Saturday, January 27, 2024

Chest pain and a computer ‘normal’ ECG. Therefore, there is no need for a physician to look at this ECG.

Written by Jesse McLaren, comments by Smith

A 55 year old with a history of NSTEMI presented with two hours of exertional chest pain, with normal vitals. Below is the triage ECG, with a computer interpretation (Marquette 12 SL) of “normal” which was confirmed by the over-reading cardiologist.

What do you think? 

Should this patient continue to stay in the waiting room, without interruption of the physician to interpret the ECG, because the computer interpretation is normal?

Interpretation by the GE/Marquette 12 SL conventional algorithm

Smith: This article, published this month (!), tells us that we physicians do not need to even look at this ECG until the patient is placed in a room because the computer says it is normal:

Validity of Computer-interpreted “Normal” and “Otherwise Normal” ECG in Emergency Department Triage Patients

I reviewed this article for a different journal and recommended rejection and it was rejected.  There were zero patients in this study with a "normal" ECG who had any kind of ACS!  This defies all previous data on acute MI which would show that even undetectable troponins do not have a 100% negative predictive value.  So this study is actually worthless.  

On the other hand, if the physician is unable to recognize subtle OMI, as is the case with the overreading cardiologist, then the conclusion would be correct.


ECG analysis

There’s normal sinus rhythm, normal conduction, normal axis, normal R wave and normal voltages. What sticks out is the ST depression in aVL, which is reciprocal to subtle inferior ST elevation and bulky T waves.  There is also a down-up T-wave in aVL, which makes aVL even more diagnostic.  This is diagnostic is inferior OMI, accompanied by inferior Q waves, and with a flat ST segment in V2 that could indicate posterior extension.


A history of coronary artery disease and a stent to the same territory further increases pre-test likelihood of acute coronary occlusion, including in-stent thrombosis. It’s also possible that the old inferior MI left residual ST elevation and reciprocal ST depression, which can difficult to differentiate from acute OMI.

The patient had a history of ‘NSTEMI’ a decade prior, with an RCA stent. Does this change your interpretation? 


Smith: Old inferior MI with persistent ST Elevation ("inferior aneurysm") has well-formed Q-waves.  In inferior aneurysm, we usually see QR-waves, whereas for anterior aneurysm, we see QS-waves (no R- or r-wave at all!).  This ECG has Q-waves, but they are not very wide nor very deep, and so I doubt that the inferior STE is due to old MI.

See these posts: 

Chest Pain, ST Elevation, and an Elevated Troponin: Should we Activate the Cath Lab?

Pulmonary Edema, Hypertension, and ST Elevation 2 Days After Stenting for Inferior STEMI


There’s limited information from the NSTEMI decade ago, except that the cath report describes a critical RCA stenosis treated with a stent, and the following discharge ECG:

This shows inferior Q and inferoposterior reperfusion, corresponding to RCA territory. So this NSTEMI was likely a STEMI(-)OMI with delayed reperfusion. A couple of years later these ECG changes resolved, except for small inferior Q waves, leaving an almost normal baseline ECG. Below are the baseline and new ECG from the current patient presentation:

This confirms that the new ECG represented acute inferoposterior OMI: the inferior ST segments have straightened, increase the area under the curve of the T waves, there is new ST depression in aVL which is highly sensitive for inferior OMI, and there’s flattening of the ST segment in V2. 

But the ECG was labeled 'normal', and the patient waited to be seen.



Two hours later the first troponin I returned at 450 ng/L (normal <26 in males and <16 in females), which flagged the patient to be seen. By this time their pain had spontaneously resolved, and another ECG was done which was also interpreted as normal. What do you think?

On its own this ECG is nonspecific, with in isolated T wave inversion in III. But compared to the previous ECG the inferior T waves have deflated, the ST depression in aVL has resolved, and the ST segment in V2 is no longer flat and has a taller T wave. This confirms that the previous ECG with pain represented inferoposterior OMI, and that the current ECG with resolved pain represents reperfusion.

So the patient had a transient acute coronary occlusion that spontaneously reperfused but is at risk for reocclusion. The patient was admitted as ‘NSTEMI’ which is supposed to represent a non-occlusive MI, but the underlying pathophysiology is analogous to a transient STEMI. Fortunately the patient did not reocclude while awaiting the angiogram. Next day ECG showed ongoing reperfusion:

Angiogram found 90% RCA in-stent thrombus corresponding to the ECG, peak troponin was 12,000 ng/L which is a sizable infarct, and echo showed new inferior wall hypokinesis. This retrospectively confirms the diagnosis of OMI, yet the patient had a discharge diagnosis of ‘NSTEMI’.

Discharge ECGs showed inferior Q waves, and now very obvious inferoposterior reperfusion T wave inversion (the same pattern as after their prior ‘NSTEMI’): TWI in III/aVF with reciprocal tall T waves in aVL, and tall T wave in V2 reciprocal to posterior TWI:

Below are the baseline ECG(#1),  ECG with occlusion (#2), and follow up ECGs with progressive reperfusion – from initial normalization (#3)to progressive reperfusion TWI (#4-5):


Another study has claimed that computer interpreted ‘normal’ ECGs can “safely wait for physician interpretation until the time of patient evaluation without delaying an acute STEMI diagnosis.” (Deutch et al. Validity of computer-interpreted ‘normal’ or ‘otherwise normal’ ECG in emergency department triage pateints. West J Emerg Med 2024). They compared computer interpretations with cardiologist interpretations and final diagnosis of STEMI, and found a Marquettte 12 12SL had 100% negative predictive value of STEMI. But in this case, the ECG did not meet STEMI criteria and therefore the patient did not get emergent reperfusion and did not have a diagnosis of STEMI, so their ‘normal’ ECG (also by Marquette 12 SL) would be considered valid despite the patient having an acute coronary occlusion that was visible on ECG. Fortunately they spontaneously reperfused, or else they could have had a worse outcome, but deferring all ‘normal’ ECGs will perpetuate delayed diagnosis and reperfusion for STEMI(-)OMI.

As Deutch et al note in the limitations section, their study “does not directly address other outcomes of interest to an emergency physician such as acute coronary occlusion MI (OMI) which may benefit from timely reperfusion therapy…Moreover, there is a growing body of literature supporting a paradigm shift from evaluating ECGs for STEMI vs no STEMI as an indicator of OMI that may benefit from emergent reperfusion to evaluating ECGs for signs acute total OMI (inclusive of STEMI negative OMI) vs non-OMI.”

This same body of literature has highlighted the hazards of computer-interpreted ‘normal’ ECGs – including dozens of cases on this blog and a 7-year retrospective review that found 4% of true positive Code STEMIs presented with an ECG labeled ‘normal’ by computer interpretation—many of which were identified in real time despite the false reassurance of the computer interpretation and had rapid reperfusion. This 4% is underestimation because it only included patients admitted as STEMI not those admitted as NSTEMI like the case above.

Rather that comparing conventional computer interpretations with STEMI criteria, the real goal standard should be patient outcome of OMI, and then AI can be trained to look for subtle signs of occlusion. I sent the first three ECGs to the Queen of Hearts: even without comparison to baseline the first ECG was identified as OMI, the second was called Not OMI because it had normalized, and the third was called ‘OMI’ because the Queen is currently trained to apply this label to reperfused OMI as well. Future versions will be able to integrate serial ECGs, and differentiate between OMI and reperfused OMI.

Let's look at explainability for that first ECG:
You can see that it is aVL which the Queen is most concerned about.
 She diagnoses OMI with high confidence.
She was correct
And this is in an EKG that the conventional algorithm diagnosed as completely normal!!  
That algorithm could at least have given a diagnosis of "Nonspecific ST-T abnormalities", but it could not even do that!!


If you want this bot to help you make the early diagnosis of OMI and save your patient and his/her myocardium, you can sign up to get an early beta version of the bot here.

Take away

1.     Computer interpretations of ‘normal’ are based on STEMI criteria, which will miss STEMI(-)OMI

2.     ‘NSTEMI’ does not differentiate between occlusion, reperfusion at risk of reocclusion, and non-occlusive MI

3.     Patterns of occlusion and reperfusion can be learned and taught, including to AI

4.  The Queen of Hearts not only recognizes this "normal" ECG as not normal, but correctly diagnoses OMI with High Confidence


MY Comment, by KEN GRAUER, MD (1/27/2023):

Superbly illustrated case by Dr. Jesse McLaren — showing that serial ECGs called "normal" and "NSTEMI" were in reality diagnostic of prior and current OMIs, including evolving patterns of reperfusion.
  • To Emphasize: In a patient with new CP (Chest Pain) — the oversight of calling the initial ECG in today's case "normal" — is a mistake that should not be made.

  • For clarity — I've labeled this initial ECG in Figure-1.

Figure-1: I've labeled the initial ECG in the ED.

KEY Findings in the History:
As per Dr. McLaren, despite a past history of prior MI — and a presenting history of 2 hours of new CP — ECG #1 in Figure-1 was interpreted as "normal" by the cardiologist overreading the computer interpretation — and — this patient was left in the waiting room without his/her ECG being immediately interpreted by the ED physician.
  • KEY Point: All patients who present to the ED for new CP should promptly have a triage ECG recorded, that is then immediately interpreted by the ED physician. Today's case illustrates what happens when this procedure isn't followed. It literally should take an experienced ED phsyician minimal seconds to interpret the initial ECG of a patient with new CP.
  • Cardiologists assigned to overreading the initial ECG of a patient with new CP should: i) Not allow themselves to be biased by what the computer says until they have independently interpreted the ECG; andii) Account for the clinical reality that a patient who presents to the ED for new CP is by definition in a "high-prevalence" group with significantly increased likelihood of having an acute event — especially if this patient has a history of a prior MI (as in today's case). As a result — even subtle ECG abnormalities have to be assumed acute — until proven otherwise (which was obviously not done in today's case).

KEY Findings in the Initial ECG:
In a patient with new CP — emergency providers should pick up on the following ECG findings within seconds:
  • The shape of the ST-T wave in lead V2 (within the RED rectangle in this lead). As we often emphasize — there normally should be gentle upsloping with slight elevation of the ST segment in leads V2 and V3. When this normal feature is lost in one or both of these anterior leads in a patient with new CP — and is replaced by an isoelectric or slightly depressed ST segment — posterior OMI should be strongly considered until proven otherwise.  
  • Not only is the ST segment in lead V2 of Figure-1 isoelectric and straightened — but there is abnormal angulation between this straightened ST segment and the taller-than-expected T wave in this lead V2. The shape of the ST-T wave in this lead V2 is a "face" that should be instantly recognized as abnormal in a patient with new CP.
  • Confirmation that the ST-T wave in lead V2 is abnormal — is forthcoming by the obviously "hypervoluminous" T wave in neighboring lead V3 (within the BLUE rectangle in this lead — with the dotted BLUE line showing that this T wave in V3 is even taller than the R wave in this lead).

Abnormal Findings in the Limb Leads of Figure-1:
  • I've enclosed within the RED rectangle in lead aVL the KEY limb lead change that should immediately catch your attention. In a patient with new CP — there is no way that the downsloping ST depression with terminal biphasic (down-up) T waves in this lead can be normal (RED arrows in lead aVL).
  • The fact that the ST-T wave in lead aVL is definitely abnormal — should heighten your attention to ST-T wave appearance in the inferior leads, since there is so often that "magical" reciprocal (mirror-image opposite) relationship between the ST-T wave deviation in lead aVL vs lead III (as well as in the other 2 inferior leads = leads II and aVF).
  • In this context — the subtle-but-real "fattening" of the peak of the T wave in each of the inferior leads (upright BLUE arrows in these leads) — with subtle straightening of the ST segment takeoff and a hint of J-point ST elevation in leads II,III,aVF is abnormal until proven otherwise.

Are the Q waves in the inferior leads of Figure-1 "significant"?
  • Over the years — various definitions have been proposed for what should constitute a "significant" Q wave, based on width and/or depth of the Q wave. The obvious implication of such definitions — is that IF a given Q wave is deemed "significant" because it satisfies a certain millimeter-based definition — that this then indicates infarction has occurred at some point in time.
  • I feel such definitions are misleading because: i) Rather than some millimeter-based global definition for Q wave "significance" — other factors defy such generalization (ie, Relative size of the QRS in the lead being looked at in context with the presence or absence of Q waves in neighboring leads)ii) Q waves are not necessarily a permanent finding following infarction — in that what used to be a very large Q wave may with time decrease in size (and even disappear); iii) Axis shift and/or inconsistant chest electrode lead placement may influence both the presence and dimensions of any Q waves seen; andiv) The angle of the bed at the time the ECG was recorded may also a influence whether or not Q waves are seen.
  • BOTTOM Line: In my opinion — there is no perfect definition for what constitutes a "significant" Q wave (ie, a Q wave that indicates infarction has occurred at some point in time). Instead — this is a qualitative judgment to be made by the experienced clinician on the basis of a series of factors. My initial impression regarding the inferior Q waves in Figure-1 (which I arrived at before having seen prior tracings on today's patient) — was that these Q waves are not "deep" (considering the relatively tall R waves in leads II,III,aVF) — but that these Q waves are wider-than-I-would-normally-expect — and therefore, provide yet one additional ECG feature consistent with inferior infarction at some point in time. 

No comments:

Post a Comment

DEAR READER: I have loved receiving your comments, but I am no longer able to moderate them. Since the vast majority are SPAM, I need to moderate them all. Therefore, comments will rarely be published any more. So Sorry.

Recommended Resources