FDA Experience and Perspective on Non-Inferiority Trials

You You ' ' ve heard a lot about NI trials already, so I ve heard a lot about NI trials already, so I ' ' ll try not to ll try not to repeat. The critical issues are, by now known to you: repeat. The critical issues are, by now known to you: 1. In most cases NI trials pose inferential problems, but 1. In most cases NI trials pose inferential problems, but you use them when you have no choice, i.e., when you you use them when you have no choice, i.e., when you simply cannot leave patients untreated (placebo simply cannot leave patients untreated (placebo-treated) treated) and must use active treatment as the control. But the and must use active treatment as the control. But the need need to use an active control does not always mean this design to use an active control does not always mean this design will be a valid test of effectiveness. will be a valid test of effectiveness. 2. The NI study seeks to show that the new drug is not inferior 2. The NI study seeks to show that the new drug is not inferior to to the standard by too large an amount; that amount is called the the standard by too large an amount; that amount is called the non non-inferiority margin, M or delta. The NI margin has two inferiority margin, M or delta. The NI margin has two determinants determinants • • Inferiority must not be greater than the whole effect of the con Inferiority must not be greater than the whole effect of the control trol (because then you (because then you' 've lost the whole effect). So you must know the effect ve lost the whole effect). So you must know the effect of the control in the new study. The whole effect of the control of the control in the new study. The whole effect of the control is called is called M M 1 1 the largest possible NI margin. the largest possible NI margin. • • Inferiority must not be clinically unacceptable. This is a clini Inferiority must not be clinically unacceptable. This is a clinical, not cal, not statistical, judgment. The largest clinically acceptable differe statistical, judgment. The largest clinically acceptable difference is called nce is called M M 2 2. It must be no longer than M. It must …