Planning and Conducting the ISCHEMIA Trial: Setting the Record Straight

David J. Maron, MD; Robert A. Harrington, MD; Judith S. Hochman, MD

Disclosures

Circulation. 2018;138(14):1384-1386. 

In This Article

Primary End Point

When writing the original grant application, we proposed and were approved for the previous 5-component primary end point. All-cause mortality would have been a preferable primary end point, but it would have required a sample size >11 000 and was not considered feasible. Cognizant that unstable angina and heart failure are more susceptible to bias in diagnosis and reporting, we proposed in the early postaward period to change the primary end point to the 2-component end point of cardiovascular death or myocardial infarction. We were aware that it is common for event rates (and consequently statistical power) in clinical trials to be lower than initially projected. We therefore prospectively developed a contingency plan that was written into the original protocol to revert to the original 5-component end point to retain power if a sufficient number of primary end point events had not accrued by a designated time point. To mitigate potential bias, we developed rigorous mechanisms to capture all events and stringent objective criteria to define unstable angina and heart failure end points reviewed by an independent blinded adjudication committee. In 2017, before accrual of 50% of projected end point events, the NHLBI convened an independent panel that recommended reversion to the 5-component end point to achieve adequate statistical power under a range of clinically important and plausible between-group differences. The NHLBI approved the panel's recommendation, retaining the 2-component end point as the key secondary end point. No one with access to unblinded end point data participated in the decision to change the end point.

Although it is not desirable to modify fundamental design elements of a clinical trial, sometimes it is necessary to adjust to postaward realities so that the aim of the trial can be preserved. Design changes after the trial is in progress should be carried out in a transparent manner that minimizes bias and is well justified by information coming from both external data sources and within the trial. This is especially critical in the case of the primary end point: the criteria and process should be prespecified in the protocol, including the condition that it is done before unblinding of events by treatment group.

As a research community, we have made great progress with the use of clinicaltrials.gov to post an overview of design features at the start and protocols and statistical analysis plans when trial results are published. Perhaps the best time for broad comment is before the design is finalized (between award and sponsor/funding agency approval of the protocol) or when results are presented (the historical model) to aid in trial interpretation. A model has not been established for the communication of important revisions to trial design while a trial is in progress. In the era of social media with rapidly spread sound bites based on incomplete information about ongoing trials, the risk of misinformation is high and could adversely affect participant recruitment and retention, undermining the trial's integrity. Rather, the best approach may be to update clinicaltrials.gov within a few months of the change to inform the research community. We trust this perspective will provide the community with further generalizable insights into the complexities of designing and executing trials that go well beyond hypotheticals and speculations. There is no such thing as a perfect trial design that can be implemented under the budgetary, regulatory, healthcare system, and time constraints that exist in the real world, in addition to satisfying the preferences of patients and physicians. The critical test is whether the trial can provide meaningful data for patients and physicians to aid in making clinical decisions. We believe the ISCHEMIA trial meets that criterion.

processing....