Robust identification of a neuromarker of methamphetamine craving

Abstract

In the January issue of Cell Reports Medicine, Tian et al.¹ apply high-density 128-channel resting-state electroencephalography (EEG) to examine the neurophysiological connectomes in individuals with methamphetamine use disorder (MUD). They identify neurobiological connectome biomarkers for craving prediction in MUD.

In the January issue of Cell Reports Medicine, Tian et al.¹ apply high-density 128-channel resting-state electroencephalography (EEG) to examine the neurophysiological connectomes in individuals with methamphetamine use disorder (MUD). They identify neurobiological connectome biomarkers for craving prediction in MUD.

Main text

The use of methamphetamine (MA) is a rapidly growing global health crisis. Recent surveys show that almost 2 million people in the United States (∼0.7% of the population) used MA, with approximately 1 million meeting criteria for MA use disorder (MUD).² Despite being a major health condition, existing research on MUD treatment effectiveness is weak, and the relapse rates are high.³ Although there is an urgent need to develop effective evidence-based treatments and interventions for MUD, such treatment development is thwarted by our limited understanding of precise neurobiological targets associated with key disease-specific clinical symptoms, such as craving in MUD.⁴

Drug craving is a complex psychological phenomenon⁵ and is often used as a target of treatments for substance use disorder. Unprovoked craving (or a strong desire for drugs) or cue-induced craving (elicited by cues associated with drugs) are strong predictors of drug use and relapse,⁶ but heterogeneity in assessing subjective craving presents a challenge for treatment-outcome studies. Recent evidence suggests that neuroimaging markers associated with craving can help uncover subtleties that may not be detectable with subjective assessment of craving.⁷

Specifically, the high spatial resolution of the functional MRI (fMRI) has been leveraged to delineate activity of specific brain regions, as well as connectivity within and between brain networks, that underlie cue-induced craving and predict clinical outcomes such as substance use, craving, and relapse.⁷ Studies have also used electroencephalography (EEG) to measure brain activity in acute substance use or substance use treatment andhave consistently found, among other effects, activity in the beta band (12–35 Hz) as an endophenotype of substance use, which may be associated with craving.⁸ Beta activity has established roles in somatosensory and motor processes and is involved with executive control, working memory, and brain network integration.⁹

Indeed, continuing advances in EEG technology and sophisticated signal processing algorithms have allowed researchers direct and millisecond-level access to regional and global neural activity for precisely localizing cortical regions and examining brain networks (using oscillatory connectomics approach) that are specific to drug cue reactivity and cue-induced craving. Nevertheless, our understanding of brain networks underlying craving in substance use disorder in general and, specifically in MUD, is still quite limited.

In their study,¹ Tian and colleagues address various existing gaps in knowledge by leveraging the temporal and spatial advantages of high-density EEG to examine source-resolved oscillatory connectivity networks. The authors inspected (1) differences in regions and networks between individuals seeking treatment for MUD and healthy controls (HCs), (2) networks associated with MA cue-induced craving in individuals with MUD, (3) machine-learning-based classification between MUD and HC groups, and (4) machine-learning-based prediction of MA craving. For this purpose, authors examined resting eyes-open and eye-closed EEG activity after the participants (101 MUD; 52 non-MA-using HCs) watched a 5-min craving-induction video related to MA use and rated their craving on a visual-analog scale. The MUD sample was divided into a subsample (n = 57) for MA-craving biomarker discovery and a subsample (n = 44) for craving prediction. Functional connectivity networks (FCNs) were generated for each EEG oscillatory band (i.e., delta, theta, alpha, beta, and gamma frequency bands) separately for eyes-open and eyes-closed conditions. Within the discovery sample, the authors reported that whereas FCNs in beta bands differentiated between MUD and HCs, FCNs in delta and beta bands, specifically those related to interoception and cognitive control, were associated with self-reported MA craving. Of these FCNs, only those in beta predicted self-reported MA craving at an individual subject level in the left out testing data.¹ Interestingly, all significant findings—that is, group differences and correlations with craving—were from the eyes-open condition.¹

These results are noteworthy because of (1) a large sample size allowing for training and testing subsamples; (2) the use of oscillatory connectomics and brain network approaches; and (3) machine-learning-based on sample prediction of MA craving. We have recently highlighted the dearth of EEG studies in MUD¹⁰ and even fewer that link EEG-based source-resolved FCN biomarkers to clinical outcomes, such as craving. Altogether, Tian and colleagues¹ have taken important steps in identifying putative MUD-specific biomarkers. Indeed, these results open possibilities for further examining whether these FCN-based neuromarkers modulate across different substance use and treatment phases, craving trajectories, inter- and intra-day variability, and subtypes of individuals. Additionally, these results set the stage for deeper mechanistic examinations of specificity of these FCNs—that is, increased beta connectivity between right insula and right superior eye field connectivity and reduced beta connectivity between left superior motor cortex and bilateral visual cortex—in relation to MUD and MA craving. Other questions include how or whether these beta FCN neuromarkers of MA craving are associated with consistently observed findings of dysregulated beta in substance use disorders or what the role of beta FCNs in relation to those of other EEG bands, or their association with other networks, is and what the appropriate interventional targets that will modulate beta FCN activity would be and whether that modulation would change clinical outcomes in MUD. It was also encouraging to see that the authors adequately acknowledged several limitations of the study, including the lack of EEG assessment before craving induction,¹ which limits the inference about the specificity of the association between beta FCNs and MA craving.

Taken together, these results set an important direction for future studies to replicate these results and to further develop techniques for more robust and generalizable detection of neuromarkers of high clinical utility in substance use disorders. However, long-term issues loom in terms of systemic gaps in treatment availability and utilization, addressing the sparsity and samples sizes of most EEG studies in substance use and treatment research, and modeling the transdiagnostic role of dysregulated beta in intrinsic functional networks found not only in various substance use disorders but also in other psychiatric conditions.¹⁰