Equivalence Trials: Understanding the Statistical and Clinical Issues Christopher Cannon, M.D. C. Michael Gibson, M.S., M.D. Brigham and Women’s HospitalBeth Israel Deaconess Med Center

Superiority Trials l When a drug in a new class is developed, it is usually tested vs. placebo, in addition to the other standard therapies, to determine if it improves outcomes. Examples: New Classes: Placebo-controlled trials 4S – statin GISSI-1, ISIS-2 – streptokinase GISSI-1, ISIS-2 – streptokinase ASSET – t-PA ASSET – t-PA EPIC, PURSUIT, PRISM-PLUS – IIb/IIIa inhibitors EPIC, PURSUIT, PRISM-PLUS – IIb/IIIa inhibitors

Superiority trials (2) l If a new drug in an existing class is developed, there are 3 ways to test it: l Superiority testing vs. placebo in a different patient population, or related indication Example: CARE, LIPID – pravastation (Pts with lower LDL levels than in 4S trial) EPILOG, EPISTENT, ESPRIT - IIb/IIIa inhibitors in low risk patients, and/or stented patients EPILOG, EPISTENT, ESPRIT - IIb/IIIa inhibitors in low risk patients, and/or stented patients

Superiority trials (3) l If a new drug in an existing class is developed, and it has properties that make it potentially superior to an existing drug (or device) in that class: l Superiority testing of new drug (device) vs. older drug l Examples: GUSTO – I: t-PA vs. streptokinase ESSENCE, TIMI 11B: enoxaparin vs. unfractionated heparin ESSENCE, TIMI 11B: enoxaparin vs. unfractionated heparin STRESS, BENESTENT: stents vs. balloon angioplasty STRESS, BENESTENT: stents vs. balloon angioplasty

Equivalence Trials – Why? If on the other hand, a new drug (or device) is felt likely to be similar to the old drug, and the hypothesis is that they are similar, the third way to compare 2 drugs is an “equivalence” or “non-inferiority” trial l This would be the case if one expects similar overall outcomes for major outcomes (death, MI etc), with perhaps some advantages on other aspects (pharmacokinetics – once/day vs. twice/day dosing, fewer side effects) l If that class of drugs is established as a standard of care, one cannot remove it from clinical care, to test the new drug against placebo.

Hypothetical Trial Drug A vs. Drug B for stenting 500 patients randomized 250 per group. 30 day rate of stent thrombosis 4 (1.6%) for Group A 4 (1.6%) for Group B Are the drugs equivalent?

Hypothetical trial - Answer NO – too small sample size. How do you tell? Confidence intervals 4/250 = 1.6%, 95% CI (0.44% – 4.0%)

Absolute Difference 04.5 Confidence Intervals to Compare Two Treatments Drug A Drug B Large trial 40/ /2500 (1.6%)(1.6%) Small trial 4/250 4/250 (1.6%)(1.6%)

Design of Non-Inferiority Trials Region of non-inferiority must be defined in advance  If the upper bound 95% Confidence Interval of the difference between two treatments lies entirely below the pre-specified boundary then these treatments may be considered clinically equivalent Region of non-inferiority must be defined in advance  If the upper bound 95% Confidence Interval of the difference between two treatments lies entirely below the pre-specified boundary then these treatments may be considered clinically equivalent

Superiority vs. Equivalence Trial Design Superiority: l Hypothesis is Treatment A is better than Treatment B l Statistical testing: Prove that Tx A is not equal to Tx B (disprove Null hypothesis) Non-Inferiority (Clinical Equivalence): l Hypothesis is Tx A is at least as good as Tx B l Statistical testing: Prove that Tx A is not worse than Tx B Equivalence: l Statistical testing: Prove that Tx B is not worse (and not better) than Tx A

Relative Risk Designing a Non-inferiority Trial Standard Therapy Aggressive Therapy Non-inferiority “clinical equivalence” 420/ /2000 (21%)(20%) Superiority 800/ /4000 (20%)(16%) Equivalence 4000/ /20000 (20%)(20%) 0.8