Can One Session of iTBS Improve Visual Memory in Methamphetamine Recovery? A 17-Man Sham-Controlled Study

Short answer: In a small randomized study of men undergoing residential rehabilitation for methamphetamine use disorder, one session of intermittent theta-burst stimulation (iTBS) over the left dorsolateral prefrontal cortex was associated with a lower visual working-memory orientation error than sham stimulation after adjustment for baseline performance (F(1,14) = 8.398; p = .012). The adjusted between-group comparison after five sessions was weaker and did not meet the conventional .05 threshold (p = .073). A mathematical model estimated that the active group’s memory capacity increased by approximately 1.176 of the four displayed items after five sessions (p = .002), while precision of the retained items did not significantly change. This experiment did not administer methamphetamine, quantify methamphetamine in a biological specimen, measure driving, or establish a person’s legal impairment.

The primary source is Yurong Sun, Huimin Wang and Yixuan Ku, “Intermittent Theta-Burst Stimulation Increases the Working Memory Capacity of Methamphetamine Addicts,” Brain Sciences 12 (2022), article 1212: https://doi.org/10.3390/brainsci12091212. The original nine-page study provides the methods, baseline group comparisons, response-error analyses, reaction-time results and mixture-model estimates addressed here.

1. What did the researchers actually test?

The central question was whether a brief magnetic-stimulation procedure aimed at a frontal cortical region could alter performance on a narrowly specified visual working-memory task in people with a documented history of methamphetamine use disorder. The researchers compared stimulation delivered over the left dorsolateral prefrontal cortex, usually abbreviated DLPFC, with a sham procedure positioned over the vertex. They assessed participants before treatment, after the first exposure and after the fifth exposure within a single week.

This is an intervention study of a computerized memory response, not a laboratory methamphetamine-challenge experiment. Its independent comparison is between two groups receiving different stimulation procedures; its outcome is angular error when participants reproduce the orientation of briefly presented visual stimuli. Broader claims about executive function, successful abstinence, relapse prevention, brain dopamine release or criminal responsibility require measurements the study did not make.

2. How many independent people were included?

Eighteen male participants were recruited from a rehabilitation center in Nanchang. All had histories of regular methamphetamine use and met the authors’ stated DSM-5 substance-use-disorder criteria. One person was excluded from analysis because the working-memory response pattern reached the authors’ chance-level exclusion rule. The resulting analyzed sample comprised 17 independent men: nine in the active iTBS group and eight in the sham group.

That denominator must not be enlarged by counting their repeated sessions, presentations or response trials as new participants. Each individual could contribute responses at baseline and two follow-up testing occasions, but these are nested measurements from the same person. The paper supplies no female participants and no multisite validation cohort. Its findings therefore need replication before extension to women, other treatment settings or a legally evaluated individual.

3. Were the two groups demonstrably identical before stimulation?

No. The authors reported conventional between-group tests that did not detect significant baseline differences in listed characteristics. That is not the same as proving equivalence, particularly with nine versus eight people. For example, the active group’s average reported abstinence duration was 272.444 days (SD 97.815), compared with 373.625 days (SD 249.258) in the sham group, with p = .277. Reported monthly drug consumption history was 13.844 g (SD 12.838) versus 6.725 g (SD 4.742), with p = .103.

Baseline orientation-response errors were 30.960° (SD 5.989) and 30.314° (SD 7.613), respectively (p = .848). Baseline reaction times were 1.813 and 1.672 seconds (p = .564). These descriptive values matter because the investigators subsequently adjusted certain comparisons for baseline performance. None is a contemporaneously quantified methamphetamine concentration.

4. What was the active magnetic-stimulation exposure?

Stimulation targeted the F3 scalp position, used as a practical approximation of left DLPFC placement. The equipment was a Yiruide CCY-IA magnetic-stimulation machine with a figure-of-eight coil. Stimulation intensity was individualized at 80% of the resting motor threshold, defined by observable right-hand finger movement in five of ten attempts when testing the left motor cortex.

Each daily iTBS procedure delivered 600 pulses over approximately three minutes: triplets at 50 Hz repeated every 200 milliseconds, with two-second stimulation trains separated by eight-second pauses. Participants were scheduled for five sessions, one per day. The comparison procedure used the same nominal stimulation schedule with the coil positioned perpendicular to the vertex. That is a procedural sham, not proof that every sensory experience or the adequacy of blinding was perfectly matched.

5. What does the visual memory test measure?

Each trial presented four oriented Gabor patches for one second, following a brief fixation interval. After a one-second blank retention period, the computer identified one of the original four locations and the participant used a mouse to reproduce that patch’s angle. The angular difference between the correct and reported orientations was the primary direct behavioral error, ranging from zero to 90 degrees. A smaller average error signifies better performance on this particular orientation-recall procedure.

The researchers recorded a reaction-time outcome separately. Baseline testing used 60 response trials, while testing after the first and fifth stimulation occasions each used 300 trials. Repeated exposure to the same general task can produce learning, practice or response-familiarity effects in either group. The sham comparison and statistical group-by-session interaction therefore matter more than simply observing improvement over time within the active group.

6. What were the main differences in orientation error?

For angular response error, the linear mixed model found an overall session effect, F(2,30) = 37.377, p < .001, and a group-by-session interaction, F(2,30) = 4.765, p = .016. The overall group main effect was not significant: F(1,15) = 0.677, p = .424. The significant interaction supports the narrower proposition that the pattern of change across testing occasions differed by assigned treatment, not that active participants were uniformly better at every observed moment.

In separate ANCOVA comparisons adjusting for prestimulation performance, the active group had less orientation error than sham after one session, F(1,14) = 8.398, p = .012. After five sessions, the adjusted contrast was F(1,14) = 3.750, p = .073. It would be inaccurate to convert the latter into a conventional statistically significant between-group result or to characterize the study as demonstrating ever-greater superiority on every direct measure.

7. Evidence table: randomized groups and directly recorded task effects

The table reports the investigators’ numerical tests rather than substituting visually estimated chart coordinates for values that were not printed as precise individual observations. Angular error, response time and model-derived capacity should not be merged into a single invented “impairment score.”

8. Did participants respond faster?

Both groups’ reaction times tended to shorten over testing. The original mixed model reported an overall session effect, F(2,30) = 3.739, p = .035. But its overall treatment-group comparison was F(1,15) = 0.694, p = .418, and the group-by-session interaction was F(2,30) = 0.211, p = .881. Neither of those latter tests demonstrated a differential active-stimulation advantage.

This separation is essential. A treatment that changes the accuracy of reproducing a visual angle does not necessarily improve the speed of making that response. Nor does faster mouse performance in a repeated computer task establish faster hazard recognition, braking, emergency decision-making, or any other real-world driving behavior.

9. What is the difference between memory capacity and memory precision?

The paper fitted angular responses using a standard mixture model. The model partitions a distribution into a component concentrated around the true stimulus angle and a more uniform component representing guessing. From the estimated guessing probability, the investigators calculated a parameter K for capacity. Because each display contained four items, the estimated quantity is tied to that four-item stimulus array; it is not a directly counted number of thoughts a person can hold in every situation.

Precision is a different parameter: the spread of accurately centered responses among items modeled as remembered. A narrower error distribution can indicate greater precision even without an increase in the proportion of items estimated to be retained. Conversely, a lower estimated guessing proportion may imply greater modeled capacity while the precision of retained orientations remains unchanged. The paper examined these endpoints separately instead of treating them as interchangeable.

10. What did the model show after one and five sessions?

Relative to the prestimulation baseline, the active group’s estimated capacity increased by about 0.676 items after the first session. The authors reported p = .051 and Cohen’s d = 0.766. This is a near-threshold observation but is not conventionally significant at .05. After five sessions, the active group’s estimated increase was about 1.176 items, with p = .002 and Cohen’s d = 1.503. The sham group’s modeled capacity did not show the corresponding reported change.

Those results are not in conflict with the fifth-session adjusted angular-error contrast of p = .073. The analyses ask different statistical questions using distinct outcomes and comparisons. The significant modeled capacity change is measured against a within-group baseline, whereas the ANCOVA comparison directly contrasts groups after accounting for prestimulation behavior. Both should be reported if a reader is evaluating the strength of the intervention evidence.

11. Evidence table: model-based capacity, precision and study boundaries

The capacity estimates depend on assumptions of the mixture model and its interpretation of guessing. A statistical model applied to behavior is not the same as an objective direct count of stored neural representations, and it does not create a validated diagnostic biomarker by itself.

12. Was stimulation better after five days than after one?

That question has no single yes-or-no answer unless the endpoint is specified. The reported active-group modeled capacity gain was numerically larger after five sessions than after one. However, the investigators did not show a progressively stronger significant adjusted active-versus-sham difference in mean angular error: the first-session between-group result was significant, while the fifth-session between-group result was borderline under a .05 rule. The source discussion expressly notes that the direct overall memory-performance benefit did not enlarge in a simple monotonic way.

It would therefore overstate the evidence to advertise “five days proved better than one day” without qualifying which outcome and statistical contrast support that claim. The source measured performance shortly after the fifth session; it did not supply a prolonged post-treatment follow-up demonstrating lasting retention, sustained abstinence or later relapse prevention.

13. Did stimulation alter dopamine in these participants?

No participant-specific synaptic dopamine measurements are reported. The paper discusses dopamine circuitry as a possible physiological rationale for targeting DLPFC, citing prior work. That background mechanism is not a directly assayed result of the 17-person experiment. Magnetic-stimulation pulse schedules, cortical targets and computational memory parameters must not be relabeled as chemical quantitation.

The experiment also did not draw venous samples to measure parent methamphetamine, amphetamine metabolites, dopamine, or any other drug concentration. The participants’ self-reported or documented historical consumption and estimated abstinence duration were descriptive intake characteristics. They are not a calibrated concentration-time profile, validated recent-use classification or exposure threshold.

14. Did the study establish a reduction in craving or relapse?

No. The authors explicitly acknowledged that they had not obtained a valid measure of participants’ methamphetamine craving suitable for relating craving change to working-memory change. The experiment likewise did not demonstrate a lower rate of subsequent methamphetamine reuse, fewer arrests, greater treatment completion, or better real-world occupational performance as measured trial outcomes.

The theoretical idea that better working memory might support self-regulation is worth investigating, but it remains a hypothesis here. A task-specific improvement after magnetic stimulation should not be treated as proof of clinical recovery or future behavioral control. Those outcomes require a different design, suitable follow-up, prespecified endpoints, and independent replication.

15. Could the outcome classify an individual as impaired or unimpaired?

No. There is no standardized individual cut point linking this orientation-reproduction task to legal intoxication, driving fitness, capacity to consent, credibility of recollection or responsibility for an event. The study involved rehabilitation participants tested on a brief visual memory procedure after magnetic stimulation. It did not test an accused person, a driver, an eyewitness at a relevant incident time, or any clinically validated individual diagnostic threshold.

The outcome also has no demonstrated one-to-one relationship with biological evidence. A specimen-specific toxicology interpretation distinguishes what was actually collected and measured from what a behavioral task might suggest. In this experiment, quantitative whole blood, separated plasma or serum, oral fluid, urine and hair methamphetamine measurements were not reported as trial endpoints. Without those assays, the study offers no laboratory basis to back-calculate a person’s drug level.

16. What are the most consequential study limitations?

The analyzed sample was extremely small and included men only, from one residential treatment setting. One randomized participant’s task results were excluded under an accuracy rule. Prestimulation and later test occasions had different numbers of trials; both groups had repeated opportunities to become familiar with the task. The sham procedure used a perpendicular coil at the vertex, and the paper does not establish complete matching of sensory experience or assess the success of masking with a reported blinding test.

The study reported multiple related outcomes and comparisons without establishing a large confirmatory replication. The after-one-session modeled capacity result was p = .051, the after-five-session adjusted between-group error result was p = .073, and the direct reaction-time interaction was null. These qualifications are as important as the significant orientation-error interaction and later model-derived capacity gain. No long-term follow-up or quantitative methamphetamine pharmacokinetics were included.

17. How should this research be used in forensic toxicology?

The defensible use is limited but instructive: cognitive-performance endpoints in people with past substance-use disorders can change after a non-drug intervention, and statistical conclusions depend on precisely defined outcomes and comparison groups. Orientation error, model-estimated capacity, response latency, clinical history and chemical drug concentration each answer different questions. A result in one category cannot silently substitute for proof in another.

When an actual case involves a drug-test result, the specific confirmation method and original laboratory record are indispensable. When the question is therapeutic monitoring, drug timing and a quantitative measured biological matrix must be identified. This study supplies no validated bridge from its visual response task to either of those forensic inquiries.

18. Frequently asked questions

1. Does one iTBS treatment help visual memory in methamphetamine rehabilitation?

The small trial reported a statistically significant reduction in adjusted orientation-recall error after one active session compared with sham (p = .012). That is preliminary evidence for this task in the analyzed 17 men, not established clinical efficacy in all rehabilitation patients.

2. Did five treatments prove a greater overall benefit than one?

Not on every measure. Modeled capacity increased more after five sessions within the active group, but the adjusted active-versus-sham orientation-error comparison after five sessions had p = .073. The two analyses are not interchangeable.

3. Was memory capacity measured directly?

The investigators directly collected mouse responses to visual orientations. Capacity K was subsequently estimated from a fitted distribution of response errors and guessing, using a four-item task; it was not directly observed as a neural item count.

4. Did stimulation make participants respond faster than sham?

The response-time analysis showed an overall session effect, but not a significant group effect or group-by-session interaction; the latter had p = .881. A specific speed advantage is not established.

5. Did the investigators measure blood methamphetamine or a drug cutoff?

No. They did not report calibrated methamphetamine concentrations in forensic venous whole blood, separated plasma, serum, urine or oral fluid. There was no dose-to-concentration conversion or impairment cutoff validated by this experiment.

6. Does the result prove less craving, relapse or unsafe driving?

No. Craving change was not validly measured, relapse was not a demonstrated follow-up endpoint, and no driving task, crash, incident-specific causation or individualized legally relevant performance threshold was tested.

7. What is the key caution when citing the study in court?

Give the correct sample denominator of 17, identify the narrowly defined visual-memory task, report both significant and nonsignificant comparisons, distinguish model-estimated capacity from direct error and reaction time, and avoid claiming that the article establishes a particular person’s methamphetamine level, impairment, last-use time or incident cause.

Primary study and interpretive boundary

Sun Y, Wang H, Ku Y. Intermittent Theta-Burst Stimulation Increases the Working Memory Capacity of Methamphetamine Addicts. Brain Sciences. 2022;12(9):1212. Primary original full study: doi:10.3390/brainsci12091212. The findings should be interpreted within its 17-man analyzed sample, short stimulation schedule, orientation-recall task, and absence of quantitative drug testing or individual legal-impairment validation.