By now, I think everyone has heard of the test threshold. It is most often discussed in the context of the workup for pulmonary embolism. (I think it is an important concept that we probably need to employ more widely in emergency medicine, but that is the topic of a future post.) At the opposite end of the spectrum, there is an equally important concept called the treatment threshold, which gets far less attention. In this post, I discuss how thinking about the treatment threshold made me more comfortable managing STEMI with thrombolytics.
There is an interesting discrepancy in medical education. Almost all education takes place in giant academic institutions – ivory towers – so most trainees never see how medicine is practiced “in the real world”. This was especially apparent to me in the management of STEMI.
Throughout my training, STEMI was managed in the cardiac catheterization lab. All you had to do was recognize the ECG, give some aspirin, and the patient would disappear for definitive management. As a result, I was a fully trained emergency doctor, working independently in the community, the first time I ever gave thrombolytics for STEMI. (That was a fun shift. When EMS arrived with the STEMI patient I was the only doctor working with 2 nurses to simultaneously manage an open fracture dislocation of the ankle made worse by a significant ketamine emergence reaction, a hypoxic COPD patient requiring BIPAP, an abdominal sepsis with a blood pressure of 60/40, and young man with pain out of proportion in his leg that turned out to be phlegmasia cerulea dolens, but that is a story for another time.)
In most cases, the process is simple. Thrombolytics are excellent drugs for STEMI, and the complications are pretty rare. (FTT 1994; Antman 2004) However, as I have gained more experience and seen more patients with borderline ECGs, I have paradoxically become less comfortable managing chest pain patients in an environment that relies on thrombolytics.
Hopefully everyone has heard or read the OMI manifesto. (If you haven’t, I have embedded a talk by Pendell Meyers below). I also hope that most people have discovered Dr. Smith’s amazing ECG blog. Combined, you quickly learn that not all acute occlusion MI (OMI) fits into guideline recognized STEMI patterns. There are a lot of subtle but extremely important ECG changes.
As a result, there is always some uncertainty in the diagnosis of STEMI. Occasionally that uncertainty leads to disagreements with cardiology. Occasionally, neither doctor knows the answer. When you are working in a system that uses PCI, this problem is mostly about inconvenience: making the cath lab team come into the hospital in the middle of the night. An unnecessary angiogram is a relatively low risk proposition. However, the use of lytics changes everything. Every patient who receives lytics has a chance of significant side effects. In this setting, if I misinterpret the ECG, the result isn’t just a grumpy cardiologist showing me a clean cath. Inappropriate lytics result in real harm to the patient, skewing the risk/benefit ratio. That harm has made me a little gun shy with borderline STEMI ECGs. That is, until I started thinking seriously about the mathematics of the treatment threshold.
What are the benefits and harms of thorombolytics?
Thrombolytics are clearly beneficial in patients with acute OMI, with STEMI being the surrogate ECG marker we generally use. Overall, thrombolytics provide about a 2% absolute reduction in mortality. (FTT 1994; Antman 2004) Earlier is probably better (although such statements are always based on problematic observational data), with the AHA citing about a 0.2% increase in mortality with every hour that passes. Put another way, benefit in the first 2-3 hours is about 3%, compared to 2.5% in the first 6 hours, and 2% in the first 12 hours, and there is no benefit by the time you pass 12 hours. (FTT 1994) There are probably other benefits as well, such as the prevention of CHF or cardiogenic shock, but those outcomes are less well described in the literature. To keep this post simple, I will just focus on the main benefit: mortality.
Of course, thrombolytics also cause harm. In the treatment of STEMI, there is about a 0.4% increase in hemorrhagic stroke and a 0.7% increase in other life-threatening bleeding. (There is a fair degree of uncertainty around these numbers, especially as they apply today, because not all trials used aspirin, let alone dual antiplatelet therapy, and about 5% of the included patients had normal ECGs.)
What is the treatment threshold?
The treatment threshold is all about deciding when the pretest probability of disease is high enough that you should just go ahead and treat, rather than doing any further testing. It is directly related to the test threshold (which calculates when the pretest probability of disease is low enough that you should stop your workup, rather than doing further testing.)
If we had perfect tests, we wouldn’t need to use this threshold approach. A perfect test is one that is 100% accurate (with no false positive, nor false negative results), as well as completely harm free, provides immediate results, and is preferably cheap. Unfortunately, no such test exists.
Thus, every time we order a test, we need to consider 2 very important outcomes. There are the false positives, which result in a treatment being given to patients who cannot possibly benefit but who will be exposed to all the harms. (This is the primary contributor to the test threshold). At the other end of the spectrum are the false negatives, which result in patients with disease being deprived of a potentially important treatment (the primary contributor to the treatment threshold). In addition to false positives and negatives, the test and treatment thresholds account the harms and benefits of treatment, and the harms and benefits of the test itself. (Things get a little more complicated when there are multiple options for testing, treatment, or both.)
I have included the full formula below, but for the purposes of this post, I am just going to use simplified estimates (which hopefully won’t upset purists too much). Let’s look at how the risk/benefit of thrombolytics changes I as become less certain about the diagnosis.
Translating statistics from a population level to the individual is somewhat difficult. An individual patient either lives or dies – there is no such thing as dying 3% of the time. For ease of reading, I just write the absolute risks as percentages, but “a 2% decrease in morality” really should be understood as “for every hundred patients like this patient, 2 fewer will die.”
If I am 100% sure that the patient has an occlusion of a coronary artery, treatment results in a 2% decrease in mortality, with only a 0.4% rate of intracranial hemorrhage, so clearly, I should treat.
If I am 80% sure that the patient has an occlusion of a coronary artery, it means that 2 out of every 10 patients I treat will receive no benefit. On average, this brings the benefit down to 1.6% (80% of the 2% benefit). The harms are unchanged (0.4%), as you are exposed to the risks of thrombolytics whether or not you are a true STEMI. Thus, even if I am somewhat uncertain about your diagnosis, the benefits of treating a high-risk potential STEMI outweigh the harms.
(Note: This math is a long way from perfect, because the trials that demonstrated the 2% benefit from thrombolytics would have included some of these borderline patients, so some of this uncertainty is already baked into the final numbers.)
20% sure – The breakeven point
If we continue to follow this logic, we will eventually come to a point that the harms equal the benefits. If only 1 in 5 (20%) patients actually have an acute OMI, the net benefit for all patients falls to 0.4%, which equals the harms. In this scenario, further testing is reasonable. (Yes, I know this is an oversimplification, because the harms – head bleeds – are qualitatively different than the benefits – mortality). However, I find this number very reassuring, because it tells me that even if there is considerable uncertainty, the benefits of treating still outweigh the harms.
50% sure – Delaying isn’t the same as forgoing
Delaying thrombolytics is not the same as forgoing thrombolytics altogether. Delaying decreases the mortality benefit but doesn’t eliminate it altogether. We would like to treat STEMI as soon as possible, but delaying thrombolytics by 15 minutes to see a second ECG only has a minimal impact on the overall benefit of treating. Even waiting an hour to see a troponin might be justified in the right patient. In fact, a prolonged delay to angiography might even be justified for the occasional patient, in the setting of significant uncertainty.
For patients who present in the first 2 hours, delaying thrombolytics by one hour could cause a 0.5% increase in mortality. Clearly, we want to avoid that delay. However, imagine the patient who presents with a borderline ECG and a reasonable story for ACS. You aren’t sure if this is STEMI. The cardiologist on the phone isn’t sure if this is STEMI. You are right on the fence – its 50/50. If you treat this patient immediately, the overall benefit is 1.5% (half of the 3% benefit seen in early presenters). On the other hand, a delay of 1 hour might increase our certainty to 100% if we see a positive troponin. This delay in treatment dropped the benefit from 3% to 2.5%, but because we are now 100% certain about the diagnosis, we expect the see the full 2.5% benefit. Therefore, waiting one hour actually increased the patient’s expected benefit from 1.5% to 2.5%.
Trying to make sense of these numbers
Playing with these numbers has made me a lot more comfortable managing STEMIs with thrombolytics. I think the most important lesson is to be comfortable treating even with a little uncertainty. If you think there is an occluded coronary artery, just give the thrombolytics. Even if you were wrong about the diagnosis 25% of the time (which is higher than my false positive cath lab rate), the net benefit still clearly comes down on the side of treating.
However, in scenarios with higher uncertainty, I don’t think we should feel bad about short delays if the delay will help make a more accurate diagnosis. A 10-minute delay for a second ECG will have a minimal impact on the effectiveness of thrombolytics. A longer delay for bloodwork might even be occasionally justified. For the right patient – say someone with only a 10% chance of acute OMI – this same line of thinking could be used to justify forgoing thrombolytics altogether, and transferring directly for PCI, even if that transfer is going to take 2 or 3 hours.
However, I worry that this line of thinking could get a little dangerous. Thrombolytics are a lifesaving treatment. They shouldn’t be delayed routinely. We don’t routinely need blood work. The benefit outweighs the harm even when there is significant uncertainty. (The breakeven point was 20% certainty.)
These examples are all somewhat oversimplified, but I think they make it clear why a threshold approach has made me more comfortable giving thrombolytics for STEMI. Nothing in life is certain. In the heat of the moment – at 4am staring at a borderline ECG – that uncertainty can feel overwhelming. Playing with these numbers has helped me tame that uncertainty. I have been reminded that thrombolytics provide significant benefit for STEMI patients, and that the benefit will be preserved even when faced with a little uncertainty.
Thank you for making it through a fairly unconventional post. I would love to hear your comments and questions.
A huge thank you to to Drs. Anand Swaminathan, Salim Rezaie, and Casey Parker for their help developing and providing peer review on this post on the treatment threshold.
Pendell Meyers discussing the OMI manifesto:
Antman EM, Anbe DT, Armstrong PW, et al. ACC/AHA guidelines for the management of patients with ST-elevation myocardial infarction–executive summary. A report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines (Writing Committee to revise the 1999 guidelines for the management of patients with acute myocardial infarction). Journal of the American College of Cardiology. 2004; 44(3):671-719. [pubmed] [free full text]
Aoki N, Sakai M, Ohta S. Test-Treatment Threshold. In: Kattan M. Encyclopedia of Medical Decision Making 2455 Teller Road, Thousand Oaks California 91320 United States . SAGE Publications, Inc.; 2009.
FTT: Fibrinolytic Therapy Trialists’ Collaborative Group. Indications for fibrinolytic therapy in suspected acute myocardial infarction: collaborative overview of early mortality and major morbidity results from all randomised trials of more than 1000 patients. Lancet (London, England). 1994; 343(8893):311-22. [pubmed]
Pauker SG, Kassirer JP. The threshold approach to clinical decision making. The New England journal of medicine. 1980; 302(20):1109-17. [pubmed]