ICDs plus CKD = no benefit but more hospitalizations

A large observational study of patients with heart failure and reduced ejection fraction (HFrEF) who had chronic kidney disease (CKD)  found no benefit from implantable cardioverter defibrillators (ICDs) and some increased morbidity (see icd in ckd more hosps jamaintmed2018 in dropbox, or doi:10.1001/jamainternmed.2017.8462).

Details:
--5877 eligible patients from Kaiser Permanente with LVEF <40% from 2005-2012 were tracked; those who received an ICD were matched (1:3) with those not getting an ICD by age, CKD status, and high-dimensional propensity scoring (see commentary below). data analyzed from 2015-7.
--1556 received ICDs vs 4321 without
--CKD defines as eGFR <60 ml/min/1.73m2
--mean age 73, 31% women, 75% white/13% black/7% asian or pacific islander/15% hispanic, LVEF mean 28.7%, 32% known CHD/7% ischemic stroke or TIA/38% atrial fib/31% mitral or aortic valvular dz/17% PAD/95% dyslipidemia/85% hypertension/53% diabetes/18% depression/39% chronic lung dz, systolic BP mostly 100-139 (only 11% out of this range), baseline hemoglobin mostly 11-15
--meds: 81% ACE-I/ARB, 87% b-blocker, 76% loop diuretics, 16% thiazides
--mean eGFR: 43.5 (51%: 45-59, 34%: 30-44, 14%: 15-29)
--significant differences between the ICD and non-ICD groups: those getting ICDs had lower LVEF (26.8% vs 29.4%), more known CHD (37.1% vs 30.1%), more on loop diuretic (82.8% vs 73.5%), but less likely to have dementia (2.8% vs 3.8%)

Results:
--in the fully-adjusted models, comparing the ICD and non-ICD groups:
    --all-cause mortality: 2541 patients(32.7%) died over mean of 3.1 yrs, nonsignificant difference between the groups, HR 0.96 (0.87-1.06)
    --hospitalization for heart failure: 1922 patients (32.7%), a 49% increase, HR 1.49 (1.33-1.60)
    --any-cause hospitalization: 4093 patients (69.6%), a 25% increase, HR 1.25 (1.20-1.30)

Commentary:
--this study, after controlling for many variables, found no mortality benefit and an associated increase in hospitalizations both for heart failure and overall, in a contemporary population
--given that this was a large observational study, they used high-dimensional propensity scoring to try to mathematically equalize these different populations of patients, some getting ICD vs not. 500 variables from different data dimensions were identified (demographic, hospitalization, procedure, outpatient care, lab and medication data). The most prevalent ones in their database were identified as covariates, these were ranked by their frequency, then were used for multivariate logistic regression. The most highly-adjusted model derived from this process included about 32 variables (eg age, sex, race, smoking, MI, unstable angina, atr fib, ventric tach/fib, valvular heart dz, dyslipidemia, htn, diabetes, depression, chr lung and liver dz, LVEF, hemoglobin, blood pressure, eGFR), plus meds (ACE-I/ARB, b-blockers, diuretics).
--the major studies on ICDs in the past did not include patients with CKD. which is a problem since:
    --CKD is common overall (about 14% of US adults)
    --cardiovasc disease is the leading cause of morbidity and mortality in those with CKD, with heart failure as one of the most common cardiac problems
    --and, of the 5.7 million adults in the US with heart failure, 30% have CKD
--there have been a couple of small studies looking at patients with HFrEF and CKD prospectively, both showing no advantage of ICDs in those with renal failure
--and a retrospective analysis of the pivotal MADIT II trial (one of the key studies finding mortality benefit from ICDs and leading to current ICD recommendations) found that those getting an ICD had a 90% higher risk of first and 74% increased risk of recurrent heart-failure related hospitalizations.
--why would ICDs have higher morbidity (ICDs prevent death from severe arrhythmias, not improvement in heart failure, but why an increase in heart failure??). not clear. perhaps a few ICD-related infections. perhaps there were unanticipated differences between the groups which led to more hospitalizations in those with ICDs. But, perhaps it was due to the increased medicalization of those with ICDs: i have had a few patients with ICDs or pacemakers who have felt afterwards that having these devices was "unnatural", they had an associated dysphoria (reinforced in the patients with ICDs who were really scared when there were perhaps unnecessary shocks!!, and who thereafter lived in constant fear), and several people certainly seemed to act frailer and more vulnerable after the device was implanted [this is all my guessing, no real data]. 

so, his brings up a few issues:
--most clinical studies are done with significant exclusions, and CKD is a frequent one.  We are often seeing patients with CKD, so it is often really unclear what to do. Unfortunately, the results of these studies are typically generalized: most clinicians are unaware of the details of the exact exclusion criteria of the study (or forget them, given the onslaught of medical studies, guidelines, etc we get daily), but the headline the “drug xxx helps in patients with yyy” is what is remembered, reinforced, and used in clinical practice (much to the delight of the drug makers).This article brings to light that there may well be very real limitations to the generalizability of their results: clinical studies have exclusions to make it easier for the researchers to interpret the findings in smaller studies (not so many variables), which basically undercuts the real utility of the results in actual patient care, which may lead to the results being inappropriately generalized. One perhaps useful approach is to read the study’s inclusion/exclusion criteria, the baseline demographics and morbidities of the patients in the study, and ponder whether this “average” study person is in any way related to the patients you are treating.​ I do not mean to be a total therapeutic nihilist, and I do continue to prescribe lots of drugs, but I do think it is important to have a healthy skepticism and constantly reassess whether you are actually doing well by your patients. It is sometimes helpful also to look at subgroup analyses, which do not have the same scientific rigor as the primary outcomes, but do raise questions about generalizability
--there are real concerns about using ICDs, with suggestion that sudden death is much less common than older data suggested, and that we should not be using the same criteria as in the MADIT studies in determining which patients get ICDs. See http://gmodestmedblogs.blogspot.com/2017/07/decreasing-sudden-death-in-heart-failure.html ​ , a meta-analysis which comments on a few items: the actual sudden death rate in more current studies is 1.0% in the 90 days after randomization (was 2.4% in the earlier trials, the basis for the ICD recommendations); and in the older studies, there was not maximal medical management (esp b-blocker or mineralocoricoid antagonist; and only one study used sacubitril/valsartan). a more recent trial (DANISH) confirmed no advantage of ICD in those with nonischemic HFrEF when added to maximal medical therapy. And there are studies showing that the reversal of LV remodeling (and improvement in LVEF) may take up to 12 months, not just the 40 days as currently recommended for ICD placement post-MI
--and, not exactly sure what this means, but the 2017 Am Coll of Cardiol Expert Consensus Decision Pathway for Optimization of Heart Failure Treatment (which really should have a clever acronym) does not even include ICD indications (see chf guidelines 2017 jacc2017 in dropbox, or go to http://www.onlinejacc.org/content/early/2017/12/12/j.jacc.2017.11.025 )
--so, i personally am hesitant to refer patients for ICDs, at least until there are studies which convincingly show that there is a subset of patients with HFrEF who really benefit...

--and this study does reinforce that we constantly review/reconsider the generalizability of the results of studies in our actual clinical practice

Comments

Popular posts from this blog

cystatin c: better predictor of bad outcomes than creatinine

diabetes DPP-4 inhibitors and the risk of heart failure

UPDATE: ASCVD risk factor critique