Understanding Non-Inferiority Trials

Introduction

• Presenter: Christine Pruschak, University of Maryland School of Pharmacy
• Objective: To understand how to interpret non-inferiority trials and differentiate them from other study types.

Types of Trial Designs

1. Superiority Trials

   • Aim to prove that the study treatment is superior to control.
   • Null Hypothesis: Treatment X is no more effective than Treatment Y.
   • Disproving the null suggests the new treatment is better.

2. Equivalence Trials

   • Aim to demonstrate that two treatments are equal.
   • Null Hypothesis: Treatment X is either worse or better than Treatment Y by more than delta (equivalence margin).

3. Non-Inferiority Trials

   • Aim to show that the treatment is not worse than the control.
   • Null Hypothesis: Treatment X is worse than Treatment Y by more than delta (non-inferiority margin).
   • Alternate Hypothesis: Treatment X is not worse than Treatment Y by more than delta.

Why Non-Inferiority Trials?

• Cost-Effective: Require fewer patients and resources.
• Convenience: Easier to conduct than superiority trials.
• Ethical Considerations: Used when treatment comparison to placebo is unethical.
  • Example: When a current treatment is the standard of care.
• Advantages of New Treatment: New treatment may offer benefits like improved safety, convenience, adherence, or cost.

Reviewing Non-Inferiority Trials

1. Delta

   • Often based on previous trials.
   • Derived from a lower bound of confidence interval or fraction of active control effect.
   • Must compare trial design to those that confirmed efficacy of the active control.

2. Active Comparator Dosing

   • Ensure dosing is optimal.
   • Suboptimal dosing biases results toward non-inferiority.

3. Blinding

   • Not always effective in removing bias.
   • Important for subjective parameters.
   • Consider blinding investigator to trial design.

4. Analysis Population

   • Intention to Treat: Includes all randomized patients.
     • May skew results based on compliance.
   • Per Protocol: Only includes patients following treatment protocol.
     • Higher likelihood of finding treatment differences.
   • Review results from both analyses for reliable interpretation.

Example Case Study: Terpaban vs. Defaultarin

• Design: Non-inferiority trial comparing Terpaban (new) to Defaultarin (standard).
• Primary Endpoint: Time to first stroke or ischemic event.
• Non-Inferiority Margin: Upper boundary of 97.5% confidence interval must not exceed 1.38.
• Results:
  • Per Protocol: Terpaban shows 1.18% per year stroke rate vs. 1.5% for Defaultarin.
  • Hazard Ratio: 0.79 with CI of 0.52 to 1.29.
  • Confidence interval left of non-inferiority margin indicates non-inferiority.

Conclusion

• Non-inferiority trials are useful to prove a treatment is no worse than control.
• Increasing in use due to advantages.
• Important to consider design specifics, population, and confidence intervals relative to delta.