Study Analyzes Appropriate Use of ICDs for Primary Prevention

Abstract & Commentary

By John P. DiMarco, MD, PhD, Professor of Medicine, Division of Cardiology, University of Virginia, Charlottesville. Dr. DiMarco does research for Medtronic, is a consultant for Medtronic, Novartis, and St. Jude, and is a speaker for Boston Scientific.

Source: Al-Khatib SM, et al. Non-evidence-based ICD implantations in the United States. JAMA. 2011;305:43-49.

This paper analyzes data from the national icd registry, which tracks ICD implantations for the primary prevention of sudden death in the United States. Data reported from virtually all U.S. hospitals that implant ICDs in adults were analyzed by the authors and compared to published, evidence-based criteria for ICD implantation. Guidelines from both the ACC/AHA/HRS Device-Based Therapy writing group and the Center for Medicare & Medicaid Services (CMS) were considered. Specifically, patients were classified as receiving a non-evidence-based ICD implant if they met any of the following criteria: 1) myocardial infarction within 40 days of implantation; 2) coronary artery bypass surgery within three months before ICD implantation; 3) New York Heart Association class IV symptoms; or 4) diagnosis of congestive heart failure within less than 3 months of ICD implantation. Patients were classified as receiving an evidence-based ICD implant if none of those criteria were met. Once the data were collected, the authors compared the number, the demographics, and the clinical characteristics between the non-evidence-based and evidence-based implant groups. Implantation patterns at various hospitals, and by implanter groups, were analyzed, as were in-hospital outcomes reported in the registry. Implanting physicians were characterized as electrophysiologists, non-electrophysiologist cardiologists, thoracic surgeons, and others. Patients who received a resynchronization-therapy implant were excluded from this analysis, and only patients with primary prevention indications were included.

The final dataset analyzed included 111,707 initial primary prevention ICD implants that occurred between Jan. 1, 2006 and Jan. 30, 2009. Of these implants, 25,145 (22.5%) did not have an evidence-based criterion for implant. This included 9,257 patients in whom the implant occurred within 40 days of a myocardial infarction (36.8% of this group), 814 in patients who received their ICD within 3 months of coronary artery bypass surgery (3.2%), 3,322 in patients with New York Heart Association (NYHA) class IV heart failure (12%), and 15,604 patients with newly diagnosed (within 3 months) heart failure (62.1%). Patients could be listed in more than one group. In comparison to the 86,502 evidence-based ICD implant recipients, patients in the non-evidence-based group were significantly older and had a more comorbid disease. Non-evidence-based ICD recipients were more likely to have heart failure, atrial fibrillation, ischemic heart disease, cerebrovascular disease, chronic lung disease, diabetes, end-stage renal disease, and to receive a dual-chamber ICD. The risk of in-hospital death was 0.57% among patients who received a non-evidence-based device, compared to 0.18% among patients who received an evidence-based device. Procedural complications were more common in the non-evidence-based ICD group. The overall reported complication rate was 3.23% in the former group vs. 2.41% in the latter group. As might be expected from these differences, the median length of hospital stay was significantly longer in patients who received a non-evidence-based ICD, compared to patients who received an evidence-based ICD (3 days vs. 1 day; p < 0.001). There was a wide variation in the distribution of the percentage of non-evidence-based ICD implants among the implanting hospitals included in the survey. Some institutions had more than 30% of their implants not meeting evidence-based criteria. When the implanting physician's specialty was examined, electrophysiologists were more likely than other groups to follow evidence-based criteria in selecting patients for implants. There was no decrease in the rate of non-evidence-based implants over the time period covered in the registry.

The authors concluded that a substantial fraction of ICD implants do not meet evidence-based criteria based on data reported to the National ICD Registry. Patients who received non-evidence-based ICDs had more complications and were more likely to receive their devices at certain sites and by non-electrophysiology certified physicians. The authors encourage feedback to hospitals and providers to improve adherence to practice guidelines and eventually practice guidelines.


Why are ICDs being implanted in patients who do not meet established criteria for ICD therapy? There is almost certainly no single answer. Some implants undoubtedly are performed in knowing violation of the published guidelines. Those with experience in the field have all heard some implanters state that the current guidelines are too restrictive and that they know better than the literature who "needs" an ICD. Data from the ICD Registry, and enforcement actions by CMS and other payers, will hopefully change the practices of those who prefer to base decisions on anecdote rather than data.

The ICD Registry data presented here estimate that 22.5% of all primary prevention ICD implants are not evidence-based. Unfortunately, although this paper highlights what may be a significant problem, I don't have great confidence in the actual value. At the time of implant, an ICD Registry form is filled out for almost every ICD recipient. In my own lab, we started out with the form being completed by the EP lab procedural staff. However, when we rechecked the forms as a quality-assurance exercise, we found multiple errors. Eventually, we instituted a program where a trained nurse and /or physician had to sign off on the form before it was submitted. Even with a sophisticated person filling out the data, it's sometimes hard to say whether a recent MI had occurred, date the onset of heart failure, or assign the current NYHA class. Is a small troponin leak a true MI? Is just a history of progressive symptoms over time before a patient was evaluated for the ICD enough to date the onset of heart failure? How do you classify someone who was in Class IV heart failure 2 days ago when he or she was admitted but is now improving and may get back to a lower baseline class. Even ejections fractions can have significant variability over time. Which value should you pick — the most recent or the one that meets criteria?

This paper focuses attention on what may be a significant quality problem in cardiology practice. Further studies that look more closely at the source data will be required before we know the true magnitude of this problem.