Thoracotomy; should ultrasound be the decision tool?

Audio clip: Adobe Flash Player (version 9 or above) is required to play this audio clip. Download the latest version here. You also need to have JavaScript enabled in your browser.

Another FOAMed post on thoracotomy, surely we’ve had enough???

Well I’d tend to agree with you………. but performing a thoracotomy is one of the most stressful and time critical procedures you will do in Emergency Medicine. It also holds some real risks, mainly to the staff involved in the procedure with some alarmingly high sharps injury rates in this relatively high risk cohort of patients. Being really clear in your mind about how you will decide whether to perform a thoracotomy, should the situation rear it’s head, is vital prior to the situation arising.

And let’s be honest, although it’s a relatively infrequent event it’s an interesting topic!

Cardiac Ultrasound in Cardiac Arrest

The use of ultrasound is growing exponentially, in 2012 Blyth and colleagues performed a Systematic Review of Bedside Focused Echocardiography as Predictor of Survival in Cardiac Arrest Patients. They found ultrasound in cardiac arrest held a positive likelihood ratio for ROSC was 4.26 (95% CI = 2.63 to 6.92), and negative likelihood ratio was 0.18 (95% CI = 0.10 to 0.31). With these figures ultrasound certainly looks to add value to prognostication but doesn’t quite possess the ability to determine independently if resuscitation is futile, it’s not a binary decision maker.

The Paper

A recent paper published in the Annals of Surgery looks at the utility of ultrasound to determine if thoracotomy will be futile. The paper has been discussed in EMRAP by Rob Orman and it’s well worth a listen as well.

I wanted to run through this paper as I think it’s one people will reference a lot and be quoted as a reason to game-change, if nothing else I hope this blog and podcast encourage you to go and read through the paper yourself as I’m sure everyone will have their own interpretation of the evidence.

So, it was a prospective study in the United States in 2 ED’s over a four year period from 2010-2014 of patients undergoing resuscitative thoracotomy in the ED in cardiac arrest.

These were normally either penetrating trauma patients with absent vital signs and blunt trauma patients with a loss of vital signs en route or in the resuscitation room, and in general patients who presented with this history would get a thoracotomy.

The primary outcome was the survival to discharge or organ donation. The FAST scan was performed either just prior to or during the thoracotomy and was performed by a PGY2-4 under direct faculty supervision. It’s important to note that all residents complete a 2 day, 16 hour ultrasound course in a addition to a minimum of 2 weeks training in POCUS.

Ultrasound findings were documented as adequate of inadequate with cardiac motion defined as organised, non-fibrillating contractions.

What did they find?

Well 223 cases underwent resuscitative thoracotomy during this period, with 83.9% having a FAST performed, which formed the study cohort (worth thinking that that is an average of 1 every second week per centre!).What we’re not able to know is how many patients had a FAST scan and had a decision then made not to proceed to thoracotomy

The sort of patients that we’re talking about- well the average age was was 31 (1-84) with nearly 85% being male, just over half were due to penetrating trauma and the most common mechanism was GSW. These also tended to be multiply injured patients with a concomitant head injury in nearly half of cases and multiple other areas tended to be involved as well.

22.5% of the patients undergoing thoracotomy in the ED had cardiac injury but none of these patients survived! There were 21 patients during this time period that were identified by FAST as having pericardial effusions from penetrating thoracic injuries but all were taken to the OR where 1/3 arrest ed en route.

Vital signs were lost on scene in roughly a half of patients, a quarter en route and a quarter in ED (NB the median duration of transport time was 33 minutes).

3/4 of patients received a left anterolateral thoracotomy.

Nearly a half of patients (48.7%) regained cardiac motion but only 3.2% survived and a further 1.6% of patients became organ donors.

Cardiac motion was detected in 28.9% of patients.

The sensitivity for FAST in identifying survivors or organ donators in this population was 100% (9 patients) and a specificity of 73.7%. The authors state that the negative predictive value of FAST in this context is 100%. There is no comment at all about the confidence intervals surrounding the sensitivity and specificity with this data set.

So what does this paper mean for me?

So this is where opinion is likely to be wildly split, without confidence intervals around the sensitivity and specificity I can’t get areal feel for what this small data set means, if only one patient had been deemed to have no cardiac motion but had gone onto survive then that would have slashed the sensitivity to a significantly lower value- around 90%. In this setting we’re looking for a test which has as higher sensitivity as possible, that way we can use it as a rule out test- if no cardiac motion then performing a thoracotomy becomes futile. So it is really likely that the 95% confidence intervals will have a large spread with such a small cohort of positive outcomes and it would be prudent to interpret these with a significant degree of caution.

This paper is certainly food for thought but with such a low pretest probability of survival in this patient cohort if you’re going to use USS in traumatic cardiac arrest to solely inform your practice and decide whether to proceed to thoracotomy then you’re going to need to at least wait for a paper with a larger patient population, accurately powered and with a population that reflects yours.

From personal experience I have used USS prior to the initiation of 2 thoracotomies, although I can’t be sure I’m pretty certain that irrespective of what I’d found on those 10 second scans I’d probably have proceeded to thoracotomy either way. What the images gave me was probably added confidence that performing a thoracotomy was the right thing to do, a cognitive safety blanket really.

As with almost every part of EM it’s piecing together the different parts of the puzzle that paints the picture and unfortunately it’s rare that one prognostic alone is enough.

It’d be great to hear your thoughts on the paper so please leave any comments at the bottom of the page.

Simon

 

References

Blyth, Lacey, et al. “Bedside focused echocardiography as predictor of survival in cardiac arrest patients: a systematic review.” Academic Emergency Medicine19.10 (2012): 1119-1126.

Inaba, Kenji, et al. “FAST Ultrasound Examination as a Predictor of Outcomes After Resuscitative Thoracotomy: A Prospective Evaluation.” Annals of Surgery262.3 (2015): 512-518.

Add this content to your CPD

10 thoughts on “Thoracotomy; should ultrasound be the decision tool?

  1. Mario Rugna says:

    I think ultrasound is a game changing tool in prehospital emergency medicine. Is of great utility in medical cardiac arrest in the determination and treatment of reversible causes and I (and the where I work) routinely use it to guide resuscitation efforts. I’m more concerned about its utility in traumatic cardiac arrest where I prefer (if decided to start resuscitation) a systematic (and total) approach to address all the reversible causes independently from diagnostic tools.
    Thank Simon for spreading your thoughts and let us comment on this interesting topic.

  2. sl says:

    Hi Mario

    Thanks for your comments.

    I agree that Ultrasound has got great utility in EM and multiple other specialties.

    I guess if there are poor prognostics that have already identified the patient as a very low risk of survival with a good neurological outcome (e.g extended down times prior to the consideration of thoracotomy) then POCUS does become useful, but completely agree that in isolation it would not be wise as the sole decision tool.

    Regards

    Simon

  3. Mario Rugna says:

    Thanks for the reply.

  4. Pete Finnegan says:

    Hi Simon,
    Great analysis and important point on the lack of published confidence intervals.

    Unfortunately the author choose to analyze results using predictive values.
    Predictive values as a statistical test performs less well the lower the prevalence.
    As the number of survivors with cardiac motion on ultrasound are low the prevalence is therefore low.

    However, I still think there’s still valuable info to be gained from looking at the survivors (all had witnessed cardiac motion on FAST)
    and the significance of this by calculating liklihood ratios.

    Likelihood ratios are independent of disease prevalence, unaffected by population and can be applied on an individual patient level.

    Calculating the likelihood ratios for FAST findings gives us a +ve of 3.84 and a -ve of 0 therefore the likelihood of a patient surviving RT following cardiac motion on FAST is increased nearly 4-fold whereas the likelihood of survival with no motion on FAST is 0!

    Witnessed cardiac activity on ultrasound does select potential survivors.
    Game changing?

    Pete Finnegan,
    ED Trainee,
    Alfred Trauma Centre,
    Melbourne

  5. sl says:

    Hi Pete

    Thanks for the comments, great to hear everyone’s opinions on the papers (that what makes FOAM so useful for learning for me)

    The point for me with this paper is that it is easy for people to draw conclusions like the negative likelihood ration is 0, but in reality the small population looked at makes this a possible oversight, confidence intervals would have shown that 0 should be taken with a pinch of salt as they would show a range which would probably spread over 0.1.

    Never the less I completely agree that if you can see activity that’s good, if you can’t then that’s pretty bad! I was conscious of the fact that other very respected & intelligent people in the FOAM community were starting to suggest that this may be used as a decision aid alone. And whilst I do think it adds a strong amount to the prognostication of patients my opinion is that at the moment it shouldn’t be used in isolation.

    Thanks for getting in touch!

    Simon

  6. Hjalti M Bjornsson says:

    Interesting and helpful study. Agree with the comments above that statistically there still is some doubt but overall I think this will make everybody more comfortable with making the decision to cut or not.

    I still however feel that the authors have completely missed one of the most important points in their data. When you read what has been published on ED thoracotomy, many have looked at the data and concluded that the time since vital signs were lost is of key importance, citing 5 min for blunt trauma and 15 min for penetrating trauma as a cut off doing an ED thoracotomy.

    With US, I think this thought should be completely abandoned. We all know how unreliable pulse detection by finger can be and that patients can be without clinical vital signs but still have a SBP of 50 and able to maintain some perfusion.
    So, it is surprising that I could not find any info in their paper on time since VS were lost, only location. Personally I feel that we should completely abandon the time criteria and base our decision on ED thoracotomy on US instead.

  7. sl says:

    Hjalti

    Thanks for the comments, really interesting to hear different takes in the topic.

    My problem with abandoning the pulse check is that almost all of the literature to date has encompassed the pulse check, rather than the use of US to assess for pseudo-PEA. If you completely abandon the pulse and timeframe as an indicator for the appropriate cohort to be performing a thoracotomy on then we know near to nothing about the group and their prognostics.

    I think without further work on your indications for thoracotomy it’s difficult to just switch over.

    Simon

Leave a Comment

Your email address will not be published.

This site uses Akismet to reduce spam. Learn how your comment data is processed.