implantable cardioverter defibrillators in the hospitalized elderly???

There was a pretty striking analysis of large numbers of elderly patientsinappropriately receiving implantable cardioverter defibrillators (ICDs) during acute hospitalizations (see chf aicd not help elderly bmj2015  in dropbox, or doi: 10.1136/bmj.h3529​)

background:

--the US implants more ICDs than any other country: 133,262 implants in 2009: 434 new implants/1M people, 1.5x higher than the second largest implanter

--the age of implantation is slowly increasing, with average now of 74 yo

--the major trials finding ICD efficacy were in outpatients with stable heart failure, with a mean age of 60 in the SCD-HeFT and 64 in MADIT II trials (in this latter study, there was no difference if symptomatic NYHA class 2-3 symptomatic or asymptomatic). The 23-31% survival benefit in these studies became apparent after 1-1.5 years.  

--but 1/3 of older Medicare beneficiaries have ICDs implanted during hospital admissions for heart failure or other acute co-morbidities

--the current indications for AICD use for primary prevention include those at high risk of life-threatening ventricular tach or fib despite optimal med therapy (b-blocker, ACE-I), such as those with symptomatic cardiomyopathy (NYHA class 2-3) and LVEF (left ventricular ejection fraction) <35%, or those 40 days post-MI who are asymptomatic with LVEF <30%

--not much data in elderly: a substudy of MADIT-II found that those 121 people randomized to an ICD who were >75yo had only a non-significant mortality benefit.

details:

--this was a retrospective cohort study of 23,111 Medicare recipients who had a history of heart failure, were hospitalized with an acute condition, and were considered eligible for ICD therapy for primary prevention, eg with EF <35%. 5258 received ICD and 17,853 did not. over 90% had the index admissions for heart failure. follow-up 2.8 years. 53% died

--mean age 80 in those not getting ICD and 75.5 in those getting one. other differences: higher likelihood of ICD if male, lower LVEF (25% vs 29%), less psych comorbidites, more diabetes, ischemic heart failure, stroke.

results:

--crude survival curves showed improvement in survival in the first few months only, no difference thereafter (see figure at the bottom). matching patients with an ICD vs not by high dimension propensity scoring revealed no statistically significant benefit from ICDs.

--subgroup analysis: no statistical difference in cardiac mortality by history of non-recent MI (though there was a 37% lower total mortality). no diff in group with LBBB or by BNP levels. those >81 yo did do better (RR 0.78, 0.65-0.93), though only 12% received an ICD

so, a few points:

--a large number of ICDs are put in during acute hospitalizations, which had not been studied. Hospitalized patients are pretty different from stable outpatients: they may be more likely to die from heart failure itself (and ICDs don't do much for that) instead of arrhythmia. this finding may be similar to the studies of patients admitted with MIs and getting early ICD placement: there was no benefit as compared to waiting 30-40 days and there were more non-sudden cardiac deaths in those getting early ICDs, those least likely to benefit from an ICD. Hence the recommendation to wait 40 days.

--the benefit for this Medicare population was entirely within the first few months after ICD implantation (as per graph below). This result differs strikingly from the VA and MADIT-2 studies showing benefit only after 1-1.5 years, suggesting that in the Medicare study above there was a significant selection bias to ICDs in healthier patients (ie, they did better in the next 4 months because they were healthier and therefore more likely to get an ICD). this also could explain the improved response in those >80 yo, where only 12% got the ICDs.

--as an aside, other data show that women get ICDs less frequently than men, but there are other studies finding that they have lower risk of sudden cardiac death and are more likely to have complications for ICD implantation.

--this study brings up a few issues: the most striking to me is the use of ICDs in acutely hospitalized patients, mostly hospitalized for heart failure, with no data to support that approach, and the observational data from this quite large Medicare study now going against it. ICDs clearly subject patients to adverse events and the system to very high costs (as another aside, I did have an 80+ year old patient who had a very large MI and the got an ICD, who for many years thereafter lived in fear of the ICD shocking him and therefore limited his activities significantly). the second big issue is the migration of indications from well-documented (those in their mid-60s) to much older individuals, based only on observational data, and not a lot of that.

​

geoff