The Mathematics of Syncope

No Comments

The PESIT trial deconstructed

Syncope is a familiar complaint for emergency physicians, accounting for 1-3% of ED visits. In general, the causes are mostly benign and self-limited; however, sometimes syncope can be a symptom of a more serious condition, such as acute coronary syndrome, malignant arrhythmia, or even pulmonary embolism (PE). Nevertheless, the ED diagnostic evaluation of syncope is often less than gratifying. A small percentage are diagnosed with something serious, while a larger group with negative ED work-ups are still thought to be high-risk and admitted for further observation and work-up. Unfortunately, there is no real gold standard test or validated decision instruments for syncope.

The PESIT Trial
In October 2016, a group of scientists from Italy published a syncope study with unexpected results in the New England Journal of Medicine (NEJM). The “PESIT” or Pulmonary Embolism in Syncope Italian Trial determined PE prevalence in 11 Italian hospitals by subjecting all admitted syncope patients to a systematic PE work-up within 48 hours of admission, based on a standard algorithm-the simplified Wells score-based on pretest probability and D-dimer results [1].

In 2584 syncope patients, 1867 (72%) were discharged. Of 717 admitted patients, 118 already receiving anticoagulation were excluded, as were 35 with prior episodes of syncope and four who refused to consent. The 560 ultimately included were mostly older adults (>75% were over 70 years of age), 355 (63%) had clinical evidence of another explanation for syncope, and in 330 (59%), PE was ruled out based on low pretest probability and a negative D-dimer, while 180 had further testing.

What was so surprising was that 97 of the 560 or 17% of admitted patients were ultimately diagnosed with PE, mostly by CT (72 patients), and less commonly by V/Q (24), and one by autopsy.

Traditionally, prevalence studies are not published in NEJM. So what was so noteworthy for NEJM to publish this study? If one in five admitted syncope patients really have PEs then we emergency physicians and hospitalists may just be missing a whole lot of PEs (Interestingly, the PE prevalence in this study contradicts several studies that have found the prevalence to be considerably lower at 5-10% [2,3]).

Order a d-dimer every time?
This begs the question: based on the PESIT trial, should we change practice and indiscriminately order a D-dimer on every single syncope patient or maybe just those admitted with syncope? The short answer is “No!” The reason why is hidden in the study’s numbers, and also provides a nice example to illustrate how to assess and interpret risk in prevalence studies.

First, patients at sufficiently low risk to be safe for discharge were not included in the study. Therefore, we can’t really say anything about the PE risk in this group; however, it is likely very low based on the results of clinical evaluation and whatever ED-based testing was conducted. This is supported by another study that followed discharged syncope patients and found very low adverse event rates after discharge, particularly in patients under 60 [4].

Second, consider the group deemed high risk enough to admit or those with a diagnosis of PE or something else in the ED. Absent any risk-stratification, the prevalence of PE was 17% in this population. Importantly, not all of these work-ups happened in the ED, so 17% may actually overestimate the rate of PE that could have been diagnosed in the ED given that immobility itself is a risk factor for venous thromboembolism.

Explaining Risk
Remembering back to that epidemiology class you snoozed through in medical school, an odds ratio (OR) determines the association between exposure and outcome. Specifically, an odds ratio is the odds of disease in “exposed” patients divided by the odds of disease in patients without the exposure.

Calculating an OR is simply done by: A x D / B x C. When applied to this paper, the “disease” is PE and “active cancer” is one of the exposures. For “active cancer,” this gives you the following 2 x 2 table:

To calculate an odds ratio, we multiply (19 x 417 / 46 x 78) = 2.21. This means that the odds of PE are 2.21 times higher in patients with “active cancer.” The authors generated a 95% confidence interval around this estimate; therefore, we are 95% certain that the actual odds ratio for active cancer and PE falls between 1.23 and 3.97. That is, people with active cancer are 1.2 – 4.0 times more likely to be diagnosed with PE when they have syncope than those without active cancer.

Several additional factors of clinical importance also increased the odds of PE significantly. Specifically, the odds of PE was higher in those with prior PE (OR 2.8), those with undetermined causes of syncope (OR 2.3), or clinical features of PE including a respiratory rate > 20 breaths per minute (OR 10.8), heart rate > 100 beats per minute (OR 2.5), systolic blood pressure < 110 mm Hg (OR 1.9), and clinical signs of deep vein thrombosis (OR 14.2).

Odds v. Probabilities
Sometimes people have a hard time thinking in terms of odds, and can more easily understand probabilities. Using the example of active cancer, the probability of PE in active cancer was 19/65 (29%) compared to 78/495 (16%) without active cancer. Similarly with the other significant factors, we can calculate the prevalence of PE in patients with and without each risk factor (see table).

What we see is that the presence of any of these risk factors in syncope significantly raises the odds or probability – or prevalence of PE. It means that all should probably prompt emergency physicians to work-up PE in the setting of syncope.

Wait. All of these are actually known risk factors for PE. However, a clinical pearl might be to explicitly review each of these risk factors in admitted syncope patients and consider ordering a D-dimer or imaging study, even when there is another explanation. This “other explanation” recommendation is supported by the study’s data: in 355 patients with potential alternative explanations for PE, 45 (12.7%) had PE. And 31 (69%) of those had a lobar or proximal PE on CT or a perfusion defect of 25% or more of the lungs. This means these are the most concerning, potentially life-threatening PEs.

Testing Thresholds for Considering ‘Occult PE’
The authors go on to provide some additional clinically useful observations about the data. Of the 97 patients with PE, 24 – or 24.7% did not have clinical manifestations of PE or signs of DVT. This can best be described as the “occult PE” rate, with 24 out of 560 = 4.3%, specifically those who were diagnosed with PE but had no risk factors. This addresses the question as to whether to work-up PE in the absence of risk factors: i.e., should we work-up an “occult PE” in everyone admitted with syncope?

To answer this question, we rely on another calculation: a formula from a 1980 New England Journal of Medicine paper by Pauker and Kassirer that is probably older than many of you reading this [5]. The paper calculates the testing threshold and the test-treatment threshold. The testing threshold is the probability below which we should not order a test because the potential harm from testing exceeds the benefits. The test-treatment threshold is the probability above which we should treat and not order test (see figure).

The central question: Is a 4.3% prevalence of PE in older adult patients without risk factors for PE above or below the testing threshold for “occult PE”? The Pauker and Kassirer formula provides a simple-ish way to calculate the testing threshold which takes into account: 1) The harm/risk of treating someone who doesn’t have the disease, 2) The harm/risk of not treating someone who actually does have the disease, 3) The harm/risk of doing the test, 4) The accuracy of the test. For the purposes of the test, we will assume that all syncope patients get a D-dimer, and if positive, a CT angiogram. The testing threshold can be calculated through the following formula:

Tt (testing threshold) = ((Ppos/nd x Rrx) + Rt) / ((Ppos/nd x Rrx) + (Ppos/d x Brx))

Ppos/nd – The probability of an incorrect test result (type one error)
Ppos/d – The probability of a correct positive test result
Brx – The benefit of treatment to patient with the diagnosis
Rrx – The risk of treating the patient without the diagnosis
Rt – The risk of performing the test

For calculating the test threshold for PE, we use D-dimer then if positive CT angiogram as the diagnostic test, similar to a 2004 recent paper by Kline et al. that calculated the testing threshold to derive the PERC criteria [6]. The risk from CT angiography comes from the incidence of life-threatening complications from IV contrast material and the small risk of cancer from the radiation exposure (Rt). Given this is an older population, we will assume this to be 0.00003 for both risks total. This is lower than the 0.00006 used in Kline et al.’s calculation of the testing threshold of PE, and lower because about half of the patients would never get a CT and would have PE ruled out by D-dimer and the risk of cancer from CT is dramatically lower in older adults. We also exclude the risk of contrast nephropathy that might occur from IV contrast because of a recent paper demonstrating no such relationship [7]. However, it is important to mention that our testing thresholds would be considerably higher if renal failure from IV contrast and subsequent complications were a concern.

We also assume that all patients with a negative D-dimer have no PE, a limitation given that D-dimer sensitivity is not 100%. For the formula, we use the sensitivity of CT angiography (Ppos/d) and assume it is approximately 85% from meta-analysis, with 10% (Ppos/nd) false positives [8]. For the benefit of anticoagulation, the only randomized trial comparing treatment for PE v. no treatment and demonstrated a 20% reduction in death; however, this paper was from 1960, and it was a small trial [9]. More recent studies have demonstrated that this is likely lower for all comers with PE given that may patients who “escaped anticoagulation” with hemodynamically insignificant PEs end up having no adverse outcome [10]. Because mortality benefit is the most questionable variable in PE, we will use 2.5%, 5%, and 10% for Brx, similar to recent decision analyses we conducted on this question [11]. The risk of major hemorrhage with treatment with warfarin is 1.7% and about 30-40% lower for new oral anticoagulant drugs used commonly these days for PE. For this we will assume a risk of 1.2% for Rrx.

Plugging in these numbers gives us:
Testing threshold at 2.5% benefit of PE treatment = 5.5%
Testing threshold at 5% benefit of PE treatment = 2.8%
Testing threshold at 10% benefit of PE treatment = 1.4%

This means that if we believe that the mortality benefit gained-specifically the mortality in treated v. untreated PE)-from treating PE is 2.5% on average in a mixed population of older adults similar to those who were seen in this trial, then we should not be testing for PE unless the population prevalence exceeds 5.5%. In this case, we should not test patients without risk factors for PE.

However, at a 5% benefit of treatment, the testing threshold is below the population prevalence of 4.3% and we should be testing for PE. Solving for the mortality benefit of anticoagulation at which the testing threshold = 4.3% gives us 3.2%. So if we believe that anticoagulation is improving mortality by 3.2% or more, we should be testing admitted patients without risk factors for PE when they are admitted for syncope. In our opinion, it is likely that the clinical benefit from anticoagulation is likely somewhere in this range, so whether to work-up occult PE should be based on clinical judgment and not done reflexively.

Key Clinical Takeaways
The PESIT trial has many limitations and should not dramatically change ED practice. However, it does provide a nice platform to explain and demonstrate the mathematics of risk. We also think it provides some good clinical take-home points when you’re evaluating patients with syncope who are being admitted to the hospital:

  1. When higher-risk syncope patients (i.e., those who you think need admission) have objective risk factors for PE, test them for PE.
  2. “Occult PE” occurred, at least in this trial, in up to one in 20 patients admitted with syncope without known risk factors for PE. This is higher than expected. However, because this prevalence is not clearly below the testing threshold based on our best clinical estimates, we recommend using clinical judgment in determining which patients should be tested for “occult PE.”
  3. Don’t stop thinking about PE once you think you’ve found the reason for syncope. Patients with a clear alternative diagnosis are still at high risk for PE (about 12%), and their PEs they tend to be more proximal and life-threatening.


  1. Prandoni P, Lensing AW, Prins MH, Ciammaichella M, Perlati M, Mumoli N, Bucherini E, Visonà A, Bova C, Imberti D, Campostrini S, Barbar S; PESIT Investigators. Prevalence of Pulmonary Embolism among Patients Hospitalized for Syncope. N Engl J Med. 2016 Oct 20;375(16):1524-1531.
  2. Soteriades ES, Evans JC, Larson MG, et al. Incidence and prognosis of syncope. N Engl J Med 2002; 347: 878-85.
  3. Sarasin FP, Louis-Simonet M, Carballo D, et al. Prospective evaluation of patients with syncope: a population-based study. Am J Med 2001; 111: 177-84.
  4. Derose SF, Gabayan GZ, Chiu VY, Sun BC. Patterns and preexisting risk factors of 30-day mortality after a primary discharge diagnosis of syncope or near syncope. Acad Emerg Med. 2012 May;19(5):488-96.
  5. Pauker SG, Kassirer JP. The threshold approach to clinical decision making. N Engl J Med. 1980 May 15;302(20):1109-17.
  6. Kline JA, Mitchell AM, Kabrhel C, Richman PB, Courtney DM. Clinical criteria to prevent unnecessary diagnostic testing in emergency department patients with suspected pulmonary embolism. J Thromb Haemost. 2004 Aug;2(8):1247-55.
  7. Hinson JS, Ehmann MR, Fine DM, Fishman EK, Toerper MF, Rothman RE, Klein EY. Risk of Acute Kidney Injury After Intravenous Contrast Media Administration. Ann Emerg Med. 2017 Jan 19. pii: S0196-0644(16)31388-9.
  8. Safriel Y, Zinn H. CT pulmonary angiography in the detection of pulmonary emboli: a meta-analysis of sensitivities and specificities. Clin Imag. 2002; 26: 101–5.
  9. Barritt DW, Jordan SC. Anticoagulant drugs in the treatment of pulmonary embolism. A controlled trial. Lancet. 1960 Jun 18;1(7138):1309-12.
  10. Calder KK, Herbert M, Henderson SO. The mortality of untreated pulmonary embolism in emergency department patients. Ann Emerg Med. 2005 Mar;45(3):302-10.
  11. Pines JM, Lessler AL, Ward MJ, Mark Courtney D. The mortality benefit threshold for patients with suspected pulmonary embolism. Acad Emerg Med. 2012 Sep;19(9):E1109-13.



HEALTH POLICY SECTION EDITOR Dr. Pines is a practicing emergency physician and a Professor of Emergency Medicine and Health Policy at the George Washington University.

Dr. Rezaie is founder and editor of R.E.B.E.L EM.

Leave A Reply