Can Cognitive Tests Predict Methamphetamine Treatment Response? A 53-Person Study and the Correction That Changes the Answer

Answer first: In a South African pilot study, 30 adults with methamphetamine use disorder underwent baseline attention and trail-making tests before eight weeks of contingency management, while 23 healthy comparison participants took baseline tests only. Among the treated adults, participants subsequently classified as treatment responders completed Trail Making Test Part B faster than nonresponders, and this difference remained statistically significant after the authors’ Benjamini–Hochberg adjustment (p = .039). A difference in Continuous Performance Test (CPT) signal-detection performance was nominally significant before adjustment (p = .034) but not significant in the printed adjusted table (p = .061). That distinction changes any simplistic claim that both cognitive measures independently predict treatment success.

These findings come from the original empirical paper by Lara J. van Nunen, Marilyn T. Lake, Jonathan C. Ipser, Dan J. Stein, Steven J. Shoptaw, and Edythe D. London, Executive Function and Contingency Management in Methamphetamine Use Disorder (2021). The original study and DOI record should be consulted for participant definitions, full statistical methods, printed tables, and the limits of the inference. This was a small exploratory group comparison, not a validated individual relapse calculator, blood-drug assay, or driving-impairment experiment.

What is the specific forensic science question?

The defensible question is whether baseline performance on two different executive-function tasks differed among people who later fell into two urine-test-defined treatment-response categories during a structured eight-week program. One task measured the ability to distinguish relevant targets from nontargets. The other measured the time needed to connect and alternate a prescribed sequence on paper. Those measurements are not interchangeable, and they are not chemically measured concentrations of methamphetamine.

A separate question is whether the observed between-group test differences establish how a particular person will behave in treatment, in traffic, or during a legal event. They do not. The source does not report the sensitivity, specificity, calibration, decision threshold, independent validation cohort, or uncertainty required to make a single-patient forecast. The trial’s response categories were defined retrospectively by follow-up urine classifications, not assigned experimentally. This article separates the reported sample-level differences from personal predictions that were never tested.

Who was studied: 53 people, but only 30 in treatment?

The analysis included 53 independent human participants: 30 with diagnosed methamphetamine use disorder and 23 frequency-matched healthy controls. All 53 completed baseline assessments. Only the 30 clinical participants entered the eight-week contingency-management treatment program. Calling the study a 53-person trial of medication or abstinence treatment would misleadingly imply that the 23 healthy controls were assigned to the same intervention.

Within the 30-person treatment group, the investigators classified 17 as responders and 13 as nonresponders according to the presence of methamphetamine-positive urine results during treatment. A responder had no methamphetamine-positive urine tests by the study’s definition; the comparison category was more than one positive result. The published criterion is a categorical treatment-monitoring endpoint rather than a direct estimate of how much methamphetamine each person ingested or how impaired anyone was at a particular time. The source’s categories should be reproduced as defined rather than silently replaced with a new conventional definition of relapse.

Repeated stimulus presentations during the CPT, successive lines drawn on the Trail Making Test, and urine specimens collected during care are nested observations involving the same people. They do not increase the study’s independent human sample from 30 patients to hundreds. The group sizes are especially important when considering the stability of a p-value close to a multiple-comparison threshold.

What is contingency management, and what did this trial observe?

Contingency management is a behavioral treatment strategy in which specified incentives are linked to predefined treatment-related behaviors, such as attendance or drug-negative testing. This pilot used an escalating reinforcement schedule over eight weeks. The present analysis focused on whether differences already observable in cognitive testing were associated with treatment-response categories formed during that program. It did not randomize participants to become cognitively impaired, nor did it give methamphetamine to test dose-dependent effects.

To interpret the outcome, one must know exactly what a urine result means in the underlying treatment protocol, how missing specimens are handled, whether results are screened or confirmed, and whether a given collection belongs to the same person and visit. The journal paper does not supply a quantitative mass-spectrometric time series from which a person’s drug concentration trajectory could be reconstructed. A program label of responder therefore cannot be equated with direct proof of biological abstinence at every unobserved hour.

How did the CPT and Trail Making Test measure different functions?

The Conners Continuous Performance Test uses repeated on-screen stimuli and the participant’s responses to test sustained attention, discrimination, vigilance, and response control. The principal value analyzed in this paper was d-prime, a signal-detection index describing how well a participant distinguished targets from nontargets. Larger d-prime generally indicates better discrimination in the administered task. It is dimensionless; reporting it as seconds, milligrams, or a blood-drug concentration would be a category error.

Trail Making Test Part B is a paper-and-pencil sequencing task requiring visual search, processing speed, mental flexibility, and alternating attention. Its principal outcome was time to completion, with a shorter duration usually indicating faster task performance under those instructions. This duration includes motor drawing, visual scanning, and sequence maintenance, so it is not a pure single-neuron or isolated inhibitory-control measurement.

Either test may be influenced by education, reading and visual familiarity, motivation, motor ability, medical history, fatigue, language, testing conditions, and practice. A difference among small groups does not establish that methamphetamine itself directly caused every measured score or that the test diagnoses acute intoxication.

What did the original statistical plan require?

The researchers assessed normality and homoscedasticity of the outcome data, then used nonparametric Kruskal–Wallis comparisons. They specified a nominal significance level of .05 and reported a Benjamini–Hochberg adjustment to control the false discovery rate across multiple comparisons. This matters because several related group contrasts can produce an apparently interesting small unadjusted p-value by chance. A statistical correction is not an optional narrative detail when the article itself reports it.

In particular, the source’s prose and abstract refer to the CPT responder-versus-nonresponder difference as significant at p = .034. But the actual adjusted column of Table 2 reports p = .061 for that same CPT comparison. At the conventional .05 threshold the adjusted result is nonsignificant. The Trail Making responder contrast, in comparison, moved from nominal p = .013 to adjusted p = .039 and still met that threshold. Readers who rely only on the abstract could incorrectly present both as surviving multiplicity adjustment.

What did the four principal cognitive comparisons actually show?

These are medians from separate comparisons and are not average paired change scores or effects of experimentally administered methamphetamine. CPT results are signal-detection values; the Trail Making numbers are task-completion times. The table also preserves the source’s corrected p-values instead of silently substituting the abstract’s unadjusted CPT result. A statistically significant difference between these selected small groups still is not a validated individual clinical decision rule.

Why does the nominal versus corrected CPT result matter so much?

A reader might see the nominal responder comparison p = .034 and announce that weaker initial attention predicts future methamphetamine-positive urine testing. The source’s own multiplicity-adjusted result of .061 does not support the same threshold claim. The numerical medians still differ, but the adjusted test did not reject the stated null hypothesis at .05. It is therefore appropriate to describe the direction of the observed difference while explicitly saying the corrected comparison was nonsignificant.

That finding should not be inverted into proof of exact equality. A small exploratory sample may fail to resolve a real modest association. Nor should the adjusted nonsignificance be dismissed as a mere technicality: correcting for multiple comparisons is part of protecting against false positive interpretations. The correct evidentiary status is that the source reported a nominal CPT finding that did not survive its printed multiple-testing correction.

This ambiguity also cautions against estimating a person’s relapse probability from a two- or three-decimal difference in d-prime. The paper never tested a prespecified cutoff separating individual responders from nonresponders, and the group medians overlap in any realistic distribution. No receiver-operating-characteristic area, predictive value, or prospective validation is supplied.

Does the Trail Making finding prove a robust predictor?

No. The Trail Making comparison did remain significant after the adjustment applied in the table: median 119.60 among responders versus 198.21 among nonresponders, corrected p = .039. This is the stronger of the two within-treatment cognitive comparisons under the authors’ stated multiplicity treatment. It supports a small-sample group association between faster baseline performance and the subsequently assigned response category.

Yet the study enrolled just 30 patients, with 17 in one outcome group and 13 in the other. Task performance is multifactorial, and the source did not resolve all potential confounders or compare a score-based prognostic algorithm against clinical judgment. There was no independent cohort to check stability of a threshold, no randomized change in task performance to test causality, and no trial establishing that coaching patients to complete the lines faster would reduce drug use.

The finding deserves attention as a research hypothesis. It does not justify labeling a patient who completes Trail Making B slowly as inevitably nonresponsive, more intoxicated, dishonest, dangerous, or more likely to cause a particular incident.

How did the MUD patients differ from healthy controls before treatment?

The two groups were frequency-matched on demographic and selected background categories, but important differences remained. The clinical group had fewer years of education and higher ratings of depression. Seven of the 30 clinical participants, and none of the healthy controls, met the study’s criterion for antisocial personality disorder. Those differences can matter for both cognitive task performance and treatment engagement, and they cannot be assumed away because a source calls its controls “matched.”

Not every demographic comparison differed. Baseline reported mean age and an intellectual-function measure showed nonsignificant group comparisons. A nonsignificant p-value is not a guarantee of exact equivalence; it means the specified statistical test did not establish a difference in this sample. Here the larger education and mood contrasts are obvious alternative context for a cautious reading of the lower group CPT d-prime.

These measurements are participant characteristics and rating instruments, not methamphetamine exposure assays. The group differences make it hazardous to attribute the CPT result solely to the chemical effects of previous methamphetamine use. Nonrandom recruitment and other unmeasured variables remain plausible explanations for part of any observed performance separation.

What does “responded” mean when urine screens define the outcome?

The investigators categorized treated participants based on methamphetamine urine testing during the program. They did not report each participant’s complete quantitative methamphetamine exposure history, controlled dose, precise last use, or independently validated abstinence status for every moment between collections. The outcome is a study-defined urine-monitoring response category over eight weeks, not a clinical diagnosis of complete and durable recovery.

Urine classifications depend on the collected specimen, collection date, target compounds and analytical decision rule. Depending on the method, a presumptive immunoassay may react with chemically related substances, while a confirmatory chromatographic procedure has different selectivity. This article does not provide the detailed original individual chromatograms, detection limits, mass-spectrometric validation, quantitative parent concentrations, or chain-of-custody documentation required for a case-specific interpretation.

Equating a categorical urine test with a measured venous blood concentration would be chemically incorrect. Urine and blood differ in formation, partitioning, temporal relationship to administration, and reporting conventions. Even a valid urine-positive classification does not automatically prove contemporaneous psychomotor impairment, and a test-negative classification does not guarantee absence of use at every other time.

Did the researchers chemically measure methamphetamine exposure?

The investigators used urine test classifications as treatment-monitoring data, but did not publish a quantitative, specimen-by-specimen chemical assay of parent methamphetamine in legal venous whole blood. They also did not quantify methamphetamine in separated serum or plasma, hair, oral fluid, sweat or vitreous humor in a way that could supply a blood concentration for the cognitive-test day. No validated analytical calibration curve, chromatographic concentration series, legal impairment cutoff, specimen-specific uncertainty, or elimination slope appears in this analysis.

That matters because a computerized d-prime score and a Trail Making completion time are not measures of a drug molecule. Drug concentration is an analytical laboratory quantity; behavioral performance is a task-dependent observation. Without a demonstrated and validated relationship, there is no scientific equation to convert a participant’s seconds or signal-detection index into nanograms per milliliter of methamphetamine.

The study also did not experimentally administer methamphetamine during testing. It observed participants with preexisting diagnosed histories and subsequent treatment monitoring. Statements about the acute effect of a known methamphetamine dose, duration of an individual’s intoxication, or measured blood concentrations would exceed what was actually done.

Is association equivalent to forecasting a new patient’s treatment outcome?

Statistical comparisons after forming responder and nonresponder groups can detect associations, but they are not the same as assessing whether an individual prognosis works when used prospectively. A usable predictive instrument would specify variables and thresholds before application; describe how missing values are handled; estimate sensitivity, specificity, calibration, and false classification rates; and demonstrate performance in an independent group of patients. The paper did not complete that sequence.

Small study size magnifies this limitation. A difference between 17 and 13 people can be sensitive to a few atypical participants, incomplete measurements, social factors, and choice of test. Because the study is exploratory, it offers potential variables for future work rather than evidence that a particular defendant, patient or employee can be assigned a scientifically established probability of a future drug-positive specimen.

It is especially important not to confuse association with intervention efficacy. The paper did not randomly assign improved cognitive skills or show that changing d-prime or Trail Making time through training causes successful contingency management. A correlation between baseline function and later treatment category could be influenced by education, psychiatric symptoms, access to care, and clinical characteristics that also affect treatment engagement.

Can the control-group result establish methamphetamine-caused cognitive injury?

Not on its own. The clinical participants had a history meeting methamphetamine-use-disorder criteria, while the healthy controls did not. That was a comparison of existing groups, not an experimental assignment to methamphetamine versus placebo. Differences in education, depression ratings, and psychiatric comorbidity were reported at baseline. A lower mean or median cognitive measure in one group cannot determine whether prior methamphetamine exposure, earlier socioeconomic disadvantage, sleep, coexisting illness, or other factors caused the difference.

Nor did the investigators perform magnetic resonance imaging, neurohistology, a direct dopamine assay, or autopsy evaluation that could establish the presence or location of an individual’s neurological injury. Their tests probed behavior. A clinician may consider such behavior with the full history, but a single group-level comparison cannot establish a lesion, irreversible deficit, or specific chemical neuropathology for one person.

For the same reason, the nonsignificant overall Trail Making B comparison between 30 MUD participants and 23 healthy controls does not prove that the two populations are exactly equivalent in executive functioning. There is uncertainty around a small heterogeneous sample, and performance can depend on how the task is administered.

What conclusions would exceed the research in a legal case?

Neither the CPT nor Trail Making Test was evaluated here as a validated roadside sobriety instrument. The paper had no controlled on-road driving, driving-simulator collision, emergency-braking, divided-attention driving task, eyewitness recall examination, or event-specific reconstruction. It did not test whether any score predicts the behavior of a particular driver, witness, or person involved in litigation.

Therefore an expert should not treat a low d-prime score as proof that a person was acutely intoxicated, that a witness was unreliable, or that methamphetamine caused a particular collision. The comparison cannot establish that a person was above a statutory drug threshold, because no corresponding quantified legal whole-blood methamphetamine measurement and validated threshold relation were reported. It also cannot establish individual last-use time or chemical impairment from a urine response category.

A proper person-specific opinion would need separate, relevant evidence: the actual specimen and analytical method if drug presence is at issue; documented contemporaneous behavior if function is at issue; the timing and circumstances of the event; and a method validated for the specific conclusion. A group study may inform hypotheses about attention and treatment engagement, but it cannot substitute for that evidence.

Why are the null findings and conflicting prose important?

Responsible forensic interpretation should separate three categories: analyses that meet the published corrected significance threshold, analyses that were performed but are nonsignificant, and issues the researchers never measured. The within-treatment Trail Making p = .039 is in the first category. The adjusted CPT responder result p = .061 and MUD-versus-control Trail Making p = .554 are in the second. Measured blood methamphetamine, real road impairment, and individual litigation outcomes are in the third.

The source’s prose sometimes speaks broadly of two baseline executive tests differentiating responders. Its printed adjusted CPT p-value provides a more restrained statistical answer. A scientist should identify the exact row and column used, rather than presenting the abstract as overriding the multiple-testing adjustment. The paper’s Discussion also blends the names of CPT and Trail Making when describing completion time. Completion time belongs to Trail Making B, not the primary CPT d-prime outcome. The underlying experimental definitions are more reliable guides than an imprecise sentence.

Nonsignificance is not proof of no effect or evidence of universal equivalence, and significance does not automatically provide clinical importance or individual prediction. Repeated use of modest samples and multiple tests increases the importance of uncertainty estimates, multiplicity, and replication. The corrected table is thus not a pedantic footnote: it guards against an exaggeration of what the data support.

What should an evidence-grounded takeaway say?

The original study provides a focused exploratory observation: in 30 adults with methamphetamine use disorder taking part in eight weeks of contingency management, those in the program-defined responder category completed Trail Making B more quickly at baseline than nonresponders, and the adjusted between-category p-value was .039. The CPT responder contrast was nominally p = .034 but had a printed adjusted p = .061, so its corrected status was nonsignificant. Comparison with 23 healthy adults showed lower MUD-group CPT d-prime p = .012 but nonsignificant Trail Making performance p = .554, in a sample with documented educational and depression differences.

The findings warrant larger, preregistered, independently validated prognostic research. They do not quantify parent methamphetamine in legal venous whole blood, prove acute or permanent person-specific impairment, establish a general addiction-risk test, or identify the cause of an individual event. The strongest use of this paper is a carefully bounded discussion of baseline task-performance differences and the vital role of checking corrected source statistics.

Frequently asked forensic questions

1. Did all 53 participants receive contingency management?

No. The study included 30 adults with methamphetamine use disorder who entered the eight-week treatment program and 23 healthy comparison participants assessed at baseline. The latter were not 23 additional randomized treatment patients.

2. Did both cognitive tests significantly distinguish treatment responders?

Not after the source’s printed adjustment. The Trail Making B responder comparison had a corrected p = .039. The CPT responder comparison had nominal p = .034 but Benjamini–Hochberg corrected p = .061, which was nonsignificant at .05.

3. What is the difference between CPT d-prime and a Trail Making time?

CPT d-prime is a dimensionless signal-detection discrimination measure. Trail Making B measures completion time for a visually guided alternating-sequence task. Neither is a directly measured drug concentration.

4. Does a positive urine-screen category reveal someone’s blood methamphetamine level?

No. Urine and venous whole blood are different matrices. The paper did not establish a quantitative, confirmed drug concentration series or a validated conversion from urine classification or cognitive score to whole-blood parent methamphetamine.

5. Can a slow baseline Trail Making score prove a patient will relapse?

No. The observed association arose from 17 responder and 13 nonresponder patients. There was no independent validation cohort, calibrated individual decision rule, prospective sensitivity or specificity analysis, or random intervention changing task performance.

6. Does nonsignificant CPT p = .061 show responders and nonresponders are identical?

No. It means the corrected comparison did not cross the stated .05 statistical threshold in this small sample. It does not prove exact equivalence, but it also does not authorize describing the adjusted result as statistically significant.

7. Can this paper establish why a particular driver, witness, or patient acted as they did?

No. It contains neither case-specific event observations nor measured drug levels and no validated single-person forensic impairment method. Person-specific conclusions require independently reliable evidence tied to the relevant specimen, event, timing and question.

Primary source: van Nunen LJ, Lake MT, Ipser JC, Stein DJ, Shoptaw SJ, London ED (2021), Executive Function and Contingency Management in Methamphetamine Use Disorder, Journal of Alcoholism & Drug Dependence 9(5):342. Read the original DOI-linked article. The source reports original study data and author conclusions; the explanations here are independently written and do not replace the full paper or an individualized expert evaluation.