Sixty-four effect sizes are not 64 independent exercise trials. In a 2025 systematic review addressing executive function in drug addiction, repeated outcomes from 11 randomized trials contributed to a three-level meta-analysis. Most listed trials involved methamphetamine dependence. The distinction matters when the length of a forest plot is used to suggest a larger body of independent evidence than actually exists.
The review reported a favorable average effect on executive-function measures. Interpreting that finding requires preserving the hierarchy of observations, the nature of the tests and several inconsistencies in the published report. These are questions about the strength and applicability of a synthesis, not reasons to replace its positive findings with an unsupported conclusion of no benefit.
What evidence was eligible?
Chen and colleagues searched Embase, Scopus, PubMed and Web of Science, updating the search through April 30, 2025. They included English-language, peer-reviewed randomized controlled trials in people diagnosed with drug addiction or substance use disorder. Exercise was the principal intervention; comparisons included health education, routine treatment or quiet sitting. Animal studies were excluded from the clinical synthesis.
The 11 trials were published between 2017 and 2024. Activities included aerobic exercise, tai chi and aerobic exercise combined with attentional bias training. Protocols ranged from a single session to 12 weeks, with sessions lasting 20–60 minutes. This was rehabilitation research, not a controlled experiment administering a known methamphetamine dose to test acute intoxication.
Why did 11 trials produce 64 effect sizes?
A trial can supply several measurements: accuracy and reaction time on the same task, different executive tasks, more than one assessment time, or separate active arms sharing a control group. Those measurements are related because some come from the same participants or comparison groups. Counting them as independent replications would misdescribe the evidence.
The authors used a three-level random-effects model to accommodate dependence. Its levels represented sampling variance, variation among effect sizes within a study and variation between studies. This is a methodological response to correlated results; it does not transform one participant cohort into multiple independently randomized cohorts.
The source’s table also changes the meaning of its count symbol. Table 1 uses k for included studies; Table 3 explicitly defines k as extracted effect sizes. Thus the 62 entries in the methamphetamine subgroup of Table 3 are effect sizes, not 62 trials or 62 patients.
What was the estimated benefit?
The overall analysis reported Hedges’ g = 0.40, with a 95% confidence interval of 0.28–0.53, from 64 effect sizes across 11 trials. Hedges’ g is a standardized group contrast. A value of 0.40 does not mean that 40% of participants recovered, that cognition improved by 40%, or that 40% achieved abstinence.
| Domain | Extracted effect sizes | Hedges’ g | 95% confidence interval |
|---|---|---|---|
| Inhibitory control | 34 | 0.25 | 0.09–0.42 |
| Cognitive flexibility | 10 | 0.58 | 0.27–0.88 |
| Working memory | 20 | 0.57 | 0.34–0.79 |
The domain moderator test was statistically significant, F = 3.360, p = 0.041. These findings support an outcome-specific account rather than a single claim that every executive ability changed equally. The tasks and their scoring remain relevant to what the pooled estimates mean.
How stable was the pooled result?
The authors identified an outlying working-memory result involving reaction time on the five-number Sternberg task after an eight-week intervention. Excluding that outlier changed the pooled estimate from g = 0.42, 95% CI 0.28–0.56, to g = 0.40, 95% CI 0.28–0.53. That modest change supports numerical stability under that particular sensitivity analysis.
It does not eliminate other uncertainty. The main overall analysis reported moderate heterogeneity, I² = 57.33%, attributable to between-study variation; the within-study variance test was nonsignificant. Four included trials were rated at high risk of bias. A statistically significant average can coexist with differences in trial quality, populations and protocols.
The multilevel Egger test did not detect publication bias, p = 0.180. A nonsignificant bias test should be reported as failure to detect the pattern tested, rather than proof that no studies or unfavorable outcomes were missing.
Why does the participant denominator need qualification?
The abstract and results describe 906 participants, while the discussion describes 809. Table 1 contains an additional inconsistency in one three-arm trial: its listed arm sizes are 24, 23 and 23, which sum to 70, while its sex column reports 60 men. These discrepancies prevent presenting one total as independently reconciled from the report.
Substance labels also conflict. The prose describes one heroin study and ten morphine studies, but Table 1 identifies the other ten as methamphetamine studies. Table 3 likewise labels its larger substance subgroup methamphetamine. This commentary follows the explicit table labels when describing the relevance to methamphetamine, while flagging the conflicting prose. It does not silently substitute morphine, an opioid, for methamphetamine, a stimulant.
Table 3 gives the methamphetamine subgroup estimate as g = 0.40, 95% CI 0.26–0.53; nearby prose gives g = 0.39, 95% CI 0.24–0.53. An exact subgroup quotation should identify which location supplied it. Resolving these differences would require clarification or examination of the underlying extraction and analyses, beyond simply copying the abstract.
Can the review determine an individual’s recovery or impairment?
No individual diagnostic threshold follows from a standardized group average. Better Stroop, working-memory or switching performance under a study protocol does not establish restoration of all everyday functioning, durable abstinence or fitness to drive. Those outcomes require their own measurements and appropriate follow-up.
The synthesis also supplies no methamphetamine blood, urine or oral-fluid concentration that diagnoses impairment. It concerns people with a diagnosed disorder in intervention studies, not a concentration–performance calibration at a specific incident time.
The useful conclusion is qualified: this historical synthesis reported an encouraging average executive-function signal, based on 11 trials with dependent outcomes, heterogeneous protocols, risk-of-bias concerns and unresolved reporting discrepancies. The number of plotted estimates should not inflate the number of independent studies or conceal uncertainty in the population being summarized.
Scientific source
Chen J, Wang X, Zhang X, Xiao W, Kabachkova A, Tan Y. Effects of exercise on executive function in individuals with drug addiction: a systematic review and three-level meta-analysis. Frontiers in Sports and Active Living. 2025;7:1646327. doi:10.3389/fspor.2025.1646327. Historical human-trial synthesis; search coverage through April 30, 2025. Numerical discrepancies described above are present in the supplied report; no participant-level reanalysis was performed.
Leave a Reply