COMMENTARY

Just Say No to the ICD in Patients With Kidney Disease

John M. Mandrola, MD

Disclosures

June 17, 2019

Nephrologists help patients in two main ways. One is with their prowess in internal medicine; kidney doctors impress me as supreme clinicians. The other is obvious: renal replacement therapy.

A bold and provocative randomized controlled trial (RCT) from Dutch investigators suggests a third way nephrologists can help: Don't let their patients near an electrophysiology device lab.

The ICD2 trial[1] is bold because it is the first RCT to test an implantable cardioverter defibrillator (ICD)  in patients with end-stage renal disease (ESRD). All previous ICD trials excluded patients on dialysis and enrolled few patients with chronic kidney disease. The ICD2 trial is provocative because it found no signal of benefit from the device—even though enrolled patients were those most likely to benefit from sudden death prevention.

The Study

The investigators asked a simple question: will a primary-prevention ICD benefit patients who are on dialysis? The primary endpoint was sudden death; overall death and ICD-related complications were secondary endpoints.

Many discussants on social media criticized the authors' decision to include patients with a left ventricular ejection fraction (LVEF) of 35% or greater. I will return to that point later.

Three background issues concerning kidney disease require comment.

  • Millions of people worldwide live with ESRD. Since mortality rates in these patients are more than 10-fold higher than those in the general population, any therapy that can help would have a huge impact.[2]

  • Death from cardiovascular causes drives much of the higher mortality. The annual rate of sudden cardiac death (SCD) in hemodialysis patients (5% to 7%) is higher than in patients with heart failure (4%) or nondialysis CKD (1.5% to 2.7%).[3]

  • Kidney failure creates special challenges for implanting an ICD. Platelet dysfunction increases the risk of bleeding; electrolyte shifts predispose to both bradyarrhythmias and tachyarrhythmias, as well as pacing and defibrillation issues; hemodialysis access precludes access on that side of the body; and, for many reasons, the rate of bacterial infection is high.[4]

The ICD2 authors planned to enroll approximately 200 patients on renal replacement therapy, but the trial was stopped early because of futility.

Over a nearly 7-year follow-up in which all patients were accounted for, SCD occurred in 11 of 97 patients (11.3%) in the ICD group and 8 of 91 (8.8%) patients in the control group. The hazard ratio (HR) of 1.32 had wide confidence intervals (CIs; 0.53 to 3.29) and a P value of .55.

For overall mortality, 52 patients in the ICD group and 47 patients in the control group died (HR, 1.02; 95% CI, 0.69 to 1.52; P = .92). The null mortality results were remarkable because nearly 14% of patients in the ICD arm received appropriate therapy for ventricular tachyarrhythmia.

ICD complications were substantial: 25 device-related adverse events occurred in 22 of 80 patients who actually received an ICD.

Comments

Let's first put these findings in the context of current evidence and expectations.

ICD practice today strictly (and I believe correctly) adheres to trial inclusion and exclusion criteria. For instance, US clinicians may not bill government payers for an ICD unless the patient was similar to those enrolled in the supporting trials. It follows, then, that because patients on dialysis were excluded from the primary or secondary prevention trials, there is no class I indication for an ICD in a patient with advanced kidney disease—no matter the ejection fraction or indication.

This notion transcends regulation; based on the pathophysiology of ESRD, not extrapolating ICD trial evidence to these patients makes clinical sense, and what little trial-level evidence there is does not support use of the ICD.

The strongest evidence against the use of an ICD in patients with even modest degrees of kidney disease comes from a patient-level meta-analysis of three pivotal ICD trials (MADIT, MADIT-2, and SCD-HeFT).[5] The authors took advantage of the fact that 36% of the nearly 2900 patients in these three trials had a estimated glomerular filtration rate (eGFR) less than 60 mL/min/1.73 m2. They found strong evidence that the survival benefit of ICDs turned on GFR. Namely, the ICD was associated with survival benefit for patients with a normal eGFR (adjusted HR, 0.49; 95% credible interval [CrI], 0.24 to 0.95), but not for patients with eGFR less than 60 mL/min/1.73 m2 (adjusted HR, 0.80; 95% CrI, 0.40 to 1.53). A contemporary matched cohort study also reported no ICD benefit among patents with ESRD.[6]

Now to the ICD2 authors' decision to include only patients with an LVEF of 35% or greater. I believe this was a wise choice, one that greatly enhances the trial's external validity.

Patients on dialysis are different from the typical patient with systolic heart failure—whose reduced ejection fraction acts as a large contributor to SCD risk. In patients with ESRD, many non–ejection fraction-related factors drive the risk for SCD. These include uremia, anemia, left ventricular hypertrophy and fibrosis, autonomic dysregulation, electrolyte fluxes, fluid shifts, and dialysis-induced hemodynamic stress. It belies common sense to use LVEF alone in this milieu.

The authors made clear that they deliberately chose a group of patients "who might benefit from ICD therapy and receive a meaningful extension of their life span." They noted that doctors were already reluctant to refer vulnerable patients to the trial and that "implanting a device in patients with significant comorbidity (heart failure and ESRD) was considered unfeasible and unethical because no pilot data were available."

In other words: patients on dialysis with low ejection fraction were simply too sick to benefit from an ICD. The landmark SCD-HeFT trial confirms this view.[7] In SCD-HeFT, a trial that enrolled young mostly male patients with minimal comorbidities, ICD benefit took almost 2 years to accrue. I reached out to Canadian nephrologist Swapnil Hiremath, MD, from Ontario, and he confirmed that a dialysis patient with LVEF of 25% will not live long enough to benefit. Indeed, patients with low EF account for only 10% to 15% of dialysis patients.[8]

Another important finding from ICD2 was the harms from putting hardware in these vulnerable patients. Despite the expertise of a single high-volume academic center, one in four patients experienced a complication.

Some electrophysiologists might argue that the now available subcutaneous ICD, which has no intravascular lead, will reduce complications. The flaw here is that the subcutaneous device has essentially no pacing support and patients on dialysis are prone to bradyarrhythmia.

Conclusion

ICD2 is an important study. We now have evidence where there was none. Knowing what does not work helps both patients and clinicians alike.

If we want to help our patients with an advanced kidney disease, the answer is clear: do not put ICDs in them.
 

Comments

3090D553-9492-4563-8681-AD288FA52ACE
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:

processing....