Doctoring ICD/Pacemaker

Are ICDs overused?

The medical news of the week nearly shocked me off my bike trainer.  It isn’t often that electrophysiology makes the major-network evening news broadcast.

The teaser proclaimed…“thousands of heart patients have unnecessary expensive cardiac devices…Should they be removed?”

They were talking about ICDs (internal cardiac defibrillators), and were referring to the widely publicized JAMA study that reported one in five (25000) ICDs implanted between 2006-2009 were outside of evidenced-based guidelines.

I am a real doctor so let’s start with a real story:

Mr [Smith], a jovial pot-bellied farmer, was in to see me recently.  His demeanor is one that exudes life.  His story is as epic as the warmth of his smile.  It’s the kind of story that sticks with you—human stories always resonate more than paired-t-tests.

About a year ago, the text message form the pacemaker nurse read: Mr [Smith] called this morning because he was shocked at 4 AM.  His transmission showed VF [VF is ventricular fibrillation–translation: sudden death.]

My response:  Can he come in now, I have a case in the EP lab soon.

A few minutes later this response came:  No. He just sat down at Denny’s with his wife.  He said he can come in after that…He feels fine.

Herein lies the problem with calling non-evidenced-based ICDs ‘inappropriate.’  Statistics are one thing, Mr Smith’s eventful morning was yet another. He was enjoying a Grand-slam breakfast with his wife, but if not for that ICD, he would have been in the funeral home.

When Mr Smith finally made his way to the office, I learned some interesting things about his past history.  I had implanted his ICD prophylactically almost a year ago.  The reasons were that he had a low-ejection fraction (weak heart) and a previous heart attack.  But here is the kicker: on one of the pre-ICD imaging studies the strength of his heart–measured by the ejection-fraction (EF)–was estimated at 38%, however, 35% is the upper-limit for placing an ICD.  Alas, his life-saving shock was delivered by a non-evidenced-based ICD!

Why would I have deviated from the guidelines?

Perhaps because of my 15 year experience base, or that the notches in Mr Smith’s resting ECG whispered ventricular tachycardia to me, or because I know that one echo-interpreters EF of 38% is another’s 33%. These vagaries are doctoring–in the real-world.

Enough storytelling, let’s get back to the study.

So why are doctors erring on the side of implanting more ICDs than statistical outcomes researchers say they should?

For the moment, let’s assume the statisticians are right; there are too many ICD implants.  Here are four possible explanations:

  • Mr Smith’s story illustrates the primary reason.  ICDs are not used to treat a malady, they are used to prevent death.  In the culture of heart doctors, death is failure, and heart doctors despise failure.
  • Most heart doctors don’t live with a palliative care specialist who tells stories of human suffering that makes the painless death from a malignant arrhythmia seem like a blessing.  Thus, they may not realize the importance of discussing end-of-life wishes with the patient before ICD surgery. (And they sure aren’t paid for their time to do so.)
  • Heart doctors are optimists.  They seem to purge negative ICD data from their minds. How many heart failure patients considered for an ICD hear about the “sobering” 168-day median survival after an appropriate ICD shock? (As published in the NEJM in 2008.) How many patients understand that an ICD–unlike an insurance policy– has substantial risks, including infection, painful shocks and manufacture recalls?
  • The cynics would say that ICDs are overused because they are easy to implant and well compensated.  That would be true, especially when one compares the ease of a thirty minute ICD implant with the angst of a thirty minute office visit to explain that a dad or mom has a terrible heart and won’t benefit from an ICD.  (That said, ICD surgery is not like spinal fusions; ICD implanters aren’t getting million dollar royalties from medical device makers.  At least none that I know are.)

Now let’s insert five real-world notions into the analysis of the ICD debate and science of this study.

  • The ICD registry data likely over represents the number of inappropriate ICDs.  For example, thirty percent of the ICDs were deemed inappropriate because of a heart attack in the past 40 days.  But how many of these hart attacks were ‘enzyme rises’ that were called heart attacks, or were minor heart attacks in a patient who previously qualified for an ICD based on a remote heart attack years ago?
  • Not all ICD implants neatly fall into guidelines.  For instance, what about the patient who presents with newly discovered heart block who may theoretically benefit from an ICD.  Would you do place a pacemaker knowing there was a good chance of needing an ICD later?
  • Indications for ICDs are misclassified. At least in my experience, ICDs implanted because of symptoms (secondary prevention) are frequently misclassified as prophylactic. ICD registry data is submitted by a retrospective review of charts by chart-reviewers not clinicians. These are not prospective trials.
  • More than half of the questioned ICD implants were in patients with newly diagnosed heart failure. Although it sounds Clinton-esque to ask, “How new is new,” concerns about timing of ICD implants are not insignificant.  Even experts aren’t sure about how long after the initial diagnosis one should wait before implanting an ICD.  These words are from the 2008 ACC/AHA/HRS ICD guidelines…

...The optimal time required for this assessment is uncertain; however, another analysis determined that patients with non-ischemic DCM (Ed note: DCM means dilated cardiomyopathy or a weak heart) experienced equivalent occurrences of treated and potentially lethal arrhythmias irrespective of diagnosis duration. These findings suggest that use of a time qualifier relative to the time since diagnosis of a non-ischemic DCM may not reliably discriminate patients at high risk for SCD in this selected population. Given these considerations, physicians should consider the timing of defibrillator implantation carefully.

  • Although there were modest increases in in-hospital complications (3.2% vs 2.4%) for non-evidenced-based ICDs, the study does not report on what really matters: what were the long-term outcomes of the two groups?  Did the non-evidenced-based group fare worse at one or two years?

In summary, this is an important trial that sheds light on an important topic.  ICDs are serious.  In selecting patients for ICDs, we can do better. We should do better.

But now I fear that the pendulum is swinging.  It’s a heavy wide pendulum that is sure to hurt when it hits. It always hurts when statistical spreadsheets and over-reaching arbitrary guidelines bully themselves into the doctor-patient relationship.

Mr [Smith] is sure glad that his care came from a doctor, not a pdf file.


12 replies on “Are ICDs overused?”

I enjoyed reading your post. But studies like the JAMA research mentioned here are usually quickly dismissed by the CRDM industry as being “out of context” or, as you imply here, failing for oversimplifying a complicated topic. In this case, a Boston Scientific’ CMO came out in an interview to remind us that more than a million patients are eligible for ICDs, but only a small percentage is getting the devices they need. It maybe true, but hard not to take it with a grain of salt.

So what is it? Do we need to implant more devices or stay within established guidelines and implant fewer of them? As you put it, the answer might be more complicated than that.

I think what we really need is to get better (much better) at identifying who is at risk of SCA or to come up with less invasive means to protect those at risk. After all, no patient wants a lifetime of unnecessary device and lead replacements and all the related adverse events that come with having a transvenous implant. I know I don’t.

But isn’t that the point of having guidelines in the first place? If evidence-based medicine can’t help us, I don’t know what can.


You are right; it’s both: we put in too many ICDs, and we put in too few ICDs.

You say we need to get better at identifying those at risk. You are absolutely right. We recommend ICDs in patients with advanced heart disease because they are at high risk. But in absolute numbers of sudden-death, many more cases occur in people with no known heart disease. How can we identify and prevent these cases? That’s the critical question.

I fully support science-based practice. The problem in ICD guidelines are that the grey areas are vast.


You wrote: <>

Quit your crying Doc ! I’m in one of the professions (not medicine) just like you. Regrettably, I don’t get paid for every nano-second of my time either. Life ain’t always fair ! Get use to the idea !

The reality of the matter is that you are a professional. Professionals aren’t a bunch cry-babies complaining about how well they get paid. If you aren’t paid well enough, it’s your own damn fault. Suck it up and work harder.

As a professional, in the United States of America, you are not only among the highest paid individuals in the United States, but among the highest paid in the entire world, even when you throw in the sun, the moon, the stars and all the other planets. You’re damn well paid. You have no right to complain about how well you are paid.

That’s not what the post was about at all. Pretty abrasive response for someone who didn’t even read the post.


Thanks for a thoughtful post.

It seems we EPs can’t seem to get out of the crosshairs.

Nevertheless, I believe there is more truth than fiction in the JAMA article. I’ve seen numerous instances of ICD implants outside guidelines over the years. Some are clearly justifiable, many are not.

I have been a staunch advocate of following guidelines for years. Often this is a significant battle. I’ll often get patients referred to me by cardiologists with the intent of getting an ICD soon after MI, revascularization or heart failure diagnosis. I’ll note that their “not ready yet” and watch them through the appropriate interval with medication optimization. One of my favorite office visits is the one when they come back with EF back up near normal. I’ll pat them on the back and tell them they don’t need us anymore.

I am well aware many doctors don’t follow this strategy. I doubt their motivations can be reduced to simple greed in most instances, though. Most implanting doctors I know are already pretty busy and not looking for more work. They are probably dealing with some degree of misplaced passion and lack discipline. Sometimes the external forces lobbying to get the implant done are significant.

My main criticism of the JAMA article is directed at the data quality. I think to a large extent, we are dealing with a “garbage in, garbage out” phenomenon. The data entry is generally not done by physicians and there is not a strong motivation to get it right. In our institution, overworked nurses who have limited access to records do their best to check off all the appropriate boxes in between actual patient care tasks. It’s hard to get an accurate history from a sedated patient and a few sheets of old records. This whole JAMA study hinges on the quality of this data which is cobbled together on the fly and not independently validated.

To answer Hugo’s inquiry above: we both implanting too many and too few ICDs. If we restrict the ICD implants to those who clearly meet guidelines, we still are only reaching a minority of the indicated population. Each of these patients who were indicated, but never offered ICD implant are at significant risk for death. Some day I’d love to hear an outcry about the epidemic of unnecessary deaths in this population. The human tragedy here far exceeds the bad that has come from “unnecessary” ICD implants. If ICDs were as cheap as aspirin, there probably wouldn’t be much argument here. No doubt, it’s mainly about the money.


P.S. We’ve been trying for years to come up with strategies to “cone down” the ICD indicated population (i.e. EP testing, Holters, T wave alternans, etc.). Most of this work has been industry sponsored. So far, the best we can do is the current guidelines. Just because an ICD doesn’t go off doesn’t mean it wasn’t a good idea to put it in. Think airbags. Are there too many of these in cars?


These are fantastic words; they echo the importance of clinical judgement, discipline, and critical thinking in evaluating science.

I agree whole-heartedly, and I especially like the airbag analogy.

Thanks a bunch.

About your site: the new layout looks great! Unlike other new-and-improved pages, yours didn’t get all confusing! 🙂 Happy New Year, we love the page!

I attended a lecture by Narayan in which he said we needed better ways to identify those who needed ICDs and had some lovely theories, but no firm practice suggestions. In the audience was a mix of private practice and academic EPs as well as fellows and general cardiologists. I asked whether Narayan used any age cut off or any co-morbid conditions to limit ICD use as these are areas where I’ve had questions and concerns about when to refer for implant vs not implant. He did not have firm guidelines or recommendations. One of the private practice EP docs stated that he’d implant a device in a pt over 80 yo if he met criteria, then went on to say ‘what if it was your father?’ If it was my 80 yo father I’d want him to leave him the heck alone because VT/VF is a great way to die. That is preferable to end stage HF.
I know an EP in private practice who has admitted that they’ve implanted devices in pts who did not really meet criteria. The implant was done because the referring doc requested it. It was easier to implant the device than document the reason for not implanting (and if declined, some other competing EP would do the implant). In the academic setting where I work we are more guideline oriented and I see less of this. But yet we still have those patients we miss who end up dying of VT; recently heard of a 55 yo pt s/p recent MI who had VT arrest and died. Didn’t meet criteria as EF >40, still < 40 days post MI. Very much a quandary.


This is fantastic. Two points stand-out.

“If it was my 80 yo father I’d want him to leave him the heck alone because VT/VF is a great way to die. That is preferable to end stage HF.”

Bravo! The essence of patient-centered care is that we present the information–the pros, cons, expectations and alternatives–and the patient chooses. Many will choose not to have an ICD.

“It was easier to implant the device than document the reason for not implanting (and if declined, some other competing EP would do the implant)”

The ease of this decision is accelerated by the fact that our present hc model rewards how much medicine we do, not how well we do it. Nuanced discussions with patients and families about what the prophylactic ICD trials really say is far harder than implanting the ICD. This is the same concept that primary care doctors wrestle with: it’s easier to write four letters–“Z-Pac”–than it is to explain that viruses do not respond to antibiotics.

Comments are closed.