A 2014 study found that an imaging response during a risk-taking task was associated with later self-reported drug use among treatment participants with methamphetamine dependence. It did not establish a clinically validated scan that can determine whether a particular person will relapse.
Start with the follow-up denominator
Gowin and colleagues scanned 68 people enrolled in treatment during early abstinence. Sixty-three were reached at approximately one year; five could not be tracked. Among the 63 followed participants, 18 reported relapse and 45 reported remaining abstinent from the drugs covered by the follow-up definition.
Thus, the reported 29% relapse fraction uses 18 of 63, not 18 of 68. The five missing follow-ups cannot be assumed abstinent. This was a selected treatment cohort, not an estimate of relapse among everyone using methamphetamine.
“Relapse” was broader than methamphetamine use alone
The follow-up asked about sedatives, hallucinogens, stimulants, marijuana and opiates; any reported use in those categories counted as relapse. The outcome should not be relabeled as exclusively confirmed methamphetamine recurrence.
Follow-up information came from interviews and was not confirmed with urine analyses. Initial screening procedures and a later interview outcome are different sources of evidence. A negative urine screen around scanning would not validate a year of subsequent abstinence.
The task contrasted safe and risky choices
Participants selected among 20-, 40- and 80-point options. The 20-point option was safe, while the larger options could produce gains or losses. Unbeknownst to participants, choosing the same option throughout would yield the same final point total.
The later-relapse and abstinent groups did not significantly differ in the frequency of risky choices, either overall or after losses. Their imaging responses nevertheless differed: the abstinent group showed a greater separation in insular activation between risky and safe decisions.
A null behavioral comparison does not demonstrate identical real-world decision-making. Equally, a regional imaging difference does not prove that one group visibly took more risks on the measured task.
Association and classification were promising but imperfect
The reported right-insula analysis had an area under the receiver-operating-characteristic curve of 0.76 in this cohort. AUC summarizes discrimination across thresholds; it is not a statement that 76% of every future population will be classified correctly.
Using a median split, the paper reported correctly identifying 14 of 18 participants who relapsed and 27 of 45 who remained abstinent. Those counts also show the errors: four later-relapse participants and 18 abstinent participants fell on the other side of the classification. The counts are clearer than treating the scan as a definitive prediction.
A Cox coefficient is not a person’s absolute probability
The time-to-event model reported an exponentiated coefficient of approximately 0.34 per standard-deviation increase in the risky-minus-safe insula signal. Although the paper used odds-ratio terminology in places, the analysis was a Cox proportional-hazards model. Its relative time-to-event association should not be treated as a 34% absolute probability of relapse.
The imaging feature and prediction analyses came from the same cohort. The paper did not establish external performance in an independent clinical population. Selection of imaging clusters, limited event counts and model complexity therefore matter when assessing possible optimism in the reported discrimination.
What the study does not justify
Co-occurring substance histories, a predominantly male sample, missing follow-ups and self-reported outcomes limit generalization. The findings do not show that an insular difference caused relapse, that treatment based on the scan prevents relapse, or that the result identifies intoxication or driving impairment.
The study offers a historical research signal worthy of further testing. It supplies no basis for treating an individual scan as proof of future conduct or substituting it for a complete clinical and behavioral assessment.
Primary scientific source
Gowin JL, Harlé KM, Stewart JL, Wittmann M, Tapert SF, Paulus MP. Attenuated Insular Processing During Risk Predicts Relapse in Early Abstinent Methamphetamine-Dependent Individuals. Neuropsychopharmacology. 2014;39:1379–1387. doi:10.1038/npp.2013.333. Original commentary on the supplied historical prospective follow-up study.
Leave a Reply