When does a trial fail?

In a clinical research trial, a clinical endpoint generally refers to occurrence of a disease, symptom, sign or laboratory abnormality that constitutes one of the target outcomes of the trial.

These endpoints have to defined before the trial is started (Check out Clinical trial.gov for examples and see if there is any manipulation of these..yes it is all trackable!) and are either primary or secondary. 

If the trial does not meet the primary endpoint it is a failure even if there are positive secondary endpoints. A second trial using the successful (secondary) endpoint as a the primary endpoint, which would have to be passed, needs to be done for it to be deemed successful.

This is to stop pharma cherry-picking and spinning the positive results.

This is especially the case because if you look at loads of outcomes by chance, one will be positive.

Is this a silly rule that causes unnecessary delays or is it there to protect you against pharma spin?

About the author



  • If the secondary endpoints data is strong enough and clinical relevant, why run a new trial? For the benefit of who? It must be in the interest of patients and neurologists to get more possible treatment options with different MOA, and it is possible to use the label to describe the indication.

  • As crazy as it sounds, I tend to put huge value in the findings outside of what was being investigated. Often there is less bias. It is the value of the blind walk. As my first boss always said, "tell me what you want to prove, and I will find the statistic to do so." That said, often these secondary findings happen as the result of a built in but unrecognized structural bias. So the findings should not be viewed as a result but rather as a sign post directing further study. The current system is as it should be.

  • It's necessary. You need to run well designed trials, otherwise all sorts of things would get approved that were neither safe nor effective.

  • If it's a pre-defined endpoint I don't really see your point. If it's datamining without a pre-defined endpoint, I see your point and agree.

By MouseDoctor



Recent Posts

Recent Comments