Should ICDs in Primary Prevention Go the Way of Aspirin?

John M. Mandrola, MD


May 19, 2019

Times change. Therapies once proven beneficial may not remain so forever.

Aspirin used to be effective for the primary prevention of cardiac events. Three large trials published last year showed that this is no longer the case.[1,2,3] Aspirin's effects in the body have not changed. What's changed is the population and the underlying medical therapy.

Two studies presented at the 2019 Heart Rhythm Scientific Sessions (HRS), together with the neutral results of the DANISH trial,[4] make me wonder whether implantable cardioverter defibrillators (ICDs) for primary prevention may suffer the same fate as aspirin.

I will make the case that we need a redo of the primary prevention ICD trials. First, a brief review of the DANISH trial, then the two studies from HRS.

Recall that DANISH randomized 1116 patients with nonischemic heart failure to an ICD or usual care. The trial's key feature was a test of ICD therapy in the context of contemporary therapy, including cardiac resynchronization therapy (CRT) which almost 60% of the participants had. Over nearly 6 years of follow-up, implanting an ICD did not result in a statistically significant reduction in the primary endpoint of death.

An outside observer might look at these results and think: given the known harms and costs of an ICD, we should de-adopt it for this indication. That is not what happened.

Instead, the electrophysiology community pointed to the DANISH subgroup analysis showing that younger patients did better with an ICD, and multiple groups published meta-analyses[5,6,7,8] showing that if you combine the old positive trials with DANISH, there was still benefit for the ICD.

The problem with these responses is that subgroup analyses of a trial with a neutral primary endpoint are fraught with problems, and combining old and new studies in a meta-analysis is akin to comparing apples with oranges.  

HRS Presentations Question ICD Net Benefit

In a late-breaking trial session at HRS, Valentina Kutyifa, MD, PhD, from the University of Rochester, presented comparative data from old vs new trials from the MADIT series  on the use of ICD in patients with heart failure.

The impetus for this research came from a meta-analysis of 12 heart failure trials spanning a 20-year period that showed a statistically significant 44% decline in the rate of sudden death from old to new trials.[9]

The MADIT post hoc study included slightly more than 4000 patients. The "old" study, the control arm of sorts, was MADIT-II, which enrolled patients between 1997 and 2001.[10] The comparator group, or the newer trials included MADIT-CRT[11] (enrolled 2004 to 2008) and MADIT-RIT[12] (enrolled 2009 to 2011).

The 3-year cumulative probability of ventricular arrhythmia and death declined across the three studies. Relative to the older trial, the rate of ventricular tachycardia (VT)/ventricular fibrillation (VF) and death decreased by roughly 30% and 60%, respectively in the newer trials.

The authors did a number of other statistical analyses, but the take-home was that the risk of death and VT/VF has declined over time. Noting these temporal changes, the authors called for "further studies on contemporary risk stratification strategies in low EF patients that will account for the use of CRT and improved medical management."

The medical-speak I put in italics disguises the real meaning: Medical therapy is better, CRT helps a lot, rates of VT and VF have declined, and, thus, the incremental benefit of the ICD may no longer be significant—as in DANISH.

Second Paper: Harms

ICDs come with harms—which act as a drag on net benefits. 

Marit van Barreveld is a Dutch PhD student who leads the DO-IT registry,[13] a nationwide prospective cohort study enrolling patients who have had ICD implants for primary prevention in 28 centers in the Netherlands.

In a poster session, van Barreveld presented the rate of major complications at 2 years of follow-up from more than 1400 patients who received their ICD between 2014 and 2016.  Their results are stunning.

They observed ICD-related complications in 195 (13.5%) patients, including 113 (7.8%) with major complications. Nearly half the complications occurred after 30 days; 106 (7.3%) patients required surgery to deal with a complication.  Five patients died of ICD-related complications.


The sobering Dutch data on complications is similar to Danish registry results.[14] Consider also that, unlike in the United States, both Denmark and the Netherlands restrict ICD implants to small numbers of high-volume centers. These high rates of complications, therefore, could be considered a best-case scenario and likely underestimate the rate of ICD complications in the United States, where many devices are implanted in low-volume centers.

ICD proponents might push back and cite a systematic review of ICD complications that found a lower 9% rate of complications in randomized, controlled trials (RCTs) and a 3% rate in the National Cardiovascular Data Registry (NCDR).[15]  Two reasons you should favor the Dutch and Danish data are that RCTs typically underestimate harms, in part because of the careful selection of "healthier" patients, and the NCDR describes only complications around the time of the implant.

The older ICD trials produced positive results because trialists enrolled patients with a high risk of death from ventricular arrhythmia but few other comorbid conditions. Most of these patients had ischemic cardiomyopathy due to completed infarcts—a substrate particularly prone to ventricular arrhythmia and sudden death.

Trial discussant Leslie Saxon, MD, from the University of Southern California, commented that the substrate for heart failure has changed over time. Better medical therapy and acute percutaneous coronay intervention has reduced the pool of patients with ischemic heart failure, and, now, heart failure is more often nonischemic—a substrate less prone to ventricular arrhythmia and sudden death.

Much like aspirin, ICDs are now being used in a different population: Better medical therapy, CRT, and temporal trends have resulted in a reduced burden of ventricular arrhythmias. And ICDs treat only ventricular arrhythmias.

At the same time, real-world data from reliable, contemporary sources suggest significant harms from the device. As the population ages and patients with heart failure live with more comorbid conditions, ICD complication rates will likely increase.

For all these reasons, I agree with a recent editorial from the University of Chicago and Oregon Health and Science University: Evidence, they write, should come with an expiration date.[16]

We need to rethink the use of primary prevention ICDs.

Electrophysiologists exist to help our patients. We embraced ICDs when the old trials showed benefit. We said, then, that this was good for our patients. We were right.

What would be good for our patients now would be to reaffirm this with new evidence.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.