PE Prognostication with Dr Anand Senthi (part 2)

Ok, its here – the second part of the PE podcast.  This is where all the new concepts are squirrelled away.  We know that our current approach to PE is imperfect and likely does harm to a good number of patients – so can we do better?

Dr Anand Senthi is back to run us through a few new concepts – the test threshold, the mortality threshold and the role of the simplified PESI score in the prognostication process.

As usual there is a lot of evidence, key papers and scoring systems mentioned – so the links to these are below.

Enjoy, it is 33 minutes and I think might need a second listen if you are new to these ideas…

There was quite twitter debate a few weeks back when we started tweeting about some of these concepts – and it was hard to explain in 140 characters, so hopefully this podcast does the concepts more justice.  Would be keen to hear your opinions on the comments or Tweet me and @DrSenthi .

Here is a rough visual roadmap through Anand’s PE work up algorithm.  Looks complex, but it is really simple to do – faster than getting consent for a CTPA I reckon.  All the data you need is in your initial clinical assessment – just plug it into MDCalc and Bob’s your uncle!

Proposed PE algorithm

Proposed PE algorithm




Dr Newman, David and Schriger “Rethinking Testing for Pulmonary Embolism: Less Is More” from Annals of EM 2011

Dr Kline’s paper from Journ. Thromb. and Haemostasis in2004 that calculated the “test threshold” for D-dimer at a pretest probability of 1.8%.

The Pauker and Kassirer method for calculating a test threshold – if you really want to know where they come up with these risk levels!

Mitchell (and Kline), Academic Emergency Medicine 2012  prospective study of the risk of contrast-induced nephropathy in patients undergoing CTPA.

HAS-BLED risk assessment tool for bleeding risk on anticoagulation (derived from AF patients – not VTE)

Outpatient Bleeding Risk Index,

1 point was scored for each of the following:

(1) 65 years or older,

(2) history of GI tract bleeding,

(3) history of stroke, and

(4) one or more comorbid conditions (recent myocardial infarction, anemia [hematocrit, <30%], renal impairment [creatinine level, >1.5 mg/dL {>133 µmol/L}], or diabetes mellitus).

The patient was low risk if the score was 0, moderate risk if the score was 1 or 2, and high risk if the score was 3 or more.

Possible PE Test thresholds – sensitivity analysis – a range of proposed test thresholds calculated using varying assumptions on risk and benefit sides of the equation.  This is from Anand’s ACEM talk

Calder, Herbert, Henderson – Annals of EM 2005  The mortality of untreated pulmonary embolism in emergency department patients – was likely less than 5%

Pines and Lessler (Acad Emerg Med, 2012)  performed the “mortality benefit threshold” calculations – i.e. how much benefit would treatment have to achieve in order to make a work-up worthwhile in a given pretest risk cohort? In the low risk group it works out at about 20%…. so you would need to have an intervention that dropped mortality by 20%….  this is not proven or even likely with current anticoagulation drugs!

The Simplified PESI (PE severity index) score on MDCalc – a tool to predict will do well after the diagnosis of PE.  If your sPESI is = 0, you are really very unlikely to die from a PE.  Jimenez et al showed the mortality risk was 1%, then subsequent prospective studies have shown even lower “treated mortality” rates

Moores, Jimenez et al (Journ Thromb & Haemost, 2010) showed the PESI score was better than troponin for predicting mortality in PE patients




  1. Minh Le Cong says:

    thanks guys!
    untreated PE mortality 5%…interesting. Then why bother investigating or treating DVT ? Arent most PEs as a result of a DVT? Are you saying mortality of untreated DVT is low too ..low enough to disregard?

    I have seen a number of deaths from PE/DVT..sure you both have too? what are you saying..that was just bad luck they were in that 5%?!

    question to both of you as you seem know the numbers..if you personally thought you had a PE , what would you do?

  2. Anand Senthi says:


    Firstly not saying anything about DVT. Let’s stick with PE.

    Secondly, yes the untreated PE mortality rate in ED patients is likely less than 5%. Deaths we see fall in that in figure. Note we are not talking about massive PE.

    In response to your comments here and twitter:

    We are not advocating not treating patients who have confirmed PE.

    What we are saying is that if untreated PE mortality is low and the patients risk of having a PE is low (eg Wells < 2 = 3% PTP) then the data suggests that it may not be in that patient's interest to be investigated due to the risks of investigation and the risks of giving AC to false positives.

    This is completely different to a person who actually has confirmed PE. Assuming it is a true positive then their risk of PE is 100% not 3% that the Well's < 2 patients have. Different calculation of risk/benefit follows.

  3. Minh Le Cong says:

    thanks Anand for responding.
    One thing still seems incongruous in the approach you and Casey suggest.

    You both admit that you would treat confirmed PE

    Yet you propose using biostats and a somewhat convoluted approach of risk stratification to reduce the need to investigate and Dx these low risk PEs.

    In effect you suggest sending some patients home who may well have a PE but you use a biostatistical approach to justify this as they are “low risk”.

    Do you not see the incongruity between treating all confirmed PE vs your low risk stratification approach?

    It seems you are really having it both ways. Which in my view , is illogical. If you dont want to test for low risk PE cases, then dont do matter how you justify it. Your justification process seems unnecessarily complex.

  4. Anand Senthi says:

    hey Minh,
    I don’t see any conflict as you suggest.

    Even assuming PE should be treated if diagnosed, this is very different to whether you should investigate someone who has:
    -- a very low risk of having PE (eg Wells <2)
    -- a very low risk of dying even if they did have PE (eg sPESI=0)

    If the investigation/treatment process was risk free it would be a different story but given the data suggests:
    -- there is significant, previously underestimated, risk of CIN/renal failure/mortality from intravenous contrast
    -- there is a 50% risk that CTPA positive is actually a false positive in low risk (Wells <2) patients
    -- there is a significant risk of harm from anticoagulation of false positives
    when you put all this together there is a PTP below which (the test threshold) it does not make sense to investigate these patients. You will be more likely to harm then help these patients.

    The current approach of finding all PE's in anyone who could have one is killing more patients than it is saving. We need an alternate approach.

    I also don't think the suggested approach is that complicated. The beauty of it is that the PERC/Wells scores and sPESI scores are easy to calculate and easily accessible on services like Mdcalc if you forget.
    Sure you could use clinical gestalt instead re PTP but Wells/PERC build this in with a little bit of objective data and the combination is better at identifying a low risk group then gestalt alone.
    You could use gestalt to determine likely prognosis but the items you would use to form such gestalt are pretty much the sPESI score -- it is super intuitive.

    So all in all I don't see much advantage to a pure gestalt process over the suggested approach. Also the whole problem with PE is that we are grossly over-investigating partly due to a fear of litigation. Therefore a more robust statistical approach with data supporting it that can be documented in the notes would hopefully provide clinicians more confidence to not investigate these low risk patients. At the very least it provides the data required to have a risk based patient centred discussion regarding the whether or not to investigate.

    Finally I just wanted to address the point you have raised here/twitter about where every diagnosed PE case be treated. Before answering I want to be very clear that this is not what we are referring to in the suggested approach to reduce our investigation of PE.
    However with that acknowledged there is a strong case that some patients with actual diagnosed PE may not benefit from anticoagulation. For a start there is not data that anticoagulation provides benefit over placebo and some data that it is equivalent to NSAIDS and we've only studied sick hospital patients not well ED patients.
    It is quite conceivable that a PE in a young ED patient with normal sats/BP/PR and no significant cardiorespiratory comorbidities (ie sPESI =0) who has no DVT on ultrasound might gain no net benefit from anticoagulation (AC). Or alternatively they may gain similar benefit from aspirin. I think it would be quite ethical and valuable to do an RCT of AC v aspirin v placebo in such patients.
    The other group of PE patients who may not benefit from treatment are those at increased bleeding risk from anticoagulation.
    Just reiterating, that not treating such patients is not what we have been talking about in the podcast. I'm just adding this into the discussion as you asked about it.


  5. Minh Le Cong says:

    Let me couch this another way. DO you think every proximal DVT should be treated?

  6. Minh Le Cong says:

    thanks Anand.
    I appreciate you both want to justify using a biostatistical approach, a targetted reduction in PE Ix for “low risk” patients in ED.

    Whether this will improve care/outcomes/complications of OverIx, is , as you say, a matter of conjecture for now.

    You offer the notion based on some articles that untreated PE carries a low mortality. Although you deny that you are advocating nontreatment of confirmed PE, it is hard not to draw that conclusion by the mere fact you raise these articles and the issue of untreated PE of “low risk” having low mortality.
    Bear in mind those articles draw those conclusions from missed PE cases with repeat presentation to ED. They did not look at, nor am I aware of any data where there was a deliberate controlled strategy of non treatment of PE.

    Indeed a RCT would answer this question but I suspect as you cite that the medicolegal implications will make such a trial untenable .

    Therefore I return to the incongruity of your approach. If you admit that for mostly medicolegal reasons you would still Rx the small PE”lung lint” if confirmed, it seems odd that you will justify sending patients home with possible PE

  7. Anand Senthi says:

    thanks Minh,
    I don’t see any incongruity at all. Rather than repeating my explanation above but let me try to make the point with a fictitious example:

    Let’s say we are talking about thoracic aortic dissection. Let’s say that CT aortogram is the only way to diagnose it and we found out tomorrow in a new massive study that contrast has a 10% risk of death. Let’s also assume aortic dissection has a 50% mortality, which I could halve by discovering it, so a 25% ARR.

    -- if diagnosed I would definite treat a dissection because their risk of having a dissection is 100% (assuming no false positives). They would be a 25% mortality reduction by treating them. It is a no brainer -- we should treat them.
    -- if however you were able to work out that a patient has a 10% risk of having a dissection using say a score called the Wells Dissect Score. If I investigate them they have a potential benefit of 10% x 25% = 2.5%. However their potential harm = 10% … this new amazingly high risk of contast.

    Therefore even though I would definitely treat a diagnosed dissection I would not investigate a person with 10% risk of dissection because the harm (10%) outweighs the benefit (2.5%).
    So you see there is nothing conflicting or incongruous about this position. Does that make sense?

    Now you can replace this fictitious example with a real example that we are dealing with right now: say a contrast risk of 0.1% and a PE treatment benefit of say 2.5% and a risk of anticoagulation of 0.2% and you end up with a similar scenario. For a patient whose risk of PE is low such as 3% (i.e Wells Score <2) it would cause more harm than benefit to investigate them given these risks.

    So you see, whether you would treat a disease is a completely different question to the question of whether you would investigate for that disease given a certain PTP and a certain set of risks of investigation.

  8. Anand Senthi says:

    btw -- I should add in response about the evidence of low mortality in untreated PE, it is worth noting that there is NO good evidence that untreated PE in well ED patients carries any higher mortality. To assume this based on flawed data from the 60’s on inpatients does our patients a gross disservice and exposes them to serious risk of harm.

    Given none of the evidence is perfect we must take a risk based approach based on the best available evidence we have right now. Often this is best done in consultation with the patient.

  9. Minh Le Cong says:

    thanks Anand.
    good attempt at elaborating on your suggested approach but still does not wash with me!

    10% risk of dissection so would not investigate?!

    What risk would you find unacceptable? 20%? 30%

    Your position seems illogical. Saying you would treat any dissection you find but not investigate for some you think the risk of Ix outweighs the possible benefit.

    How do you know that 10% risk of dissection is acceptable to the patient? You refer to this in your subsequent comment but herein lies another undiscussed aspect of your suggested approach. You both seem to discuss the biostats as if they are clean data, black and white and useful for prediction. Try explaining your approach to a layperson !

    If they find 10% or whatever % is unacceptable, then what is your position? Refuse to investigate on biostatistical principle?!
    And if you relent and agree to investigate on patient wishes then really what use is your suggested approach?

    I suggest it may be misleading and potentially more harmful to try to counsel a patient on your suggested biostatistical risk prediction…and the result maybe the patient chooses to have test anyway as not reassured by 10%..or even 5 % risk you give them.

    Fact is the public have all heard horror stories of hospital patients dropping dead from DVT/PE.
    If you counsel that that Dx is a possibility but you think it low therefore not investigate, .you leave yourself open if this is clearly your stated regular approach. Sure if the patient has been advised of your calculated risk stratification, accepts your recommendation of not investigating and followup is arranged then that is ideal outcome for your suggested approach. Accept though, that you will miss some PE with this approach. Accept that some of those PE will have bad outcomes. Accept that this may be awful for all!
    This is why you both still accept that confirmed PE require Rx as standard or care
    I think it better that you make a gestalt risk stratification, decide if testing needed and if not then advise the patient of such.

    You still have not addressed the question regarding proximal DVT.

  10. Anand Senthi says:

    Hi Minh
    -- on the fictitious aortic dissection example I simply can’t see why you don’t understand the rationale. I think you may have been thrown by my use of aortic dissection as the example. The point was it was a fictitious case where contrast was super deadly (10% risk). It was also fictitious in that there are actually other ways to diagnose dissection. However it was fictitious to make a point.

    You should not think of an investigation in any different way to a treatment.

    If you had a person who was having a STEMI and they were also having a haemorrhagic stroke at the same time, would you thrombolyse them? Of course not -- the risks of treatment outweigh the benefits. No sane doctor would do so.
    Where the risks of investigation outweigh the benefits the same applies. It is basic mathematical risk weighting we do everyday.
    I’m unsure why you are treating the risk/benefit calculation that I’m sure you do when you choose treatments differently to the same risk/benefit calc you should perform when you are choosing investigations.

    So now it is your turn to please answer the following 2 questions:
    1. would you treat a patient with disorder X with treatment Y if the risk of harm from treatment Y was 10% and the risk of benefit was 2.5%?
    2. would you investigate a patient with a disorder X with test Z if the risk of harm from test Z was 10% and the risk of benefit from proceeding with investigation was 2.5%?

    Let me know your answers so we can progress this discussion.

    Re prox DVT this is not relevant because PE and DVT are not the same. However yes most proximal DVT’s would be treated, the exception being patients with an increased bleeding risk.

  11. Anand Senthi says:

    in addition:
    -- in response to your point about using Gestalt alone, as explained above it appears at least with PE investigation, gestalt is not as good as Gestalt + a little patient data combined. i.e PERC and Wells <2 identify a very low risk group much better than Gestalt alone. Why use an inferior method when you can use a superior method?

    -- as you know I am all for involving the patient in the decision whether or not to investigate. I won't go over this all again since we discussed this at length in the Unstuffing the Turkey clinical case
    however in summary I think patient's should be involved in key decisions in their care and the amount you involve them will depend on how grey the decision is.
    i.e if I have a 20 year old low gestalt risk and PERC negative patient who had a clear history and exam supporting a chest wall injury I would not involve them in the discussion not to investigate for PE. However if the decision to investigate involved a decision whether clinicians might differ in approach (eg a Wells <2 patient) I will usually involve the patient, inform them of the risk v benefit and come to a joint decision.
    For the most part I think PE investigation is an area where the patient should be involved and they should be given the best data we have to make the decision because there are several grey risk v benefit decisions. This is unlike most other diagnostic processes in other conditions where things are usually much more clear cut.

  12. Minh Le Cong says:

    thanks Anand!

    In answer to your questions, of course if those statistics are correct then it would be illogical to do something deliberately harmful to the patient!

    But thats not the complete story is it?
    you are playing the numbers..even higher risk pts will be harmed
    If you truly believe those statistics you quote then why does the situation change when you have confirmed PE and you both admit you would still offer anticoagulation!!??

    Are you sitting on the fence entirely due to medicolegal reasons?

    Most patients dont want to be treated as “numbers” . As you know many will ask a doctors opinion to do whats “best”
    I support the motion to reduce PE testing
    how you do that, is where we disagree
    you claim your approach is superior but it has not been validated, has it?

    it is artificial and arbritrary to say DVT and PE are not the same illness
    is not a PE a really proximal DVT!
    do not many PEs arise from DVT?
    do they not have the same Rx?
    are they not in the venous system?

  13. Anand Senthi says:

    In response to this:
    “If you truly believe those statistics you quote then why does the situation change when you have confirmed PE and you both admit you would still offer anticoagulation!!??”
    That’s easy. The numbers change like this (in the scenario we are referring to):
    -- in the person with a 3% risk of PE they have about a 1/1000 risk of death from PE (i.e 3% x <5%). To make the diagnosis you must expose them to further risks that likely outweigh this.
    -- in a person with an actual PE they have a <5% risk of dying and you don't have to expose them to any tests.
    That's the crux of the discussion right there. That's exactly why it is completely different if a person actually has a diagnosed PE.

    If you agree with those 2 questions above, then we must agree.

    Agree my suggested approach is not robustly validated. It is merely a suggestion of how I have interpreted the data to provide the patient with their best risk v benefit analysis based on the available imperfect evidence. By contrast the current approach that most docs apply has absolutely no basis in evidence. That's my main point. There is not need to conclusive prove the alternative when the status quo is completely unproven. One merely needs to compare the 2 and choose which appears to have the greatest foundation in science to advise the patient as accurately as we can about their risk.

    Finally, PE and DVT are certainly not the same illness. They share similarities but are different. eg if someone has a small PE but no DVT how big is their risk of a further PE compared to a person with a proximal DVT???

  14. Minh Le Cong says:

    thanks Anand.
    No I dont accept those
    Are you claiming that all PEs afford the same risk of complications?
    This article

    suggests that PE by itself does worse than DVT by itself and the combination of the two is even worse 3 month outcomes!

    In other words PE with a DVT = not good!
    And PE worse than DVT

    SO logically if you think we should Ax/Rx DVT, then why would you NOT do same for PE given above!

  15. Anand Senthi says:

    thanks Minh.
    I believe you are still missing the central theme.
    just because you would treat a diagnosed disease it does not follow that you should investigate any person who as any risk of disease because the risk of investigation may outweigh the benefit from investigation. This is the TT.
    Obviously you would not Ix a person with a 1 in million chance of PE.
    You would not investigate a person with a 1 in a 1000 chance.
    Would you investigate a person with a 1 in 100 chance? That’s below even Klines TT threshold from 10 years ago. A low gestalt PERC negative person is such a patient.
    Would you investigate a person with a 3% chance of PE? The updated TT suggests this person is also below the TT.

    The bottom line is there is a TT below which any doc would not investigate. What can be argued is what that TT is and why it is that number. That is definitely up for debate.

  16. Anand Senthi says:

    sorry I should add the difference between DVT & PE in regards to your question is that investigating for DVT is essentially harmless (other than cost) while investigating for PE is potentially quite harmful.
    That is why the TT for DVT I(x) would be exceedingly low, while the TT for PE I(x) is much higher. Make sense?

  17. Minh Le Cong says:

    thanks Anand
    I am not missing your central arguement. I support the motion to reduce OverIx of PE

    I am unconvinced that your suggested hybrid approach is any better than the existing approaches. By your own admission, your suggestions are not validated and are an amalgam of previous ideas . I dont mind inventive thinking but it must follow that your discussion with the patient addresses the issue that you are following as yet unvalidated medical process. Thats fine as long as its declared.

    I support your drive to prove your theory and await your published work on this.

  18. Anand Senthi says:

    thanks Minh
    absolutely my approach is unvalidated. But the status quo is completely unvalidated. My contention is that the data supports my approach better than the status quo which is grossly flawed.
    When I discuss with patients I make them very aware about the uncertainty in the data, the status quo and my approach. Patients have a right to know.
    thanks for the discussion.

  19. Anand, is there any way to get hold of your ACEM talk/slides on test thresholds relating to this. Thanks

    • HI GAvin
      The “TEST Threshold” figure summarises the various calculations of a possible TT in the blog post accompanying the podcast is taken from Anand’s slide set with his permission.
      I leave it up to him to distribute more…

Speak Your Mind