Evaluation of Syncope in the Emergency Department

Authors: Theodore Segarra, MD; Lee Grodin, MD; Taylor Conrad, MD; Raymond Beyda, MD
Editors: Michael C. Bond, MD FAAEM and Kelly Maurelus, MD
Originally Published: Common Sense January/February 2018

As syncope is a common yet nebulous complaint, evaluation of the patient with syncope presents a unique challenge. Syncope is defined as a brief loss of consciousness and postural tone with rapid return to baseline mentation. Yet, rather than having a single underlying cause, syncope itself is a syndrome with many potential etiologies. Some identified causes are arrhythmia, myocardial infarction (MI), cerebrovascular accident (CVA), hemorrhage, and pulmonary embolism (PE).[1] In this edition of RJR, we review the potential etiologies of syncope, the utility of risk stratification tools in the workup of syncope, and the prevalence of atypical causes of syncope.

S Safari, et al. Comparison of Different Risk Stratification Systems in Predicting Short Term Serious Outcomes of Syncope Patients. J Res Med Sci. 2016 Aug 1;21:57. eCollection 2016.

Given the potential for serious underlying cause in syncope, EPs must quickly assess, risk stratify patients, and determine appropriate disposition. To this end, several risk stratification scores were developed to identify patients at moderate and high risk of serious negative outcomes. Yet there exists significant variability between these rules regarding their criteria, outcomes, and applicability in various populations. Safari and colleagues address this issue in their study by comparing the short term (one week) diagnostic accuracy of four of the most well-known predictive models for patients with syncope. These models included the San Francisco Syncope Rule (SFSR), the Osservatorio Epidemiologico sulla Sincope nel Lazio (OSEIL) model, the Boston model, and the Risk Stratification of Syncope in the Emergency Department (ROSE) score.

The authors conducted a prospective, non-blinded, observational diagnostic accuracy study involving 187 patients chosen by convenience sampling and screened by EM residents and attendings at teaching hospitals in Tehran, Iran from October 2013 to October 2014. Exclusion criteria were age less than 18 years, pregnancy, drug/alcohol/substance abuse, and known underlying cause of syncope. Mean age of participants was 64.2 +/- 17.2 years, and 64% were male. Patients were screened for all potential risk stratifying criteria based on a pre-designed checklist and were followed for one week to determine rates of three primary outcomes: mortality, MI, and CVA. Models were compared using receiver operator curve (ROC) analysis for relative diagnostic accuracy in predicting the primary outcomes.

The one-week incidences of mortality, MI, and CVA were 19 (10.2%), 12 (6.4%), and 36 (19.2%), respectively. The authors determined the area under the ROC curve for each outcome. The authors found there was extremely low accuracy for all four models in predicting mortality, MI, and CVA. Further, they found no significant difference between the four models in their predictive accuracy for any of the three outcomes. The authors then combined all criteria in a single pooled model, and found similarly poor predictive capability. The authors concluded that all four models, as well as the pooled model, are unable to accurately predict one-week mortality, MI, or CVA in syncope patients.

Unfortunately, several critical flaws plague this study. Though the choice of a one-week time point is useful for the EM provider by providing a unified and short time frame for comparison, the problem is that each model was designed and powered to predict different outcomes over different time frames (SFSR: serious outcomes at 1 week to 1 month; OSEIL: mortality at 1 year; Boston: serious outcomes at 1 month; ROSE: serious outcomes at 1 month).1,6-9 Therefore, it is not surprising that models like OESIL, Boston, and ROSE may not be useful in predicting short term risk. Moreover, models like the SFSR were designed to be applied to patients with no obvious cause of syncope after full evaluation in the ED. However, in this study, the criteria were applied to all patients prior to an initial evaluation. Furthermore, prior validation studies revealed that the most commonly identified cause of syncope after initial evaluation was arrhythmia, which was not assessed by the current study.[6]

The study’s design leads to significant bias due to its observational, non-blinded (Hawthorne effect) nature. In addition, the small sample size, with less than a 30% incidence of each negative outcome, results in a study that is underpowered to identify differences between the four models. Also, the use of convenience sampling leads to significant sampling bias, particularly given the small sample size and unblinded nature. Furthermore, as a diagnostic accuracy study, the authors failed to include controls and to describe their methods for determining true positive rate and false positive rate. As a result, it remains unclear how they determined the true diagnostic accuracy of each model. Due to these significant flaws, the study’s results should not alter current practice.

V Thiruganasambandamoorthy, et al. Predicting Short-Term Risk of Arrhythmia Among Patients with Syncope: The Canadian Syncope Arrhythmia Risk Score. Academic Emergency Medicine. 2017 Nov;24(11):1315-1326.

In patients with no identifiable cause of syncope, the possibility of paroxysmal arrhythmias which may not be detectable on initial evaluation should be considered. As a result, newer clinical decision rules have focused increasingly on identifying patients at risk of arrhythmia. Thiruganasambandamoorthy and colleagues recently conducted a multi-center prospective cohort study in order to identify variables from the clinical evaluation and ED testing to derive a prediction model for clinically important arrhythmias related to syncope, the Canadian Syncope Arrhythmia Risk Score (CSARS). The outcomes evaluated were death due to arrhythmia or unknown cause, arrhythmia, and procedural interventions to treat arrhythmias within 30 days of the ED evaluation and disposition. Patients with pre-existing arrhythmias or other serious conditions were excluded in the study. Thirty-nine predictor variables were identified from statistical analysis of the cohort, and from that initial group eight predictors were selected to be included in the risk score: vasovagal predisposition, history of heart disease, ED systolic pressure >180 or <90mmHg, troponin elevation (>99th percentile), QRS duration >130ms, QTc >480ms, ED diagnosis of vasovagal syncope, and ED diagnosis of cardiac syncope. The score ranges from -2 to +8, (vasovagal predisposition and ED diagnosis of vasovagal syncope confer -1 point). It is important to note the authors mention the troponin value, although included in the score, was not present for the majority of patients, and further, values could not be compared across sites due to differing laboratory tests. As such, the >99th percentile cutoff was chosen, and a missing troponin was assumed to be below that percentile.

Of the 5,010 patients included in the final analysis, 106 (2.1%) patients experienced a study outcome with 0.9% of the outcomes occurring outside of the hospital. No subsequent analysis was performed on this sub-grouping. After analyzing the sensitivity, specificity, and positive and negative predictive values for each score, the authors assigned an estimated 30-day risk percentage with a score of ≤0 associated with <1% risk of outcome. Scores of 1-3 are associated with a 1.9-7.5% risk and scores 4-8 associated with a 14.3-22.2% risk. Not calculated in the paper, the negative likelihood ratio for a score ≤0 is 0 to 0.054 and the positive likelihood ratio for a score >0 is 4.2 to 24.5. As the prevalence for death due to unknown cause and arrhythmia in the population is low, the greatest utility may be that a low score (≤0) has a very good negative predictive value.

The authors discuss that this analysis represents a unique reporting of predictors for short-term arrhythmia or death following initial evaluation in the ED for syncope. These predictors are also consistent with past studies evaluating one-year outcomes. Interestingly, two predictors not included in the final CSARS mode (advanced age and absence of a prodrome) did show significant association with outcomes but when adjusted for other variables were not included in the final score calculation. A major strength that this score provides is that it parses out abnormalities in ECG, specifically QRS prolongation and QTc prolongation, rather than grouping the characteristic as “abnormal EKG” as most other prospective studies have done.[10-12] A major limitation was that one fifth of eligible study participants were not enrolled as the providers did not complete study forms during their evaluation. With regards to the large number of missing troponin values and assumption of normal results, the authors argue that the populations missing these values were substantially younger with fewer comorbidities and that the presumed normal value was reasonable even though this data was not specifically reported. Lastly, it is important to note that the score does include two, arguably subjective, predictors (final ED diagnosis of cardiac or vasovagal syncope). The predictor with the highest odds ratio was ED diagnosis of cardiac syncope (4.29) and the lowest odds ratio was ED diagnosis of vasovagal syncope (0.27). The investigators address this point noting that many currently used decision-making tools for venous thromboembolism and chest pain contain similar predictors and that when removed from the score, the remaining predictors provide sufficient risk stratification.

While CSARS still needs validation from other studies, it could offer additional information by providing estimated risks for patients. However, one can argue that some of the most important predictors remain the result of a provider’s clinical examination and history taking to differentiate vasovagal syncope from cardiac syncope which can be challenging.

P Prandoni, et al. Prevalence of Pulmonary Embolism Among Patients Hospitalized for Syncope. N Engl J Med. 2016 Oct 20;375(16):1524-1531.

This study sparked significant controversy and received attention last year when it was published in the New England Journal of Medicine. In this paper, the authors attempted to study the prevalence of PE in patients admitted to the hospital for syncope. Their main conclusion was that nearly 1 in 6 (approximately 17.3%) patients admitted for syncope had a PE. This figure is astounding as it goes against much of what many EM providers likely encounter in actual clinical practice.

This study was a cross sectional study which took place in 11 participating Italian hospitals. Importantly, the participating hospitals screened 2584 patients presenting with syncope during the study period. Of this original study population, 1867 were discharged, and 717 were admitted. Of the 717 admitted 157 were excluded for various reasons, leaving a final group of 560 admitted syncope patients for analysis. All 560 patients were evaluated with the simplified Wells Score to assess the pretest probability of PE, using dichotomous unlikely (if less than or equal to 4) versus likely (greater than 4) results to drive further testing. All patients underwent D-dimer testing as part of their initial syncope workup. Of the 560 patients evaluated, 330 had low pretest likelihood and negative D-dimers, and 230 had either a high likelihood, positive D-dimer, or both.

Of the 230 patients with a positive D-dimer or high likelihood of PE, one died and was found at autopsy to have a PE (proximal and bilateral), 180 had a CT scan, and 49 had a VQ scan. Out of the initial 229 patients who underwent advanced pulmonary imaging, 97 PEs were found. When compared to the initial number of patients who presented to the ED for syncope and were screened, this is a prevalence of PE in the undifferentiated syncope population of 3.8% (97/2584). A significant selection bias, selecting only admitted patients to run the tests, is contributing to the conclusions that were found. The rate of PE in the patients that were discharged is not known, and they were excluded from the final data set.

Why was the prevalence of PE so high in the admitted patients? The average age of these patients was 76 with no significant difference between those with or those without PE. But there were some very important key differences between those diagnosed with PE compared to those with negative imaging. Those diagnosed with PE had significantly higher rates of previous venous thromboembolism (VTE) (11.3% vs 4.3%), higher respiratory rates (45.4% vs 7.1%), more tachycardia (33% vs 16.2%), higher rates of having a blood pressure less than 110 mmHg (36.1% vs 22.9%), more clinical evidence of DVT (40.2% vs 4.5%), and a higher rate of active cancer (19.6% vs 9.9%). Basically, patients found to have PEs also had many other worrisome signs and symptoms to suggest they may indeed have a PE, and likely would have been worked up to exclude PE. The very fact that they were admitted suggests that their presentation was more worrisome to the EM provider.

What about the PEs that were found? Of the 180 patients who underwent CT, 72 were detected, with 30 in the main pulmonary artery, 18 in a lobar artery, 19 in a segmental artery, and 5 were sub-segmental. Of the 49 patients who underwent VQ scanning, 24 PEs were diagnosed with 4 involving more than 50% of the area of both lungs, 8 involving 26 to 50% of the area of both lungs, and 12 involving 1 to 25% of the area of both lungs.

While this study did follow a systematic protocol-based approach to exclude PE based on assessment of pretest probability, it did so for a subset of very high-risk syncope patients. Many had other risk factors and clinical features concerning for PE. Applying nonspecific clinical prediction scores and blood tests can lead to over-diagnosis and overestimation of the prevalence of disease, which likely accounts for the high rate reported in this study. As the authors correctly note, PE should be on the differential diagnosis of patients presenting with syncope; however, testing in the undifferentiated syncope patient can lead to misdiagnosis and medical misadventures. Therefore, evaluating for VTE or PE should be driven by the history, physical examination, risk factors, and validated clinical prediction scores applied to appropriately selected patients.

A Frizell, et al. Prevalence of Pulmonary Embolism in Patients Presenting to the Emergency Department with Syncope. Am J Emerg Med. 2017 Jul 31. pii: S0735-6757(17)30639-3.

The authors performed a retrospective, secondary analysis of prospectively gathered data from patients presenting with syncope to an academic ED from July 2010 to December 2015. Prospective subjects were asked, “Have you passed out in the last 24 hours?” A total of 778 patients were screened, with 348 eligible for inclusion. Of these, two were excluded because they were diagnosed with PE in the ED. Researchers calculated the PERC score for all enrollees. The patients were followed for 30 days after the visit. Telephone outreach successfully reached 68.4% of patients. Of the remainder, the electronic medical record system was reviewed for subsequent health care encounters. The end point was diagnosis of PE in any setting in the 30 days following enrollment. Data was analyzed using descriptive statistics and presented as percentages for categorical variable and means for continuous variables. Pearson’s chi-square test and Student’s t-test were used.

Demographic and historical attributes were collected including the prevalence of VTE (15.8%, CI:12.3-20). Patients reported symptoms of shortness of breath (54.3% CI: 49-59.5), chest pain (49.1%, CI:43.9-54.4), and calf pain/swelling (17%, CI:13.4-21.3). While in the ED, 22.7% of patients underwent PE testing by D-dimer, V/Q testing, or CTPA. In the study patients, 50.1% underwent further hospital evaluation (23.9% placed in observation unit). None of the patients who remained in the hospital were diagnosed with PE outside of the two patients diagnosed in the ED. By telephone follow-up, three patients (0.9% CI: 0.3-2.5) reported a diagnosis of PE in the follow up period. Overall the rate of PE in patients presenting to the ED with syncope was 1.4% (95% CI: 0.6-3.3).

The authors discuss that their findings differ significantly from those of Prandoni, et al. Significant differences between the two studies help explain this. The Prandoni study did not include all patients with syncope as it excluded those who had syncope attributed to vasovagal events, pharmacologic iatrogenesis, situational provocation, or hypovolemia. Therefore, the Prandoni study evaluated the prevalence of PE in only a subset of patients with syncope. This work up was done on the inpatient units. In contrast the study by Frizell, et al., relied largely on the discretion of the providers for work up, and telephone and EMR review for follow-up.

While it is convenient to partially disregard the high prevalence of PE in patients presenting with syncope found by Prandoni, we must do so with caution. It is perhaps more palatable to the ED provider to accept a lower incidence given the commonality of syncope as a chief complaint, but must remember that patients that are sick enough to warrant an admission are likely to have more risk factors that should put a PE on the list of possible causes for the syncope.

Conclusion
Syncope is a complex and difficult chief complaint for the EM physician to evaluate, mainly due to the multitude of potential underlying etiologies and the lack of any single well-validated risk-stratification tool. Though several risk scoring criteria have been shown to be helpful, at the end of the day, clinical judgment and maintaining a broad differential remain the key components to appropriately managing these patients. In the patient with no obvious cause of syncope after thorough evaluation, it appears that arrhythmia is the most common cause of syncope symptoms, and that PE is likely not as prevalent a cause for syncope as was reported by Prandoni. As EM providers, we must continue to strive to both protect our patients while simultaneously utilizing our resources appropriately and avoiding unnecessary admissions and undue burdens on the health care system.

References:

1. R Saccilotto, C Nickel, H Bucher, E Steyerberg, R Bingisser, M Koller. San Francisco Syncope Rule to Predict Short-term Serious Outcomes: A Systematic Review. CMAJ. 2011 Oct 18; 183(15): e1116–e1126.

2. S Safari, A Baratloo, B Hashemi, F Rahmati, M Forouzanfar, M Motamedi, L Mirmohseni. Comparison of Different Risk Stratification Systems in Predicting Short Term Serious Outcomes of Syncope Patients. Journal of Research in Medical Sciences. June 2016. 21:57.

3. V Thiruganasambandamoorthy, I Stiell, B Rowe, M Mukarram, K Arcot, K Kwong, A McRae, G Wells, M Taljaard. Predicting Short-Term Risk of Arrhythmia among Patients with Syncope: The Canadian Syncope Arrhythmia Risk Score. Academic Emergency Medicine. August 2017. doi: 10.1111/acem.13275.

4. P Prandoni, AW Lensing, MH Prins, M Ciammaichella, M Perlati, N Mumoli, E Bucherini, A Visonà, C Bova, D Imberti, S Campostrini, S Barbar; PESIT Investigators. Prevalence of Pulmonary Embolism Among Patients Hospitalized for Syncope. New England Journal of Medicine. October 2016. 375(16): 1524-31.

5. A Frizell, N Fogel, J Steenblik, M Carlson, J Bledsoe, T Madsen. Prevalence of Pulmonary Embolism in Patients Presenting to the Emergency Department with Syncope. Am J Emerg Med. 2017 Jul 31. pii: S0735-6757(17)30639-3.

6. J Quinn, I Stiell, D McDermott, K Sellers, M Kohn, G Wells. Derivation of the San Francisco Syncope Rule to Predict Patients with Short-term Serious Outcomes. Annals of Emergency Medicine. 2004 Feb;43(2):224-32.

7. F Colivicchi, F Ammirati, D Melina, V Guido, G Imperoli, M Santini; OESIL (Osservatorio Epidemiologico sulla Sincope nel Lazio) Study Investigators. Development and Prospective Validation of a Risk Stratification System for Patients with Syncope in the Emergency Department: The OESIL Risk Score. European Heart Journal. 2003 May;24(9):811-9.

8. S Grossman, C Fischer, L Lipsitz, L Mottley, K Sands, S Thompson, P Zimetbaum, N Shapiro. Predicting Adverse Outcomes in Syncope. Journal of Emergency Medicine. 2007 Oct;33(3):233-9.

9. M Reed, D Newby, A Coull, R Prescott, K Jacques, A Gray. The ROSE (Risk Stratification of Syncope in the Emergency Department) Study. Journal of the American College of Cardiology. 2010 Feb 23;55(8):713-

10. Del Rosso A, Ungar A, Maggi R, Giada F, Petix NR, De Santo T, Menozzi C, Brignole M. Clinical Predictors of Cardiac Syncope at Initial Evaluation in Patients Referred Urgently to a General Hospital: The EGSYS Score. Heart. 2008;94:1620-1626.

11. F Sarasin, B Hanusa, T Perneger, M Louis-Simonet, A Rajeswaran, W Kapoor. A Risk Score to Predict Arrhythmias in Patients with Unexplained Syncope. Academic Emergency Medicine. 2003;10:1312-1317.

12. T Martin, B Hanusa, and W Kapoor. Risk Stratification of Patients with Syncope. Annals of Emergency Medicine. 1997;29:459-466.