Friday, December 2, 2011

Not just S1-Q3-T3: Pulmonary embolism and the ECG

When it comes to pulmonary embolism, there is more to the ECG than the presence or absence of S1-Q3-T3. Hard to believe, but true! Unfortunately, none of the other signs get quite as much press.
Not helping my case.

Let me give you the conclusion of this rant up front, in case you have better things to be doing.
When you are evaluating a patient for a possible PE, there are 4 things you should be looking for on the ECG:
  1. Tachycardia
  2. Incomplete or complete RBBB
  3. T-wave inversions in V1 through V2, V3, or V4.
  4. S1-Q3-T3
 Probably there are two reasons for this relative ignorance of the utility of the ECG in this setting. The first is a sense of nihilism. For example, how often have you received a "pimping" question from an attending along the lines of "What's the most common ECG finding in PE?" The attending-dependent answer is either tachycardia, or no ECG findings. Either answer may be true, if appropriately phrased and qualified, but really, you can't win.

Reprogram the scenario?

The implicit lesson is that the ECG will not be helpful in evaluating for PE.

A second reason, at least at Yale, is that we have been so well tutored on the use of electrocardiography to assist in the diagnosis of PE that the ECG seems quaint by comparison. After all, why would you care about indirect assessments, when you can just look at the heart?

Really, which of these technologies would most appeal to the Zune generation?
The essential study to know is the Daniel paper, from Chest in 2001. You should print it up and hit a coffee shop, beacause a quick glance at the abstract is not going to do it justice, so go ahead and download it here. Now, let me set the context, because studies on PE are nuts: the evolutions in technology and prevalence are just moving along so fast, it's really hard to keep up (Like with Zunes!).

The study used ECGs that were obtained on patients who were seen in the ED, and were suspected of having a PE. In current studies, that means a d-dimer was ordered, but in this study, back in the period 1997-1999, it meant patients who had a pulmonary angiogram. The rate of positive angiograms was 43%, showing that this was a group with a fairly high pretest probability of PE.

The authors developed an ECG scoring system, based on prior literature and case series, and wanted to see how well it could predict the diagnosis of PE. The nice part is that they used EM physicians (attendings and senior residents) as the readers, not cardiologists, giving it some real-world applicability.

The first important finding is that the ECG is not likely to be the best tool for diagnosing or ruling-out PE. Tachycardia was the most significant abnormality found on everybody, PE or no. A complete S1Q3T3 was found in only 7% more people with PE than without - not real helpful.

We already knew this, but I'm just going to stop there and reemphasize this results: You can't just look at any one sign on the ECG,  S1-Q3-T3 included. It won't help you.

They ran through the results a few different ways, and it's interesting to check out all the plots and bar graphs. The money graph, however, shows how calculating the total score can be used.

Put into English:

A score of ≥10 has a specificity of over 97% for the combined presence of a pulmonary embolism with severely elevated pulmonary artery pressures. 

That's a sick, sick patient. Of course, sensitivity stinks, less than 25%, but hey.

At least one case report has highlighted this use of the ECG - when you have a high Daniel score, and for various reasons you can't get imaging, it helps justify the empiric use of anticoagulants.

Well, this study doesn't really resemble what we do today, with al these pulmonary angigrams and such, so should the results inform our current practice? Marchick decided to run all the ECGs in Kline's PERC study through the Daniel score, to see how it could help.

Remember, to get into the PERC study, a patient just had to have some test for PE ordered - CT, V/Q, US, or a d-dimer. Not surprisingly, the rate of PE found in that cohort was below 6%. This is a far cry from the Daniel study, where 43% were positive!

Using a slightly modified Daniel Score, they found that the ROC curve for the score had a area of 0.61, consistent with the original study. More importantly, they looked at the individual elements:

Both S1-Q3-T3 and T-wave inversions from V1 to V4 have an LR+ of 3.7, which makes them fairly good, but not high enough to be decisive on their own. Fortunately, per their logistic regression...

... we learn that these two elements are independently predictive, and so we can use those, plus tachycardia, to suggest a PE. (Of course, this has to be in the right context. A patient with known pulmonary hypertension from other causes will show these same signs of RV strain).

So what? you say, holding your phased array probe over the patient, blue goop running down your wrist. If you can visualize RV dilitation, paradoxical septal motion, or perhaps even tricuspid regurgitation, what use is the ECG?

Echo: Useful in the evaluation of rigor mortis.
Some folks in the ED in Florence , Italy put down their espressos long enough to investigate the incremental value of the ECG and the echo in the prognosis of PE (not diagnosis - all the patients were known to have a PE).

It gets a little complicated for my little brain, but they essentially took a group of people who had been diagnosed with PE, and looked at 3 things:
  • Which of the patients had signs of RV strain on the ECG, by Daniel criteria.
  • Which of the patients had signs of RV dysfunction on echocardiography.
  • Which of the patients had clinical deterioration or died.
So, when they boiled down the data, they found 2 important results. First, that ECG signs of RV strain are an independent predictor of clinical deterioration. It isn't a surrogate marker for echo data. Second, it appeared that signs of RV dysfunction by echo only predicts deterioration if there are already ECG signs of RV strain. 

The graph below show that echo evidence of RV dysfunction did not predict a bad prognosis unless there were coincident signs of RV strain (RBBB, T-wave inversions, etc.) on the ECG.

So, conclusion:

If you are looking at the ECG, and PE is in the differential, you need to look for all the possible indicators of right ventricular strain on the ECG. It gives you both diagnostic and prognostic information. No one will be pimping you on calculating the exact Daniel score, but you should be scanning for all the elements.


  1. Here's to warning medical students everywhere of the pitfalls of the dreaded ECG in PE question!