Dynamic Substates of the Default Mode Network Are Associated with Mystical Experiences and Clinical Outcomes after Magnesium-Ibogaine Treatment in Veterans with Traumatic Brain Injury

Psychedelics have shown promise for a range of psychiatric and neurological conditions, and earlier neuroimaging research has linked their effects to changes in the default mode network (DMN), a network involved in self-related thought, autobiographical memory, and mind-wandering. However, previous findings have been mixed, and most studies have relied on static connectivity measures that average brain activity across an entire scan. The authors also note that ibogaine, an atypical psychedelic, has only recently begun to show preliminary therapeutic potential, including in veterans with traumatic brain injury (TBI), and that its neural effects on the DMN had not yet been studied dynamically. Shinozuka and colleagues set out to test whether magnesium-ibogaine treatment alters DMN connectivity in veterans with TBI, and whether any neural changes relate to mystical experience and later clinical outcomes. They aimed to compare static and dynamic measures of DMN function across baseline, immediately after treatment, and one month later, with the broader goal of identifying more precise neural correlates of ibogaine’s subjective and therapeutic effects. The study builds on the same MISTIC cohort by focusing on temporal patterns within the DMN rather than treating the network as uniform across time.

The researchers used data from the prior MISTIC open-label observational study. The sample comprised 30 male U.S. veterans with mild-to-moderate TBI, each of whom received a single dose of ibogaine (12.1 ± 1.2 mg/kg) together with intravenous magnesium sulphate given before and after ibogaine (1 g each time) to reduce cardiovascular risk, particularly QT interval prolongation. Resting-state fMRI was acquired at three main timepoints: baseline, immediately post-treatment (3-4 days after ibogaine), and one month post-treatment. The extracted text reports usable fMRI data for 29 participants at baseline, 28 at the immediate post-treatment timepoint, and 25 at one month, with some missing data because of MRI safety issues or other logistical reasons. The researchers also used clinical and subjective measures collected in the parent study, including the Revised Mystical Experience Questionnaire (MEQ-30), CAPS-5 for PTSD, MADRS for depression, HAM-A for anxiety, and WHODAS-2.0 for disability. Long-term clinical follow-up was also mentioned at 3, 6, 9, and 12 months, and baseline TBI burden was characterised using the Boston Assessment of TBI-Lifetime. For static functional connectivity, the brain was parcellated with the Schaefer-200 atlas and grouped into the seven Yeo networks. The researchers calculated Fisher-transformed Pearson correlations between regions, averaged them within and between networks, and tested timepoint effects with linear mixed-effects models with subject as a random effect. They corrected timepoint contrasts using Benjamini-Hochberg false discovery rate (FDR) correction. They also correlated MEQ-30 scores with changes in within-DMN connectivity, and explored associations between clinical outcomes and static DMN connectivity. For dynamic analyses, the researchers restricted the HMM analysis to the 37 DMN parcels, fitting Gaussian Hidden Markov Models across K = 2 to K = 10 states and selecting the number of states using the elbow of the free-energy curve; K = 5 was chosen. They quantified fractional occupancy and switching rate, where switching rate refers to how often the brain transitions between substates and therefore inversely reflects how long a state persists. Mixed-effects models were again used to test timepoint effects, with additional models adjusting for head motion, age, and number of TBIs. They also correlated session-normalised changes in DMN substate 3 switching rate with MEQ-30 scores and with clinical outcomes, applying FDR correction. In a further exploratory analysis, the researchers fit HMMs to the whole-brain time series after principal component reduction to 15 dimensions. To interpret one DMN substate, the researchers constructed an internal-to-external cognition gradient using Cognitive Atlas terms, Neurosynth-derived activation maps, and a semantic embedding approach. They then tested whether the spatial pattern of DMN substate activity aligned with this gradient, using a null model that preserved spatial autocorrelation.

Static connectivity analysis did not show a significant mean change in within-DMN functional connectivity after ibogaine at either immediate post-treatment or one month. The reported estimates were β = -0.016, pFDR = 0.438 at the immediate timepoint and β = -0.028, pFDR = 0.175 at one month. The only significant static within-network change was a reduction in salience network connectivity immediately after treatment (β = -0.064, pFDR = 0.016), which was not present at one month. There were no significant between-network connectivity changes. Although mean within-DMN connectivity did not shift significantly, individual differences mattered. MEQ-30 mystical experience scores were positively correlated with changes in within-DMN connectivity at one month (r = 0.478, pFDR = 0.036), with a weaker trend immediately after treatment (r = 0.336, pFDR = 0.087). However, static within-DMN connectivity was not significantly associated with any of the clinical outcomes reported in the Methods at either post-treatment timepoint. The dynamic HMM analysis identified 5 DMN substates as optimal. Ibogaine was not associated with a significant change in fractional occupancy of any substate, but it was associated with reduced switching rate, meaning longer duration, of DMN substate 2 immediately after treatment (β = -0.010, pFDR = 0.020) and DMN substate 3 at both immediate post-treatment (β = -0.006, pFDR = 0.038) and one month (β = -0.007, pFDR = 0.020). These findings remained significant after adjusting for head motion, age, and number of TBIs. Head motion itself was not correlated with switching rates. For DMN substate 3, the session-normalised switching rate at the immediate post-treatment timepoint was negatively correlated with MEQ-30 score (r = -0.457, pFDR = 0.033), indicating that stronger mystical experiences were associated with a greater reduction in switching rate, or longer state duration. The one-month correlation remained negative but was not significant (r = -0.271, pFDR = 0.200). This substate showed higher activation in lateral than medial DMN regions, and this pattern aligned significantly with a gradient from externally oriented to internally oriented cognition (r = 0.346, pspatial = 0.045). The significance of this gradient was reported to persist at a 15% selection quantile but not at 5% or 20%. Exploratory analyses found that the change in session-normalised switching rate of DMN substate 3 at one month was positively correlated with all four clinical outcomes at that timepoint: CAPS-5 (r = 0.460, pFDR = 0.048), MADRS (r = 0.567, pFDR = 0.015), HAM-A (r = 0.700, pFDR = 0.001), and WHODAS-2.0 (r = 0.548, pFDR = 0.018). The authors interpret these positive correlations as meaning that larger increases in the duration of DMN substate 3 were associated with greater clinical improvement. The same switching-rate change also correlated with later improvements at 12 months in HAM-A (r = 0.627, pFDR = 0.015) and WHODAS (r = 0.551, pFDR = 0.035). Baseline TBI burden was not significantly correlated with changes in this switching rate. Whole-brain HMMs identified states that resembled canonical networks, including the DMN, but none of these whole-brain states showed significant switching-rate changes after ibogaine, and the whole-brain DMN-like state was not significantly correlated with MEQ-30 or clinical outcomes. Overall, the dynamic DMN substate 3 was the main neural feature linked to both mystical experience and clinical improvement.

The authors argue that this is the first study to describe post-acute effects of ibogaine on DMN dynamics, and they present three main interpretations. First, dynamic measures were more sensitive than static measures for detecting ibogaine-related changes and for relating brain changes to clinically meaningful outcomes. In their view, static connectivity still captured some individual variability, especially its positive association with mystical experience at one month, but it did not identify the treatment-related temporal reorganisation that dynamic analysis revealed. Second, they interpret the reduction in switching rate of DMN substate 3 as a specific post-treatment alteration that tracked both mystical experience and clinical improvement. They contrast this with earlier psychedelic studies, which often measured the brain during the acute drug experience and reported decreases in within-DMN connectivity. The authors note that their fMRI scans were collected 3-4 days and one month after treatment, so the post-acute pattern they observed may differ from acute psychedelic effects. They also note that the salience network showed a transient reduction in static connectivity immediately after treatment, which they suggest could relate to reduced hypervigilance in PTSD, but this did not persist at one month. Third, they argue that DMN substate 3 is spatially meaningful because it showed relatively higher activity in lateral DMN regions and lower activity in medial regions, and this pattern aligned with a gradient from externally to internally oriented cognition. They suggest, cautiously, that this may fit the phenomenology of mystical experience, in which self-focused awareness diminishes and engagement with the surrounding environment increases. However, they state that the functional meaning of this gradient was not directly experimentally verified in the study and requires further validation. The authors acknowledge several limitations. The treatment was open-label, so expectancy and context effects cannot be separated from ibogaine itself, especially because the treatment setting included other wellness activities such as sweat lodge, massage, yoga, reiki, breathwork, and meditation. The sample was all male and almost entirely White, limiting generalisability to women, non-White populations, non-veteran patients, and other veteran groups. They also did not test alternative methods for estimating DMN dynamics, such as leading eigenvector dynamics analysis, co-activation patterns, or sliding-window connectivity. Finally, they note that the gradient result was sensitive to the choice of selection quantile, and that future studies should collect fMRI both during and after the ibogaine experience to compare acute and post-acute neural effects directly. In terms of implications, the authors suggest that psychedelic neuroimaging should not treat the DMN as a single temporally uniform system. They propose that specific dynamic substates may be more informative than static averages for understanding both subjective experiences and therapeutic change after ibogaine.

The authors conclude that, in veterans with TBI, dynamic analysis revealed reorganisation of DMN activity after ibogaine that was not captured by static connectivity measures. They state that ibogaine specifically lengthened the duration of one DMN substate, and that this change was associated with mystical experience and improvements in PTSD, depression, anxiety, and functional disability. They also conclude that psychedelic neuroimaging is likely to be more informative when it examines specific DMN substates rather than treating the DMN as temporally uniform.