Clinical Utility of fMRI in Evaluating of LSD Effect on Pain-Related Brain Networks in Healthy Subjects

Faramarzi and colleagues situate their study within renewed scientific interest in classic psychedelics and an expanding literature that explores their therapeutic potential, including for chronic pain. Earlier functional neuroimaging work has characterised LSD's acute effects on emotion, self-processing and perception, and separate fMRI studies have detailed network alterations in chronic pain, but the specific effects of LSD on the pain neural network (PNN) have not been examined. The introduction therefore frames a gap: how LSD modulates the brain circuits that process pain in healthy humans remains unknown. This study set out to examine LSD-induced changes in brain regions implicated in pain processing using resting-state fMRI. Using multiple analysis techniques—amplitude of low-frequency fluctuation (ALFF), independent component analysis (ICA), seed-based functional connectivity and dynamic causal modelling (DCM)—the investigators aimed to map regional activity and both functional and effective connectivity among key PNN nodes (including left anterior insula, thalamus, secondary somatosensory cortex, anterior cingulate cortex and dorsolateral prefrontal cortex) in a within-subject crossover comparison of intravenous LSD and placebo in healthy volunteers. The authors present the work as a comprehensive fMRI biomarker study to improve understanding of LSD’s effects on pain-related brain architecture.

Twenty healthy volunteers were enrolled after medical and psychiatric screening and provided written informed consent. Main exclusion criteria included age under 21 years, any diagnosed psychiatric illness or family history of psychotic disorders, prior experience with classic psychedelics, psychedelic use within the preceding 6 weeks, pregnancy, problematic alcohol use (>40 units/week) and other medical contraindications. The extracted text also states that data were obtained from the OpenNeuro repository, but details of what data were added from that source are not clearly reported. A balanced-order, within-subject crossover design was used to compare intravenous LSD and matched saline placebo. Each participant completed two scanning sessions at least 2 weeks apart, although one passage in the extraction ambiguously states “one day placebo and the next day LSD”; this inconsistency is present in the extracted text. LSD was administered intravenously as 75 µg in 10 mL saline over 2 minutes, placebo was 10 mL saline, and participants were blind to session order while investigators were not. Resting-state data were acquired with eyes closed. The imaging protocol included arterial spin labelling (ASL) scans 100 minutes after dosing and two 14-minute BOLD resting-state fMRI scans and two MEG resting-state scans acquired at approximately 135 and 225 minutes post-infusion; total scan sessions also included further MEG and ASC questionnaire assessments. Preprocessing removed the first five frames, applied realignment (excluding runs with displacement >1.5 mm), tissue segmentation, co-registration of anatomical to functional images, normalisation to MNI space and spatial smoothing with a 6 mm Gaussian kernel. ALFF and regional homogeneity (ReHo) metrics were calculated using the CONN toolbox; ALFF was computed by band-pass filtering the voxel-wise BOLD time series and taking the root-mean-square. ICA was performed with 64 components at group level to identify resting-state network maps focusing on pain-related nodes (insula, thalami, frontal pole, ACC, parietal operculum). Functional connectivity was assessed using multivariate ROI-to-ROI connectivity (mRRC) in CONN, with permutation testing (10,000 permutations) and false discovery rate (FDR) correction at P ≤ 0.05. Effective connectivity was modelled using dynamic causal modelling (DCM) following Friston’s Bayesian framework. Regions chosen for DCM (based on prior literature and neuroscientist input) included ACC, SII, dlPFC, left anterior insula (lAIC) and left posterior insular cortex (lPIC). Time series were extracted from ROIs and used to compare connection strengths between sessions; model estimation and session-level analyses applied Bayesian parameter averaging and FDR correction (P = 0.05). Statistical comparisons used Kolmogorov–Smirnov tests for normality, t-tests for normally distributed measures and Mann–Whitney U tests when distributions were non-normal, with α = 0.05. The extracted text indicates visual analogue scale (VAS) ratings of pain intensity were collected during scans and the 11-item altered states of consciousness (ASC) questionnaire completed after each session, but the methods do not report detailed timing or analytic steps for correlating imaging markers with these behavioural measures.

ALFF results showed that, across pain-related regions, the placebo session yielded more activated voxels than the LSD session. The extracted text reports specific voxel counts for some regions under placebo: frontal pole right 42 voxels, frontal pole left 31 voxels and ACC 559 voxels (P < 0.05). The authors state that spontaneous activity increased in both LSD and placebo sessions in several pain-related areas, but the comparative ALFF maps indicated greater regional activation under placebo in most PNN nodes. ICA decomposition identified differential spatial patterns between conditions. Placebo produced more active voxels than LSD in ACC, left and right thalamus, right insula, left and right parietal operculum and right frontal pole. Conversely, LSD showed more activated voxels than placebo in left frontal pole and left insula. The group ICA was conducted with 64 components and pain-related ROIs were selected by neuroscientist input; statistical thresholds reported indicate P < 0.05 for reported contrasts. Functional (seed-to-seed) connectivity analysis using mRRC revealed distinct patterns in the two sessions. In the placebo condition, left thalamus was functionally connected to parietal operculum and left parietal operculum showed connectivity with left and right frontal pole and left insula. Under LSD, left parietal operculum retained functional links with left and right frontal pole but the connectivity with left insula seen in placebo was not reported for LSD. DCM-based effective connectivity comparisons between sessions identified significant differences in three connections: lAIC to itself (lAIC–lAIC), lAIC to dlPFC, and SII to dlPFC (P < 0.05). Between-session statistical tests included two-sample t-tests and Mann–Whitney U tests; Bayesian parameter averaging was used in session comparisons. The extracted text mentions that correlations between fMRI biomarkers and clinical pain criteria were calculated, but it does not clearly report numerical correlation results or statistical values for those analyses.

Faramarzi and colleagues interpret their findings as evidence that LSD alters activity and connectivity within a set of pain-related brain regions, and they present this work as the first comprehensive fMRI biomarker study of LSD’s impact on the pain neural circuitry in healthy subjects. The authors report that placebo produced greater regional activation than LSD across many PNN nodes in ALFF and ICA analyses, while LSD produced relatively greater activation in left insula and left frontal pole. They suggest these patterns indicate differing influences on emotion, attention and response selection during pain processing, with LSD hypothesised to better modulate pain-related emotions and attentional control. The discussion links observed network changes to established pain pathways: the thalamus as a hub for sensory relay, ACC and dlPFC for cognitive and emotional appraisal and endogenous analgesia, and insula and parietal operculum for emotional responses to nociceptive input. Authors propose that increased activity in ACC and insula under either condition may engage endogenous analgesic pathways (PFC to PAG), and that altered PO–insula connectivity under placebo may represent a mechanism contributing to reduced pain sensation. Effective connectivity differences involving lAIC and dlPFC, and SII–dlPFC pathways, are highlighted as evidence that LSD modulates directional interactions within the PNN. The authors situate their results alongside broader psychedelic literature, contrasting LSD’s effects on thalamocortical and visual processing with other agents (psilocybin, MDMA) and noting overlapping but distinct network impacts. They acknowledge ongoing uncertainties about brain functions in pain processing and the need for further research. Methodological or interpretative caveats are discussed at a general level: the authors call for larger-scale computational approaches, improved noise-reduction methods (including fuzzy techniques), inclusion of additional brain regions, and shorter or less uncomfortable scan protocols to enhance data quality and clinical applicability. The extracted text does not clearly report whether the authors explicitly discuss potential biases introduced by investigators being unblinded to condition order or other specific limitations such as sample size constraints.

The authors conclude that multimodal fMRI analyses (ALFF, ICA, functional connectivity and DCM) provide a useful framework for characterising how LSD modulates activity and connectivity in pain-related brain networks of healthy volunteers. They suggest that the observed differences between LSD and placebo—particularly in insula and frontal pole activation and in effective connectivity involving lAIC and dlPFC—offer a promising basis for future clinical and mechanistic studies of psychedelics in pain modulation. Practical recommendations include expanding computational resources for fMRI analyses, applying noise-reduction techniques such as fuzzy methods, sampling additional brain regions for a more complete mapping of pain processing, and minimising participant discomfort during scanning to improve data quality. The authors position their work as an initial step towards using fMRI biomarkers to inform future clinical applications and policy decisions regarding psychedelic research in pain contexts.