The answer is no unless a more stringent criterion is applied. Let the null hypothesis in a non-inferiority (NI) trial state that Ho:C-T>6 against H1:C-T=< 6. C and T denote the control and treatment population means and 6 is the NI margin.
NI says that if 0< C-T< 6, C is not superior to T. If C is slightly better than T, T is NI to C. But if T is slightly better by the same amount T is superior. This interpretation is common and is noted in the FDA NI guidance document.
By allowing the margin to equal 6 for NI and 0 for superiority for T but not for C leads to an illogical conclusion. Suppose the 95% CI is (-5.9, -0.5). If the trial is conducted by T's sponsor, T is superior to C (because the upper limit < 0), but if conducted by C's sponsor, C is NI to T (because the lower limit > -6). The two interpretations are illogical. Arguments claiming that T has other benefits (e.g. lower AE rates) require a separate evaluation, and not a compromise in evaluating efficacy.
For T to claim superiority should require that the upper limit of the estimate of C-T< -6. For C to claim superiority should require that the lower limit of the estimate of C-T > 6.
|