A brain-connectivity model classified response to abstinence-based treatment, but its response definition was a reduction in craving. The model should not be described as a validated scan predicting who would remain free of relapse.
Du and colleagues studied 82 methamphetamine users and 68 healthy controls. Baseline functional MRI and clinical measures preceded long-term abstinence-based treatment, followed by reassessment of craving. Response required more than a 30% reduction in the craving scale. Principal-component regression with leave-one-out cross-validation identified network patterns associated with the response measure.
Thirty-nine participants met the response criterion and 43 did not. Connectivity patterns involving several large-scale networks were associated with response. The authors also related the imaging pattern to external gene-expression atlas data. Those atlas associations are not evidence that the same gene expression was directly measured in each participant, and internal cross-validation is not the same as testing a model in an independent clinical cohort.
The selected sample completed long-term follow-up under abstinence-based care. Generalization to people with different comorbidities or treatment settings therefore requires additional validation. Some comorbidity screening relied on verbal assessment rather than standardized scales. Above all, a craving-change endpoint is different from observed relapse, toxicologically confirmed abstinence or functional recovery. The study offers a candidate prediction framework and mechanistic hypotheses, not an individual clinical classification ready for routine use. The outcome definition must remain attached to any claim about prediction accuracy.
See an imaging study that actually followed reported relapse.