Can Neurofeedback Plus Swimming Reduce Methamphetamine Cue Craving? A 32-Man EEG Trial

Short answer: In a small randomized study of 32 men in a methamphetamine treatment camp, ten weeks of combined neurofeedback and supervised swimming was associated with lower picture-elicited, self-reported craving than neurofeedback alone, swimming alone, or the comparison condition. The combined group also differed on resting scalp-electroencephalogram (EEG) power measures, including the delta-to-alpha ratio. These are promising group-level rehabilitation findings, not proof of reduced methamphetamine exposure, prevention of relapse, or an individual’s fitness to drive. The investigators did not quantify methamphetamine or its metabolites in venous whole blood, plasma, serum, urine, or any other specimen as a reported study outcome.

The source is Fadaei, Najafi, Miladi-Gorji, Tajik-Mansoury, and Afkar (2025), Effect of Neurofeedback Training and Swimming Exercise on the Electroencephalographic Changes and Craving in Methamphetamine-dependent Patients, Basic and Clinical Neuroscience 16 (special issue), 179–192, original primary research doi:10.32598/bcn.2023.886.4. The evidence concerns a specific inpatient treatment population, four small randomized groups, before-and-after EEG recording, and a laboratory picture-response task. It should not be conflated with studies that measure drug concentration or objective real-world relapse.

What was actually tested?

Investigators examined whether neurofeedback (NFB), swimming, or their combination was accompanied by differences in EEG spectral power and responses to pictures associated with methamphetamine use. All participants were adult men with a history of methamphetamine dependence and were residing in a treatment camp in Taybad, Iran. The study was conducted between 2020 and 2021 and was reported in 2025. The article describes randomization to four treatment conditions, preintervention testing, ten weeks of the assigned regimen, and postintervention testing.

The comparison was not a controlled administration of methamphetamine. There was no random allocation to active drug exposure, no measured acute dose, and no documented postdose blood concentration. The experimental interventions were behavioral training and exercise during a period described by the authors as abstinence or treatment. This distinction matters because even a convincing EEG or craving change cannot independently establish what drug was circulating in a particular person’s blood at a particular time.

How many independent people were studied?

The researchers initially screened 114 individuals. Thirty-one declined to participate, and 43 did not meet inclusion criteria, leaving 40 eligible people randomized into four conditions. Eight participants later did not complete sufficient participation and were excluded from the final analysis. The reported analyzed sample was 32 independent men, eight per group. The article does not supply a treatment-arm-specific attrition breakdown sufficient to establish why each exclusion occurred or whether withdrawal was associated with prognosis.

Each participant contributed multiple EEG signals, time epochs, electrodes, images, and pre/post measurements. These observations provide information about an individual but are not independent human sample members. An analysis that presents thousands of EEG time points as if thousands of patients had completed the trial would seriously exaggerate the strength of the evidence. At eight men per treatment arm, even large-looking average differences require careful attention to uncertainty, selection, baseline values, and protocol balance.

What were the four treatment conditions?

The control condition involved the camp’s usual arrangements without the study’s active neurofeedback or swimming protocol. The neurofeedback group attended three 30-minute sessions per week, using EEG-contingent visual and auditory feedback. The swimming group undertook three supervised 60-minute sessions per week, with warm-up, swimming, and cool-down. The combination group underwent both training modalities on alternating days. The planned duration was ten weeks for all groups.

These are not equal-duration intervention arms. The combination condition could entail approximately six scheduled training days per week, compared with three in either single-modality group. More contact, activity, structure, novelty, or social interaction might contribute to observed differences. Consequently the results do not isolate a pharmacological effect, and they do not establish that simultaneous neurofeedback and swimming create a unique biological synergy beyond the additional intervention exposure.

Trial design and the actual measurements

What did the EEG electrodes measure?

Recordings were obtained from a 19-electrode scalp montage. The investigators analyzed resting eyes-closed EEG after filtering and artifact screening, calculated spectral power using Fourier-based methods, and summarized delta, theta, alpha, and beta frequency bands. For the offline EEG analysis, the original methods specify delta approximately 0.5–4 Hz, theta 4–8 Hz, alpha 8–12 Hz, and beta 12–30 Hz. Absolute power has physical voltage-squared units; relative power describes a frequency band’s proportion of total power. A ratio of delta to alpha power is a derived mathematical measure.

Electrical activity recorded at the scalp is a physiological signal arising from distributed neuronal populations and recording conditions. It does not identify methamphetamine, its metabolites, or the time of last use. Nor can changing a scalp EEG band be directly equated to altered dopamine concentration, cellular regeneration, reversal of neurotoxicity, or clinical recovery. Such mechanistic and functional conclusions would require independent experiments and validated clinical endpoints that this protocol did not provide.

How did neurofeedback work?

The neurofeedback intervention used a commercial ProComp system. The reported central-region training attempted to reinforce sensorimotor rhythm while discouraging selected other frequency bands; a parietal alpha/theta component used auditory feedback. In the visual component, an image changed when participants met EEG-defined signal thresholds. Participants were encouraged to find mental strategies that improved their reward feedback. In the auditory component, musical tones provided contingent reinforcement as participants kept their eyes closed.

The apparatus generated feedback from brain-electrical measurements. It did not administer, metabolize, or assay methamphetamine. Moreover, feedback-training bands and offline analysis bands were defined somewhat differently in the original Methods. Translating the intervention into a standardized neurophysiological dose therefore requires more information than simply naming a frequency. The original investigation did not include a sham-neurofeedback comparator able to separate real EEG contingency from expectancy, attention, therapist contact, and practicing a mental task.

How was the swimming program implemented?

The swimming regimen took place in a supervised pool at a reported water temperature around 32°C. Sessions comprised about ten minutes of warming up, forty minutes of swimming, and ten minutes of cooling down. The authors described the target as moderate intensity based on an age-predicted maximum-heart-rate method. The intervention was repeated three times each week for ten weeks. The trial did not report continuous individual swimming metabolic costs or objectively measured differences in delivered exercise workload for every participant.

That limitation matters when assessing whether the combination condition truly received the same swimming intensity as the swim-only arm. Exercise behavior can also change sleep, daily structure, mood, social contact, and the participant’s reaction to drug-related cues. Those pathways may be scientifically plausible but were not individually isolated as mediators of the EEG or craving findings. A treatment component associated with improvement is not automatically the sole causal mechanism.

What happened to absolute EEG power?

The researchers reported a significant overall pre/post-by-group effect for absolute power using a multivariate analysis (Wilks’ lambda 0.039; reported F=11.22; p=0.0001). They then reported postintervention group differences in alpha, beta, theta, and delta absolute power. The original across-condition F statistics were 68.04 for alpha, 7.25 for beta, 6.62 for theta, and 5.10 for delta, with reported p values of 0.0001, 0.001, 0.002, and 0.007 respectively.

All three active groups had lower reported postintervention alpha power than controls. The combined condition had lower alpha power than the swimming-only arm in the reported pairwise comparison (p=0.0001). For beta power, the combined condition differed from both neurofeedback alone (p=0.044) and swimming alone (p=0.049). Changes in theta and delta power likewise differed in specified comparisons with controls. These are EEG spectral group contrasts, not drug toxicology results. Nor does lower absolute power, by itself, demonstrate superior cognition, sobriety, or a clinically validated normal EEG.

What about relative alpha, relative delta, and their ratio?

The source distinguished a frequency band’s absolute power from its proportion of the recorded total. Relative alpha power was lower after treatment in the swimming (p=0.002) and combined (p=0.0001) groups than in controls. Relative delta power was higher in swimming (p=0.022) and combination (p=0.029) than in controls. These findings can coexist with a reduction in absolute delta power because the denominator—the total spectral power—can also change. Relative percentage and absolute voltage-squared are different quantities.

The investigators additionally found an overall treatment-condition comparison for the delta/alpha ratio (F=7.61; p=0.001). The combined group had a higher postintervention ratio than controls (p=0.0001) and neurofeedback alone (p=0.024). The ratio is not a blood toxicology marker or a diagnostic cutoff. It is sensitive to both numerator and denominator and to recording and processing decisions. In particular, a larger delta/alpha ratio cannot independently establish the absence of intoxication, reduced craving, improved attention, or lowered relapse probability.

Which EEG and craving comparisons were significant?

What did the craving task actually show?

Investigators displayed methamphetamine-related photographs and neutral material, asking participants to rate the strength of cue-elicited desire. The Methods describe a visual analog rating from 0 to 100 for individual pictures. In the published figure, the total score spans a much larger numerical scale, consistent with aggregation across items, although the exact aggregation is not fully explained. The important reported inferential result is that the combined neurofeedback-plus-swimming arm had lower postintervention total picture-induced craving scores than control, neurofeedback-only, or swimming-only arms, with reported pairwise p=0.0001.

Self-report during a picture task measures a response in that setting. It is not a biological measurement of methamphetamine ingestion, a verified moment of relapse, or a prospective count of drug-free days. The report does not provide a complete externally validated transformation of the aggregate craving score into an individual’s relapse risk. The correct conclusion is narrower: combined intervention participants reported less craving when presented with study pictures at posttest.

Did response times improve as well?

Yes, in specific laboratory comparisons. The investigators reported a treatment-related comparison for reaction time to picture-induced craving (F=4.41; p=0.01). Postintervention responses were faster for the swimming arm than controls (p=0.004) and for the combined arm than controls (p=0.0001). The original graph depicts group averages with error bars, but its reaction-time axis does not clearly identify the unit. A numerical interpretation of the graphical bars as milliseconds would therefore exceed the source description; the reported test probabilities and direction of effect are safer statements.

Faster responses to a static picture are not the same outcome as hazard perception while driving, sustained vigilance, divided attention in traffic, lane position, or braking under time pressure. Nor does a shorter response time necessarily mean fewer mistakes or better decision quality. None of those driving outcomes was evaluated here. The study offers a signal about the picture task, not a transferable legal threshold for adequate psychomotor performance.

Which reported findings were genuinely not significant?

Baseline demographic comparisons included age (p=0.646), years of methamphetamine-use history (p=0.948), BMI (p=0.485), and estimated VO₂ max (p=0.615), none statistically significant by the investigators’ comparisons. The article also states that pretreatment EEG absolute and relative spectral power, the delta/alpha ratio, and picture-induced craving did not differ significantly among the study arms. Specific probabilities for every one of those baseline physiological and task comparisons were not supplied, so numerical p values should not be invented.

A nonsignificant baseline test is not proof that the small arms were identical. Conversely, outcomes that were never measured—methamphetamine concentrations, objectively verified relapse, adverse roadway events, laboratory sensitivity to identify impaired drivers—should not be reported as if they yielded negative statistical tests. The right categories are statistically significant, tested but nonsignificant, and not investigated. Keeping these categories separate prevents a plausible research finding from becoming an unsupported forensic assertion.

Can the combination be credited to a neurofeedback–exercise interaction?

The combined arm outperformed single interventions on the reported picture-elicited craving measure and on some EEG comparisons. That finding justifies future factorial, well-powered studies. It does not establish that neurofeedback plus swimming has a unique interaction effect, because the combined arm appears to have received both sets of sessions. Unequal contact days, coaching, routine, movement, motivation, and expectancy could all influence outcomes. The article reports comparisons among four conditions but does not present a decisive independently replicated mechanistic test of combined treatment’s active ingredient.

A rigorous next study might equate the total time and contact among conditions, use credible sham neurofeedback, measure attendance and exercise intensity prospectively, prespecify a narrow primary endpoint, and follow objective clinical outcomes. Here the proof remains at the level of a small completed trial, with only eight analyzed participants in each arm and no clear relapse follow-up. A promising association is not the same as demonstrated long-term clinical effectiveness.

How much do dropout and sample selection matter?

The study’s pathway from 114 initially screened people to 40 randomized and 32 analyzed people is not a minor bookkeeping detail. The eight postallocation exclusions represent one fifth of eligible randomized participants. If withdrawing participants differed in severity, response, side effects, motivation, or ability to attend training, complete-case estimates might not reflect the effect in everyone offered the intervention. The publication states that exclusions involved low motivation to continue but does not provide the arm-level missing-outcome detail needed to fully evaluate differential attrition.

All analyzed patients were men in a particular residential treatment setting. That design may reduce some uncontrolled environmental exposure but restricts generalization to women, adolescents, people receiving concurrent pharmacotherapy, active drivers, people with serious medical comorbidities, and patients treated in community clinics. The trial also excluded individuals with several clinical histories. No posttrial longitudinal surveillance establishes whether treatment benefits persist after the structured camp schedule ends.

Why do scalp EEG frequency bands not identify a drug?

EEG frequency power is a many-to-one biological observation: sleepiness, arousal, movement artifact, eye closure, medication, age, task, electrode impedance, and disease can influence a recorded spectrum. Methamphetamine use history may also correlate with EEG patterns without making any single EEG feature chemically specific. Chemical identification instead requires a validated analytical method directed at a drug or metabolite in a defined physical specimen, with appropriate controls and interpretation of the reporting range.

For example, an assay of parent methamphetamine in unseparated venous whole blood answers a different question from an assay of a metabolite in urine. Plasma and serum are separated blood components, not interchangeable concentrations with whole blood. Oral fluid, hair, vitreous, and exhaled breath have still different collection and disposition properties. The experiment here collected scalp electrical recordings and self-reports, not those matrices. For general interpretation of such distinctions, see the published biological matrices guide.

Could these EEG patterns prove acute intoxication or the timing of use?

No. The trial supplied neither an individually quantified methamphetamine concentration nor a validated numerical classifier of acute drug influence. Duration of prior use was part of the clinical description, not a laboratory estimate of last use. Resting EEG band measures also lack the required drug-specific toxicokinetic information. A claim that a particular person’s higher or lower alpha power proves they consumed methamphetamine within a specified period would add facts not measured by this research.

The same applies to assertions about dose, route, chronic tolerance, and metabolism. In forensic work, even a chemically confirmed analyte result must be interpreted in light of specimen, collection time, quantitation, pharmacology, and evidence relating to function. EEG can be relevant to a clinical assessment for other reasons, but this small rehabilitation study was not a toxicological identification or quantitation validation study. A summary of what chemical confirmation actually entails appears in the published confirmatory testing guide.

Can the study establish individual driving impairment?

It cannot. The investigators did not administer a standardized driving simulator, field sobriety battery, attention task validated against crash outcomes, roadway hazard test, or individual impairment classification. Neither study arm underwent measured blood sampling alongside a real driving event. A lower average craving score does not determine whether a person made safe decisions at a particular intersection; a changed delta/alpha ratio does not establish reaction time during an alleged collision.

This is especially important when group means are cited in expert testimony. Statistical differences between rehabilitation programs say little by themselves about whether any named individual was drug-affected, truthful, negligent, capable of forming intent, or responsible for an event. Such propositions would require independently admissible case facts, appropriate validation, and an inference chain specific to the alleged issue. A favorable EEG group finding cannot close those evidentiary gaps.

What does the study contribute to forensic-toxicology research?

Its useful contribution is a controlled, albeit small, comparison of behavioral interventions using a common pre/post physiological recording protocol and a specific picture-cue craving outcome. The report provides concrete evidence that the combination arm’s EEG spectral proportions and cue-related self-report responses differed from the comparator arms. The results generate testable questions about whether structured exercise, contingent feedback, and their combination may change responses to drug-associated stimuli during treatment.

The study also illustrates why forensic toxicology must distinguish observed effects from explanations. A trial can demonstrate a statistically detectable difference in a surrogate endpoint without measuring the toxicant. Conversely, a laboratory drug identification cannot by itself establish clinical effect. These two directions of inference demand different validation evidence. The published therapeutic drug monitoring guide offers background on how real chemical concentrations differ from physiological or behavioral observations.

What is the most defensible takeaway?

For the 32 analyzed men in this one ten-week intervention trial, combined neurofeedback and swimming was associated with a stronger reduction in picture-elicited craving than either component alone in the reported comparisons, alongside lower absolute power in selected EEG bands and a higher delta/alpha ratio in certain contrasts. Baseline demographic differences were not statistically demonstrated, although the small sample does not establish equivalence. The combined regimen involved more training exposure than a single modality, and dropout plus limited follow-up constrain interpretation.

Those conclusions should remain tethered to what was physically and statistically evaluated: EEG voltage-derived measures, pictures, button or response timing, and subjective ratings. The trial did not quantify methamphetamine or metabolites, objectively prove no relapse, establish a blood-drug cutoff, or test whether any individual was intoxicated or impaired during a disputed incident. Research that respects those limits is more useful than a broader claim that the source cannot sustain.

Frequently asked questions

1. Did neurofeedback and swimming eliminate methamphetamine craving?

No. The combination group had a lower average picture-induced craving score than the other groups after treatment, with reported pairwise p=0.0001. The outcome was a self-reported response to pictures under study conditions, not proof that all craving disappeared. The source does not provide a validated threshold for clinical remission or an objectively measured future relapse endpoint. Individual responses may differ from the group pattern.

2. Were 32 participants assigned to each arm?

No. The analyzed trial included 32 people in total, allocated as eight men each to control, neurofeedback, swimming, and combined treatment. The report describes 40 eligible people initially randomized and eight later excluded from analysis. Ten weeks of visits, 19 electrodes, and numerous EEG epochs are repeated observations, not additional independently randomized people. Treating those repeated measures as a large participant cohort would inflate precision.

3. Does a higher delta/alpha EEG ratio identify recovery from methamphetamine?

Not by itself. The combined arm’s ratio was higher than controls and neurofeedback alone in the reported posttest contrasts. But the ratio is derived from scalp-recorded EEG frequency power, and many physiological or technical factors can affect it. The study did not validate a delta/alpha cutoff against biological drug clearance, clinically verified recovery, or relapse risk. It is a research endpoint, not a chemical confirmation of abstinence.

4. Was methamphetamine measured in venous whole blood or serum?

No quantified parent-drug or metabolite assay in venous whole blood, separated serum, plasma, or another biological matrix was reported as an outcome of this trial. Its experimental procedures concerned behavioral training, EEG recording, and picture-elicited craving. Residence in a treatment camp and a history of dependence are not substitutes for a matrix-specific validated chemical result at an identified collection time.

5. Could swimming alone change the picture-response task?

Yes. Swimming alone was associated with a shorter response time than the control condition in the report (p=0.004), and the swimming group also had several EEG power differences. The report did not show that swimming alone reduced craving scores as strongly as the combined program in the reported pairwise comparisons. Response speed, subjective craving, and EEG spectral power are distinct endpoints; improvement in one does not automatically establish improvement in another.

6. Was the combined program conclusively better because of biological synergy?

No conclusive synergy was demonstrated. The combination condition received both exercise and feedback sessions, potentially more total intervention days than either single-modality group. A factorial mechanistic trial that balances contact time and includes sham feedback would be needed to test the combination effect more sharply. The current results support follow-up research but do not establish a uniquely additive neurobiological mechanism or generalized superiority across patient populations.

7. Can the findings establish DUI impairment, accident causation, or recent drug use?

No. The experiment lacked quantified forensic drug samples, individual driving assessment, real roadway observations, toxicokinetic timing analysis, and actual collision-event data. A group-level EEG or craving contrast cannot be mapped to a particular person’s impairment, intent, credibility, last use, or accident causation. Those questions require independent, validated evidence addressing the relevant person, specimen, time, task, and event.

Original empirical source: Fadaei A, Najafi M, Miladi-Gorji H, Tajik-Mansoury MA, Afkar M. (2025). Effect of Neurofeedback Training and Swimming Exercise on the Electroencephalographic Changes and Craving in Methamphetamine-dependent Patients. Basic and Clinical Neuroscience 16 (special issue), 179–192. Primary study DOI: 10.32598/bcn.2023.886.4.