Monte Minute‎ > ‎Monte Minute‎ > ‎

Non-Inferiority Trials - 9/17/2012

posted Sep 17, 2012, 3:35 PM by Rohit Das   [ updated Dec 27, 2012, 6:54 AM by Purnema Madahar ]

Ahhhhh – IPRO curtailed one of our clinical conferences today, which was going to serve as the topic of today’s daily. Instead, I’m going to write briefly about a topic that is impossible to explain via a blog, but wish to throw the concept into your head – noninferiority trials.

This topic came up in resident journal club today, when we were talking about a study in 2006 that compared Lamivudine and Entecavir in the treatment of chronic Hepatitis B. As part of the analysis, the investigators first did a noninferiority analysis and subsequently did a superiority analysis. Both turned out favorable for the powers that be.

Noninferiority trials are becoming commonplace in the literature. As medicine is evolving and therapeutics are becoming increasingly effective, it is difficult to market drugs as being MORE EFFECTIVE than current therapy – such studies require a higher power, and therefore, a ton of people and longer follow-up. So, drugs are more often being marketed as being better than current therapy in other ways – less frequent dosing, less frequently required monitoring, less cost, less adverse effects…etc.

The goal of a noninferiority analysis is to simply prove that the new drug is “not significantly worse” (meaning not worse enough to a degree of clinical significance) than the active control. However, there are several aspects of this premise that must be taken into account when critiquing this kind of analysis:

  • The concept of “assay sensitivity” dictates that a trial has the ability to differentiate between a new, effective intervention versus an ineffective intervention. This is fairly easy to guarantee when comparing to placebo. However, in a noninferiority trial, one must assume that the active control is indeed “effective,” which isn’t always a correct assumption.
  • On a similar note, the study population in a given noninferiority trial may not be the same as the one initially studied when the active control was deemed to be effective. Furthermore, study design will differ with different investigators. The premise that the efficacy of an active control is maintained through different trials is called the “constancy” assumption.
  • Finally, I hope everyone is able to recall the idea of “intention to treat (ITT).” In efficacy analyses, ITT biases towards the null, and positive findings are more significant. However, in a noninferiority study, biasing towards the null is actually FAVORABLE, and the motivation to keep patients in their respective arms, from a statistical standpoint, is removed. Nevertheless, ITT remains a part of noninferiority analyses, mainly for the purposes of maintaining randomization and minimizing confounders.

Now that I got the concept in your head – solidify it. I’ve attached a nice review article on the subject. With efficacy studies becoming more difficult to conduct in the face of already existing good treatments, noninferiority studies are becoming more common, and it’s important to understand them.


More clinical stuff tomorrow...

Through the Looking Glass: Understanding Non-Inferiority
Schumi et. al., Trials 2011, Volume 12:106