Last week, while Iain was giving a keynote lecture in Austria, I was also heading for distant shores to speak about chest pain evaluation at a special conference set up by the legendary Amal Mattu: the University of Maryland Emergency Cardiology Symposium. You’ll know, of course, about Amal Mattu and his reputation as a one of Emergency Medicine’s all time best educationalists and a guru in emergency cardiology. Well, Amal and his awesome team in Maryland (which also includes the likes of Mike Winters and Haney Mallemat) have recently implemented both the HEART score and a shared decision making pathway to help evaluate their patients with suspected ACS. From this came his idea to run a conference. You can find his highly entertaining intro to the conference and the expert faculty he had along right here. (It’s #FOAMed, of course)…
I was honoured to open the day’s proceedings with my talk: ‘A war on two fronts – how to get your chest pain evaluation right’. The aims of my talk were to bring everyone up to speed with the best evidence about how factors from your evaluation of a patient with chest pain influence the probability that the patient actually has an acute coronary syndrome, but then also to understand what constitutes an acceptable risk of a missed diagnosis and why it’s important that we’re not too over-cautious.
If you’d rather watch my talk than read a summary, you can catch a full recording for FREE – courtesy of Amal Mattu’s team – right here…
In the first half of my talk, I described how the patient’s history and physical exam might influence the probability that the patient is having an acute MI. A lot of this was based on the research I did in my PhD from 2005 to 2009, in which we recruited just over 800 patients with chest pain, recorded their symptoms and evaluated their outcome. Here’s a summary of some of our key findings:
- The character of a patient’s chest pain really doesn’t change the probability of an acute MI. If the pain is heavy or crushing in nature, the probability of an acute MI shifts only from 19% to 22%. That doesn’t really help us. What’s more, patients who’s pain felt like ‘indigestion’ have exactly the same probability of acute MI (22%).
- The most contentious point of the day was that risk factors for heart disease (hypertension, hyperlipidaemia, smoking, diabetes and family history) do not change the probability that a patient has an acute MI. Patients with no risk factors are as likely to have acute MI as patients with three risk factors. This doesn’t sit comfortably with everyone and some people didn’t want to believe the evidence – even though it’s been replicated several times. Maybe this is because the most dangerous risk factor is one that nobody knows you have – because those risk factors haven’t had any treatment
- Our clinical judgement is limited too. Taken alone, we found (in a study of almost 500 patients) that our ‘gestalt’ by itself can neither ‘rule in’ nor ‘rule out’ ACS. In fact, if doctors thought the diagnosis was ‘definitely not’ ACS, there was still a 9% chance that they were wrong (without doing any tests). If the docs thought it definitely was ACS, they were right – half the time!
In fact, the evidence clearly shows that the patient’s history and physical exam aren’t that helpful. There are some features that change the probability:
- If the patient vomited with their pain, there’s a 41% chance that they’re having an acute MI
- If the patient looks sweaty, there’s a 59% chance that it’s an acute MI – that’s more predictive than a single hs-troponin test at the regular cut-off!
But the bottom line is that, if your patient has symptoms that could be compatible with ACS and you haven’t got another clear cut cause, then you ought to be thinking about investigating for ACS. Anything else puts your patients at risk – there’s good evidence for that.
BUT – HANG ON – ARE YOU SAYING WE SHOULD JUST INVESTIGATE EVERYONE?
The danger here is that we stop thinking – and just investigate everyone for ACS. But we can’t do that. There will always be subjectivities, grey areas, patients where we’re really not sure. We can never get to zero risk of missing an ACS diagnosis without completely overwhelming our healthcare systems and being totally nihilistic. If we’re really going to do this well, we need a validated early rule out strategy. But even a validated strategy won’t have zero risk either – often far from it. So what is an acceptable risk of missing ACS?
In my talk, I asked the audience exactly that question. If you were going to send home a patient with chest pain and reassure them that they don’t have ACS, what risk would you accept that they would develop a major adverse cardiac event (i.e. die, have an acute MI or need coronary revascularization) within the next 30 days? As usual, people’s opinions were split – just as they were when Martin Than did an international survey asking just that question.
In Martin Than’s survey, just 40% of doctors would accept a 1% risk of a major adverse cardiac event within 30 days. The rest would accept much less than that. In fact, you don’t get a majority until you get down to 0.1% risk.
Maybe that’s helpful to us when we’re deciding if an ‘early rule out’ strategy should be used in practice. Maybe it tells us that any decision rule or algorithm won’t be accepted unless there’s less than 0.1% risk of a missed diagnosis (or a negative predictive value of >99.9%). But, then again, have you ever read a paper about any diagnostic test where the NPV is >99.9%? (Especially when you take account of the confidence intervals). I know the answer – it’s no.
Maybe asking doctors what we think is an acceptable risk is the wrong way to go about this. After all, if you ask me what risk I’d accept personally that I’d miss a case of ACS in practice, I’ll say ‘zero’. I never want to miss a case of ACS. Imagine what it would do to my reputation as a so-called expert! And I think all of us feel the same… Missing ACS hits our reputation, our self-image and we worry that it could cost us our licence or get us sued. With all those interests at stake (not just the patient’s wellbeing), we’re naturally going to be risk averse when it comes to missing ACS.
So maybe we need a better way. It could be about cost. Steve Goodacre’s work from Sheffield shows that we admit so many patients who don’t have ACS that even a strategy that isn’t very accurate (e.g. a single troponin test on arrival – even with a high sensitivity assay) could be deemed to be cost-effective! This shows how important it is that we have an early rule out strategy. But we want to do better than a single troponin test on arrival, right? We want to get people home and not miss ACS.
It seems that this is about balancing risks. As well as the risks of missing ACS, there are risks of over-investigation, after all. Patients get infections in hospital. Our tests (e.g. CT, angiography) use ionising radiation, which has risks. And we like to treat patients when we admit them with a provisional diagnosis – and that has risks too. In the ACCOAST trial, as many as 2% of patients given prasugrel had a major bleed within days of starting treatment.
In fact, Jeff Kline estimated that if the pre-test probability of ACS is less than 2% then we shouldn’t do any further investigation – because the risks of pursuing the diagnosis exceed the potential benefits of treatment. So maybe that’s as close as we have as a way of balancing risks. Maybe we need a rule out strategy that gets us to less than 2% probability – a negative predictive value of >98%.
That sounds like as good an answer as we’ll get about what constitutes an acceptable risk. But hang on a minute… Aren’t we missing something here? So far, we’ve had a really interesting, in depth discussion about the acceptable risk of missing ACS. But isn’t there someone else we need to involve in this decision about what’s acceptable?
Is the doctor’s opinion the only one that matters in this situation? What about the patient? How about what they think about risks and benefits? They have preferences and values too, after all. Shouldn’t we be asking them what they think?
Maybe the future is shared decision making – involving the patient in deciding the most appropriate way forward and giving them the options. Erik Hess from the Mayo Clinic in Rochester has done a lot of work on this. His Chest Pain Choice trial from a few years ago showed that patients who were offered shared decision making took a more active role in their care, had greater knowledge and more often chose to go home rather than have more tests.
Maybe this is the future. And maybe we’ll find better ways to engage patients in the decision making process, to understand risk and to tell us what they think is actually acceptable. If we can achieve that, our job is simply to help them to make the decision that they’re most happy with. Erik is continuing to work on this – and so am I – watch this space!
The slides from my talk at #UMECS16 are here…
And you can find full recordings of all the talks, including Simon Mahler on a modified HEART score protocol following his successful RCT; Barbra Backus on the original HEART score; Louise Cullen on the ADAPT protocol and making changes in practice; Bill Brady on the ECG in low risk chest pain; Judd Hollander on coronary CT and high sensitivity troponins; and a Q&A session in which we dealt with a number of very pertinent and practical questions from the audience. Amal Mattu did a superb job of putting this together – it was an honour to join.
And, having been introduced as the ‘choir boy’ of low risk chest pain (by virtue of my organ playing abilities – one of my interests outside medicine), I’m so glad that I seem to have managed to change Amal’s opinion by the end…
— Amal Mattu (@amalmattu) April 1, 2016
Fantastic, I’ve gone up in the world from ‘choir boy’ to ‘rock star’. Thanks for everything, Amal! https://t.co/4GYSaG5hm7
— Rick Body (@richardbody) April 1, 2016