Expect Intellectual Fireworks At FDA Advisory Panel On Ezetimibe

You can expect a lot of high-powered intellectual fireworks at next Monday’s FDA advisory panel considering Merck’s application to upgrade the indications for Vytorin and Zetia. Both brands contain the  cholesterol-lowering drug ezetimibe, which was studied exhaustively– in many senses of the word– in the IMPROVE-IT trial.

The discussion will likely focus on what to do about missing data and how to use subgroup analyses. In the end the panel will likely support the expanded label, though it will also make clear that the clinical benefits of adding ezetimibe to statins are modest at best and perhaps only relevant to a small subset of patients.

The advisory panel participants include well-known and outspoken cardiologists and statisticians including Sanjay Kaul, Milton Packer, and Thomas Fleming. Presenting the IMPROVE-IT data for Merck will be Chris Cannon and Eugene Braunwald, a co-chair of the trial. (The other co-chair, Rob Califf, is not scheduled to speak. Califf has been nominated to be the next FDA Commissioner.)

The existing labels for Vytorin and Zetia state that the drugs have not been shown to improve cardiovascular morbidity or mortality. The proposed new label would state that the drugs are indicated for the reduction of cardiovascular events (cardiovascular death, nonfatal myocardial infarction, nonfatal stroke, hospitalization for unstable angina, or need for revascularization) in patients with coronary heart disease (CHD). With the exception of statins no other cholesterol drugs have received this indication.

The FDA review largely supports the main findings of the previously published results from IMPROVE-IT. In the 18,1444 patients with acute coronary syndrome randomized in the trial, cardiovascular events were reduced from an event rate of 30.2% in the simvastatin arm to 28.4% in the ezetimibe and simvastatin arm. In the main intent-to-treat analysis this worked out to a 6.4% reduction in relative risk (HR 0.94, CI 0.89-0.99; p=0.016). The FDA clinical reviewer calculated that 50 people would need to take ezetimibe in addition to simvastatin for 7 years in order to avoid 1 primary endpoint event.

Probably the most difficult issue before for the committee will be how to handle the important  subgroup analyses from the trial. As the IMPROVE-IT investigators reported at the ESC meeting earlier this year, the beneficial effects of ezetimibe occurred almost exclusively in  the 27% of patients in the trial who had diabetes. The FDA analyses also note that the beneficial effects were also only apparent in the 15% of the study population who were at least 75 years old.

Here’s how one of Monday’s panel members, Sanjay Kaul, commented about the diabetes finding during the ESC meeting: “The clinically relevant and statistically significant findings in the diabetic subset help optimize the use of ezetimibe in clinical practice. Given these results, it would be difficult to support its use in non-diabetics, nearly three-fourths of the overall cohort in IMPROVE-IT.”

“There was,” summarized one of the FDA reviewers, “no effect seen among the under 75, non-diabetic subgroup, which were a majority of the study population.”

Another difficult issue for the reviewers will be what to do about missing data, since 11% of randomized patients had missing data for the primary endpoint.

Solved: The Mystery Of The Missing Simultaneous Publication

The FDA documents shed light on a lingering mystery. At the time of the original presentation of IMPROVE-IT at the American Heart Association meeting in November 2014, many observers wondered about the absence of a simultaneous publication. The FDA documents offer a look behind the sausage-making process of clinical trials and reveal that only weeks prior to the AHA the trial investigators ran into some tricky problems:

In October 2014, in connection with study close out and prior to the planned database lock, the Applicant identified reports of non-cardiovascular hospitalizations without corresponding SAEs reported in the clinical database. This issue was identified only weeks prior to scheduled database lock (October 21, 2014). According to a statistical analysis plan addendum dated October 21, 2014, it was found late in the process of data cleaning “…that the hospitalizations previously identified as not potentially an endpoint event, AE or SAE of special interest or cancer associated were being under reported as SAE’s…. The time required to complete this additional cleaning requirement was incompatible with locking the data base on October 21th as planned. All other data cleaning criteria for efficacy endpoints and predefined safety endpoints have been fully met and CEC adjudication has been completed. Importantly, a commitment had been made to present the trial results at the scientific sessions at the AHA [American Heart Association] on November 17th, 2014.”

The sponsor decided to perform two separate locks of the database instead of one. Prior to the first data base lock, it was decided that the final analysis would be performed on these data, supporting both the presentation of the trial results at the AHA and for preparation of the manuscript. Subsequently, the hospitalizations/SAEs would be reconciled, and any additional clinical events, protocol-defined safety endpoints, or other safety events of concern would be submitted for review by either the CEC or the SAE safety desk. Any such events were to be flagged as “post-study events” that would be excluded from the final analysis.

The second lock occurred on January 23, 2015, after the reconciliation was completed. The second lock was used to support safety reporting (except for the selected safety events of interest).

The FDA reviewer commented: “Locking a clinical trial database with knowledge of incomplete data cleaning is highly unusual.” But a detailed analysis by the FDA suggests that this unusual sequence of events did not have an important impact on the outcome of the trial. But I suspect this probably explains the unusually long delay between the initial presentation of the trial in November 2014 and the final publication in June of this year.