November 10th, 2009
Deciding Who Gets Prophylactic ICDs — We Need a Better Way
John Spertus, MD, MPH
The ACC/AHA guidelines endorse prophylactic ICDs for NYHA Class II/III patients with an LV ejection fraction ≤35%, but I’m reluctant to recommend this to my patients who are doing really well. Why put them through the hassle and risk? Will they really benefit, especially if I think they’re just as likely to get an inappropriate shock as a life-saving one? A recent paper in Circulation has caused me to reconsider my approach.
Wayne Levy and colleagues used a modified version of the Seattle Heart Failure Model to identify which patients meeting the ACC/AHA criteria will benefit most from prophylactic ICDs versus medication alone. They found that ICDs offered substantial benefit for patients at low risk for overall mortality — but provided no benefit for those in the highest quintile of risk (i.e., patients with a projected annual mortality rate approaching 20%). This really impressed me, because I’ve been especially reluctant to recommend ICDs to patients who I perceive to have the best prognosis.
Two important insights from this work:
We often presume that the greater a patient’s underlying risk, the greater the benefit of a given intervention. But, in the setting of heart failure, where ICDs protect only against arrhythmic death and not against death from pump failure, the competing risk of death from pump failure can swamp the potential benefits of protection from arrhythmic death. Confirming this was the observation that in the highest-risk patients, there was, in fact, a nonsignificant 24% reduction in sudden cardiac death, despite no benefit in overall mortality.
Risk-prediction models can give us a better sense of how individual patients might respond to a given treatment, so that we can tailor therapy to their individualized estimate of therapeutic benefit, as well as their personal goals and preferences. To do this, however, we need to start developing the infrastructure to implement validated, evidence-based risk stratification at the time of medical decision-making.
Until this additional research is accomplished, how will you apply the insights of Dr. Levy’s study in your practice? Do you think the modified Seattle Heart Failure model is the best way to stratify risk, or are there other mortality estimates that you might use? If so, which ones? How will you routinely incorporate these into your practice?
7 Responses to “Deciding Who Gets Prophylactic ICDs — We Need a Better Way”
NEJM Journal Watch — Recent Cardiology Articles- To Lower Stroke Risk in Patients Undergoing Atrial Fibrillation Ablation, Is a Device or a Medication Better?
- Intensive vs. Standard Blood Pressure Lowering in Patients with Diabetes
- Anticoagulation in Patients with Cancer-Associated Pulmonary Embolism: How Long is Long Enough?
- Spironolactone Use After Percutaneous Intervention for Acute Myocardial Infarction
- CRISPR/Cas9 Gene Editing for Transthyretin Cardiomyopathy
Sometimes Patients Know Best…
Very glad you posted this. I saw a patient in clinic yesterday who has ischemic cardiomyopathy with an ejection fraction of 30%, but uses his treadmill every day and recently did 13 METs on a Bruce. Earlier on, just after his 4-vessel bypass surgery, he wasn’t doing quite as well and we had strongly recommended a prophylactic ICD, which he refused. Now, 2 years later, he’s a model patient and is probably proving to us that he had been right all along. So I’m pretty sure I don’t need to the Seattle Heart Failure Model to feel better about this particular patient, but I like the idea of having a tool to help with decision making — or at least conversations with patients — in cases where symptoms are present…
Isn’t this the challenge??
Your experiences are exactly mine. It is tough to feel like he would benefit from an ICD when he is doing so well, but based on the SCDHeFT data, it seems as though he would benefit from a higher probability of a longer life with one, given that his greatest risk of dying is from SCD, which is exactly what ICDs protect against. What are your thoughts? was the patient right? will the ‘average benefit’ from prophylactic ICDs be proven right? What if he gets a life-saving shock within a year after it is inserted? What if he gets one and gets inappropriate shocks? Hmmmm…..
Time will tell, or will it be too late?
Great points, and all well taken. I’ll admit I’m influenced by many stories of inappropriate shocks and procedures gone bad, and then there’s the hassle you mentioned and the cost. With my patient, I want to think he’s active with great exercise tolerance, so his personal decision was ok. I used to think it was also better that he had a narrow QRS, but the data didn’t support that. Then there’s the longer term data from MADIT II (was NNT really 6?) and the Levy article you mentioned above — it all makes sense, but for some reason still hard for me to intuit and in turn explain to the patient. Every clinic visit with him, I still go through the motions of discussing the ICD and documenting “offered and refused” but without really believing it’s absolutely the right thing (beyond, of course, knowing the data). Do you think we’ll get better at SCD risk prediction any time soon? If not, I guess we have to go with the data we have and hope that more studies like Levy’s help to guide the way. So based on the data that’s out there, do you think our line of conversation should be something like “you’re more likely to benefit because you’re doing so well”. On the other hand, for the patient with severe symptoms, in Levy’s worst quintile (even if their true life expectancy is >1 year), do we venture to say the opposite?
As they say in Dragnet: Just the Facts, Mam, nothing but the facts…
I think the best strategy is to just present the best estimate of the data that we have. So, if this patient was in the lowest risk quintile of SD-HEFT (actually, his risk is likely much lower than any of the patients in SCD-HeFT given that he is Class I or better) he would have about a 50% reduction in the risk of death. If you estimated his 1-year mortality to be 4%, then with the ICD it would likely be only ~2%. Presenting this information to him (either as risk of dying or, conversely, saying that you guess that 96% of patients like him are alive at one year and among those who get an ICD, 98% are alive at 1 year) would be the best way to help him understand his individual benefits from the procedure. Among the risks – as you point out – would be a ~7% risk of an inappropriate shock. What we need are tools that allow us to translate risk prediction models to more accurately estimate patients’ individualized, evidence-based, estimates of risks, benefits and complications of treatment so that we can have more informed conversations with our patients.
We have developed a tool for doing this for the choice of bare metal and drug eluting stents at the time of PCI. We are studying the impact of this on patients’ choices and their long-term adherence to dual anti-platelet therapy. I look forward to sharing the results, when they are available. In the mean time, it would be great to get access to high-quality models for primary prevention of ICDs so that we can try this approach in this setting as well.
Competing Interests: We have a patent on a technology, PRISM, that executes multivariable models at the point of care with individual patient data to inform doctors’ and patients’ medical decision-making.
Just the numbers…
I really like the way you pull it all together with numbers, albeit estimates, that would really make sense to patients. I would love to have these sorts of point-of-care tools at my fingertips for my practice. Although it doesn’t sound like they yet exist for deciding on ICDs, could you direct us to where we might access your tool for deciding on stent choice for PCIs? Really looking forward to hearing more about the tool for antiplatelet therapy also…
This study identifies those who may not benefit from an ICD and highlights the continued need to tease out those individuals that may benefit from those that may experience adverse events and no benefit.
I am currently seeing a thirty-year-old presenting with monomorphic VT and mildly elevated troponin with a severely depressed ejection fraction in the setting of an MI likely within the past few weeks. The patient is also undocumented. He unfortunately does fit well into the current guidelines and recent studies. I am curious to see what others would choose regarding an ICD in this individual.
Eugene, What a tough case! I am not totally sure that i understand what is undocumented, his rhythm or the patient himself, as the latter would clearly bring up challenging economic issues about the device, its costs and the need for follow-up. Without knowing more (e.g. his NYHA, renal function, etc.), but given his young age, i would guess that his mortality would not be high enough that there would be little benefit from an ICD. Regardless, these kinds of situations clearly underscore the need for more explicit risk-prediction models of outcomes with and without the device. If given that data, making these types of decisions would like by much clearer.