Can Theta-Burst Stimulation Reduce Methamphetamine Cue Craving? What a 57-Man Trial Found—and Why Its Prediction Scores Need Caution

Short answer: In a randomized rehabilitation-center study, 57 men with methamphetamine use disorder completed four weeks of active or sham intermittent theta-burst stimulation (iTBS). The 24 active-treatment completers showed greater improvement than the 33 sham completers in their self-rated cue-induced craving and several self-reported executive-function dimensions. The comparison for craving was statistically significant (p = .045), as were the inhibition and working-memory rating-scale comparisons (p < .001 and p = .042). But other executive-function dimensions did not meet the conventional significance threshold, nine initially randomized active participants were missing from the final comparison, and the study’s reported prediction confidence intervals are internally inconsistent. These results do not establish an individually reliable relapse test, an objective neurocognitive recovery measurement, or a concentration of methamphetamine in legally collected venous whole blood.

The empirical source is Wang and colleagues’ 2021 randomized clinical trial, Predictive Role of Executive Function in the Efficacy of Intermittent Theta Burst Transcranial Magnetic Stimulation Modalities for Treating Methamphetamine Use Disorder, published in Frontiers in Psychiatry, volume 12, article 774192. The complete original study is available through its primary research DOI. Its methods and six numbered tables support a narrower conclusion than the claim that one questionnaire can reliably predict whether a particular patient will recover.

What was the treatment question actually being tested?

The researchers asked whether stimulation of a frontal cortical target might change two related but different outcomes during methamphetamine recovery: participants’ ratings of executive-function difficulties and their ratings of craving after viewing methamphetamine-associated images. A related exploratory question was whether improvement on the executive-function ratings tracked or classified improvement in craving. Magnetic stimulation was the randomized intervention; methamphetamine itself was not experimentally administered. The comparison therefore concerned a non-drug treatment delivered to people with histories of methamphetamine dependence, not the acute pharmacology of a controlled methamphetamine dose.

This distinction matters for interpretation. A reduction in a cue-elicited urge can be meaningful to a treatment program, but it is not direct observation of the patient’s future behavior. A lower score on a questionnaire about memory or inhibition is not identical to a verified improvement on an independently administered objective cognitive task. Nor does a relationship between the two questionnaires establish which outcome caused the other. The paper is most useful when its intervention results, correlation analyses, and exploratory receiver-operating-characteristic calculations are read as three separate layers of evidence.

How many independent participants were studied?

At baseline, the investigators assessed 66 men with methamphetamine use disorder and also recruited 65 male community controls for a separate descriptive comparison. The controls were not randomly assigned to receive active or sham stimulation. The trial randomized the 66 rehabilitation patients equally: 33 to active iTBS and 33 to a sham procedure. Nine patients assigned to active stimulation were transferred to other facilities before completing the course and were omitted from the analyzed outcome dataset. Thus, the four-week active-versus-sham findings involve 57 independent trial completers: 24 active and 33 sham. The additional 65 controls do not make the randomized treatment analysis a 122-person experiment.

All participants were men and the recruited treatment cohort came from a single compulsory rehabilitation center in China, with reported detoxification within three months. The original article states that the excluded active patients and finishers did not have considerable differences in selected recorded baseline features. That observation cannot rule out unmeasured differences arising from transfer or other reasons for attrition. The final groups were also unequal in size despite initially equal allocation. An analysis restricted to observed completers answers a different question from an intention-to-treat estimate for everyone randomized.

What did four weeks of iTBS involve?

Active stimulation targeted the left dorsolateral prefrontal cortex, localized by the F3 position on a scalp cap. Participants were scheduled for five sessions a week across four weeks, totaling 20 treatment sessions. Each daily procedure lasted approximately three minutes and supplied 600 pulses in triplets at 50 Hz, with bursts repeated at 5 Hz, two seconds on and eight seconds off. Stimulation intensity was set at 100% of the person’s resting motor threshold. These figures describe magnetic pulses and stimulation timing, not milligrams of a substance or a chemically measured exposure.

The sham group received a similar nominal session schedule but with the coil rotated 90 degrees away from the head. The report describes patient blinding and an assessor who was not told the assignment. However, physical sensation, discomfort, expectations, and treatment context may differ between active and rotated-coil conditions. A statement that participants said they believed stimulation was real does not demonstrate that every sensory or expectation feature was equivalent. The study recorded adverse events throughout treatment and reported no observed seizures or other serious reactions in this relatively small trial; that is a limited observation, not a population safety guarantee.

What was actually measured: executive function, craving, or a drug level?

The principal executive-function instrument was the Chinese-language Behavior Rating Inventory of Executive Function, Adult Version (BRIEF-A). It contains 75 questionnaire items organized into dimensions such as inhibition, shifting, emotional control, self-monitoring, initiation, working memory, planning, task monitoring, and organization of materials. The Behavioral Regulation Index, Metacognition Index, and Global Executive Composite aggregate related parts of that inventory. The scores are behavior ratings, not laboratory response times or assays of neurotransmitter quantity. Higher scores in the source convention correspond to more reported difficulties.

Craving was assessed using a visual analog scale after presentations of methamphetamine-related scenes and paraphernalia. The investigators described approximately 34 drug-associated images, repeated presentations, and a line ranging from no craving to extreme craving. They defined change in craving as the pretreatment rating minus the post-treatment rating, so a positive number generally indicated a reduction in reported urge. Neither an image rating nor BRIEF-A scale score measures whether a participant actually consumed methamphetamine later, whether relapse occurred, or whether a specific incident was caused by a drug. The paper contains no contemporaneous quantitative methamphetamine testing in human venous whole blood or a validated conversion from scale points into ng/mL.

Were the groups comparable on the reported baseline measures?

The two final treatment groups had similar measured baseline values under conventional significance tests, although a nonsignificant comparison does not prove clinical equivalence. For example, the source lists baseline craving means of 2.848 for sham and 3.515 for active treatment, with p = .405. Global Executive Composite ratings averaged 106.393 versus 110.636, respectively, with p = .380. Baseline inhibition scores were both approximately 12.818, while reported prior methamphetamine consumption averaged 0.481 versus 0.454 grams per use episode, with p = .735.

That last quantity is a history variable supplied in the treatment assessment. It is not a dose administered by the investigators, not a laboratory-measured concentration, and not a reliable timestamp of use. Similar measured means at entry do not remove every possible confounder, particularly after the nine active-assignment exclusions. The appropriate question is therefore whether the analyzed groups changed differently over the intervention period, while acknowledging the design’s limits. It would be inaccurate to portray the treatment and sham participants as scientifically proven identical before intervention.

Did the active group report less cue-induced craving than sham?

Yes, within the analyzed completers the reported distributions of change in craving differed. The median change was 1 scale point, with a parenthetical spread of 3, in the active group, compared with 0 and a parenthetical spread of 2.5 in the sham group. The original Mann–Whitney comparison gave Z = −2.112 and p = .045. This is a nominally statistically significant treatment-group contrast in reported cue-induced craving change, not in independently observed drug consumption. It is fairly close to the conventional .05 threshold and comes from a trial examining numerous related outcomes.

The direction of change is plausible within the intervention’s proposed behavioral rationale, but the statistical result should not be enlarged into a claim that every participant benefited or that relapse was prevented. A scale point can have different meanings for different people and contexts. The source did not follow discharged participants long enough to determine whether the four-week rating advantage produced sustained abstinence, fewer real-world drug exposures, improved driving, or better legal outcomes. A temporary difference in urge after pictures cannot replace physical toxicology or longitudinal behavior data.

Which executive-function ratings favored active stimulation?

The researchers calculated a percentage reduction score for each BRIEF-A dimension using a formula dependent on whether the participant’s score increased, decreased, or stayed constant. It is important to treat these as derived rating reduction percentages, not as the percent of neurons restored or percentage improvement on a validated neuropsychological performance task. On inhibition, the median reported reduction was approximately 38.750% in active treatment versus −7.692% with sham; p < .001. Working memory ratings showed medians of 13.393% versus 0%, with p = .042. Global Executive Composite reduction was 30.308% versus 2.393%, with p = .002.

Other nominally favorable comparisons included shifting, emotional control, self-monitoring and the Behavioral Regulation Index. The source’s wording sometimes groups these findings together as executive-function improvement. A closer reading shows that the clinical scales do not all agree statistically, and these reductions remain questionnaire-derived. The design cannot determine whether objectively measured working-memory storage capacity, emergency response speed, complex multitasking, or real-world behavior would improve by the same amount. A claimant or expert should specify the exact named scale, comparison, and denominator rather than converting all executive domains into an undifferentiated improvement claim.

Evidence table 1: four-week craving and executive-rating contrasts

All values are from the source’s original Table 4. The number of independent people is 57, not the number of questionnaire items or repeated presentations. The paper labels the parenthetical quantities as quartile spacing; they are not standard deviations, confidence intervals, or direct point estimates of the treatment effect. Multiple nominal significance tests were performed; the table does not present an independently validated familywise adjustment.

Which reported changes were not statistically significant?

Four of the clearly reported BRIEF-A composite or subscale contrasts in the table above did not reach p < .05: planning/organization (p = .086), task monitoring (p = .052), organization of materials (p = .253), and the Metacognition Index (p = .065). These are substantive outcomes, not missing or irrelevant footnotes. For task monitoring, a numerical median favoring active treatment is insufficient to call the between-group test statistically significant when the printed p is .052.

The distinction is especially important because a reader might assume that a significant Global Executive Composite means every constituent executive skill improved. Composite findings can arise from selected domains without a corresponding significant result on every member. The study also did not apply performance-based tests sufficient to equate the BRIEF-A working-memory subscale with objective memory span, information retention, multitask correctness, or occupational competence. A nonsignificant result does not prove exact equality, but it does bar a confident affirmative claim based solely on that specific test.

Did inhibition or working-memory change correlate with craving change?

Within the 24 active-treatment completers, the authors reported Spearman rank correlations between the amount of BRIEF-A rating reduction and the amount of VAS craving reduction. Inhibition reduction correlated with craving reduction at r = .44, p = .03; working-memory reduction at r = .58, p = .003; and Global Executive Composite reduction at r = .41, p = .04. These correlations describe how two changes varied across a selected, small subgroup. They are not proof that improved executive control mediated the intervention’s biological action or caused abstinence.

The corresponding plotted correlations in the sham condition were near zero or nonsignificant, which makes the treatment subgroup pattern potentially interesting. However, correlations between two self-reported instruments can also share response tendency, expectations, context and unmeasured factors. The source did not demonstrate prospective external validation in different clinics, women, or individuals outside the treatment program. Nor can a group correlation be converted into a certainty about one person’s current state, retrospective intentions, or future risk. A scientifically useful question is whether the relation survives rigorous replication with objective tasks and longitudinal outcomes.

What did the stepwise regression add—and what did it not add?

The exploratory stepwise regression used craving change as the outcome and considered executive rating reductions and patient histories as candidate explanatory variables. For active participants, the printed standardized coefficients were 0.523 for inhibition reduction (p = .009), 0.500 for working-memory reduction (p = .013), and 0.414 for Global Executive Composite reduction (p = .044). Each coefficient describes a fitted association under the model’s selected predictors, rather than the effect of an experimentally manipulated questionnaire score. None is the fraction of participants correctly classified or the probability of an individual’s recovery.

These predictors are conceptually and mathematically related: the Global Executive Composite includes domains such as inhibition and working memory. With only 24 treated completers, stepwise selection across correlated scale variables can be unstable, and conventional significance levels may not capture selection uncertainty. A large-looking coefficient in this small sample does not resolve causal direction or ensure predictive calibration. The experiment randomized stimulation, not separately randomized improvement in inhibition, working memory, or composite score. Those distinctions should remain visible whenever its exploratory regression is discussed.

Evidence table 2: associations, prediction metrics, and negative findings

The association coefficients, standardized regression coefficients and areas under a receiver-operating-characteristic curve answer different questions and are not interchangeable effect sizes. The source’s Table 6 reports confidence-interval bounds for working-memory and global-composite AUC estimates that do not contain their corresponding printed point estimates. The published precision cannot be accepted uncritically; no corrected intervals, validation cohort, or externally verified classification accuracy is supplied in the article.

Can the reported ROC results serve as a clinical prediction rule?

Not on the evidence available. The researchers dichotomized change in cue-craving VAS at a difference of two: changes above two were classified as an effective response and changes of two or less as ineffective. That is a chosen internal response definition, not demonstrated long-term freedom from methamphetamine use. Under this classification, the paper reported a working-memory scale reduction AUC of .874 with 100% sensitivity and 70.6% specificity, and a Global Executive Composite AUC of .761 with 57.1% sensitivity and 94.1% specificity. These statistics come from the same small cohort in which the measurements and cutoffs were studied.

More importantly, the printed 95% confidence intervals for the two AUCs are inconsistent with the corresponding point estimates. An AUC of .874 cannot be inside an interval reported as .220–.631, and .761 cannot be inside .209–.659. No reader should silently substitute a conjectured corrected interval or claim calibrated reliability. The authors’ inhibition classifier was also nonsignificant by its printed test (p = .112), notwithstanding a reported AUC of .710. The sound interpretation is exploratory discrimination requiring source reconciliation and new external validation, not a ready-to-use individual decision threshold.

Was the treatment a measured biochemical restoration of the brain?

No direct chemical measurement of frontal cortical neurotransmitter concentrations, dopamine release, GABA, or other brain metabolites was reported in this experiment. The paper discusses mechanisms by which theta-burst stimulation might alter cortical excitability and executive functions, referring to other published research. Such physiological explanations can motivate a hypothesis, but they cannot be transformed into measurements from these participants. Coil placement, pulse frequency, self-rated task ability and cue-induced craving are different classes of data from biochemical assays or tissue studies.

In particular, no actual methamphetamine or amphetamine metabolite concentration was quantified in a legally collected human venous whole-blood sample. The source did not report chemically assayed separated plasma, serum, oral fluid, hair, sweat, or vitreous concentrations from treatment outcome visits either. Historical drug-use amounts and a detoxification history are not quantitative forensic testing. Readers who need to evaluate laboratory evidence should distinguish different biological matrices and what they measure. A self-report cognitive change cannot be converted into an individualized drug concentration, time of use, or validated impairment cutoff.

What do the attrition, scale design, and follow-up limitations mean?

The central concern is that nine people assigned to active stimulation—more than one quarter of that arm—did not complete the outcome assessment, while all 33 sham-assigned participants appear in the final arm comparison. The source attributes their departure to transfer between facilities for criminal reasons, not documented lack of efficacy, but treatment assignment and institutional circumstances could still affect observed samples. Excluding them rather than analyzing an appropriate randomized intention-to-treat population makes the treatment estimate more vulnerable to selection bias and limits its generalizability.

The trial was restricted to male rehabilitation residents, relied primarily on one executive rating inventory, and lacked long-term relapse follow-up. The sham procedure could feel different from active treatment. The authors evaluated multiple BRIEF-A subscales, composites, correlations and classifiers, increasing the importance of multiplicity and prespecified hypotheses. Their source limits explicitly include the single executive-function scale, blinding difficulties, and short predictive horizon. The plausible result is a hypothesis that selected self-rated executive dimensions may move with cue-craving relief during structured care. It is not a general claim about permanently reversed neurotoxicity or a forensic method for identifying mental state.

Does this trial establish who is safe to drive, reliable as a witness, or likely to relapse?

No. There was no instrumented on-road driving evaluation, crash investigation, stop-signal driving simulator outcome, standardized legal impairment classification, witness-memory examination, or person-specific event reconstruction. The paper did not follow an individual into a relevant collision or litigation event and did not measure when methamphetamine was most recently consumed with a chemical assay. Neither a favorable BRIEF-A reduction nor an unfavorable cue-craving response establishes what a named person perceived, intended, remembered, or caused at some separate time.

Even an accurately quantified drug result would have to be interpreted with its matrix, validated method, sampling time and case context. The requirements of confirmatory drug testing are therefore distinct from this study’s scale-based behavioral intervention. Its figures do not provide a validated relationship between methamphetamine ng/mL and a questionnaire score. Treatment research may be valuable without being an evidentiary shortcut for individual criminal or civil causation.

How should the findings be summarized in forensic-science writing?

The strongest narrow summary is that among 57 analyzed men in one rehabilitation-center randomized study, four weeks of active left-DLPFC iTBS produced greater reductions than sham in picture-cued craving and selected BRIEF-A executive-function ratings. Several other subscales and composites remained nonsignificant, and correlations and exploratory ROC analyses among only 24 active completers did not validate an individual prediction algorithm. The source’s inconsistent confidence-interval reporting adds a particular reason not to assert precise predictive accuracy. It is also important to keep the 65 baseline healthy controls separate from the 57 randomized-treatment completers.

For scientific reporting, name the instrument, endpoint, biological denominator, comparison group, and statistical test each time. Do not call the 600 magnetic pulses a drug dose, the working-memory subscale an independently confirmed neural capacity test, or the rating of craving a verified relapse event. Report null outcomes alongside positive ones. The original empirical paper is accessible at its published primary-study record, which should be consulted for its complete sample-flow description, original Tables 1–6, Figure 2 correlations, and disclosure statements. This measured precision preserves the useful experimental observations without claiming a legal or clinical conclusion the investigators did not test.

Frequently asked questions

1. How many patients were analyzed for the treatment comparison?

Fifty-seven men completed the four-week comparison: 24 in active iTBS and 33 in sham. Sixty-six men were originally randomized, and nine assigned to active treatment were not included in the reported final analysis. The additional 65 community controls were used for a separate baseline comparison and did not undergo random treatment assignment.

2. Did iTBS significantly reduce methamphetamine cue craving?

Within the study’s final completers, the active-versus-sham comparison for change in picture-elicited VAS craving was statistically significant at p = .045. That conclusion concerns a self-rated response to cues over four weeks; the researchers did not demonstrate lower subsequent drug-use frequency or relapse incidence in the same analysis.

3. Was every executive-function domain significantly better?

No. Although inhibition, working-memory rating reduction and the Global Executive Composite favored active treatment, planning/organization p = .086, task monitoring p = .052, organization of materials p = .253, and Metacognition Index p = .065 were not significant under the source’s p < .05 rule.

4. Does a working-memory BRIEF-A score prove objective memory capacity improved?

No. BRIEF-A is a behavioral rating inventory. The source did not administer an independent battery establishing improved capacity for objective span tasks, real-world multitasking, driving hazards or detailed witness recollection. A questionnaire domain and a directly performed neuropsychological task are not interchangeable measures.

5. Can the reported AUC of .874 predict one person’s treatment success?

Not reliably on this source alone. It was an internal exploratory ROC result from only 24 active completers, based on a chosen two-point craving-change classification. Its printed confidence interval does not contain its own AUC point estimate, and no external individual-prediction validation was reported.

6. Did researchers measure methamphetamine in venous whole blood?

No quantitative methamphetamine or amphetamine concentration in human forensic venous whole blood was reported as a treatment endpoint. No paired serum, plasma, urine, oral-fluid or hair drug assay established a personal concentration-time relation. Historic consumption reports, magnetic pulses and self-rated craving cannot substitute for direct drug quantitation.

7. Can this article establish an individual’s intoxication, relapse, or incident causation?

No. It studied group changes in treatment-related ratings and exploratory statistical relationships. It did not measure a specific driver’s behavior, a witness’s actual memory, a known injury event, chemical last-use time, prospective relapse, or causal responsibility. Those questions require separate validated evidence appropriate to the particular individual and event.