An end point that involves objectively determinable events is the best way to assess cardiovascular outcomes for obesity drugs, committee members of the Endocrinologic and Metabolic Drugs Advisory Committee agreed at the panel’s March 29 meeting.
But if a cardiovascular outcomes trial (CVOT) has two analysis time points – one prior to approval and one after approval – an end point that includes more subjectively determined events could be used in the early assessment, the two statisticians on the panel agreed.
The assessment of a suitable end point came during a full day of discussion about when and how CV outcome trials should be conducted for weight-loss drugs.
On the previous day, the panel heard presentations from the Food and Drug Administration and a series of guest speakers on the disease, current therapies, past trials, and agency guidance for assessing the CV safety of diabetes drugs. The diabetes guidance calls for a meta-analysis of phase II and III results to look at CV risk, with an outcomes trial required before approval, after approval, or not at all, depending on what the meta-analysis finds.
On the lone voting question the FDA put to the panel, members agreed 17-6 that sponsors should conduct a meta-analysis or CVOT for drugs that do not have a theoretic risk or signal for CV harm. In explaining their yes votes, 13 members said that a two-stage assessment process would be acceptable. This would involve one assessment before approval and the second post approval.
FDA officials had already assured panel members that outcomes trials must be conducted pre-approval for obesity drugs that do have a CV signal. The issue of the appropriate end point arose in the context of design parameters for those studies.
The panel was asked whether the primary end point should be major adverse cardiovascular events (MACE) – identified as CV death, nonfatal myocardial infarction, and nonfatal stroke – or MACE-Plus, which includes those events and others, such as hospitalized unstable angina and emergency coronary revascularization.
Noise a Problem With More Subjective CV Events
Expanding the MACE composite end point could result in more CV events, facilitating the detection of a safety signal. Panel members, however, were concerned about the statistical "noise" created by including events that are determined by a physician’s judgment and are not objectively measured.
Gaining the additional events may not be worth it, Dr. William O. Cooper of Vanderbilt University in Nashville, Tenn., pointed out. Hospitalization for angina could depend on whether the patient is very vocal about his or her condition or has access to health care or who the treating physician is. The additional events "are much more subjective, and I think we’d have difficulty both in terms of adjudication and what they might mean," he said.
Enrichment of a study to find a signal is desirable, noted Dr. Sanjay Kaul of Cedars-Sinai Medical Center, Los Angeles, but "minimizing or reducing bias is equally desirable, if not more so. And if you add end points that are somewhat subjective ... you’re adding noise and thereby biasing the results toward the null, which will drive you to a false sense of reassurance about ruling out cardiovascular risk."
Limiting the end point to MACE, whose ascertainment and adjudication is less subjective, "is the way to go," he said. If more events are desired, he suggested using objective end points that can be standardized and adjudicated and are less prone to bias, such as ischemia-driven revascularization.
"If there are other end points that might be impacted by the test drug, it might make sense to assess them in an additional analysis, but not as part of an expanded MACE," said Dr. John H. Alexander of Duke University in Durham, N.C. "As a general rule, I would stick to MACE as our main end point."
Two-Step Analysis Can Use Two End Points
The panel previously discussed the two-stage assessment process for drugs with a CV signal. Dr. Allison B. Goldfine of the Joslin Diabetes Center in Boston suggested that if this option were used, the end point could be different in the preapproval and postapproval analyses.
"When you’re setting up your preapproval [study], you really want the larger safety net and the larger number of observations of events to actually help inform about potential risk," she said. "Within the preapproval window, I would be more inclined to allow documented ischemia or ischemia-driven revascularization" as part of a MACE-Plus end point.
Expanding MACE to include events along the same pathophysiological pathways would facilitate acquiring the number of events needed to assess safety, Dr. Goldfine noted. There would be the potential for noise, but investigators would still have a good idea of the safety signal. The analysis after approval would use "only the harder MACE," she suggested, asking the panel’s statisticians for input.