A team led by Kanhao Zhao and Yu Zhang at Lehigh University has identified a distinctive pattern of brain connectivity that both differentiates cocaine use disorder from healthy controls and forecasts who will benefit from a non-invasive brain stimulation therapy aimed at curbing cravings. The researchers used resting-state brain scans from participants in Mexico to train a machine learning model, and then tested the pattern on participants in the United States and Mexico. The findings were published in Nature Mental Health.
Cocaine use disorder is characterised by compulsive drug seeking and high relapse rates, with current psychosocial treatments helping some patients but leaving others untreated. Medical professionals have long sought objective biological markers to improve diagnosis and tailor therapies.
To probe brain function, the study relied on functional magnetic resonance imaging, or fMRI, which measures blood flow to reveal which brain regions are communicating. Resting-state scans, taken while participants lie still without performing tasks, allow researchers to map functional connectivity across networks and assess how these networks interact.
Previous imaging work has shown structural and task-related brain changes in cocaine addiction, but resting-state connectivity can offer a more sensitive read on brain dysfunction. The researchers sought a reliable connectivity signature that could both identify the disorder and indicate likely treatment response to repetitive transcranial magnetic stimulation, or rTMS, a non-invasive therapy that uses magnetic fields to modulate brain activity in targeted regions.
In the discovery phase, the Mexican cohort comprised 129 individuals — 71 with cocaine use disorder and 58 healthy controls. Resting-state scans were analysed to map functional connections across brain regions, and a machine learning model was trained to separate the two groups.
The model identified a signature of cocaine use disorder with an overall accuracy of 83 per cent. Specific patterns included heightened communication between networks responsible for visual processing and attention, and between networks governing goal-directed behaviour and sensory attention. Conversely, connectivity between the brain’s default mode network and the limbic network, which processes emotions, was reduced.
To test whether the signature was robust beyond the initial sample, the researchers applied the same model to an independent replication group of 82 patients with cocaine use disorder and 81 controls from sites in the United States and Mexico. The signature correctly classified individuals in this group at a rate of about 70 per cent, with the same connectivity differences observed.
The team then explored whether the cocaine-use signature could predict treatment outcomes. They analysed data from a small randomised trial involving 45 patients with cocaine use disorder, of whom 25 received active rTMS and 20 received a sham procedure. Using baseline brain scans, they isolated the signature’s component connections and asked whether their strength could forecast changes in self-reported cocaine cravings after two weeks.
The predictive model successfully forecast craving reduction for those who received active rTMS, with particular frontal-lobe connectivity patterns linked to better outcomes. The model did not predict changes for the sham group, suggesting the signature’s predictive power was tied specifically to active treatment.
Additionally, the researchers observed that active rTMS appeared to normalise connectivity between networks involved in cognitive control and motor planning after the two-week course.
However, the authors caution that the study remains observational. While the findings establish associations between brain connectivity and cocaine use disorder, they cannot prove that the signature causes addiction. It remains possible that prolonged cocaine use alters brain networks, or that certain pre-existing patterns predispose individuals to addiction. Experimental manipulation of the implicated circuits would be needed to clarify causality.
The researchers also note limitations in participant demographics and data collection. The study included substantially more men than women, raising questions about whether the signature generalises across sexes. Differences in how functional MRI data were collected across sites and scanners could also affect results, underscoring the need for standardised procedures and broader geographic representation in future work.
If validated, the connectivity signature could eventually serve as a clinical screening tool, helping doctors determine whether a patient is a suitable candidate for rTMS or whether alternative therapies might be more appropriate.
The study, Discriminative functional connectivity signature of cocaine use disorder links to rTMS treatment response, lists Kanhao Zhao, Gregory A. Fonzo, Hua Xie, Desmond J. Oathes, Corey J. Keller, Nancy B. Carlisle, Amit Etkin, Eduardo A. Garza-Villarreal and Yu Zhang as authors.
