At EMSy, we have been discussing the transition from the STEMI/NSTEMI paradigm to OMI/NOMI for some time. Not as an academic provocation, but as a clinical necessity: the OMI Manifesto has opened a path, now daily practice needs to walk on it.
When Ilaria Costantini proposed this article to us, we immediately understood it was the right piece at the right time. Because it doesn't just argue that STEMI criteria are insufficient. It puts them to the test with a systematic review and meta-analysis, published in Emergency Care Journal. And the numbers that emerge are those that, deep down, we all suspected: one in three coronary occlusions escapes STEMI criteria.
Ilaria is an Emergency Medicine Physician at the Emergency Department of Arzignano (Vicenza), passionate about cardiovascular emergencies, ultrasound and Evidence-Based Medicine, with a declared weakness for emergency toxicology and how we reason under pressure. She is co-author of the Handbook of Emergency Toxicology. She conducted this research independently, without institutional funding, with only the strength of method and clinical curiosity.
When she's not in the ED dismantling dogmas, she runs in the hills and dives underwater. In both cases, she goes deeper than expected.
Simon Grosjean
The clinical case: a hiker with chest pain
I was on duty in the emergency department when a man in his fifties arrived, apparently healthy, athletic. During a mountain hike, and we're talking about approximately 20 kilometers, not a simple walk downtown, he had developed oppressive-type chest pain.
The classic pain that we all recognize as "typical."
On arrival he was symptomatic, frankly suffering. Yet, he had no risk factors: not diabetic, not hypertensive, no significant family history. He did regular physical activity, indeed intense, and had never reported symptoms in the preceding days.
As always in these cases, the first thing we did was an ECG.
The report? "Not STEMI."
Yes, sure. I didn't have a "textbook" ST segment elevation, the kind that everyone immediately recognizes as STEMI. But neither was it a normal ECG.
What struck me most was this: in the precordial leads there was ST segment depression, and tall symmetric T waves. In aVR instead there was ST segment elevation, but only in aVR... essentially the patient did not meet STEMI criteria for sending to the cath lab... but is that right?
From STEMI/NSTEMI to OMI/NOMI: the paradigm shift
Emergency physicians who are absolute ECG nerds have already understood that we are facing a de Winter pattern. In our heads we have already made the paradigm transition from STEMI/NSTEMI to OMI/NOMI and we have already understood and accepted that "bad infarctions" exist even if the ECG does not meet STEMI criteria.
In fact, these are OMIs (Occlusion Myocardial Infarction), a more clinical and pathophysiological definition compared to STEMI (ST Elevation Myocardial Infarction) which has a purely electrocardiographic definition. It follows that by logical rigor a STEMI can be an OMI and an OMI can also be STEMI... or even something else.
For a detailed discussion I refer to the OMI Manifesto [1].
The Bayesian reasoning we do
Even though we physicians don't love numbers and tend to see probability or statistics as something abstract and distant, in reality, in front of every patient, we always do Bayesian reasoning.
We start from a sign, from a symptom, from the clinical picture as a whole. We begin to ask ourselves which chapter of Harrison it corresponds to. We form an idea. This is our initial probability. Then we do tests or investigations (diagnostic tests) that based on their outcome allow us to exclude or include some causes... that is, they modify our initial probability into something more accurate.
In our case, this patient, for the type of pain he reported and how he presented, in our heads had a certain probability of infarction. We did the ECG and became even more convinced it was an infarction. If it had been an ECG with STEMI characteristics, the game would have been closed.
But a diagnostic test is never a perfect test.
To be truly useful in clinical practice, it must be applied to the right patient, at the right time. Not "blanket," as we well know.
Moreover, to interpret it correctly, we must know its performance: sensitivity, specificity, limitations.
The question no one had asked
In clinical practice, I began to recognize cases of infarction that would have previously escaped me. "Real" infarctions, from occlusion, but without STEMI criteria.
And at that point a seemingly trivial question arose: what is the diagnostic accuracy of ECG STEMI criteria for occlusive myocardial infarction? I expected to find a clear, quick, certain answer. In fact, I felt embarrassed not to know it. I asked myself why I hadn't asked this question. For D-dimer I asked it many times.
Because STEMI criteria are not improvised: they have been defined, validated, integrated into guidelines for over twenty years. And on this distinction rests the entire organization of the time-dependent system for chest pain.
No answer, literature gap. I found two meta-analyses that tried to estimate the prevalence of occlusive infarctions among NSTEMIs [2, 3], but on the performance of the "ECG STEMI criteria" test, nothing.
Yet we all believe in it. We are facing what we can define as a dogma, but "ST elevation is not the disease."
The study: method and definitions
At that point we tried to give an answer to this question ourselves. That is: how accurate is the STEMI paradigm in recognizing a true coronary occlusion?
In our work, we tried to quantify this [4].
We conducted a systematic review of studies published from 2000 onwards, selecting those that reported both ECG data and angiographic results, so as to be able to directly compare STEMI and actual occlusion.
To do this, we first clearly defined the disease of interest:
OMI = acute coronary occlusion, defined as TIMI flow 0-2, or TIMI 3 associated with very high troponin values, indicative of a real occlusion with significant myocardial damage (angiographic criteria).
This was our "reference truth."
At that point we compared:
- ECG (STEMI vs non-STEMI)
- with angiography (OMI vs non-OMI)
The problem of heterogeneity
Surprisingly, we did not find many studies. In total, only nine papers were eligible for analysis. And this is already a finding.
Because we are talking about one of the pillars of emergency medicine: the ECG in STEMI. Yet, the studies that truly evaluate its diagnostic accuracy with respect to coronary occlusion are few.
The included studies were very heterogeneous. Different in the population studied, in how the disease state (OMI) could be defined, and in study design.
For some a purely angiographic definition was possible (TIMI flow 0-2), others a composite definition, also including elevated troponin or echocardiographic data.
And this has a direct impact on the results.
For those paying closer attention, a crucial point immediately emerges: the oldest study is Koyama's from 2002, and STEMI criteria have undergone variations (although not dramatic) since then. All this introduces an important source of heterogeneity.
The results
We nevertheless tried to answer the initial question, with the data we had available.
Figure 1. Forest plot of the sensitivity of ECG STEMI criteria in recognizing occlusive infarction (OMI), stratified by subgroups.
Figure 2. Forest plot of the specificity of ECG STEMI criteria in recognizing occlusive infarction (OMI), stratified by subgroups.
Figure 3. SROC curve (Summary Receiver Operating Characteristic) with AUC 0.76 (0.71-0.83).
And the results were, in part, surprising.
Approximately one-third of patients with coronary occlusion were STEMI-negative.
One patient in three.
Is that a lot? The candy metaphor
The question, at this point, is no longer technical. It's clinical. Is that a lot? Is it a little?
To understand, we can change perspective.
Imagine a jar with 10 candies. Three or four of these contain a deadly poison. Would we trust taking one at random?
Probably not.
Yet, it's exactly what we do when we use STEMI as the only filter to decide who to send to the cath lab immediately. STEMI criteria are our triage with which we decide who to treat immediately and who can wait instead.
But in fact a STEMI-negative does not exclude an occlusion.
Are we really willing to accept "missing" one patient in three?
Because it means that STEMI is not an exclusion test. And a test used at Triage must be highly sensitive so as not to leave anyone behind.
Bayesian reasoning, again
We can no longer stop at "it's not STEMI so I'll wait for troponin."
We must start asking ourselves: how likely is it still that this is an OMI?
And here Bayesian reasoning returns.
We start from an initial probability based on the clinical picture. Then we add the ECG. STEMI yes or STEMI no, but with sensitivity around 60%, means that a negative result doesn't lower the probability enough. So then we can take our ECG back in hand and review it from other points of view looking for STEMI equivalent patterns or other patterns that are emerging as indicative of OMI.
How much can we trust?
We said that our study is affected by significant heterogeneity: different studies, different populations, different definitions of OMI. All factors that, theoretically, could distort the real estimate.
And so the question is inevitable: how much can we trust these results?
Surprisingly, a lot. Because when we compare our data with those in the literature, we find remarkable consistency. Our results are substantially overlapping with those of another meta-analysis that evaluates the diagnostic accuracy of ECG STEMI criteria for OMI among patients with ROSC [5]. They are also in line with the results achieved by meta-analyses that estimated the prevalence of OMI among NSTEMIs [2, 3].
And when different results, obtained with different approaches, converge at the same point, they become strong. For this reason, even if imperfect, our estimate is plausible and clinically relevant.
The epilogue
But let's return to that patient.
Typical clinical picture, ongoing pain, ECG with de Winter pattern... and he was one of that 30-40% of patients... at angiography TIMI 1 at the LAD.
References
- Meyers HP, Weingart SD, Smith SW. Dr. Smith's ECG Blog: The OMI Manifesto. [https://hqmeded-ecg.blogspot.com/2018/04/the-omi-manifesto.html](https://hqmeded-ecg.blogspot.com/2018/04/the-omi-manifesto.html)
- Khan AR, Golwala H, Tripathi A, et al. Impact of total occlusion of culprit artery in acute non-ST elevation myocardial infarction: a systematic review and meta-analysis. European Heart Journal. 2017;38(41):3082-3089. doi:10.1093/eurheartj/ehx418
- Hung C-S, Chen Y-H, Huang C-C, et al. Prevalence and outcome of patients with non-ST segment elevation myocardial infarction with occluded "culprit" artery – a systemic review and meta-analysis. Critical Care. 2018;22(1):34. doi:10.1186/s13054-018-1944-x
- Costantini I, Breglia A, Pilav M, Zanatta M, Ventura MV, Ricci G. Missed occlusions: diagnostic accuracy of ECG STEMI criteria for identifying occlusive myocardial infarction. A systematic review and meta-analysis. Emergency Care Journal. 2025. doi:10.4081/ecj.2025.14233
- McFadden P, Reynolds JC, Madder RD, Brown M. Diagnostic test accuracy of the initial electrocardiogram after resuscitation from cardiac arrest to indicate invasive coronary angiographic findings and attempted revascularization: A systematic review and meta-analysis. Resuscitation. 2021;160:20-36. doi:10.1016/j.resuscitation.2020.11.039
