Implantable defibrillators : complications under-estimated and benefits over-estimated

II - Evidence of efficacy of ICDs

In a systematic review (1) of adult patients with LV systolic dysfunction (regardless of whether the patients have had heart failure symptoms), 86% of whom had New York Heart Association class II or III symptoms, ICDs reduced all-cause mortality by 20% (95% CI, 10% to 29%) in RCTs largely because of a 54% relative reduction of sudden death and by 46% (CI, 32% to 57%) in the observational studies. Thus the efficacy of ICD therapy is definitely proven. The resulting years of life gained and the costs per year of life saved in some major RCTs have been recently calculated (2); they are reported in
Table 1.

Table 1 : Cost-effectiveness of ICD in different populations (from Sanders, N Engl J Med 2005; 353: 1471-80) 

The observational studies demonstrated a reduced frequency of non-cardiac death in ICD recipients (RR, 0.74). This fact suggests that clinicians select healthier patients for ICD insertion, and this probably accounts for the larger apparent benefit from ICDs on all-cause mortality in observational studies than in RCTs.

While there is a definite benefit in patients with a history of ventricular tachyarrhythmias and in those in NYHA class II or III, a post hoc meta-regression using aggregate trial data from 12 RCTs showed no significant association between the proportion of patients with class I symptoms and reduction in mortality (P = 0.13). Similarly, the results of the trials are inconclusive for patients in class NYHA IV (p=0.62) (Table 2). 
 
Table 2. Which patients should be considered for ICD.

2 - Underestimation of “real-world” complication rates

In a systematic review (1),

death associated with implantation of ICDs occurred in 1.2% (CI, 0.9% to 1.5%) of procedures and occurred mainly in patients with left ventricular dysfunction;  mechanical complications occurred in 5.3% (4.6-6.2). The implantation success rate and safety of ICDs were similar in clinical practice and RCTs.  Among 30,984 Medicare patients (3),

10.8% of patients overall suffered one or more complications during the hospital stay during which time their ICD was implanted. The overall complication rate was similar whether an ICD or CRT-D device was implanted.

The frequency of post-implantation complications per 100 patient-years included 1.4 (CI, 1.2 to 1.6) device malfunctions, 1.5 (CI, 1.3 to 1.8) lead problems, and 0.6 (CI, 0.5 to 0.8) site infections. 

Rates of inappropriate discharges was surprisingly high, being 19.1 per 100 patient-years (CI, 16.5 to 22.0) in RCTs and 4.9 per 100 patient-years (CI, 4.5 to 5.3) in observational studies.

These shocks are commonly due to double counting, oversensing, ectopy, and supraventricular tachycardias, ranging from sinus tachycardia to atrial fibrillation. ICD shocks have consistently been demonstrated to reduce overall quality of life and increase the incidence of depression and anxiety while obligating strict driving restrictions (1).  Aside from morbidity, these shocks may have attendant lethal risks as patients receiving inappropriate shocks in the SCD-HeFT study (4) were at higher risk for death (HR: 1.97) as well those in the MADIT II (5) (HR: 2.29). The insertion of the device may be directly or indirectly proarrhythmic (6).  There are numerous speculated mechanisms by which an ICD may promote arrhythmogenesis including device malfunction, induction of arrhythmias from inappropriate shocks, pacemaker-facilitated triggers and reversal of activation wavefronts from epicardial resynchronization increasing dispersion of refractoriness. Additionally, local lead effects with mechanical irritation and late fibrosis may be a potential mechanism for VT.

The long-term reliability of implantable cardioverter-defibrillator leads has become an increasing concern. In a study (7), 15% leads failed during the follow-up. The estimated lead survival rates at 5 and 8 years after implantation were 85% and 60%, respectively. The annual failure rate increased progressively with time after implantation and reached 20% in 10-year-old leads (P<0.001). The major lead complications were insulation defects (56%), lead fractures (12%), loss of ventricular capture (11%), abnormal lead impedance (10%), and sensing failure (10%). Patients with lead defects were younger and more often female. These findings raise important issues for patients with expected longer life span at time of ICD implantation, such as younger patients, patients with preserved left ventricular function, or patients who have a prophylactic ICD.
 
Despite advances in ICD system design and manufacturing, devices remain imperfect. Structural failure of an implanted device has tremendous adverse effects on patient morbidity, both medically and psychologically. In 2001, Maisel et al. (8) reported a recall rate of 16.4 per 100 person-years, with 54% for hardware malfunctions and 41% for programming malfunctions. In 2005, all 3 of the major ICD manufacturing companies (St. Jude Photon Atlas, Guidant Ventak Prizm/Contak Renewal, Medtronic Marquis) issued advisories on the potential for ICD malfunction. Finally, although rare, apparently sudden or arrhythmic death events were associated with high-voltage lead failure and other deaths were related to pulse generator failure (9).

In summary, ICD insertion is unlike an “insurance policy,” as patients who do not benefit from proper therapy following the ICD insertion are still exposed to procedural and device-related complications.

3 - Overestimation of ICD efficacy in current guidelines

It is the entry criterion of the trials used to make recommendations and not the group actually studied that has driven practice guidelines. This liberal cutoff may be a movement in the wrong direction, as the average EFs were substantially lower than the inclusion cutoffs. For example, the average EF for patients enrolled in the AVID, MADIT I, MADIT II, SCD-HeFT, and MUSTT studies was 27% whereas guidelines report a cut-off of 35% (10).  In SCD-HeFT (4), the patients with an EF between 30% and 35% had no benefit from ICD which was restricted only to those with an EF<30%.  Most guidelines (see for example ref 10) rank as class I recommendation NYHA class I patients if they have an EF<30% even if the evidence of efficacy is inconclusive. 

Moreover, most of the trials had peculiar inclusion criteria that were not considered in the recommendations of the guidelines. For example, in COMPANION (11), inclusion criteria also included an acute episode of heart failure during the last year requiring hospitalisation which was not utilised in the guidelines, in DINAMIT (12) the post-MI patients had not only an EF<35% but also an impaired autonomic tone, etc.  In MADIT II (13) the benefit of ICD was observed only in those patients who were implanted late (>18 months) from their myocardial infarction but not in those who were implanted before 18 months. A screening log was not available and patients were not consecutive, so some data which could help to understand how they were selected is missing.

As a consequence, most patients currently implanted with an ICD never receive a therapeutic discharge but are exposed to the risks of ICDs outlined in this report. Three quarters to two thirds of ICD recipients in the observational studies received no therapeutic ICD discharges, and only 5% to 12% of trial participants received an appropriate shock per year (14).

ICD shocks: “appropriate” does not equal necessary. An examination of randomised trials for primary and secondary prevention has shown that the number of appropriate shocks consistently exceeds the sudden death and overall mortality rate in the control group (6). Thus, ICD therapies may not be a surrogate for sudden cardiac death, as many episodes may have been nonsustained nonfatal events. This suggests that a distinction needs to be made between shocks that are appropriate and shocks that are necessary.

Furthemore, in the SCD-HeFT trial (15) ICD shock, as compared with no appropriate shock, was associated with a significant increase in the subsequent risk of death from all causes (hazard ratio, 5.68). In the DINAMIT trial (12), the prevention of arrhythmic death with ICDs (HR: 0.42) was counterbalanced by excess death from nonarrhythmic etiologies (HR: 1.75).  In the MADIT II study (16 ), the mortality rate of patients receiving therapy for VF was over 50% at 2 years. The most common cause of death among patients who received any ICD shock was progressive heart failure. The authors speculated that successful abortion of sudden cardiac death merely shifted the mode of death to pump failure.

Conclusion:

It is ethically imperative that we are honest with the data, so that we can be honest with our patients. A reappraisal of the benefits and potential hazards of ICD therapy will enable physicians to a have a more mutually informed and balanced dialogue with their patients.