the networks rich-club coefficient related to alprazolam. While observing a general decrease in regional homogeneity in high-level brain networks in the alprazolam condition, we simultaneously could detect an increase in regional homogeneity and resting-state network coherence in low-level sensory regions. Further we found a general increase in the low-frequency compartment of the BOLD signal. In the etifoxine condition, participants did not report any significant side effects compared to the placebo, and we did not observe any corresponding modulations in our fMRI metrics. Our results are consistent with the idea that sedation globally disconnects low-level functional networks, but simultaneously increases their within-connectivity. Further, our results point towards the potential of TSPO ligands in the treatment of anxiety and depression.

Fig. 1. Analysis strategies used in our study to characterize activity dynamics in rs-fMRI.

A illustrates how resting-state FC between different ROIs was computed. We used an atlas to subdivide the cortex into segregated ROIs and correlated the average BOLD time courses within each ROI, yielding a functional connectivity matrix. Using different graph theoretical measures allowed us to analyze changes of the region-wise or whole-brain connectivity. B shows how ReHo maps were derived from rs-fMRI data. For each vertex on the surface the neighbouring vertices within a pre-defined radius were selected to define a local neighbourhood (marked in blue in this example). By computing the similarity between all time courses within this neighbourhood, we obtained a ReHo value for each vertex on the surface. C illustrates the steps to compute a fALFF map from rs-fMRI data. For each vertex, the power spectrum of its BOLD time course was computed, and fALLF resulted from the ratio of the low-frequency power to the power of the entire frequency range. D shows the basic principle of ICA of rs-fMRI data. Spatial-temporal activity maps acquired during a session at different timepoints (t) were decomposed by ICA into a set of spatially independent components (ICs), representing coherently activated resting-state networks.

Fig. 2. Results of the side effects questionnaire for the different interventions.

Alprazolam produced stronger side effects such as fatigue, sleepiness, concentration impairment, dizziness, and confusion compared to etifoxine and placebo. Error bars represent 95% confidence intervals across subjects. Significant differences between treatments: *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001, ****p ≤ 0.0001, ns not significant, (ns) not significant after false discovery rate correction.

Fig. 3

A–C Average edge density, global efficiency and local efficiency of the FC network under different medications andl for different correlation thresholds σ. We found reduced edge density and reduced global and local efficiency in the alprazolam condition compared to baseline, and the placebo and etifoxine conditions. In (D) a reduction of the rich-club coefficient due to administration of alprazolam can be observed. Regions, where the difference to alprazolam becomes significant, are indicated in bars above, with the respective colours for the placebo, etifoxine and baseline condition. Error bars represent 95% confidence intervals across subjects. Significant differences between conditions in (A–C): *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001, ns not significant, (ns): not significant after false discovery rate correction. Bars in (D) indicate regions where a significant difference to alprazolam can be observed with p ≤ 0.05.

Fig. 4. Comparison of rs-fMRI measures between alprazolam and placebo.

The first row (A) shows the regions in the left (L) and right (R) hemisphere where FC degree was significantly lower after administering alprazolam compared to placebo. The second row (B) depicts several regions in which local connectivity, as defined by ReHo, was lower after administering alprazolam compared to placebo, except for one region in superior/medial aspects of somato-motor cortex. The third row (C) shows that low-frequency amplitudes, as derived from fALFF, were higher after administering alprazolam compared to placebo. The fourth row (D) illustrates alterations in ICA-based resting-state networks, which display higher connectivity after the administration of alprazolam within the temporal, occipital and right somatosensory cortex. Yellow and blue colours depict t-values (from blue, alprazolam < placebo, to yellow, alprazolam > placebo).

Fig. 5. Comparison of rs-fMRI measures between alprazolam and etifoxine.

The first row (A) shows that FC degree was significantly lower after administering alprazolam compared to etifoxine. The second row (B) depicts several regions in which local connectivity, as defined by ReHo, was lower after administering alprazolam compared to etifoxine. The third row (C) shows that low-frequency amplitudes, as derived from fALFF, were higher after administering alprazolam compared to etifoxine. The fourth row (D) illustrates alterations in ICA based resting-state networks, which show higher coherence in the alprazolam condition in the temporal, occipital and right somatosensory cortex. Yellow and blue colours depict t-values (from blue, alprazolam < etifoxine, to yellow, alprazolam > etifoxine).