PE and Evaluation of Risk: Pik Mukherji (AllNYCEM6)
- Mar 13th, 2014
- Manny Singh
This is a summary of the talk given by Dr. Pik Mukherji (@ercowboy) at the 6th All NYC EM Conference, culled from notes & tweets (#allnycem6). Dr. Mukherji is EM/IM trained and is the EM/IM residency director at Long Island Jewish Medical Center.
Dr. Mukherji began by describing the “chasm” of clinical judgment that exists when it comes to assessment of risk. Physicians can have widely variable estimations of mortality from disease entities like ACS or PE, which can have huge implications on decision making. In part, this has to do with differences in the weight we give to the data points we have available, and how this informs our gestalt. If we’re all weighing the variables differently, we’ll never be on the same page. Bottom line: Docs are crazy when it comes to evaluating risk.
The pitfalls that are commonly encountered in risk assessment include:
- Overestimation of disease risk
- Underestimation of treatment risk
- Overestimation of treatment benefit
When contemplating whether you’re going to investigate for a PE, understanding these risks and benefits is crucial to figuring out how hard you’re going to look.
In epidemiological terms, PE is a huge problem. The incidence quoted by Dr. Mukherji is 600k/year, although he stated that 10 years ago, same slide would have cited 400k/year. That being said, the incidence and prevalence are much less important for PE than mortality. Really, we’re just diagnosing more (and treating more), but the same number of people are dying from this disease every year.
So Dr. Mukherji concluded that we over-workup patients with suspected PE. He then mentioned the algorithm posted at emcrit.org, which begins with: “Do you think this could be a PE? No really. No really, do you?”
One distorting factor in assessing risk is the propagation of inapplicable data. He cited a paper from 20101, which states that “PE is fatal in up to 30% of patients, but the death rate can be reduced to 2% to 10% if PE is diagnosed and treated promptly.” This should make us pause, and ask ourselves if these numbers ring true.
He followed this up by mentioning another paper2 referring to these same oft-cited data, which states: “these figures originate from investigations that have little relevance to modern emergency medicine, which include studies dating to the 1940s, many of which have significant methodologic pitfalls. These data are also based primarily on either inpatient or autopsy populations, neither of which is representative of patients treated in the emergency department (ED). Analysis of untreated or missed pulmonary embolism in ambulatory patients reveals mortality and recurrence rates of less than 5%.”
What about determining risk of PE in pregnancy? Pregnancy increases risk of PE four-fold, but this is essentially taking the risk from “super duper low to just super low.” Furthermore, pregnant patients are far from a homogenous group. 85% of the time, PE will occur in peri-partum patients, while those in the first trimester are essentially at baseline risk. So why treat them the same?
As a general rule, pregnant women don’t have PE. Evidence is emerging to support this – Dr. Jeffrey Kline (@klinelab) tweeted: “Submitting a meta-analysis of over 25,000 ED patients evaluated for PE: pregnancy does NOT increase the risk of diagnosis of PE; RR=0.50!” One might interpret this to mean that pregnancy is protective, which doesn’t make any sense, right? In reality, its “more of a story of over-testing than risk.”
Then there’s the overestimation of benefit/underestimation of risk for PE treatment. He cited a study3 looking at patients with DVT and silent PE, and the influence of anticoagulation on the resolution rate of PE. The patients were randomized into groups either receiving anticoagulation or not, and at 3 month follow-up there was no difference between the groups.
Is anticoagulation harmful? In the real world, the actual rates of significant bleeding are higher (1-7%) than they tend to be in well-controlled trials (0.5%).4,5 But does it harm more than help? He mentioned an editorial by David H. Newman, MD in Annals,6 which concluded that our current testing model actually leads to 6 times increased mortality in treated vs. untreated patients.
Less sick patients get less aggressive workup and intervention – you wouldn’t perform an intervention that’s more harmful than the disease itself. An “incidentaloma PE” isn’t the same entity as a PE causing hypotension. But the decision to anticoagulate is usually based on the very presence or absence of PE, rather than for the effect of the clot burden.
Does size matter? Estimating the size of a thrombus or of overall clot burden can lead to the danger of “pictures driving therapy.” This approach can be flawed, and Dr. Mukherji provided a few examples: One paper7 found that cancer patients with sub-segmental PE had similar mortality to those with more proximal PE (6.5% to 10.7%). He contrasted this with a case study describing a patient who was met PERC and had low risk Geneva and Wells scores, but was still found to have a saddle embolus.8
What does matter in assessing risk? Right ventricular dysfunction, troponin, BNP, and an EKG with right heart strain. In a patient without shock/hypotension, RV dysfunction, or myocardial injury, short-term mortality from a PE is low (<1%).9
His take home points:
- Your patient probably doesn’t have a PE.
- Yes, you need to be more risk tolerant.
- No, we’re not sure what to do with low-risk (NOT small!) PE’s