Nov 14, 2016
Welcome to Circulation on the Run, your weekly podcast summary and backstage pass to the journal and it's editors. I'm Dr. Carolyn Lam, Associate Editor from the National Heart Center and Duke National University of Singapore. In today's podcast interview we will be discussing the ruling in and ruling out of myocardial infarction with the European Society of Cardiology 1-hour algorithm. Stay tuned for a discussion of new data and controversies on this hot topic. Now, here's a summary of this weeks issue.
The first paper brings us one step closer to the ultimate goal of cardiac tissue engineering. That is to replicate functional human myocardium in vitro. In this study, by first author Dr. Ruan, corresponding authors Dr. Murry and Regnier from the Institute for Stem Cell and Regenerative Medicine and University of Washington, authors recognize that human-induced pluripotant stem cells, or iPSC-derived cardiomyocytes, really provide a cell source for cardiac tissue engineering. However, their immaturity limits their potential applications. Hence, they sought to study the effect of mechanical conditioning and electrical pacing on the maturation of iPSC-derived cardiac tissues.
They found that after two weeks of static stress conditioning, the engineered myocardium demonstrated increases in contractility, tensile strength, construct alignment, cell size, and SERCA2 expression. When electrical pacing was combined with static stress conditioning the tissue showed an additional increase in force production and further increases in expression of RyR2 and SERCA2. These studies really demonstrate that electrical pacing and mechanical stimulation promote the maturation of the structural, mechanical, and force generation properties of iPSC-derived cardiac tissues and constitute a really important contribution to cardiac tissue engineering.
The next study is the first large-scale, nationwide, population-based investigation of the association between congenital heart defects and any placental measure. This study by Dr. [Matheson 00:02:27] and colleagues from Aarhus University Hospital in Denmark, included all 924,422 live-born Danish singletons from 1997 to 2011. Congenital heart defects was present in 7,569 newborns. The authors compared the mean differences in placental weight between newborns with and without congenital heart defects and found that only three specific subgroups of congenital heart defects were associated with measures of impaired placental growth. These included Tetralogy of Fallot, double outlet right ventricle, and major ventricular septal defects. In these subgroups, the mean deviations from the population mean head circumference and birth weights were reduced by up to 66%, with adjustment for placental weight. In other words, up to two thirds of the deviations in fetal growth, including fetal cerebral growth, may be related to the impaired placental growth. The present work provides an important contribution to the existing knowledge on the association between congenital heart defects and placental anomalies as well as the possible importance for fetal growth in this population.
The next study provides an up-to-date evaluation of the cost effectiveness of antibiotic prophylaxis in the prevention of infective endocarditis. In this study by first author Dr. Franklin, corresponding author Dr. Thornhill, and colleagues from the University of Sheffield, the cost effectiveness of antibiotic prophylaxis, namely single dose amoxicillin or clindamycin, in patients at risk of infective endocarditis. They did this using, firstly, recent estimates of the effect of antibiotic prophylaxis on infective endocarditis in the English population; secondly, rates of antibiotic adverse drug reactions; and thirdly, estimates of the probability of developing infective endocarditis following dental procedures derived from French data. All this as foundation for analysis of cost and health benefits.
A decision analytic cost effectiveness model was used based on the decision model by the National Institute for Health and Care Excellence, or NICE, that was used to inform the 2008 guidelines. The authors found that antibiotic prophylaxis was less costly and more effective than no antibiotic prophylaxis for all patients at risk for infective endocarditis. In fact, if antibiotic prophylaxis was reinstated in England for those at moderate or high risk of infective endocarditis, it could save 5.5 to 8.2 million pounds and result in health gains of more than 2,600 quality-adjusted life years. Antibiotic prophylaxis was even more cost effective for those at high risk of infective endocarditis, being cost effective even if only on 1.44 cases of infective endocarditis was prevented per year. In summary, these updated findings really support the cost effectiveness of guidelines recommending antibiotic prophylaxis use, particularly in high risk individuals.
The last study provides data on long term cardiac mortality among survivors of cancer diagnosed in teenagers and young adults in the largest population-based cohort to date. Furthermore, the study provided, for the first time, risk estimates of cardiac death after each cancer diagnosed between the ages of 15 to 39 years. For example, survivors of Hodgkin lymphoma, lung cancer, acute myeloid leukemia, non-Hodgkin lymphoma, and CNS tumors experience 1.3 to 3.8 times the population-based mortality rates. This study provides important insight into the cardiotoxicity of the treatments given in the past to teenagers and young adults with each individual type of cancer and importantly, provides an initial basis for developing evidence-based follow up guidelines.
Those were you summaries. Now for our feature interview.
Our feature paper today discusses the hot and controversial topic of ruling in and ruling out myocardial infarction with the European Society of Cardiology 1-hour algorithm. I'm so excited to have with us the corresponding author of the paper that really represents the first multi-center external validation of these ESC guidelines for MI and the first multi-centered direct comparison of the performance of the algorithm with high-sensitivity troponin I and high-sensitivity troponin T assays. This would be Dr. Martin Than from Christ Church Hospital in New Zealand. Welcome Martin.
Thank you very much. It's a great pleasure for me to be able to join everybody and talk here.
It's great to have you. We also have with us the editorialist on this paper, Dr. Allan Jaffe from Mayo Clinic, Rochester, Minnesota. Allen, it's so good to hear your voice again.
Good to talk to you again too, Carolyn.
Finally, we have Dr. Deborah Diercks, Associate Editor from UT Southwestern. Welcome Deb.
Oh, it's good to be here and I'm looking forward to the conversation and what we're going to learn from these two gentlemen.
Absolutely. You know what? I'm going to start with Martin. I love the way to set up your paper. You very correctly pointed out that there's a tension in that ED physicians require really high sensitivity to confidently rule out MI and send patients home, whereas cardiologists do not want high proportion of false positives because we don't want false high risk to lead to invasive testing. I just love, if you could start by telling us how the ESC 1-hour algorithm fits into all this and what you were trying to do in your study.
I heard Deb Diercks on the phone as well, who's a very respected emergency physician in this area, and I think we would both say that we have a certain bias in our perspective on this, which is of course we are the people at the end of the day that have to send people home when they present with chest pain and possible myocardial infarction. We are also, of course, the people that take the fall if there are any mistakes made. Historically, people have not been very kind to emergency physicians who miss such a diagnosis. It's an extremely high source of medical legal action in the United States and, in fact, worldwide. So we're somewhat paranoid as a speciality about missing cases of myocardial infarction because at the end of the day, the worst thing that can possibly happen is for you to send someone home who comes to harm from the very clinical complaint for which they came to you for help. We want to avoid that at all costs and that was the basis behind us trying to put together this paper.
Soon after the ESC guidelines come back and I returned from London, where they were announced at the conference, to New Zealand, I received quite a lot of phone calls and correspondence saying, "Okay, we see these new ESC guidelines are out. When are we going to start introducing them?". I immediately wanted to say, "Well, the key thing is to understand how they would work, how they would be implemented, and whether they'd work in my own setting" because if we want to implement them in New Zealand or Australasia, we would want to double-check on that first. That's the basis and the philosophy behind the manuscript.
Tell us what you found.
As Allan will be the first to point out, I think there are a number of flaws in the data we had available to us that allowed us to do this analysis, but based on the concept that when we've surveyed emergency medicine physicians, the sensitivity that was wanted was at least 99% if not higher. We found that neither of the algorithms produced that level of sensitivity, although the algorithm based on hsTnI was very close. I think it's 98.8%, so that was very good. Reasonably wide confidence intervals on that. The hsTnT algorithm performed slightly less well with a sensitivity around 97%. I guess, if I was to start with an a priori question, which is did we reach a standard of 99%, then our answer to this was, in one case, not quite, and the other case, no, we probably didn't. We said that if you wanted to use a metric of negative predictive value, which I know a lot of people do, then there was actually very good negative predictive value in the high 99 percentage range for both pathways.
Do you mind if I stretch you a little bit and ask you to describe exactly what you did in the cohorts? You were saying that there were some imperfections. Maybe you'd like to tell us a little bit about that.
Absolutely. As always, when you're writing a paper, you look back and you always feel there are far too many imperfections, but I guess the principle one I would say that's been noted is that we had samples done on arrival and the algorithm itself specifies a [inaudible 00:11:43] one-hour second sample. We didn't have those specimens, so we had to base our data analysis on samples done either at 90 minutes afterwards or two hours afterward. It's clearly not being tested exactly as it was written, although one could argue that that slightly delayed sampling is potentially reflective of real life, where it's very hard to hit a one hour mark in a busy emergency department, and two, where the slight delay in getting the samples would actually allow more time for a troponin to rise and therefore give a chance of providing a better sensitivity.
I think the other I guess key flaw is that of course, the people present to emergency departments at different time frames following the onset of their symptoms. There's been some valid concern raised that algorithms may not necessarily perform as well in very early presenters. In fact, that is something that's being emphasized now in the ESC guidelines.
Right. Allan, I loved your editorial. You did mention a couple of these points. Would you like to maybe clarify your view of this?
I think that there are two or three terribly important issues. We all would like to have very facile algorithms. Particularly given removing the high sensitivity, the idea would be gee, wouldn't it be nice to have something really simple that works perfectly? If you look at the validation and the way the algorithm has been put together, immediately there are some concerns that people ought to have and that at least we tried to point out, that were important. One of them Martin has already discussed a little bit, which is one looks at most of the validation studies. There are very few patients who are evaluated very early after the onset of their symptoms. That's a potential problem because the overlap, since they use very low values or very small change, that there could be, with people who have real disease, is in those very early presenters. The initial algorithm from the ESC used both a very low level troponin and a set of change criteria. Actually when they published those criteria, they changed that and eliminated, at least for the first three hours, the very low values. If one looks at Martin's study, it was again, the very early patients who potentially may have been missed. I think we need more data before we go ahead and acknowledge that this will be working for those early presenters.
There are two other problems with the population that we need to be careful about. It's been well known that when you have a negative troponin at six hours all the way back to [Chrisann's 00:14:26] original article in the '90s, that you're pretty safe. The population that you'd like to look at really are the patients who, after two hours in Martin's study, since he took a little bit longer given the logistics that were there in New Zealand and Australia, is the patient who came in at four hours because by six, they're actually meeting that six-hour criteria. When you have a large number of other such patients, you simply add noise and it makes you sensitivity look better, but it's not necessarily the case that that give you that same degree of reassurance that ED physicians would like.
The third population-related issue is that you'd like to do this in all-comers. The protocol was developed for chest pain patients, but there are a variety of patients in whom we evaluate myocardial infarction in, who may not qualify for that. The patients who are critically ill, for example, who may have Type 2 infarctions. The individuals who may come in who are very elderly, who often don't have chest pain so we don't identify them necessarily as a rule out. Interestingly, if you start thinking about those groups, they tend to have much higher troponin, so they may well skew the cut-offs that are used and change the algorithm.
In truth, we don't want more than one way of defining myocardial infarction. We only want one algorithm for ruling in and ruling out. Having an all-comers study, in my way of thinking, would be important. In that same regard, let me point out that you can rule out myocardial infarction because you don't have an acutely changing pattern of troponin elevations, but what we really rule in myocardial infarction? You rule in acute cardiac injury. Could be myocarditis, could a apical ballooning. There are a whole variety of other types of disease entities that could be involved and the arbitrary value of 52 that was put in the algorithm really, I think, is much too low for two reasons. One reason, because it didn't include all-comers. A second reason is because of the way in which the comparison between troponin T and I were done. I'll talk about that in just a moment. I would point out that using a different assay, the troponin I assay, in another set of studies, another group from Hamburg has suggested that very different metrics would be much better.
The final thing to say about extrapolation between the assays, and then I have some suggestions about what would make this better if you want to go there now or we can wait, is the comparison and the way in which the metrics for troponin I were developed really weren't by using troponin I as a gold standard. It was by taking and using troponin T as the gold standard for the diagnosis, then thawing samples many years later, running troponin I, and then extrapolating from the gold standard of troponin T to troponin I. Well, there's several problems with that. Number one is that appropriate comparisons should be fresh samples. Fresh samples. In addition, we believe, from the way in which we think about high sensitivity, which may not be correct, that the troponin I assay should be more sensitive and in [inaudible 00:18:05] fact, in the papers that were done validating this approach or attempting to describe the approach, troponin T was wildly more sensitive than was troponin T. We're extrapolating some data that doesn't sort of fit the way in which the information we have, it would mean all of the troponin I validation studies are incorrect.
That's where those numbers came from and even more problematic are the change numbers, which are very low. For the troponin T assay, they're three in five between ruling in and ruling out, which if you look at the assay imprecision, is something the assay can't do. Now you're extrapolating them in a very, very loose manor to troponin I and making them even lower. Those are not doable sorts of things. There's a real problem with the way in which the metrics for troponin I, even though it performed well in this circumstance, ended up being developed. I think all of those things need to be taken into account when we look at the results of the study. The results that Martin and his group got are very similar to the other validation studies that have been done because they've all done it pretty much that same way. There's not a surprise that their validation is similar, but I think unfortunately, we didn't have an opportunity to unmask, in a data-driven way, the problems that I just described.
Thank you Allan. Deborah, if you could share your thoughts on this.
Martin raises some valid issues. That if something goes out as an algorithm, people want to use it. That use needs to be predicated on does it work in their patient population and is it feasible in the time frame and can it be adopted safely and what the indications are. In the emergency department, the value really is the negative predictive value because we want to be able to safely send people home. That's where rapidity of an evaluation is very important.
The other issue raised was exactly what Dr. Jaffe talked about. Does the algorithm itself reflect what we really need? Can you validate something that was created by the scientific way, but really a combination of a lot of information? Are the thresholds really valid themselves? That's the challenge with it. I think what you heard here are kind of two issues we struggle with it. We have a very respectable organization putting out an algorithm that is scientifically based and we want to adopt early, but there are questions on both sides of the issue on whether it can be adapted into real-world clinical practice on a global nature where prevalence of disease is different and the patients it'll be applied to vary, whether it's been on time of presentation or overall demographics.
Also on the scientific side, on the assays itself, are we using the right cutoff? Especially when we're looking at deltas and looking at such a rapid change. It's very nice to hear both of those points so eloquently described today during the discussion.
Thanks Deb. I fully agree. Hence, again, the importance of this paper. Martin, I'd love to hear your responses to Allan's comments and then also share with us, what's the take-home message for you as a clinician? How are you applying what you just found?
The guidelines are good on the right line, it's just as I said, they may not necessarily translate to all other environments. I guess that's my take-home message to myself, which was if I were to look at my own data from my own center, in Christ Church, and the way it's applied here, if I had applied the ESC guidelines and it had met the metrics which I was satisfied with, which I guess would be a very high sensitivity for me in terms of rule out, then I would actually seriously consider implementing it in my own center. It didn't reach that threshold so now I want to try and refine or explore further how I could allow the guidelines to do that. For example, one way that, and this is in the guidelines, but not necessarily in the flow chart, is the importance of applying clinical judgment and clinical findings with the results of the algorithm. I think that's a very important step in it. For example, if I was going to apply this in my own center, I'd want to be setting out clearly for the doctors concerned, how one would incorporate clinical judgment rather than it being a very subjective thing, which might vary significantly between a junior doctor or a far more experienced one.
I guess the take home message for me is this. The ESC guidelines are a very important piece of work. They've been robustly developed. For people who want to implement them, I'm no saying don't use them at all. I'm just saying that, you know, just think about carefully how you would use them and check whether you think they're appropriate for your setting.
That's great. Allan, what about you? What are your thoughts on how this may be applied in clinical practice and what more needs to be done?
I think we need to have a real trial where patients are managed based on the results of these approaches rather than more observational studies. I would argue that those management trials that involve an all-comers sort of population, so we are comprehensive, and should also interrogate whether or not the protocol itself is adequate or whether or not it requires follow-up to meet the metrics that have been proposed. I would point out that in the past, in the studies from the group from New Zealand and Martin Than particularly, have had very, very good follow-up. One at least needs to ask the question whether or not the algorithms that are proposed work perfectly without any follow-up or whether or not follow-up is an important component. We don't know that yet.
Thanks Allan. I'd love to give the final words to Deb. Take home messages?
You know, I think that we need to look at this as a positive in that we're looking at time frames that provide a rapid evaluation and the discussion is around safety. As long as we keep focused on appropriate evaluations for the patients and applying the right algorithm to the right patient, we're going to benefit the care of those we're really concerned about. I appreciate the work that both Martin and Allan both have done on really pointing out how we can do that in a great manor.
Thank you, all of you, for joining us today. I mean, it's been such an enlightening conversation. I'm sure the listeners have enjoyed it and thank you listeners for tuning in. Don't forget to tune in again next week.