An improvement during treatment is not automatically an effect of the assigned medication. A placebo-controlled comparison helps distinguish a shared study-course change from the added effect of the drug.
The between-group result
Coffin and colleagues’ 2013 double-blind trial randomized 90 actively using adults with methamphetamine dependence, 45 per arm, to oral aripiprazole or placebo for 12 weeks with weekly counseling. The registered schedule increased from 5 mg/day in week one to 10 mg in week two and 20 mg thereafter.
The primary urine-positivity comparison was nonsignificant: RR 0.88, 95% CI 0.66–1.19, p=0.41. In the placebo arm, positive specimens declined from 33/45 at baseline to 18/40 at the final visit; corresponding aripiprazole figures were 34/44 and 20/35. Electronic-container and self-reported adherence estimates were 42% and 74% overall. Akathisia, fatigue and drowsiness were reported more frequently with aripiprazole.
Why the denominators should accompany the percentages
The final percentages concern available observations from fewer participants. A percentage alone can conceal that change. Preserving the counts prevents an apparent trend from being mistaken for a complete series of measurements on everyone originally assigned.
The randomized model, rather than a visual comparison of two before-and-after percentages, addresses the primary treatment question. A decline in each arm can reflect features common to both arms. It cannot, by itself, establish which component of the study produced the improvement.
Did low adherence make the trial positive in principle?
Limited adherence constrains an exposure-based interpretation. It does not establish that the medicine would have worked if every dose had been taken. That is a further causal claim requiring supporting evidence.
Electronic-container openings and self-report also measure different things. An opening is not direct observation of swallowing; a report is not a continuous biological record. Disagreement between them should prompt attention to the measurement process, without automatically assigning one estimate the status of exact ingestion.
The confidence interval preserves another boundary: the null comparison is not proof of precisely zero effect. It also does not establish the favorable effect suggested by the point estimate alone. The uncertainty and the prespecified comparison belong together.
This historical trial did not demonstrate a methamphetamine-use benefit under its tested conditions. It does not decide whether aripiprazole is appropriate for a separate psychiatric indication, which would require evidence about that condition and the individual clinical context.
Sources
Coffin PO, et al. Aripiprazole for the treatment of methamphetamine dependence: a randomized, double-blind, placebo-controlled trial. Addiction. 2013;108:751–761. doi:10.1111/add.12073. Full text. The dosing schedule is also documented in the trial registration, NCT00497055.