Diagnostic-based enrichment strategies for clinical trial success and demonstration of comparative effectiveness: are these two goals the same?
*Robert Temple, FDA 


First consider trials intended to show a difference between 2 treatments: i.e., drug vs placebo or vs an active control. Enrichments intended to decrease heterogeneity and noise, increase study power for any kind of difference-showing trial. Prognostic enrichment, finding patients with a high rate of events or who are very symptomatic (leaving room for improvement) will increase the chance of showing a difference between treatments if there is one, whether the comparator is placebo or active control. It is predictive enrichment, however, that leads to the most dramatic possibilities, some obvious, some not so obvious.

If a drug treats only a subset of a disease (CF, hep C) a trial in the subset (CF type 1, only 4% of CF) is much more likely to succeed; indeed, with only 4% of CF patients, it would be hard to show anything in an unselected population. In comparative studies, a drug that treats a broad range of subtypes (all NSCLC) with a given effect will be superior to a drug that treats only a subset (EGFR & NSCLC), perhaps making the case that the drug with broader effect is easier to use.

A comparative effect of great interest is studying patients not responding (empirically) to a particular therapy. The study in non-responders will randomize patients to the new drug and the failed drug, showing, one hopes, that those patients respond better to the new drug.