A new antibiotic is tested against the standard treatment for pneumonia. The non-inferiority margin is set at 10 percentage points for cure rate. The study finds the new drug's cure rate is 5 percentage points lower, with a 95% confidence interval of [-9, -1]. Is non-inferiority demonstrated?
ANo — the new drug is 5 points worse, which proves inferiority
BYes — the entire 95% confidence interval lies above the non-inferiority margin of -10, so the new drug is not worse than the standard by more than 10 points
CNo — the confidence interval is entirely below zero, proving the new drug is inferior
DThe result is inconclusive because the interval includes negative values
Non-inferiority is demonstrated when the lower bound of the confidence interval for the treatment difference exceeds (is less negative than) the non-inferiority margin. Here, the lower bound is -9, which is above -10. We can conclude with 95% confidence that the new drug is not worse than the standard by more than 10 points. Note that the drug IS worse (the CI is entirely negative, so it would be declared inferior in a superiority test), but it is not worse enough to exceed the pre-specified acceptable margin. If the drug offers advantages (fewer side effects, oral instead of IV), this small efficacy loss may be acceptable.
Question 2 True / False
In a non-inferiority trial, the intention-to-treat analysis is the primary analysis, just as in superiority trials.
TTrue
FFalse
Answer: False
In superiority trials, ITT is conservative (non-compliance dilutes the treatment effect, making it harder to reject the null). In non-inferiority trials, this logic reverses: non-compliance, treatment crossover, and poor adherence make treatments appear more similar than they truly are, biasing TOWARD non-inferiority. The per-protocol analysis, which includes only patients who adhered to the assigned treatment, is the primary analysis because it provides a more honest comparison of the treatments as actually received. Both ITT and per-protocol should be reported, and non-inferiority should ideally be demonstrated in both.
Question 3 Short Answer
Explain why choosing the non-inferiority margin is the most critical and controversial decision in non-inferiority trial design.
Think about your answer, then reveal below.
Model answer: The margin determines what counts as 'not meaningfully worse' and directly controls the clinical interpretation. Too large a margin allows clinically important efficacy losses to be declared non-inferior — the new drug could be substantially worse than the standard but still pass. Too small a margin requires enormous sample sizes and may be unachievable. The margin should be smaller than the active control's effect over placebo (otherwise the new drug could be no better than placebo and still pass) and should represent a loss of efficacy that preserves a clinically meaningful proportion of the active control's benefit. There is no statistical formula for choosing it — it requires clinical judgment and is often the subject of regulatory negotiation.
The margin is typically set at no more than 50% of the active control's historically demonstrated effect over placebo. This ensures the new drug retains at least 50% of the known benefit. But if the active control's effect has been measured imprecisely or has changed over time (assay sensitivity concerns), the margin may be based on outdated evidence. The FDA requires a thorough historical review (meta-analysis of prior placebo-controlled trials) to justify the margin — a process called 'constancy assumption' analysis.