EVIDENCE LIBRARY
Does Early Methamphetamine Abstinence Affect Trail Making Performance? A 79-Person Study of Speed, Errors, and Route of Use
A 79-person primary study found slower Trail Making performance in early methamphetamine abstinence, with complex-task errors, real nulls and forensic limits.
Quick answer: In a 2026 observational comparison of 49 residential-treatment participants with methamphetamine use disorder and 30 community controls, the methamphetamine group completed computerized Trail Making Test conditions more slowly, even after the investigators adjusted for several measured demographic and clinical characteristics. It also made more errors on the more demanding alternating-number-and-letter condition, but not on the simpler number-connecting condition. An exploratory association between historical intravenous use and more complex-task errors appeared among the 23 women with methamphetamine use disorder, not among the men. These are group-level cognitive-test findings; the study did not chemically measure a standardized postuse blood drug concentration, determine precise time since last use, diagnose Parkinson’s disease, or validate an individual intoxication or driving-impairment test.
The original empirical report is Dumitrescu and colleagues, Trail Making Test performance in early abstinence from methamphetamine: human evidence for a drug-associated Parkinsonian-like phenotype, Frontiers in Psychiatry (2026), DOI: 10.3389/fpsyt.2026.1773668. The distinction between a measured task result and an unmeasured forensic conclusion matters throughout this analysis.
What is the principal scientific finding?
The investigators compared two existing groups rather than randomly assigning methamphetamine exposure. Their central result is a reproducible-looking difference within this sample and analysis: the treatment cohort required longer adjusted completion times for each Trail Making condition, while error-count differences were concentrated in the cognitively demanding parts of the task. An overall multivariate group-by-condition test was significant, F(4, 446) = 3.666, p = .006, with a reported small partial eta-squared value of .032. The group effect for time, F(1, 224) = 41.988, was reported at p < .0001; the corresponding error-count group test, F(1, 224) = 16.058, was also reported at p < .0001.
Those tests describe an association between treatment-group membership and performance. They do not resolve whether exposure itself, underlying vulnerabilities, education, mental-health histories, withdrawal stage, different test-taking experience, or combinations of those factors produced the observed group separation. The investigators statistically adjusted for some measured differences, but adjustment is not the same thing as randomization.
Who were the 79 independently studied people?
The clinical group comprised 49 independently enrolled adults with DSM-5-consistent methamphetamine use disorder, including 23 women and 26 men. They were recruited from two residential treatment programs in northern Louisiana and had resided in treatment for at least seven days. The comparison group consisted of 30 independently recruited community controls, including 14 women and 16 men. The study therefore involved 79 people, not 79 multiplied by three Trail Making conditions, not a separate cohort for each reported comparison, and not a random sample of all people who have ever used methamphetamine.
The design matters. Participants in treatment were already in early abstinence, and the researchers did not administer methamphetamine to them. The control group was recruited separately. The researchers used interview and clinical information to characterize histories and psychiatric conditions; they did not establish identical educational, socioeconomic, sleep, nicotine-use, or psychiatric profiles between groups. The original demographic table reports a statistically significant difference in educational attainment and higher depression, anxiety, and stress scores in the methamphetamine group. Those differences are possible sources of bias and are part of the interpretation, rather than incidental background.
What does early abstinence mean in this experiment?
The phrase early abstinence refers to participants living in a monitored residential setting, not to a laboratory-verified fixed interval after a known last dose. Admission histories included self-reported use information, but abstinence duration was not recorded consistently enough for a reliable individual time-since-use variable. Some admission testing existed at one site and testing could occur when recent use was suspected, yet the study did not provide a systematic toxicology dataset suitable for its cognitive models. Thus the findings cannot be arranged into a validated number of days after the last methamphetamine ingestion, and the exact abstinence stage of each participant cannot be reconstructed from this study alone.
The self-reported grams-per-day variable described typical consumption during periods of regular use. It was not a measured administered trial dose, a chemical mass recovered from a biological specimen, or an estimate of drug present in the brain during testing. Calling these quantities a quantified biological exposure would materially misstate the experiment.
How does the computerized Trail Making Test work?
The researchers used Inquisit 5 software, a laptop, and a mouse. In TMT-A, participants connected numbers 1 through 25 in ascending order. This condition requires visual search, sequencing, attention, and mouse responses. In TMT-B, participants alternated between ascending numbers and letters, for example 1-A-2-B-3-C, placing added demands on maintaining and switching a response rule. TMT B−A was used as a derived condition or shifting-cost comparison. The research recorded task-completion times in seconds and ordering errors as counts.
A longer time is not a blood concentration. An error count is not a toxicological detection. These measures reflect multiple overlapping cognitive and psychomotor functions. TMT-B generally requires more complex set-shifting than TMT-A, but its difference from TMT-A is not a chemically selective assay for dopamine, methamphetamine, or Parkinsonian neuropathology. The task supplies a carefully delimited behavioral observation.
What did the adjusted completion-time results show?
The adjusted marginal means in the original article provide the most concrete scale for the difference. For TMT-A, the community controls averaged 38.845 seconds and the methamphetamine group 51.906 seconds; the adjusted comparison was p < .001. For the more demanding TMT-B, adjusted means were 51.018 versus 72.425 seconds, respectively, with p < .0001. The source’s separately reported B−A marginal means were 11.503 versus 19.891 seconds, p = .016. These are estimated adjusted group means, not every person’s raw score, and the source’s B−A estimated value should not be silently replaced by subtracting two other modeled means.
The broader finding is that the treatment cohort was slower on each specified comparison. The especially demanding alternating condition also yielded more errors, as shown below. The result is relevant to research on executive functioning during recovery, but it cannot establish the mechanism or permanence of any particular person’s slower performance.
Original quantitative evidence table: adjusted group means
| Task and endpoint | Controls, n = 30: mean (SE) | Methamphetamine group, n = 49: mean (SE) | Reported between-group p |
|---|---|---|---|
| TMT-A time, seconds | 38.845 (2.675) | 51.906 (2.065) | < .001 |
| TMT-B time, seconds | 51.018 (2.675) | 72.425 (2.065) | < .0001 |
| TMT B−A time, seconds | 11.503 (2.675) | 19.891 (2.065) | .016 |
| TMT-A ordering errors, count | 0.940 (0.374) | 1.200 (0.289) | .591 — nonsignificant |
| TMT-B ordering errors, count | 1.406 (0.374) | 3.323 (0.289) | < .0001 |
| TMT B−A error measure, count | 0.573 (0.374) | 2.098 (0.289) | < .002 |
These are source-reported adjusted marginal means, not individual observations. The original article reports correction for age, education, psychiatric conditions, tobacco use, depression, anxiety, and stress; p-values are described as Bonferroni-adjusted pairwise contrasts. “Nonsignificant” means that the specific statistical comparison did not establish a difference at the prespecified threshold; it does not prove identical population performance.
Which error findings were significant, and which were genuinely null?
The error pattern is more discriminating than a blanket statement that methamphetamine history makes people inaccurate. In the simpler TMT-A, the adjusted mean counts were 0.940 in controls and 1.200 in the treatment group, with p = .591. This is a genuine reported nonsignificant group difference. A headline asserting that the treatment group committed significantly more errors on every Trail Making condition would be incorrect.
By contrast, the more difficult TMT-B had adjusted means of 1.406 versus 3.323 errors, with p < .0001; the B−A-derived error measure likewise differed, 0.573 versus 2.098, with p < .002. The investigators reported group-by-condition interactions for time, F(2, 224) = 4.048, p = .019, and errors, F(2, 224) = 3.571, p = .030. These support the notion that the amount of separation varied with task condition. They do not independently validate clinical neurological injury or toxicological timing.
Did biological sex change the results?
Within the methamphetamine group, the reported combined sex-by-TMT-condition multivariate interaction was not significant, F(4, 278) = 0.240, p = .915; the sex main effect on combined outcomes approached, but did not cross, the conventional threshold (p = .056). The route-of-administration covariate in that particular combined analysis was also nonsignificant (p = .272). Subsequent univariate testing reported a sex difference in completion time (p = .016), especially for TMT-B (p = .047), but not in overall error counts (p = .452). Those statements concern different specified models; they are not interchangeable.
Among community controls, no significant sex-by-task multivariate interaction was found (p = .945), and the multivariate sex main effect was nonsignificant (p = .863). Taken together, these results offer limited evidence for some sex-dependent completion-time differences in the treatment sample, but not for an across-the-board sex-specific effect on every endpoint. Small strata, nonrandomized histories, and multiple models make the exact distinction important.
What did intravenous versus inhaled use predict?
The original treatment cohort’s history was unevenly distributed by sex: among 23 women, 17 reported inhalation and six reported intravenous use; among 26 men, 11 reported inhalation, 14 intravenous use, and one oral use. This imbalance limits simple cross-sex attribution. In the researchers’ exploratory stepwise regressions, route of administration predicted TMT-B errors among women, with F(1, 21) = 11.370, p = .003, reported standardized coefficient 0.593, and change in R2 of .351. For the B−A error measure, F(1, 21) = 6.100, p = .022, coefficient 0.474, and change in R2 of .225. Corresponding models were not significant among men.
This is a historical route-of-use association among a small subgroup. Six female participants reporting injection cannot support a general causal claim that the act of injection, rather than correlated exposure histories, clinical severity, comorbidity, or selection factors, produced a neurological outcome. “35.1% variance explained” describes the fitted exploratory statistical model; it is not a percentage of women who had an injury, a probability of developing Parkinson’s disease, or a validated individual’s risk.
Original quantitative evidence table: tests, subgroup findings and limits
| Original analysis | Reported statistic | p-value | Appropriate inference |
|---|---|---|---|
| Group × task condition, combined time/errors | F(4,446) = 3.666 | .006 | Group separation varied by task |
| Group × task, completion time | F(2,224) = 4.048 | .019 | Time contrast depended on condition |
| Group × task, error count | F(2,224) = 3.571 | .030 | Error contrast depended on condition |
| Within treatment group, sex × condition | F(4,278) = 0.240 | .915 — null | No demonstrated combined interaction |
| Within treatment group, overall sex error test | F(1,140) = 0.568 | .452 — null | No demonstrated sex difference in errors |
| Women in treatment, route predicts TMT-B errors | F(1,21) = 11.370; ΔR² = .351 | .003 | Exploratory subgroup association |
| Women in treatment, route predicts B−A errors | F(1,21) = 6.100; ΔR² = .225 | .022 | Exploratory subgroup association |
The article reports multivariate and follow-up statistics with different degrees of freedom. They are presented exactly as published, not recomputed as if every task condition were an independent additional person. These exploratory and repeated-condition models warrant independent statistical scrutiny before any use as a diagnostic rule.
Why do statistical adjustment and study design matter?
The investigators adjusted their main group analyses for measured age, education, psychiatric conditions, tobacco use, and three negative-emotion subscale scores. This reduces some specified observed differences but cannot guarantee that the two nonrandomized groups would have performed identically without methamphetamine history. Recruitment source, treatment setting, baseline performance, sleep conditions, co-occurring nicotine and cannabis factors, and unmeasured exposures can still matter. Also, a regression coefficient can be numerically precise within a fitted model while the underlying clinical inference remains uncertain.
The reported analysis uses three related conditions from the same persons, and the source presents some model degrees of freedom substantially larger than the number of biological participants. Such degrees of freedom may reflect repeated observations or the specific modeling structure. They should not be interpreted as evidence for hundreds of independent patients. The source does not make participant-level datasets publicly available for reconstructing the models. Statistical significance does not substitute for model validation in a different population.
Does a Parkinsonian-like cognitive pattern mean Parkinson’s disease?
No. The phrase Parkinsonian-like phenotype describes overlap in selected executive-function patterns that are also studied in Parkinson’s disease. The experiment did not conduct a neurological diagnostic workup establishing Parkinson’s disease in these 79 people, did not measure nigrostriatal neuronal loss in living subjects, did not use dopamine-transporter neuroimaging to verify a clinical pathology, and did not follow participants prospectively to determine who developed Parkinson’s disease. A deficit on a sensitive but nonspecific neuropsychological task cannot identify its cause by itself.
The authors place their findings in a wider literature connecting chronic methamphetamine misuse with frontostriatal function and possible later neurological vulnerability. That context may justify a hypothesis and future testing, but it does not make a cross-sectional Trail Making comparison equivalent to an incidence study or a pathological diagnosis. The most defensible wording is that the researchers observed a pattern of test differences compatible with aspects of cognitive inflexibility, not that they detected Parkinson’s disease or demonstrated progression.
Was methamphetamine quantified in forensic blood or any other specimen?
No standardized quantitative drug-assay results were provided for the cognitive analysis. The paper described some possible admission toxicology at one program and screening when use was suspected; those data were not systematically collected, modeled, or sufficient to determine precise last-use time. The investigators did not report a measured parent-methamphetamine or amphetamine-metabolite concentration in legally relevant venous whole blood, and did not report comparable quantitative separated plasma or serum, urine, oral fluid, sweat, hair, vitreous humor, breath, or tissue results for these participants. There was no validated patient-drug LC–MS/MS or GC–MS calibration range, detection limit, quantitative concentration, or drug-to-test pharmacokinetic calculation in this study.
These distinctions are more than terminology. A historical report of intravenous injection is not a laboratory specimen; an observed Trail Making latency is not a parent-drug concentration; and a screening test suspected or performed in a treatment facility is not an analyte-specific measurement available for independent forensic review. For background on how measured compartments differ, see the site’s separately published guide to biological matrices in forensic toxicology.
Can the study identify acute intoxication or the time of last methamphetamine use?
It cannot. There was no controlled acute dose, no validated blood–effect time course, no synchronized drug concentration at test time, and no consistent measured abstinence duration. Slower completion can occur for many reasons beyond the presence of an active drug. Conversely, a participant could complete a task quickly while carrying biologically measurable drug; this study did not evaluate that discordance either. The paper therefore cannot provide a forensic formula converting seconds or errors into ng/mL or a rule identifying a specific recent-use window.
Analytically confirming the identity and quantity of a compound requires a tested specimen, an appropriate method, chain of custody and specimen-handling evidence where relevant, and a validated interpretive scope. The site’s published overview of confirmatory drug testing addresses that separate analytical problem. Nothing in the Trail Making paper replaces such evidence.
Could these data support an individual driving or workplace impairment conclusion?
Not on their own. The experiment did not test actual highway driving, a validated driving simulator, occupational hazard response, eyewitness accuracy, or an individualized intoxication diagnostic threshold. It also lacked quantitative contemporaneous blood-drug results. A statistically significant average difference between two groups is not a validated classifier for deciding whether a specific driver’s faculties were impaired at a particular time. Even the more complex TMT-B result is affected by task experience, motor control, vision, language familiarity, clinical history and test circumstances.
A defensible forensic review would separately examine the subject’s specific behavior and timeline, observed neurological and medical signs, co-exposures, toxicological specimen and validated chemical method where available, and the purpose for which any test was administered. It must also distinguish an observed functional deficit from its hypothesized cause. A useful additional comparison is the site’s published guide to timed therapeutic drug monitoring, which illustrates why timing and matrix assumptions cannot be silently transferred between studies.
What does this study contribute to treatment research?
The study gives a concrete description of executive-task performance among adults early in residential treatment. Because TMT-B imposes alternating sequencing rules, the observed difference may be especially relevant when evaluating the accessibility of cognitively demanding counseling, planning, or skill-based interventions. That is a potential clinical research application, not a demonstrated treatment effect. The investigators did not randomize people to cognitive rehabilitation, prove that an intervention restored performance, measure subsequent retention in treatment, or establish a causal pathway to relapse reduction.
The female route-of-use pattern likewise suggests a focused future-study question: does a prospective, adequately powered sample with verified exposure, relevant health histories, and prespecified sex-by-route interactions replicate the observed error difference? Replication would need to account for six female injection-history participants in this sample and for the distribution of smoking versus injection among men. Such future research could refine treatment planning; it cannot be treated as already accomplished here.
How should a forensic reader summarize the evidence?
A careful summary can state the original denominators, the source and timing of recruitment, and the endpoints actually observed: 49 residential-treatment patients with methamphetamine use disorder and 30 community controls completed computerized Trail Making tasks. The adjusted mean time was longer in the treatment group across A, B and the published B−A measure. Error counts differed substantially on B and B−A, but the A-condition error comparison was nonsignificant. The within-treatment sex-by-condition combined interaction was also nonsignificant, while a small, exploratory female route-of-use analysis associated self-reported injection history with complex-task errors.
The next sentence should supply the boundary: there was no standardized, contemporaneous quantitative methamphetamine drug assay, reliable individual last-use timing, direct measurement of Parkinson’s disease, or validated person-specific driving impairment result. This preserves the study’s value as behavioral research without turning its results into a toxicological detection method or case-specific causal opinion.
Frequently asked questions
1. How many independent people were included?
Seventy-nine people: 49 adults in residential treatment for methamphetamine use disorder and 30 community controls. The multiple Trail Making conditions, error scores, and regression models reuse observations from those same participants; they are not new biological subjects.
2. Which Trail Making result was not statistically significant?
The adjusted between-group TMT-A error comparison was p = .591. The methamphetamine-group sex-by-task multivariate interaction was also nonsignificant (p = .915), and the within-group overall sex comparison in errors was nonsignificant (p = .452).
3. Did the researchers prove that intravenous methamphetamine caused neurological injury in women?
No. They found an exploratory statistical association between self-reported historical injection route and more TMT-B errors within a small female treatment subgroup, including only six women with injection history. The design cannot separate route from numerous correlated histories or establish causation.
4. Were the controls and treatment participants randomly assigned to methamphetamine?
No. The researchers observed already-existing groups. Methamphetamine was not administered as a study drug. Statistical covariate adjustment cannot convert an observational comparison into a randomized exposure trial.
5. Does the reported Parkinsonian-like pattern diagnose Parkinson’s disease?
No. The study examined task performance that may overlap with aspects of cognitive inflexibility seen in Parkinsonian disorders. It did not establish clinical Parkinson’s diagnoses, measure direct neurodegeneration in the participants, or track later Parkinson’s incidence.
6. Can task-completion seconds estimate a blood methamphetamine level or last-use date?
No. The study lacked standardized quantitative human venous whole-blood measurements and a reliably recorded interval since last use. Seconds and task errors have no validated conversion into individual methamphetamine concentrations or use-time estimates.
7. Could a court use the reported group means as a driver-impairment cutoff?
No scientifically validated cutoff was derived. The participants did not undergo a driving validation study, and the results concern adjusted group distributions rather than person-specific impairment at an incident time. Individual forensic opinions require separate tested evidence and appropriate validation.
Primary research source and interpretive scope
Dumitrescu AM, Vest MF, Fleckenstein AE, Patterson JC II, Ledbetter C, Goeders NE, Loftis JM, Thompson E, Lofton K, Murnane KS. Trail Making Test performance in early abstinence from methamphetamine: human evidence for a drug-associated Parkinsonian-like phenotype. Frontiers in Psychiatry. 2026;17:1773668. Read the original empirical study: DOI 10.3389/fpsyt.2026.1773668. This original analysis is educational; its adjusted association estimates cannot stand in for evaluation of an actual person’s specimen, test validity, neurological diagnosis, or case-specific circumstances.