Non-inferiority trial
"Just as good" treatment effects are established in non-inferiority trials
In a non-inferiority trial, researchers are only focused on the margin of non-inferiority ABOVE the cut-point or population value and the upper limit of the 95% confidence interval. With non-inferiority trials, researchers are not trying to prove that the treatments are different (such as with inferential statistics) and they are not trying to prove that the treatments are equal (as with equivalency trials). Rather, researchers are trying to prove that a treatment is not inferior or JUST AS GOOD.
The margin of non-inferiority needs to be stipulated in an a priori fashion. The cut-point and margin both need to be identified a priori. After observations are collected, the means and standard deviations (continuous outcome) or the odds ratios with 95% confidence intervals (categorical outcome) are compared to the cut-point or population value and the margin of non-inferiority.
The margin of non-inferiority needs to be stipulated in an a priori fashion. The cut-point and margin both need to be identified a priori. After observations are collected, the means and standard deviations (continuous outcome) or the odds ratios with 95% confidence intervals (categorical outcome) are compared to the cut-point or population value and the margin of non-inferiority.
When to use non-inferiority trials
In the classic article by Emmanuel Lesaffre*, the reasons for employing non-inferiority trials were specified:
- It is not ethically possible to conduct a placebo-controlled trial.
- The experimental treatment is not expected to be better than the control treatment on a primary efficacy endpoint, but is better on secondary endpoints or the treatment is safer.
- The experimental treatment is not expected to be better than the control treatment on a primary efficacy endpoint, but is cheaper to produce or easier to administer.
- The experimental treatment is not expected to be better than the control treatment on a primary efficacy endpoint, but compliance to the treatment will be better.
Interpreting non-inferiority trial outcomes
The statistical reasoning in non-inferiority trials is also different from equivalency trials due to the specific interest in just one side of the margin of non-inferiority and the confidence interval. Instead of a two-sided test with an alpha value of .05 to achieve significance, we are using a one-sided test with an alpha value of .025 to assume statistical significance.
So, if the upper limit of the confidence interval does NOT cross over the upper limit of the margin of equivalence, you CAN assume non-inferiority and your p-value will be LESS THAN .025.
However, if the upper limit of the confidence interval crosses over the upper limit of the margin of equivalence, you CANNOT assume non-inferiority and your p-value will be MORE THAN .025.
So, if the upper limit of the confidence interval does NOT cross over the upper limit of the margin of equivalence, you CAN assume non-inferiority and your p-value will be LESS THAN .025.
However, if the upper limit of the confidence interval crosses over the upper limit of the margin of equivalence, you CANNOT assume non-inferiority and your p-value will be MORE THAN .025.
Click on a button below to continue.
Statistician For Hire
DO YOU NEED TO HIRE A STATISTICIAN?
Eric Heidel, Ph.D. will provide statistical consulting for your research study at $100/hour. Secure checkout is available with PayPal, Stripe, Venmo, and Zelle.
- Statistical Analysis
- Sample Size Calculations
- Diagnostic Testing and Epidemiological Calculations
- Psychometrics
*Lesaffre, E. Superiority, equivalence, and non-inferiority trials. Bull NYU Hosp Jt Dis, 2008; 66(2):150-154.