Did Working-Memory Training Change Basal Ganglia Volume During Methamphetamine Treatment? A 66-Man MRI Cohort

Short answer: The 2016 MRI study found statistically significant changes in brain-region volume estimates over four weeks of inpatient methamphetamine treatment, including basal-ganglia clusters in men receiving additional working-memory training. Its corrected group-by-time analysis identified putamen and cerebellar regions. But the authors’ direct follow-up voxel-based comparison between training plus usual care and usual care alone reported no significant volume difference in either direction. These are related, not interchangeable, findings. None measures methamphetamine concentration in a forensic venous whole-blood specimen or establishes a particular person’s drug effects, driving performance, or causal responsibility.

The sole primary empirical paper is Brooks, Burch, Maiorana, Cocolas, Schiöth, Nilsson, Kamaloodien and Stein (2016), Psychological intervention with working memory training increases basal ganglia volume: A VBM study of inpatient treatment for methamphetamine use, NeuroImage: Clinical 12:478–491, DOI 10.1016/j.nicl.2016.08.019. The analysis below separates recruitment, completed questionnaires, structural imaging contrasts, and drug testing, rather than treating every recorded observation as a new participant.

What exactly did the primary study investigate?

The investigators asked whether four weeks of residential treatment for methamphetamine dependence, with or without an additional computerized working-memory exercise, coincided with regional changes in structural brain MRI. Their emphasis was voxel-based morphometry, or VBM: a method that estimates regional grey-matter-volume-related signal after image segmentation, registration, smoothing, and statistical comparison. This is an anatomical inference obtained from image processing, not a direct histologic count of neurons, dopamine terminals, drug molecules, or synaptic connections.

A secondary set of observations described questionnaire changes and practice accuracy. These measures can contribute context, but they should not be folded into a single claim that the intervention caused superior clinical recovery. The imaging contrasts answer questions about group patterns at two visits; the questionnaires capture reported mood, impulsivity and regulation; and the screening urine result concerns eligibility during abstinence. They do not answer the same forensic question.

Who was recruited and what is the true independent sample size?

The authors recruited 66 different men: 41 men with methamphetamine as their primary substance problem, admitted to inpatient treatment in Cape Town, and 25 healthy comparison volunteers. The 41 patients were initially divided into 17 receiving treatment as usual and 24 receiving usual care plus cognitive training. The control volunteers contributed comparison information but did not enter either four-week rehabilitation arm. Therefore 66 is the overall recruitment count, not the number who completed four weeks of training or provided usable paired MRI data.

Attrition, inadequate image quality, failure of scanner-log collection and other exclusions reduced the denominators in particular analyses. The paper’s printed psychological tables use 13 paired usual-care observations and 15 paired training observations for specified repeated-measures comparisons, rather than all 41 patients. Other baseline and follow-up questionnaire columns have their own Ns. The precise complete-case count must be tied to the individual analysis; it is not defensible to multiply the enrolled patients by two scanning occasions or by thousands of brain voxels and call that a larger human sample.

Was treatment assignment randomized?

No conventional random-number sequence is described. The Methods state that a researcher alternated patient allocation in the order of admission to the inpatient clinic. This may produce two broadly comparable cohorts, but admission order can be influenced by calendar, referral patterns, treatment conditions or characteristics correlated with outcomes. Neither sequence concealment nor equal exposure to extra staff attention follows automatically from alternation.

The study therefore offers informative prospective, longitudinal structural observations with a nonideal allocation procedure, not the same causal assurance as a sufficiently powered, independently randomized, blinded intervention trial. Participants necessarily knew whether they were performing additional n-back exercises. Differences in adherence, expectations, engagement and other clinical services could contribute to the observed time course.

What constituted treatment as usual?

Both clinical arms stayed in a residential substance-use treatment program. Established care included psychological and group rehabilitation approaches rather than an absence of therapy. The additional training was therefore tested as an adjunct to active inpatient care, while the comparison group received that active background care alone. Patients were recruited after admission and assessed after an initial abstinence period, so much of the follow-up also overlapped ordinary early recovery.

Why this matters: brain-structure estimates could change with the passage of time, sleep, nutrition, medical status, substance abstinence, practice at cognitive testing, or the program’s usual behavioral support. A within-person MRI difference does not, by itself, attribute that change specifically to the new computer exercise. Comparing the adjusted training and usual-care trajectories is analytically more relevant than simply showing that the training group changed from its own baseline.

How was the working-memory exercise delivered?

The additional intervention used a computerized adaptive n-back letter task. A participant indicated whether a letter matched one presented a specified number of steps earlier. Increasing from zero-back to one-, two- and three-back required holding and updating information across more items. The original protocol targeted approximately 30 minutes per day on weekdays across four weeks, with an approximately 80% accuracy criterion before advancing difficulty.

Numbers of letters, trials, days practiced and task repetitions describe measurement density and dose of practice, not independent participants. Accuracy on an exercise practiced repeatedly is a near-transfer endpoint; generalized executive improvement would require independent and appropriately controlled evidence. The study also measured self-report questionnaires and Trail Making rather than deriving any chemical concentration from these tasks.

What specimens were used to determine methamphetamine abstinence?

The investigators described clinical confirmation that the patients had already been abstinent for about two weeks when entering this protocol, using urine sampling at the residential clinic. This was not controlled methamphetamine dosing, a parent-drug pharmacokinetic experiment, or a fully documented forensic quantitation method. The publication does not report a numerical postdose concentration of methamphetamine, its amphetamine metabolite, or a d/l-isomer ratio in a legally collected human venous whole-blood specimen.

No validated concentration–effect relationship, assay calibration data, instrumental identification criteria, blood detection limit, quantitation limit or individual last-use interval is provided for such a legal sample. Screening urine and legally collected venous whole blood answer different questions. Even a genuine confirmed urine result would not be a measured whole-blood concentration, and a clinical abstinence statement cannot substitute for specimen-specific forensic drug confirmation.

How did the researchers measure brain volume?

The imaging procedure used a 1.5-tesla Siemens Magnetom Allegra scanner and structural T1-weighted images. Voxel-based morphometry combines image preprocessing, registration into a standardized space, tissue classification and statistical tests across many locations. Brain regions named in the source include the putamen, caudate, amygdala-adjacent clusters, cerebellum and selected frontal regions. Montreal Neurological Institute coordinates and voxel cluster sizes locate group-analysis signals; they are not toxicology units or direct measurements of cellular growth.

A corrected VBM cluster can support the statement that the modeled regional structural signal differed under specified image-processing assumptions. It cannot independently identify what microscopic process caused the difference: neuronal size, vasculature, hydration, glial contributions, registration behavior, motion and other determinants may matter. The study did not use chemical imaging to identify individual methamphetamine, quantify striatal dopamine or certify a driver’s psychomotor fitness.

What statistical questions should be kept separate?

At least four distinct comparisons appear in the article: differences between groups at baseline; changes between baseline and four-week follow-up within each clinical arm; group-by-time interactions across the two clinical arms; and direct cross-sectional differences between the treatment arms at follow-up. They are not equivalent hypotheses. A significant within-training before–after contrast does not establish that training outperformed usual care. A significant interaction can identify different changes across time even when an isolated follow-up comparison is not significant.

The authors applied a family-wise-error correction to principal VBM findings and expressly label some exploratory frontal associations as uncorrected. Correction matters because whole-brain maps repeatedly test many spatial locations. A source result that survives the stated correction can be discussed as a corrected group-level finding; an isolated uncorrected voxel or a visually striking colored figure cannot be promoted to an individually validated diagnostic marker.

Which group-by-time structural findings were statistically significant?

The corrected two-by-two analysis included a right putamen cluster extending toward the amygdala, a left putamen cluster with adjacent extension, a left cerebellar cluster, and a right middle frontal gyrus cluster. These are not four independent cohorts, and the voxel counts within each cluster are not extra participants. The right and left putamen findings are anatomically relevant to frontostriatal working-memory hypotheses, but the clinical meaning of an MRI segmentation difference remains uncertain.

The primary numerical findings below are newly arranged summaries of the article’s original Table 4b. They retain the authors’ MNI coordinates, Z statistics and cluster-level family-wise-error p values. Their display does not reproduce the publisher’s graphical figures or source table layout.

Did cognitive training show a direct MRI advantage at follow-up?

Not in the reported direct follow-up VBM comparison. The source’s Table 4b states no significant volume difference for training versus usual care at baseline. It also states no significant volume difference for training versus usual care at the four-week follow-up, regardless of which direction the contrast was expressed. This genuine negative finding needs to appear prominently beside the positive interaction and within-group maps.

Some regional trajectories nevertheless differed in the repeated-measures model. The appropriate reading is that one type of statistical test identified variation across time and group, while another did not find an isolated corrected spatial difference between the arms at the final visit. Neither outcome proves exact equivalence; a small sample may lack power for direct cross-group effects. But it also does not permit claiming a definitive corrected between-arm MRI superiority when the authors explicitly report none.

What happened within the usual-care group?

For patients receiving only the existing treatment, a within-arm longitudinal comparison showed a corrected right-putamen cluster at follow-up relative to baseline, with a reported cluster of 966 voxels, Z = 6.26 and family-wise-error p = .04. The group also showed decreases in cerebellar regions over the same interval. These findings matter because they establish that some structural differences appeared even without the extra computerized working-memory task.

A change during usual care could reflect several overlapping experiences, including continued abstinence, routine psychological interventions and time. It does not establish that the same regions would return to a pre-methamphetamine baseline or that any brain condition was restored to normal. Without each participant’s brain scan before stimulant use, an improvement or normalization label remains a hypothesis about the direction and meaning of group structural patterns.

What happened within the added-training group?

Within the group receiving additional n-back practice, corrected follow-up-versus-baseline clusters included a right putamen region extending toward the amygdala/hippocampus (1,023 voxels; Z = 5.69; p = .003), left putamen with extension (3,033 voxels; Z = 4.90; p < .001), and right caudate with extension (1,132 voxels; Z = 4.36; p = .021). Bilateral cerebellar decreases were also described. These are within-arm maps and do not supersede the direct follow-up between-arm null.

The reported anatomical extensions are useful descriptions of spatial VBM clusters, not evidence that researchers measured microscopic hippocampal neurogenesis, exact synapse counts, restored dopamine transporter density or a higher drug-free survival rate. The imaging protocol and clinical outcomes were not designed to measure those proposed mechanisms. Corrections for the imaging comparisons do not remove selection bias or the limitations of small nonblinded clinical groups.

What do the positive, negative and descriptive results show together?

Placing the printed outcomes side by side prevents a one-sided summary. The study contains corrected positive interaction and within-arm imaging findings, yet also a genuine direct follow-up between-arm negative comparison. Several self-report measures changed, but a change in subjective response is not equivalent to external task transfer or confirmed substance-use outcomes. A right-putamen Z statistic is an index of the imaging test and should not be presented as an estimate of treatment effect in cubic centimeters.

The table below uses independent participant denominators and selected source statistics to clarify which comparisons answer which question. It deliberately includes the null rather than pretending every analysis favored the additional intervention.

Did the questionnaire findings independently establish reduced impulsivity?

The investigators obtained mood, perceived self-control, Barratt impulsivity, self-regulation and Hospital Anxiety and Depression Scale responses. In the printed Table 3 training subgroup, some changes were reported for working-memory task accuracy and specific self-report outcomes. However, other measures, including the objective Trail Making response-time comparison, did not demonstrate a significant improvement in the same within-arm analysis. Interpretation also depends on the questionnaire denominator, correction rule and the precise contrast.

A participant who reports better control at the end of a residential stay may indeed feel better. That observation should not be recast as a validated demonstration of safer driving, truthfulness, legal capacity or persistent abstinence. There were no individually linked whole-blood methamphetamine measurements with which to build a forensic concentration–response relation. Nor was there a separate follow-up design that established a durable individual reduction in actual drug-related harm from these imaging changes.

Were working-memory scores correlated with the measured brain changes?

The authors investigated relationships between working-memory accuracy, questionnaires and regional brain estimates. The Discussion states that significant correlations between questionnaire measures of impulsivity or self-regulation and changes in brain volume were not observed. Some exploratory relations involving right dorsolateral prefrontal cortex and working-memory accuracy appeared only without family-wise-error correction. They are hypothesis-generating, not additional confirmed regional changes.

This is important when interpreting an attractive mechanism. Even if the training group shows an adjusted cluster difference and reports more self-control, the data do not establish a validated chain from increased neuronal volume to improved inhibition to reduced consumption. Correlation, time association, anatomical localization and experimentally identified mediation require distinct analyses. The original results do not fill every link in that chain.

What was different about the healthy comparison group?

The 25 healthy volunteers were not a separate randomized training control. They provided baseline clinical and imaging comparisons to men admitted for methamphetamine treatment. The original demographics reported no statistically significant mean-age difference between healthy controls and baseline patient group; however, education differed, and the clinical and recruitment environments differed fundamentally. Those differences can affect cognitive scores, physical health and even aspects of neuroimaging.

Consequently, the healthy group should not be treated as a perfect untreated version of the same patients or as a population-level diagnostic reference for methamphetamine users. A group-average VBM finding cannot classify a random new scan as the product of drug use. Normative variability, comorbid exposure, motion, equipment and scanner-processing methods matter, as does the possibility that an apparent difference predates substance-use history.

How should corrected MRI findings be understood?

Family-wise-error correction reduces the probability that a large family of spatial tests produces a spurious significant cluster under the specified model. It does not certify that the putamen literally grew by a known number of neurons or that the treatment produced the effect. Nor does a statistically significant cluster prove that all individuals moved in the same direction. Cluster coordinates express group-model locations, not individualized diagnoses or measurements of chemical toxicant burden.

Here, the authors also conducted multiple analyses involving healthy controls, baseline patients, two clinical subgroups, repeated occasions and working-memory accuracy. Some reported exploratory effects were not family-wise-error corrected. Any claims about isolated frontal findings should preserve that qualification. A well-controlled replication with prospectively specified imaging contrasts and a better allocation process would be necessary before translating these anatomical patterns into strong clinical advice.

Can this MRI study tell whether methamphetamine improved a patient’s brain?

No. Methamphetamine exposure was historical and nonrandom, and patients entered an inpatient recovery program rather than receiving a trial methamphetamine dose. The structural comparisons concern the period after patients were no longer actively using the drug under clinic monitoring. Observed changes therefore cannot support an assertion that methamphetamine was beneficial, that it caused the imaging change or that a person’s neural function was restored. Clinical recovery is a separate question requiring independently validated outcomes.

The distinction is especially important in forensic testimony. A defendant, injured worker or driver could have a similar MRI appearance for reasons unrelated to the event in dispute. The study does not provide a calibrated dose-response model, a test sensitivity/specificity for intoxication or an individual causal attribution algorithm. Applying its group patterns to a person without a case-specific exposure record, alternative hypotheses and clinical evidence would be a substantial extrapolation.

What is the difference between urine eligibility screening and forensic whole blood?

The clinical urine confirmation described during early abstinence establishes the researchers’ intended entry conditions, subject to what the original paper actually reports about its method. Urine can contain drug metabolites for a different period than circulating blood and often is screened qualitatively. Whole venous blood includes plasma and cellular constituents, while separated serum/plasma excludes most cells. A concentration from one compartment cannot be assumed equal to another without an appropriate verified transfer or paired analysis.

No quantitative LC-MS/MS whole-blood parent methamphetamine result, analytical uncertainty, matrix-specific detection/quantitation limit or related contemporaneous impairment assessment was supplied by this MRI paper. A structural MRI signal cannot fill that evidentiary absence. For comparison of biological matrices in forensic cases, see the separately published guide to biological matrices and the published guide to confirmatory drug testing. These contextual pages do not alter the primary study’s measurements.

Does the study establish duration of methamphetamine action or a legal cutoff?

The original publication does not establish how long a particular patient’s methamphetamine concentration remained above a specified forensic threshold. It does not collect serial chemically confirmed postdose blood samples and does not test controlled drug administration at specified oral, smoked or intravenous doses. The study’s four-week treatment interval is a period of clinical intervention, not a blood-drug detection window or the duration of acute methamphetamine psychomotor effects.

A legal cutoff, when applicable, belongs to the relevant jurisdiction and validated testing context; one cannot derive it by dividing voxel-cluster size by time in treatment or by inferring an exposure from symptom improvement. Even a directly observed urine screen cannot establish a legally reliable driving-impairment threshold. Those would require different original studies and case-specific facts.

What are the principal scientific limitations?

The key constraints are modest and unequal group sizes, allocation by alternation rather than concealed independent randomization, attrition or quality exclusions, absence of an attention-matched sham-training group, and overlapping ordinary clinical recovery with extra training. The research reported several post hoc and exploratory imaging comparisons, as well as outcomes with different participant denominators. The participants were all men in a particular inpatient program, limiting extension to women, community samples or people examined shortly after use.

Equally important is the gap between group MRI statistics and patient-level outcomes. There were no legally drawn quantitative methamphetamine whole-blood samples, no validated individual impairment classifier, no observed driving test and no reconstruction of a collision or other litigated event. A corrected right-putamen group interaction does not solve these missing questions. The source itself reports an important absence of direct between-arm volume differences at follow-up.

What could a stronger study test next?

A larger preregistered trial could use genuinely randomized allocation, masked MRI image processing, an attention-matched control activity, documented complete-case imaging flow and prospectively defined region-of-interest versus whole-brain comparisons. Longer follow-up could distinguish short-term adjustment from persisting change and test whether a putative MRI trajectory predicts clinically meaningful outcomes. Researchers might compare objective task performance, independently validated abstinence and functional measures rather than relying on questionnaire improvement alone.

For a forensic exposure question, a separate study would need appropriate sampling of the specified biological matrix, validated chemical identification and quantitation of parent methamphetamine and relevant metabolites, analytical uncertainty and independently obtained behavioral observations. That remains independent of whether cognitive training changed a regional structural-MRI statistic. The right experimental design follows the question rather than forcing a diagnostic conclusion from available MRI images.

Seven frequently asked questions

1. Did all 66 men complete the four-week MRI comparison?

No. Sixty-six were recruited across patients and controls, and the 41 inpatient patients were initially assigned to the two clinical arms. Follow-up and imaging analyses had exclusions and smaller analysis-specific Ns. The printed Table 3 paired questionnaire example is n13 usual care and n15 training. Repeated MRI voxels or sessions cannot expand the independent human sample.

2. Did the cognitive-training group have a significant volume advantage over usual care at the end?

The original direct follow-up VBM comparisons in Table 4b say no significant volume difference in either direction. Corrected time-by-group interactions and within-group cluster differences are distinct inferential tests and do not cancel that reported negative result.

3. Which brain structures showed corrected interactions?

The original Table 4b reports corrected signals involving both putamina with extension toward the amygdala, the left cerebellum and right middle frontal gyrus. A VBM cluster is a statistical regional tissue-signal estimate; it is not proof of the microscopic identity or mechanism of a change.

4. Could the improvement simply reflect normal treatment and abstinence?

It could contribute. Usual-care participants also showed a within-group right-putamen difference, and both arms received residential care. Alternated assignment, different engagement and absence of an attention-matched comparator limit any attempt to isolate the specific effect of extra n-back training.

5. Was methamphetamine quantified in legal venous whole blood?

No. The paper mentions clinical urine-based confirmation of abstinence, but it does not report actual numeric whole-blood parent methamphetamine concentrations, a validated forensic assay, detection limits or a cross-matrix conversion. Imaging data and clinical urine information cannot be substituted for those measurements.

6. Did a larger putamen prove improved driving or fewer drug relapses?

No. There was no tested driving endpoint, no independent proof that structural differences mediated clinical outcomes, and no validated person-level model predicting relapse or legal impairment from brain volume. A positive group VBM interaction is not an objective forensic functional clearance.

7. Can an expert infer intoxication, last drug use or incident causation from this study?

No. Historical substance exposure, a four-week residential course, self-report measures and group MRI estimates do not identify a specific person’s current drug concentration, last-use time, ability to drive or causal role in an event. A case-specific inference would require directly relevant toxicological, behavioral and temporal evidence.

Primary evidence and bottom line

The empirical findings support a carefully delimited statement: four-week inpatient treatment was associated with structural MRI differences, and an added working-memory group participated in corrected group-by-time interactions involving the putamen and cerebellum. The same original source explicitly reported no corrected direct follow-up volume difference between the training and usual-care groups. Results with different tests, sample sizes and correction rules should remain distinct.

In a forensic setting, the paper has no valid blood-drug measurement, drug-impaired-driving threshold, individual intoxication test or person-specific causation conclusion. Its strongest value is illustrating how anatomical imaging and clinical rehabilitation research can be informative without substituting for drug quantitation. Primary original reference: Brooks SJ, Burch KH, Maiorana SA, Cocolas E, Schiöth HB, Nilsson EK, Kamaloodien K, Stein DJ (2016), NeuroImage: Clinical 12:478–491, https://doi.org/10.1016/j.nicl.2016.08.019.