Two analyses can use the same participants and still estimate different quantities. A comparison of final-month drug-use days and a comparison of improvement from baseline should not be collapsed into a single verdict about treatment effectiveness.
The trial and its two comparisons
Ling and colleagues’ 2014 double-blind trial randomized 110 adults with methamphetamine dependence, 55 per arm. Oral sustained-release methylphenidate increased from 18 mg/day in week one to 36 mg in week two and 54 mg during weeks 3–10. Both arms received counseling and incentives; four subsequent weeks used single-blind placebo.
The primary comparison of reported methamphetamine-use days in the final 30 days of active treatment was nonsignificant (p=0.22). A planned secondary change-from-baseline analysis favored methylphenidate: reductions of 6.56 versus 3.82 days (p=0.05). Urine outcomes at week ten did not distinguish groups. The authors identified dropout, small sample size and adherence measurement as limitations.
Why the baseline question matters
A final-period value describes the outcome at the chosen endpoint. A change score includes both the starting value and the endpoint. Those questions can yield different results without either calculation necessarily being erroneous.
The appropriate summary identifies which analysis was primary and which was secondary, then describes the evidence for each. A favorable secondary comparison can support further investigation. It cannot silently replace the prespecified primary comparison after the results are known.
Baseline severity also matters when comparing studies or discussing subgroups. A participant’s opportunity to reduce reported use depends partly on their starting frequency. The relevant analysis must distinguish that structure from a claim that a particular severity category has a proven biological response to medication.
Reported days, urine results and treatment phase
Self-report and urine testing supply related but different observations. A day-count endpoint should remain a day-count endpoint, and the urine comparison should be reported on its own terms. Agreement should be demonstrated rather than assumed.
The switch to placebo after active treatment is another important boundary. An observation in that later phase should not be described as a result during ongoing active medication. A case review or trial comparison should preserve when each outcome was collected relative to the regimen.
This historical study provides a useful example of how outcome selection affects interpretation. Its mixed pattern supports a precise account of the tested regimen and analytical hierarchy, without establishing a universal methylphenidate treatment effect, an individualized prognosis or a current dosing recommendation for methamphetamine use disorder.
Source
Ling W, et al. Sustained-release methylphenidate in a randomized trial of treatment of methamphetamine use disorder. Addiction. 2014;109:1489–1500. doi:10.1111/add.12608. Full text.