very carefully,” he said. “It’s clearly a very extensive, detailed, effortful report … so there will be a lot that we want to think about.”
Discussion following the presentations focused on two main topics. The first was the definition of the term risk biomarker, as used in the two presentations from representatives of CFSAN and employed in its charge to the committee. As previously noted, the committee did not use this term, due to its potential for confusion with the precisely defined term risk factor. Elizabeth Yetley, a consultant to the National Institutes of Health (NIH) Office of Dietary Supplements and to the committee, asked whether CFSAN considered a risk biomarker to be a type of surrogate disease biomarker. Speaker Paula Trumbo, supervisory biologist at CFSAN, replied, “risk biomarkers can be surrogate endpoints, but not all risk biomarkers are surrogate endpoints.”
This exchange prompted Thomas Fleming to ask whether CFSAN’s goal was to be able to determine whether a given biomarker is a reliable way to assess the level of risk a patient has for a certain event, or whether they seek biomarkers that can represent reliably whether a treatment will alter a patient’s risk for such an event, and can therefore serve as a surrogate endpoint. “The former [case] simply requires a biomarker to be a correlate,” he said, but the latter requires the effect on the biomarker to reliably predict the full effect on the true clinical endpoint. Dr. Fleming further explored this distinction in his presentation, which is summarized in Chapter 7 in this volume.
A second topic of discussion focused on whether a surrogate endpoint must have biological plausibility: that is, that the biological connection between the biomarker and the process it represents is known, if not fully elucidated. Stephen Williams noted that, in an article he and John A. Wagner had written recently reviewing the history of surrogate end-points, they had argued against requiring biological plausibility for surrogate endpoints (Lathia et al., 2009). They reasoned that highly plausible biomarkers had failed as surrogate endpoints (for example, ventricular arrhythmia, as previously discussed by Ronald Krauss), and also that this criterion would exclude some types of biomarkers (for example, microRNA) for which plausibility might be difficult or impossible to determine.
However, Dr. Williams added, an editorial in the same issue of the journal in which their article was published disputed their conclusion (Gobburu, 2009). Therefore, he asked the FDA representatives how they viewed the role of biological plausibility as a criterion for selecting surrogate endpoints. The committee’s report did not specifically recommend that surrogate endpoints have biological plausibility, Dr. Williams noted.
“From my point of view, plausibility is always helpful,” Dr. Walton stated. “For one, it provides enough enthusiasm to expend the significant