Psychedelic Resting-state Neuroimaging: A Review and Perspective on Balancing Replication and Novel Analyses

This paper (2022) by some of the most well-known authors in the field of psychedelics, reviews 42 studies where resting-state functional magnetic resonance imaging (fMRI) was used and assessed methodological variations across studies. Nearly all studies varied in data processing and analysis methodology. Guidelines are proposed to ensure consistency across the field.

Abstract

“Clinical research into serotonergic psychedelics is expanding rapidly, showing promising efficacy across myriad disorders. Resting-state functional magnetic resonance imaging (rs-fMRI) is a commonly used strategy to identify psychedelic-induced changes in neural pathways in clinical and healthy populations. Here we, a large group of psychedelic imaging researchers, review the 42 research articles published to date, based on the 17 unique studies evaluating psychedelic effects on rs-fMRI, focusing on methodological variation. Prominently, we observe that nearly all studies vary in data processing and analysis methodology, two datasets are the foundation of over half of the published literature, and there is lexical ambiguity in common outcome metric terminology. We offer guidelines for future studies that encourage coherence in the field. Psychedelic rs-fMRI will benefit from the development of novel methods that expand our understanding of the brain mechanisms mediating its intriguing effects; yet, this field is at a crossroads where we must also consider the critical importance of consistency and replicability to effectively converge on stable representations of the neural effects of psychedelics.”

Authors: Drummond E-W. McCulloch, Gitte M. Kundsen, Frederick S. Barrett, Manoj K. Doss, Robin Carhart-Harris, Fernando E. Rosas, Gustavo Deco, Morten L. Kringelbach, Katrin H. Preller, Johannes G. Ramaekers, Natasha L. Mason, Felix Müller & Patrick M. Fisher

Summary

Clinical research into classic psychedelic drugs is expanding rapidly, and resting-state fMRI has emerged as a commonly used brain imaging strategy to identify associated neural mechanisms in both clinical and healthy populations.

Here we review the existing psychedelic resting-state fMRI literature and suggest that the field is at a crossroads where it must consider the critical importance of consistency and replication to effectively converge on stable representations of the neural effects of psychedelics.

Psychedelics have re-emerged in research since the early 2000s. They include psilocybin, LSD, ayahuasca and monoamine oxidase inhibitors.

Small clinical trials with psilocybin and ayahuasca have shown promising results in treating major depressive disorder, obsessive compulsive disorder, smoking addiction, alcohol abuse, demoralisation, and depressive and anxiety symptoms associated with diagnosis of terminal cancer.

Psychedelics bind multiple receptor targets, including alpha-adrenergic receptors, most serotonin receptors, trace amine receptors and sigma receptors, but the psychoactive effects in humans appear to be driven primarily by agonist effects at the serotonin 2A (5-HT2A) receptor.

Psychedelics last 4-6 hours orally and 812 hours intravenously, and are characterised by profound alterations in affect, perceptual alterations and synaesthesia, and mystical-like experiences. These effects have motivated research into how the brain processes psychedelics.

Resting-state functional Magnetic Resonance Imaging (fMRI) measures brain function and connectivity while the participant is lying still in the scanner and not engaged in a specific task. The primary analysis endpoint is functional connectivity.

Resting-state fMRI provides a passive framework for observing brain function during psychedelic sessions, and is therefore a particularly appealing framework for evaluating psychedelic effects on the brain.

In 2012, Carhart-Harris and colleagues published the first study investigating the effects of psychedelics on resting-state functional connectivity measures.

Sixteen datasets investigating resting-state fMRI have been collected from participants given psychedelics. These datasets range in size from 9-24 participants and were collected in clinical cohorts and healthy volunteers.

Three studies investigated the effects of drug administration within 24 hours, two studies investigated effects one week after, one study investigated effects three months after, and eleven studies investigated effects at any time after drug administration.

Eleven datasets have yielded one publication, three have yielded two publications, two studies have yielded nine and thirteen publications, and two publications draw data from both of these datasets. We encourage other groups in this space to make their data and analysis scripts available to encourage collaboration.

The Carhart-Harris 2016 dataset contains three 7.2 minute eyes-closed resting-state acquisitions between 70 and 130 mins after the IV administration of 75 g of LSD. Music played during the second acquisition was analysed in one study.

Across 16 unique datasets, participants were instructed to close their eyes during the resting-state session. Four studies performed multiple resting-state scans during the psychedelic sessions, and most of the data collected so far has opted for eyes-closed imaging.

The 37 published studies differ substantially in how they analyse BOLD data, with most studies using static functional connectivity analyses.

15 articles used seed-based connectivity analysis, 13 used small localised seeds only, one used both small and network seeds and one used network seeds only. The most commonly selected seeds were the medial prefrontal cortex, posterior cingulate cortex and thalamus.

Twelve articles investigated the connectivity of networks using a network-based approach. Of these, seven investigated within- and between-network connectivity, one studied the connectivity of the DMN with all other networks, one study performed gradient-based connectivity mapping and one study investigated the global connectivity of all networks defined.

Four studies investigated ROI to ROI connectivity, one study used complex network theory, one study calculated grand-average static functional connectivity, five studies investigated whole brain effects, and one study presented a voxel wise contract of psilocybin vs placebo of brain activity.

There are several system-based models of psychedelic effects on consciousness, including the DMN disintegration theory, the thalamic gating theory, and the claustro-cortical model. These theories can be reconciled using novel data.

The burgeoning interest in psychedelic effects has facilitated the application of novel neurocomputational models to the two IV datasets. These models include temporal variability, homological scaffolds, gradient-based connectivity, retinotopic coordination, whole-brain functional connectivity dynamics, connectome-harmonic decomposition, and leading eigenvector dynamic analysis.

Six studies have investigated the “entropy” of the RSFC signal, but they have calculated the entropy in different ways. Tagliazucchi, (2014) calculated the Shannon entropy of dynamic functional connectivity states, Lebedev, (2015) calculated the Shannon entropy of functional connectivity between first two, then 200 ROIs with 5 different networks, Viol, (2017) calculated the multi-scale sample entropy.

Although different studies may use different analysis methods, they may be measuring similar phenomena. Therefore, it is important to align in terms of how this phenomenon should be measured.

Spatial parcellation is important for replicable results in seed-, network-, or ROI-based analyses. Atlases and independent component analyses can be used to parcellate data.

24 studies used atlases derived from functional data, 14 studies used atlases derived from anatomical data, and two studies used atlases that took into account both structural and functional data.

Of the studies that used predefined atlases, five used the Harvard-Oxford Probabilistic anatomical atlas, four used the Automated Anatomical Labelling (AAL) atlas, six studies used the Yeo 2011 atlas, and three studies used the 17 network parcellation.

The twelve studies investigating the effects of psychedelics on network connectivity use six different atlases of networks, which vary in number of networks described and labels and descriptions of voxels that constitute each network.

Psychedelic drugs have been shown to have persistent effects on mood and personality in both healthy and patient population groups. More data on the persistent effects of psychedelics on neural function is needed to understand the remarkable therapeutic effects of psychedelics as well as highlighting potential risks.

Neuroimaging studies have been performed on LSD, psilocybin and ayahuasca, but there are many more untested psychedelic drugs that may be 5-HT2A receptor agonists and provide candidates for investigation as therapeutic agents.

Of 37 published studies, no pair of studies used the same core data processing and analysis strategy. This limits the ability to leverage early observations and hypotheses across datasets.

Data and analysis scripts should be shared on online repositories, plasma drug levels should be considered during acute imaging sessions, resting-state scans should be clearly demarcated from “resting-state”, and unique analyses should be replicated with independent data.

This paper has reviewed the heterogeneity within psychedelic resting state neuroimaging and how this leaves the field vulnerable to generating non-reproducible features of psychedelic effects.