Does Theta-Burst Stimulation Change Methamphetamine-Word Attention? A 49-Person EEG Trial

Short answer: A sham-controlled trial of 49 analyzed adults with severe methamphetamine use disorder found that four weeks of active intermittent theta-burst stimulation (iTBS) over the left dorsolateral prefrontal cortex changed the trajectory of errors when participants identified the color of methamphetamine-related words. The drug-word error interaction was F = 4.24, p = .04. Investigators also reported time-by-treatment interactions for selected scalp EEG N1 amplitude and P3 latency differences, plus a frontal beta-power cluster. Yet the key reaction-time contrast was null (interaction p = .98), and several within-group EEG post-hoc tests were nonsignificant. This was an intervention on attention processing—not proof that EEG measures drug concentration, current intoxication, driving impairment, relapse, or an individual’s incident-related behavior.

The source is Tianzhen Chen, Hang Su, Lihui Wang, Xiaotong Li, Qianying Wu, Na Zhong, Jiang Du, Yiran Meng, Chunmei Duan, Congbin Zhang, Wen Shi, Ding Xu, Weidong Song, Min Zhao, and Haifeng Jiang (2021), Modulation of Methamphetamine-Related Attention Bias by Intermittent Theta-Burst Stimulation on Left Dorsolateral Prefrontal Cortex, Frontiers in Cell and Developmental Biology, volume 9, article 667476, original primary DOI. The study is best understood by keeping separate the randomized intervention, the computerized color-naming outcomes, subjective cue-induced craving, and electrophysiological recordings. Each captures a different part of the experiment, and no one outcome can be substituted for another.

What was the precise question in this randomized intervention?

The researchers asked whether repeated stimulation directed at the left dorsolateral prefrontal cortex would modify how adults with severe methamphetamine use disorder process drug-related words, compared with an inactive stimulation procedure. A second question concerned electrical brain responses temporally aligned with those words. A third concerned whether changes in these signals related to participants’ own reports of craving. This matters because an apparent treatment effect might be limited to response accuracy, one brief stage of electrical processing, or one subjective rating instead of an across-the-board restoration of cognitive control.

The design examined an intervention over four weeks in two rehabilitation facilities. It did not randomly expose participants to methamphetamine, evaluate the acute effects of a measured stimulant dose, or compare them with drug-free controls receiving a controlled challenge. Its interpretation is conditional on a pre-existing substance-use-disorder population, background rehabilitation, the specific stimulation protocol, and a constrained word-color laboratory task. Treating it as a general test of methamphetamine intoxication would change the scientific question after the fact.

How many independent people contributed data?

Fifty-seven adults entered randomization: 35 assigned to active stimulation and 22 to sham. Eight later failed prespecified EEG quality requirements, five from active and three from sham. The analyzed sample was therefore 49 independent adults, with 30 active and 19 sham. One subject undergoing multiple treatment sessions, scalp recordings, trials, or craving ratings remains one biological participant. Neither the twenty stimulation visits nor the many word presentations justify inflating the denominator beyond forty-nine.

The initially randomized group sizes and final analytic sizes are both important. An analysis restricted to participants with adequate EEG recordings need not reproduce an intention-to-treat analysis of every randomized person. Exclusions can matter if recording quality is related to participant characteristics or outcomes. The trial did not supply an independent external cohort to demonstrate that its EEG patterns reliably distinguish individuals outside the rehabilitation setting. The observed group averages should not be equated with an individual prediction rule.

What exactly did active and sham participants receive?

The active procedure was intermittent theta-burst stimulation at a scalp target estimated using the Beam F3 method for the left dorsolateral prefrontal cortex. In the methods, the treatment comprised bursts of three 50-Hz pulses delivered every 200 milliseconds, in two-second trains with eight-second pauses. It was administered at 100% of the participant’s resting motor threshold, with 900 magnetic pulses during an approximately five-minute session. Five sessions per week over four weeks yielded a planned total of twenty sessions.

The sham arm used the same nominal schedule with the coil turned away from the skull. Both groups could hear the device, and participants were instructed not to reveal assignment to outcome raters. Nevertheless, similar sounds do not by themselves prove identical scalp sensation or perfect masking. Both groups also received the rehabilitation centers’ standard therapeutic program. There was no experimentally assigned methamphetamine quantity, and the article says no medications were used throughout the stimulation treatment. Magnetic pulses should not be mistaken for milligrams of a drug.

How did the methamphetamine-word Stroop task work?

The computer displayed eight methamphetamine-related and eight neutral words in colored lettering. Each word was presented sixteen times, creating repeated opportunities to respond to four possible text colors. Participants pressed a key associated with the color while instructed not to focus on the word’s meaning. They completed the task before intervention and after four weeks, during 64-channel EEG recording. Each test session generated many observations nested within one participant, not a series of newly randomized subjects.

The investigators calculated reaction time on correct trials and error rate for methamphetamine-related and neutral-word presentations. They also calculated the difference in reaction time between methamphetamine and neutral conditions. It would be a category error to describe an error proportion as a drug concentration, a diagnostic psychosis score, or a quantitative measure of real driving mistakes. A word stimulus has no demonstrated one-to-one relationship with a roadside hazard, an accident, a witness observation, or any specific act alleged in litigation.

Did stimulation change errors on methamphetamine-related words?

For the methamphetamine-word condition, the researchers reported a statistically significant group-by-time interaction: F = 4.24, p = .04. The active group’s average error proportion changed from 0.10 at baseline to 0.05 after four weeks. The sham group’s corresponding estimates changed from 0.05 to 0.07. These are error proportions for repeated computer trials, not estimated percentages of people who committed a real-world error. Importantly, the same model’s overall time effect was F = 0.06, p = .81, and its overall group effect was F = 0.07, p = .79—both genuine null findings.

The interaction means the pattern of change differed by treatment group under the study’s model. It does not establish that all active participants improved, that all sham participants worsened, or that a separate post-treatment classifier could identify assignment or substance-use status. The active group also began with a numerically greater error rate, making the change-from-baseline comparison particularly relevant. Baseline differences, testing familiarity, multiple endpoints, and selective EEG inclusion temper the generality of the result.

What happened with neutral-word errors and reaction time?

Neutral-word error rates did not provide a conventionally significant treatment-specific interaction: F = 2.94, p = .09. The investigators reported a time effect at a boundary value, F = 4.11, p = .05, but the overall group comparison was null, p = .32. A result at the printed threshold should not be described as an unequivocal demonstration of improvement specific to active treatment. It is essential to report the drug-word and neutral-word comparisons together rather than highlighting the favorable one alone.

The difference in reaction time between drug-related and neutral words also failed to show the corresponding treatment interaction. The mean difference went from 11.12 to 3.32 milliseconds in active participants and from 10.01 to 0.74 milliseconds in sham participants, with the group-by-time test printed as F = 0.00, p = .98. Time effect p = .23 and group effect p = .39 were also null. Thus the study’s positive error-rate interaction did not demonstrate a matching treatment-specific advantage in this reaction-time metric. An expert should not collapse error frequency and reaction latency into a single invented impairment construct.

Evidence table 1: observed behavioral results and genuine nulls

The numerical observations above are transcribed as reported in the original study’s Table 2. A significant interaction must not be relabeled as a significant main effect; a nonsignificant result does not prove exact equality. The measured outcomes remain computerized task responses rather than clinically validated observations of driving, intoxication, or criminal conduct.

How was craving measured over the four weeks?

Craving was assessed with a 0-to-100 millimeter visual analog scale after participants viewed methamphetamine-associated paraphernalia for five minutes and recalled prior use. Ratings were obtained at baseline and weekly during the intervention. In the active group, mean reported craving fell from 62.29 at baseline to 14.46 after four weeks. Sham-group means were 50.41 initially and 52.18 at week four, with intermediate fluctuations. The reported time-by-group interaction was F = 29.24, p < .01; the time effect was F = 20.60, p < .01, and the group effect was F = 5.16, p = .03.

These observations support a treatment-associated difference in reported cue-induced craving within this setting, not proof that a biological craving substance was measured. Self-report can change with context, expectation, concurrent rehabilitation and the specific stimulus prompts. The paper did not demonstrate in this analysis that a score of fourteen or fifty predicts actual subsequent methamphetamine use, successful abstinence, dose taken, or the cause of an individual incident. Craving is a meaningful outcome, but it has to be identified as a psychological rating.

What did the scalp EEG actually record?

The researchers recorded electrical potential fluctuations from a 64-electrode cap while participants viewed the words. Data were referenced, filtered, cleaned of ocular and other artifacts, segmented relative to word presentation, and averaged to generate event-related potentials. Separate time-frequency analysis considered oscillatory electrical power, including a beta-frequency band. An electrode voltage in microvolts and a response latency in milliseconds are direct or derived physiological measurements of electrical dynamics, not chemical tests identifying molecules in a specimen.

The analysis contrasted the brain response to methamphetamine-associated words with the response to neutral words. Thus the reported ERP variables are differences between two stimulus conditions, and then comparisons of how those differences evolved in active versus sham groups. A significant contrast depends on task design, word selection, preprocessing, epoch quality and the chosen scalp sites. Nothing in this apparatus physically quantified amphetamine or methamphetamine in blood, saliva, hair or brain tissue.

Which N1 and P3 findings were statistically significant?

For the early N1 component, the group-by-time interaction in the methamphetamine-minus-neutral amplitude contrast was F = 4.43, p = .04. For P3, the analogous interaction in latency was F = 5.67, p = .02. These are the principal interaction findings in the electrophysiology results. They are appropriately described as group differences in trajectories of specified stimulus-locked EEG quantities. They do not imply that N1 or P3 invariably moves in a specified direction for every participant or on every electrode.

Crucially, corrected post-hoc within-group contrasts were not significant for N1 amplitude in active participants (p = .284) or sham participants (p = .29). For P3 latency, the active within-group post-hoc p was .22 and sham p was .17. The investigators also reported no significant effects for N1 latency, P3 amplitude, or the studied N2 and P2 components. A significant omnibus interaction can coexist with nonsignificant follow-up contrasts because they test different hypotheses under different variability and correction assumptions. Omitting these nulls would mischaracterize the strength and scope of the neural claim.

Did beta oscillations change after stimulation?

The study’s time-frequency analysis identified a cluster of reduced 16–30 Hz beta-band electrical power during drug-cue processing after active stimulation, with the largest reported change around scalp electrode AF4. A corresponding significant cluster was not detected in the sham group, and the authors described an active-versus-sham interaction contrast in their second figure. That is evidence about stimulus-linked electrical oscillation power, not a direct assay of neurotransmitter concentration, drug transport, receptor occupancy, or cortical drug distribution.

The article’s figure-based cluster result should not be converted into a fabricated exact p value, a clinical decision cutoff, or a quantitative percentage of individual recovery; the principal text does not provide such a universal numerical threshold. Nor does a beta-band change prove a measured alteration of GABA or dopamine. The discussion refers to those systems as possible mechanisms based on other work. In this trial, the relevant observed data were EEG voltages, computed power and behavioral ratings, not chemical assays of neurotransmitter concentrations.

Evidence table 2: electrophysiology and craving, including the negative results

Each statistic addresses the outcome named in its row. EEG interaction tests, within-group post-hoc tests, correlation coefficients and subjective-scale change are not duplicates of one result; they should not be averaged to invent a single efficacy index. The original article does not establish a validated individual threshold on any of these measurements.

What does the N1–craving correlation establish?

Within the active group, a change in the drug-minus-neutral N1-amplitude difference correlated with change in reported cue-induced craving: r = .48, corrected p = .03. This suggests that participants with more pronounced changes in a specified early electrical response also tended to report particular changes in craving during the intervention. It does not determine which change caused the other, whether a third factor influenced both, or whether a similar correlation would reproduce outside this selected treatment group.

Not all ERP–craving relationships were significant. The active-group P3-latency change correlation was r = −.13, p = .49. In the sham group, N1 change and craving had r = .34, p = .16, while P3 change and craving had r = −.32, p = .18. Those are substantive negative findings that limit any proposed universal EEG proxy for craving. A group association does not supply sensitivity, specificity, calibration error, an individual-level diagnostic cutoff or validated prospective prediction of relapse.

Were the groups identical at baseline?

The source reports no conventional significant differences between analyzed groups in several measured baseline characteristics. Their average age was 29.66 years in active participants and 30.73 in sham participants (p = .51), and their reported mean historical abstinence durations were 2.63 and 3.05 months (p = .32). Baseline craving means differed numerically, 62.29 versus 50.41 (p = .19). These are selected group summaries, not a demonstration that the groups were biologically identical, nor a laboratory measurement of recent exposure.

In a study of 30 versus 19 analyzed participants, nonsignificant baseline p values do not prove absence of meaningful imbalance. Some descriptive entries in the original demographic table also require careful reconciliation with the printed group totals. A rigorous reviewer should preserve source-reported denominators, refrain from correcting numbers by speculation, and avoid treating the historical time-since-use response as a validated toxicology clock. The relevant controlled assignment was magnetic stimulation, not the underlying disorder or the amount of methamphetamine formerly used.

What was not measured in blood or another toxicology specimen?

The investigators did not report quantitative measurement of methamphetamine, amphetamine, or their metabolites in an actual forensic venous whole-blood sample. They also did not present paired plasma or serum concentrations, urine results, oral-fluid concentrations, sweat-patch amounts, hair concentrations, vitreous levels, or exhaled-breath analyte results as endpoints of this intervention. There was no GC–MS or LC–MS/MS concentration result enabling a person-specific ng/mL value, no analytical collection-time validation, and no conversion ratio permitting the measured EEG endpoints to be translated into blood concentrations.

A participant’s historic reported use amount cannot substitute for a chemically measured exposure. A brain electric potential is not an assay for a molecule; a visual craving score is not a pharmacokinetic curve; and a computerized response error is not an observation of intoxicated driving. One cannot infer the legally relevant timing of drug ingestion or individual impairment by chaining these measures together without independently validated steps. The absence of an assay is an unmeasured outcome, not a statistically nonsignificant blood-drug comparison.

What did the study not prove about addiction recovery?

The study measured cue-related performance and physiological responses at baseline and after intervention, with weekly cravings during treatment. It did not demonstrate long-term persistence of the Stroop or EEG effect, independently confirmed reduction in methamphetamine use after discharge, prevention of relapse, reduction in criminal recidivism, fewer crashes, or improved occupational decision-making. The authors explicitly identified the absence of post-treatment Stroop follow-up as a limitation.

Repeated residential-care contacts and psychological treatment were supplied to both groups, and multiple outcomes were assessed in a modest selected sample. The significant craving interaction and drug-word error interaction provide reasons for replication, not permission to assert durable recovery in any individual. Mechanistic discussion about dopamine, GABA, or particular brain circuits must also remain distinct from experimentally measured neurotransmitter endpoints. The EEG method has excellent temporal resolution, but limited anatomical specificity and no molecular identity measurement in this design.

How should forensic experts use—and not use—these findings?

This paper can help explain why a task-specific treatment difference, an electrical potential change, and a person’s subjective craving are scientifically different categories of evidence. In a legal evaluation, the question may concern identity of an analyte, amount present in a properly collected specimen, pharmacological relevance at a given time, functional behavior, or event causation. This study answered none of those case-specific questions through chemical analysis. Its study-level iTBS observations should be kept within the tested rehabilitation population and outcome definitions.

The defensible summary is that active iTBS was associated with a differential trajectory in methamphetamine-word error rates, selected EEG contrasts, beta oscillation power and cravings under the study’s specific protocol, while reaction-time interaction and several component-level or correlation comparisons were null. That is narrower than proving objective current intoxication or a validated EEG addiction biomarker. The source’s licensed publication can be consulted at the original study DOI; its own methods, reported exclusions, and complete tables remain indispensable for expert assessment.

Frequently asked questions

1. Did iTBS reduce mistakes on methamphetamine-related words?

Across the analyzed groups, the time-by-treatment difference in errors on methamphetamine words reached p = .04. Active mean error proportions changed from .10 to .05 and sham means from .05 to .07. This does not prove every active participant improved or that the test predicts real-world mistakes.

2. Did the trial show faster recognition or responses to drug-related words?

No treatment-specific reduction in the methamphetamine-minus-neutral reaction-time difference was established. The interaction p was .98, and the source reported nonsignificant time and group main effects for that contrast. Accuracy and response speed therefore must be evaluated separately.

3. Do significant N1 and P3 interactions prove both components changed significantly within the active group?

No. The N1 amplitude interaction was p = .04 and P3 latency interaction p = .02, but corrected active-group post-hoc tests were p = .284 and p = .22, respectively. Different tests yield different conclusions; all should be reported.

4. Was the beta-band result a blood test or a dopamine assay?

No. It was a computed EEG power contrast in the 16–30 Hz electrical frequency range during word processing. It did not quantify methamphetamine, dopamine, GABA, or a legally collected biological drug specimen.

5. How many people actually contributed to the analysis?

Fifty-seven entered randomization, but eight were excluded under EEG signal quality rules. The published analytic groups were 30 active and 19 sham participants, for forty-nine independently analyzed adults—not hundreds or thousands of independent observations.

6. Does the craving correlation validate EEG as a personal relapse predictor?

No. The active-group N1-change association with craving change was r = .48, corrected p = .03, while P3 and sham-group corresponding correlations were nonsignificant. A selected group association is not a validated individual diagnostic, relapse forecast, or legal threshold.

7. Can this trial determine whether a specific driver had recently used methamphetamine?

No. It measured neither drug administration nor contemporary quantitative methamphetamine in venous whole blood or other biological specimens and contained no real driving incidents. It cannot establish last use, concentration, individualized legal impairment, or causation in a particular case.