Two treatments may look similar under one outcome definition and different under another. A 2011 reanalysis comparing lisdexamfetamine with mixed amphetamine salts illustrates how changing a response threshold changes the scientific question.
The original comparison
Fifty-two children aged 6–12 entered a crossover classroom study, and 50 completed it. An open-label period first optimized mixed amphetamine salts extended release at 10, 20 or 30 mg/day. Children then received that formulation, approximately corresponding lisdexamfetamine doses and placebo during randomized blinded periods.
These were repeated observations in the same children. They were not three independent groups of 50. The optimization procedure also selected doses using one active formulation before the comparison.
What did the primary outcome show?
Both active treatments improved classroom deportment relative to placebo. The original primary efficacy comparison did not establish a difference between the two active treatments. Their mean deportment scores were similar.
The later report examined clinician global-improvement ratings. Combining “very much improved” and “much improved” produced response percentages of 74% with lisdexamfetamine, 72% with mixed salts and 18% with placebo.
What happened with a stricter threshold?
The post hoc analysis isolated “very much improved,” the highest improvement category. Sixteen of 50 children, 32%, met that threshold on lisdexamfetamine, compared with eight, 16%, on mixed salts.
Ten children met it only with lisdexamfetamine, two only with mixed salts, six with both and 32 with neither. The paired McNemar comparison was p = 0.0386. The discordant pairs, rather than treating all observations as independent, determine that test.
What does this difference mean?
It supports a difference under this particular post hoc clinician-rating threshold in this small crossover sample. It does not overturn the null active-treatment comparison on the original primary classroom outcome.
The category boundary was examined after reviewing the data. Multiple possible outcomes and thresholds increase the importance of replication and prospective specification. Approximately equivalent dosing also does not make the formulations pharmacokinetically identical.
Why not declare one medicine generally superior?
Different measures captured different aspects and times of improvement. A single global rating and repeated classroom observations are not interchangeable. Small sample size, short periods, prior dose optimization and industry involvement further limit broad conclusions.
For interpretation, state the exact threshold and design: a post hoc difference in the highest clinician-improvement category, not universal superiority across all outcomes. The study provides no individual impairment or driving-safety determination.
Scientific source
López FA, Scheckner B, Childress AC. Physician perception of clinical improvement in children with attention-deficit/hyperactivity disorder: a post hoc comparison of lisdexamfetamine dimesylate and mixed amphetamine salts extended release in a crossover analog classroom study. Neuropsychiatric Disease and Treatment. 2011;7:267–273. doi:10.2147/NDT.S17002. Original commentary on the supplied paper.
Leave a Reply