Challenging a standard
The FDA had encouraged Novartis to submit the supplementary application for the HFpEF indication and even suggested some of the post-hoc analyses, the agency staff said in a briefing document for the meeting.
PARAGON-HF illustrated some of the agency’s concerns about missed opportunities in general in large research trials. Some events of interest in studies may be miscounted due to a lack of information such as a requirement for the presence of physical examination findings that are not documented in the patient’s dossier, the FDA staff said.
“We would like to consider giving ‘partial credit’ to events based on the level of evidence provided, e.g., use of an ordinal variable rather than a dichotomous ‘yes’ or ‘no,’ ” the staff said in the briefing document.
At the panel meeting, Norman Stockbridge, MD, PhD, director of the Division of Cardiology and Nephrology in the FDA’s Center for Drug Evaluation and Research, told the panelists there is no basis in law for setting a P value of .05 as the benchmark on whether to declare a trial a success or failure.
“I wanted to take a few minutes and make sure that you on the committee understands what flexibility you have in addressing the case study in question,” Dr. Stockbridge said at the start of the meeting.
He then reviewed cases where the FDA had approved claims for cardiac medicines that had not shown desired results in key tests. These include approval of enalapril for use in asymptomatic left ventricular dysfunction on the basis of the SOLVD-Prevention trial, approval of digoxin for heart failure on the basis of the DIG study, and approval of carvedilol for reduced ejection fraction following myocardial infarction on
the basis of the CAPRICORN study, Dr. Stockbridge said.
In reviewing the data for sacubitril/valsartan, FDA staff noted a similarity between investigator-reported and adjudicated results, Stockbridge said.
“This suggested that there were events that did not need all evidentiary criteria as qualified events, but likely were nonetheless,” he said. “This is an example of dichotomization of events being wasteful of information.”
Post-hoc exploratory analyses in PARAGON-HF were able to meet the commonly used standard, according to the FDA briefing document. Among the key findings of these analyses were:
- An analysis of investigator-reported events for the primary composite endpoint of total hospitalizations for heart failure (HHF) and cardiovascular (CV) death demonstrated a rate ratio (RR) of 0.84 (95% confidence interval 0.74- 0.97; P = .01).
- Investigator-reported events added 226 and 290 HHF events but decreased CV death by 56 and 58 events in the sacubitril/valsartan and valsartan arms, respectively. Hence, a net 170 and 232 events were added to the clinical endpoint committee–reported primary composite endpoint leading to a P value of .01, without a significant change in RR.
- Analysis of investigator-reported expanded primary composite endpoint events including total HHF, urgent HF visits, and CV death demonstrated a RR of 0.83 (95% CI, 0.73-0.95; P = .006. There were 136 and 173 investigator-reported urgent HF events in sacubitril/valsartan and valsartan arms, respectively.
Opening a floodgate?
Cynthia L. Chauhan, MSW, of Wichita, Kansas, who served as the consumer representative on the panel, questioned whether a decision to revisit the data on PARAGON-HF might lead drugmakers to seek to repurpose other failed trials.
“Are we opening any kind of floodgate for other researchers to go back and see this is an invitation to try to, for want of a better term, back-door their way into some approvals?” Ms. Chauhan asked.
Dr. Nissen assured her that this concern was valid and would be considered. The goal would be to allow some flexibility in cases that merit further consideration, while preventing companies from data mining until they find some evidence to support an FDA application, he said.
Re-analyzing trials “should be done carefully, conservatively, and only when it really is compelling that the public interest supports it,” Dr. Nissen stressed.
Panelists reported no conflicts of interest related to the topic of the meeting.