[Skip to Navigation]
Comment & Response
February 20, 2019

Evaluating Treatment Effect of Transcatheter Interatrial Shunt Device Using Heart Failure Event Rates—Reply

Author Affiliations
  • 1Division of Cardiology, Department of Medicine, Northwestern University Feinberg School of Medicine, Chicago, Illinois
  • 2Associate Editor, JAMA Cardiology
  • 3Evanston Hospital, NorthShore University Health System, Evanston, Illinois
  • 4Boston University School of Public Health, Boston, Massachusetts
JAMA Cardiol. 2019;4(3):299-300. doi:10.1001/jamacardio.2019.0004

In Reply We thank McCaw et al for their careful review of our article1 describing the 12-month outcomes for patients enrolled in the Reduce Elevated Left Atrial Pressure in Patients With Heart Failure (REDUCE LAP-HF I) trial, which randomized patients with heart failure (HF) and left ventricular ejection fraction of 40% or greater to an interatrial shunt device (IASD) or a sham control procedure. The authors of the letter have 2 concerns. First, they are concerned that the Kaplan-Meier curves and P values for these curves (both for major adverse cardiovascular, cerebral, and renal [MACCRE] and HF events) include all events (even the recurrent events) and, if so, that the figures and analysis approach are incorrect. However, the figures are indeed only plotting time to first event using the usual Kaplan-Meier methodology, so the figures and P values are themselves correct. The authors’ concern stems from the fact that the number at risk (shown below the curves) does not decrease right after an event. This has to do with the way the independent data coordinating center (Baim Institute for Clinical Research, Boston, Massachusetts), which performed the statistical analyses for the REDUCE LAP-HF I trial, calculates numbers at risk, which is different than the typical reporting of numbers at risk for Kaplan-Meier curves. As an example, in Figure 2,1 there are 2 events for the IASD treatment group between 180 and 270 days. The way that this is often handled is that at 180 days, there would be 21 at risk, but at 270 days, there would be 19 at risk (ie, 21 minus the 2 events that occurred between days 180 and 270). Our analysis is based on the reasoning that since the 2 events occurred in the period between 180 and 270 days, the number at risk at some point during that period is 21 and so is how we represented it at the end of that period (at day 270). In other words, we showed the number who were at risk for at least some portion of the period ending at day 270 (the same applies to other periods as well), whereas the number shown is typically the number at risk for the beginning of the period.

Add or change institution