Neurology · Neurology · 2023
The summary pairs the fatigue result with another score’s margin of error, 3.4 times too narrow.
The drug’s effect on each score, in points. Dot: result. Bar: margin of error.
The paper says
A trial compared elamipretide injections with dummy injections in a genetic disease starving muscles of energy. For fatigue, a main goal, the summary reports −0.07 questionnaire points (minus: less tired) and a margin of error, where the true effect likely lies, of −0.10 to 0.26.
Its own data show
Yet −0.07 sits near the bottom of that span, whose middle is +0.08. The results section gives the same −0.07 with −0.69 to 0.54, 3.4 times wider, matching each group’s average change and its uncertainty. −0.10 to 0.26 is exactly the span the paper’s online table gives a lesser goal, a most-bothersome-symptom score: +0.08.
A margin of error should sit evenly around its result, like a poll’s ±3 points; narrower means surer. The summary’s span rules out the drug easing fatigue by over 0.10 points, so the trial looks more precise than it was. The real span allows 0.69, over half the 1.1-point fall in fatigue both groups had. Neither shows a clear benefit.