John Mandrola, MD


August 28, 2016

The European Society of Cardiology (ESC) 2016 Congress started with a shocker. The first trial presented in the first hot-line session will change the way cardiologists think about implantable-cardioverter-defibrillator (ICD) use in patients with heart failure.

The Danish ICD Study in Patients With Dilated Cardiomyopathy[1] (DANISH) delivers two important lessons: the specific one deals with ICD use in patients with heart failure not due to ischemic causes, but another important message relates to medical evidence and its translation into practice.

A nonpacing ICD has but one purpose—to prevent sudden death from arrhythmia. In selected patients with heart failure due to ischemic disease, implanting a prophylactic ICD reduces arrhythmic death, which then improves survival.

The evidence for prophylactic ICD implantation in patients with heart failure not due to ischemic heart disease (nonischemic cardiomyopathy or NICM) is far less clear—dubious even. And, if you add shaky evidence to the background of improving heart-failure therapy, you will not be surprised by the results of the investigator-initiated randomized controlled trial.

DANISH was conducted at all centers in Denmark at which ICDs were implanted. Industry partially supported the trial but had no role in the design, conduct, data analysis, and manuscript writing of the study.

DANISH enrolled 1116 patients with nonischemic systolic heart failure and an elevated N-terminal probrain natriuretic peptide (NT-proBNP); 556 patients were assigned to receive an ICD and 560 patients were assigned to receive usual therapy. CRT was used in 58% of both groups. Investigators recorded the qualifying LVEF and NT-proBNP after patients had been taking maximal doses of guideline-directed medications.

Background heart-failure therapy was robust. More than 90% of both groups took beta-blockers, ACE inhibitors, or angiotensin-receptor blockers (ARBs), and nearly 60% of both groups took mineralocorticoid inhibitors. Investigators implanted CRT in 93% of enrolled patients who had left bundle branch block and QRS ≥150 ms.

DANISH was a long trial. It began in 2008 and enrolled patients until 2014. Median follow-up was 67 months. The results were clear:

  • For the primary outcome, death from any cause, 120 patients (21.6%) died in the ICD group and 131 patients (23.4%) died in the control group. This was not significantly different.

  • Death from cardiovascular causes occurred in 77 patients (13.8%) in the ICD arm and in 95 patients (17.0%) in the control group. Again, not statistically significant.

  • Sudden cardiac death occurred in 24 patients (4.3%) in the ICD group and in 46 patients (8.2%) in the control group. This 3.9% percent difference did reach statistical significance.

  • Analysis of the results by subgroups mostly followed the trial's main results—with one exception. Fewer patients under the age of 59 died in the ICD group. The hazard ratio was 0.51 (95% CI 0.29–0.92).

  • Safety events were not different in either group. Device infections, a prespecified safety outcome, occurred in 4.9% and 3.6% of the ICD and control groups, respectively.


The decidedly negative results of DANISH should not surprise us. Here are the three main reasons:

First is that relative to ischemic cardiomyopathy, patients with NICM are less prone to arrhythmia. Dr John McMurray (University of Glasgow, Scotland), writing in an accompanying editorial, added that patients with NICM also have lower mortality from all causes.[2] Competing causes of death are critical in the ICD decision. Noncardiac causes of death accounted for 31% of the deaths in DANISH. This, the authors write, "is not surprising in an elderly population, and it highlights the importance of carefully selecting patients for ICD implantation."

The second reason DANISH was negative is improved background medical treatment for heart failure. For instance, in the SCD-HEFT trial,[3] the RCT that led to class I guideline recommendations for ICDs in patients with heart failure, 69% of enrolled patients took beta-blockers and only 20% took mineralocorticoid blockers. Contrast that with the DANISH trial, in which virtually all patients took beta-blockers and nearly 60% took mineralocorticoid blockers. High use of beta-blockers has been offered as a reason the DEFINITE trial, another randomized clinical trial comparing prophylactic ICD therapy in patients with NICM, did not find a statistical benefit for ICD therapy.[4]

Use of CRT is the third reason ICDs proved nonbeneficial in DANISH. This gets to the basic rule of diminishing returns: the lower the event rate, the harder it is for any additional therapy to improve outcomes. It is the reason we don't implant ICDs in lower-risk patients—because the harms of the device outweigh its benefits.

DANISH should increase the use of CRT-P (pacing) devices relative to CRT-D (defibrillator) devices. Investigators in DANISH looked for a relative effect of an ICD between patients who received CRT and those who did not. They found none. They cited the CERTITUDE observational study of CRT-P vs CRT-D, which suggested excess mortality in patients with CRT-P was mostly due to non–sudden death.[5]

Those who favor CRT-D over CRT-P may be surprised by comments from the DANISH authors in regard to the landmark COMPANION trial.[6] COMPANION, a trial of either CRT-P or CRT-D against medical therapy in patients with heart failure, did not perform statistical tests on the overall difference in mortality between the CRT-P and CRT-D groups. Years later, independent authors looked at this difference and calculated a crude P value of 0.12 for overall mortality in CRT-D vs CRT-P groups.[7,8]

The positive subgroup result (age-related ICD benefit) in the setting of a negative trial deserves comment. In the press conference, lead author Dr Lars Køber (Copenhagen University Hospital of Rigshospitalet, Denmark) was asked about this statistical issue. He replied "[You] don't believe in the subgroup [analysis] in principle. But guidelines are based on subgroup analyses and meta-analysis of subgroups. DANISH does not stand alone. We now know that the benefit [of ICDs] might be lower in NICM. I'm not saying the subgroup result for this trial is true; I'm just taking the totality of the evidence." A recent update of the European heart-failure guidelines gives a nod to the age issue. "Patients with serious comorbidities who are unlikely to survive substantially more than 1 year are unlikely to obtain substantial benefit from an ICD."[9] One take of DANISH will be that it shows the importance of better patient selection. That's true, but the positive subgroup analysis should not overshadow the overall results of this trial.

Finally, DANISH forces a rethink of guidelines and how we translate evidence into practice. Before DANISH, ICD use in patients with NICM rested on less than the highest level of evidence. DANISH represents the highest level of evidence. And it was negative. As was DEFINITE, another pure randomized clinical trial in patients with NICM. What's more, editorialist McMurray made a stinging point when he wrote that DANISH "probably represents the most optimistic estimate of ICD therapy in patients with nonischemic heart failure who receive evidence-based treatment."

Incorporating DANISH into practice exposes the problem of reliance on static guidelines. The practice of medicine changes faster than ever. DANISH not only informs the narrow decision to implant ICDs in patients with heart failure, it should also intensify our focus on the quality of evidence we use.

I'm not suggesting this trial should completely stop the use of ICDs in patients with nonischemic heart failure. Rather, it reinforces a basic principle: our patients need doctors, not installers of devices.


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.
Post as: