Oh JH, Hanusa BH, Kapoor WN. Do Symptoms Predict Cardiac Arrhythmias and Mortality in Patients With Syncope?. Arch Intern Med. 1999;159(4):375-380. doi:10.1001/archinte.159.4.375
Patients with syncope frequently present with multitude of other symptoms but their significance in predicting morbidity or mortality has not been previously studied.
To determine if certain symptoms can be used to identify syncope patients at risk for cardiac arrhythmias, mortality, or recurrence of syncope.
Patients and Methods
From August 1987 to February 1991, we prospectively evaluated patients with syncope from outpatient, inpatient, and emergency department services of a university medical center. These patients were interviewed, charts were reviewed, and detailed information on 19 symptoms and comorbidities was obtained. A cause of syncope was assigned using standardized diagnostic criteria. All patients were followed up at 3-month intervals for at least 1 year for recurrence of syncope and mortality. Patients in whom the cause of syncope was determined by medical history and physical examination alone were not included in our analysis.
History and physical examination led to the cause of syncope in 222 of 497 patients enrolled. In the remaining 275 patients, the absence of nausea and vomiting before syncope (odds ratio, 7.1) and electrocardiographic abnormalities (odds ratio, 23.5) were predictors of arrhythmic syncope. Underlying cardiac disease was the only predictor of 1-year mortality. No symptom remained as independent predictor for 1-year mortality or syncope recurrence.
Symptoms, although important in assigning many noncardiac causes, are not useful in risk-stratifying patients whose cause of syncope cannot be identified by other history and physical examination. Triage decisions and management plans should be based on preexisting cardiac disease or electrocardiographic abnormalities, which are important predictors of arrhythmic syncope and mortality.
SYNCOPE IS a common disorder accounting for 1% to 6% of the medical admissions and up to 3% of emergency department visits.1 Syncope is difficult to evaluate and as a result patients with this symptom often are subjected to many diagnostic tests and hospitalization. Previous studies2,3 have shown that medical history and physical examination can be used to diagnose many of the noncardiac causes of syncope (eg, micturition syncope and orthostatic hypotension). When medical history and physical examination do not lead to a cause of syncope, coexisting symptoms such as chest pain, palpitations, or absence of prodrome are frequently used to plan diagnostic testing and to identify high-risk subsets of patients. Patients with syncope often present with a multitude of symptoms before, during, or after the syncopal episode and may also have many other symptoms that are not temporally related to the loss of consciousness (LOC). Symptoms have been used in prior studies4,5 in an attempt to separate vasovagal syncope from cardiac causes of syncope. Martin et al6 found a shorter duration of prodromal symptoms in patients with cardiac cause of syncope, but a cardiac cause was present in only 4% of those patients (much lower than in other cohort studies).2 Calkins et al5 found that sex, age, number of syncopal episodes, and duration of prodrome were predictive of arrhythmic cause. However, the population analyzed was restricted to those with known diagnosis of arrhythmia or vasovagal syncope, so its applicability in predicting cardiac causes in unselected patients at the initial presentation is unclear. The importance of symptoms in predicting the risk of mortality and recurrences has not been previously studied.
The objectives of this study were to describe the symptoms and characteristics associated with syncopal episodes and to determine if they can be used to identify patients at risk for cardiac arrhythmias as a cause of syncope, and to predict mortality, or recurrence of syncope.
Syncope was defined as a sudden transient LOC with an inability to maintain postural tone.6- 9 Patients who had symptoms compatible with seizure disorder, vertigo, dizziness, coma, shock, or other states of altered consciousness were not included. Patients must have regained consciousness spontaneously. Those who required pharmacological or electrical cardioversion at initial presentation were excluded. Index syncopal episode was defined as the last syncopal episode before enrollment in the study.
Patients aged 18 years or older with possible symptoms of syncope were identified through an exhaustive search, between August 1987 and February 1991, from 3 different sources at the University of Pittsburgh Medical Center, Pittsburgh, Pa. Daily review of inpatient admission lists, emergency department, and outpatient visit logs, as well as contact with faculty and house staff at these services, were used for the identification of patients with syncope. After an interview and chart review, those who met our criteria for having had a syncopal episode were asked to participate in the study.
The revised Declaration of Helsinki and the Guidelines for Good Clinical Practice were followed. This study was approved by the institutional review board. Written informed consent was obtained from each patient before his or her enrollment in the study.
All patients enrolled in the study underwent a standardized evaluation that consisted of the following: a complete history and physical examination; baseline laboratory examination (complete blood cell count, electrolytes, serum urea nitrogen, creatinine, and glucose); a 12-lead electrocardiogram (ECG); and prolonged ECG monitoring of at least 24 hours (Holter or bedside telemetry monitoring). Patients in whom initial evaluation suggested a cause for syncope underwent further definitive evaluation. Extensive consultations with cardiologist and neurologist were used, and further diagnostic tests such as electrophysiological studies and tilt-table test were performed at the discretion of the patient's physician or consultants.
Initial ECGs were classified into 3 groups: (1) normal was defined as an ECG with no abnormalities, or the presence of sinus bradycardia or sinus tachycardia only; (2) NST was defined as the presence of nonspecific ST- and T-wave abnormalities only; and (3) abnormal was defined as the presence of any rhythm abnormalities (eg, atrial fibrillation or flutter, multifocal atrial tachycardia, junctional or paced rhythms), conduction disorders (eg, left axis deviation, bundled-branch block, or intraventricular conduction delay), atrioventricular block (eg, complete atrioventricular block, Mobitz II, or Mobitz I with other abnormality present), short P-R interval (<0.10 second), signs of old myocardial infarctions, left or right ventricular hypertrophy, and frequent or repetitive premature ventricular beats (including ventricular tachycardia).
For the purpose of this study symptoms were classified in 2 groups: (1) immediately before the LOC or (2) during or immediately after the LOC. Symptoms during and after the LOC were combined because patients rarely could distinguish these periods. The following 20 symptoms were assessed: dizziness, diaphoresis, nausea and/or vomiting, generalized weakness, visual changes, flushing, dyspnea, headache, chest pain, abdominal pain, palpitations, tingling, vertigo, lethargy, confusion, incontinence, aura, neurologic deficits, pruritus, or tonic-clonic movements. Prodrome was defined as having 1 or more of the above symptoms before LOC. Dizziness was excluded from our definition of prodrome because most patients experience it before LOC. Any symptom reported by fewer than 5 patients (2%) was excluded from our analyses. The excluded symptoms were aura, neurologic deficits, and pruritus before or after LOC; confusion, incontinence, and tonic-clonic movements before LOC; and abdominal pain and flushing after LOC. Information about frequency of previous episodes, estimated duration of LOC, position before syncope, and trauma during the event were also obtained. Witnesses were questioned whenever available.
Information about the incidence of trauma during the syncopal episode and the duration of LOC was also obtained. Trauma was classified as (1) major trauma, defined as any fractures, head injury, or internal organ damage, or syncope leading to an automobile crash; or (2) minor trauma, defined as any bruise, cuts, or soft tissue injury. Duration of LOC (obtained from both patients and witnesses) was classified into short (<2 minutes) or long (≥2 minutes).
Because of the heterogeneity of the underlying cardiac diseases in patients presenting with syncope, a heart disease classification using the presence of cardiac disease and functional limitations to categorize cardiac comorbidities was developed and has been reported previously.10 The 5 classes of heart disease classification were class 0 (no cardiac disease), class 1 (heart disease without limitation to ordinary physical activity), class 2 (heart disease with limitation to ordinary physical activity), class 3 (heart disease with limitation to less than ordinary physical activity), and class 4 (heart disease and symptoms at rest). Heart disease was defined as coronary artery disease, valvular heart disease, cardiomyopathy, congestive heart failure, or other evidence of organic heart disease clinically or by invasive or noninvasive test results.
A cause for syncope was assigned using a modified version of previously reported diagnostic criteria.6,8 These criteria were based on review of pertinent literature and have been used in other studies of syncope. Modifications in the diagnostic criteria for arrhythmias as a cause of syncope were (1) definite diagnosis of ventricular tachycardia was used if ECG monitoring showed sustained (≥30 seconds) or symptomatic ventricular tachycardia, or there was sustained monomorphic ventricular tachycardia on electrophysiologic studies; and (2) definite diagnosis of sick sinus syndrome was given when sinus pauses were concurrently symptomatic or lasted 3 seconds or more. Unsustained ventricular tachycardia or sinus pauses of 2 to 2.9 seconds were classified in the "possible" category in the absence of another cause by clinical or laboratory testing. Diagnostic Interview Schedule modules for major depression, generalized anxiety disorder, panic disorder, somatization disorder, and alcohol and substance abuse were used for assessment of psychiatric comorbidity as has been previously reported.11- 13
All patients were followed up by telephone interview at 3-month intervals for information about the recurrence of syncope, new cardiovascular events, and mortality. When patients could not be reached, a family member and/or primary physician of the patient was contacted. All subsequent evaluations, hospital admissions, and other medical events were reviewed when records could be obtained.
Blinded reviews of the summary of information available from the circumstances surrounding the death, medical records, death certificates, records of the Bureau of Vital Statistics, and autopsy reports, when available, were used to assign the cause of death based on the World Health Organization guidelines incorporating an underlying cause and an immediate cause of death.14 Cardiac death was defined as underlying cause of death from coronary artery disease, valvular heart disease, congestive heart failure, pulmonary hypertension, congenital heart disease, restrictive cardiomyopathy, and sudden death.10
The primary focus of the analysis was to identify symptoms and characteristics of the syncopal episode that were significantly associated with any of the 3 measured outcomes: cardiac arrhythmia as cause of syncope, time to death in the first year following enrollment, and time to recurrence of syncope. These symptoms or characteristics in conjunction with previously reported predictors of these outcomes were then entered in multivariate analysis models to assess their relative statistical power. Secondary analyses were also carried out for cardiac death as the outcome.
For univariate predictors of arrhythmias, χ2 statistics were used for categorical and t test for continuous variables. Symptoms significant in the univariate analysis at the 0.1 level and previously reported correlates of arrhythmias (age >45 years, sex, race, history of congestive heart failure, myocardial infarction, or ventricular arrhythmias, and ECG abnormalities)15 were used for multivariate modeling with logistic regression analysis.
For predictors of mortality and recurrence, univariate analyses were performed using Kaplan-Meier methods and Wilcoxon statistics. Symptoms significant in the univariate analysis and previously reported correlates of mortality (age >45 years, sex, history of congestive heart failure, diabetes mellitus, coronary artery disease or ventricular arrhythmias, heart disease classification, absence of prior syncopal episodes, and left ventricular hypertrophy on the ECG or abnormal ECG)10,15 and recurrences (age <45 years, ≥4 previous episodes, diagnosis of syncope of unknown cause, and psychiatric comorbidities)16 were used for multivariate modeling using Cox proportional hazards methods.
A total of 668 patients who met the criteria for syncope presented to the medical center during the enrollment period. Of these patients, 127 declined to participate, 38 could not be located after the initial presentation, and 6 failed to provide all necessary symptom information. The remaining 497 patients were enrolled in a prospective study. The patients who did not consent to participate in the study were significantly younger (age, 46.4 years) than the cohort analyzed (age, 57 years; P<.01), but they were similar in other demographic characteristics such as sex (56% vs 58% women) and race (19% nonwhites in both groups).
Cardiac causes for syncope were found in 76 patients (15%), of whom 55 had cardiac arrhythmias. Noncardiac diagnosis was assigned to 268 (54%) and syncope of unknown cause to 153 (31%). The causes of syncope were similar to other studies2,6- 8 published in the literature (Table 1). History and physical examination led to the cause of syncope in 222 (45%) of the patients. These diagnoses included situational syncope (such as micturition or cough syncope), clinical vasovagal syncope, orthostatic hypotension, and others (Table 1). The analyses for the predictors of high-risk patients were performed with the remaining 275 patients in whom the cause of syncope was not diagnosed by the history and physical examination since diagnosing arrhythmia or cardiac death is often a concern in this group of patients.
The characteristics of the 497 patients with syncope and the subset of 275 patients included in the main analysis are shown in Table 2. The mean age of the cohort analyzed was 57 years, 58% were women, and 81% were white. This was the first syncopal episode in 34% of patients and 14% had more than 4 events during the year before enrollment. Hypertension was present in 32% of the patients, and history of myocardial infarction, diabetes, and congestive heart failure was present in 10% to 11%.
The most frequent symptoms (Table 3) immediately before the LOC were dizziness (50%), diaphoreses (20%), nausea and/or vomiting (17%), and generalized weakness (15%). The most common symptoms during or after the LOC were generalized weakness (33%), confusion (21%), diaphoreses (16%), and nausea and/or vomiting (15%). Most patients were standing (48%) or sitting (46%) before the event, and the LOC was of short duration (<2 minutes, 72%). Trauma occurred in 88 patients (32%) during the index syncopal episode, but only 14 patients (5%) had major trauma.
Cardiac arrhythmias were diagnosed in 55 patients in this study. Symptoms that were significant univariate predictors of cardiac arrhythmia (P≤.05) were absence of nausea and/or vomiting or prodrome before syncope, and headache before or after syncope (Table 4). In multivariate logistic regression analysis that included age, number of previous syncopal episodes, cardiac comorbid conditions, presence of trauma, duration of index syncopal episode, and initial ECG, only the absence of nausea and/or vomiting before syncope and having an abnormal ECG remained in the model as significant independent predictors of cardiac arrhythmias as the cause of syncope (Table 5).
The mean follow-up period was 34.3 months, and 99% of the analyzed cohort had at least 12 months of follow-up. In the total cohort of 497 patients, 42 died during the first year of follow-up (15 patients had a cardiac cause of death). In the analyzed cohort of 275 patients, 24 patients died (11 patients had a cardiac cause of death). The symptoms of dyspnea and flushing before syncope were found to be significant univariate predictors of time to death (Table 4). In Cox proportional hazards model that included age, number of previous syncopal episodes, heart disease classification cardiac comorbid conditions, presence of trauma, duration of index syncopal episode, and initial ECG, only the heart disease classification class remained a significant predictor of mortality (Table 5). The heart disease classification was also the only independent predictor of 1-year mortality when cardiac death was used as the outcome.
The presence of vertigo before syncope was a univariate predictor of recurrence (Table 4). In Cox proportional hazards model with age, number of previous syncopal episodes, cardiac comorbid conditions, presence of trauma, duration of index syncopal episode, and initial ECG, only age younger than 45 years, history of 4 or more syncopal episodes in the past year, and more than 1 psychiatric diagnosis remained as significant predictors of recurrence (Table 5).
Prior descriptive studies show that symptoms associated with syncope can be important in assigning many noncardiac causes of syncope. For example, entities such as micturition, cough, and defecation syncope are only diagnosed by a careful history taking. Our study shows that in patients without these diagnoses, symptoms surrounding LOC are not helpful in identifying patients for arrhythmic syncope, 1-year mortality, or recurrence.
The management of patients in whom a noncardiac diagnosis is established by history and physical examination is relatively straightforward. However, in a large subset of patients (55% of the patients in our cohort) the cause of syncope was not clear after the history taking and physical examination. Because of the concern for mortality or cardiac arrhythmias as a cause of syncope, these patients often undergo a lengthy and costly workup.2 Symptoms and other characteristics of the syncopal episodes such as chest pain, palpitations, trauma, duration of LOC, or even absence of a prodrome are frequently used to identify high-risks subsets of patients for cardiac arrhythmias as a cause of syncope as well as mortality. To our knowledge, this is the first study to systematically evaluate the usefulness of these symptoms and characteristics in identifying high-risk patients. Our study shows that many symptoms associated with the LOC were not useful in predicting arrhythmic syncope. Only absence of nausea and vomiting was an independent predictor in identifying those patients at higher risk for arrhythmic syncope after controlling for known predictors of cardiac arrhythmias. A possible explanation for this finding is that nausea and vomiting before LOC may be an indicator for vasovagal or psychogenic syncope, which are not easily diagnosed clinically. However, we do not have evidence from our study to suggest this possibility. Duration of LOC also failed to predict cardiac arrhythmias as a cause of syncope. We recognize that it is difficult to define the duration of the period of LOC. Even when witnesses are present it may be hard to determine the exact onset and/or the end of the LOC period. However, neither short duration (<2 minutes) nor long duration (≥2 minutes) proved to be useful in predicting any of our predefined outcomes.
In this study symptoms and characteristics of syncope also failed to predict 1-year mortality, which was largely determined by the presence and severity of cardiac comorbidities. Prior studies10,15 have also shown that, in patients with syncope, mortality was predominantly attributable to cardiac comorbidities, but these studies have not assessed the role of symptoms.
Patients at risk of recurrence have reported functional declines as significant as severe rheumatoid arthritis, back pain, and psychiatric disorders.17 Historical features such as younger age, multiple syncopal episodes in past year, and psychiatric illnesses at presentation were useful in predicting recurrence. No individual symptom was found to be a predictor of recurrence; however, our previous study16 showed that multiple somatic symptoms are associated with the presence of psychiatric illness in patients with syncope.
Three prior studies have addressed the significance of clinical history in establishing the cause of syncope. The first study was a retrospective review by Wayne4 in the early 1960s in an attempt to describe the clinical characteristics of syncope according to its cause. This article provided extensive description of clinical features of various entities but did not analyze the relationships of symptoms to predict high-risk subsets. Martin et al6 have prospectively evaluated the clinical history in 170 patients presenting to the emergency department with syncope. There was a shorter duration of warning symptoms for those patients with cardiac cause of syncope when compared with those with vasovagal syncope. However, cardiac causes were present in only 7 patients and multivariate analyses were not performed to control for other factors. Furthermore, the reliability of duration of prodrome has not been tested. Our results contrast with the findings from the study by Calkins et al,5 who found that male sex, age older than 54 years, 2 episodes of syncope or less, and duration of warning of 5 seconds or less were predictive of syncope due to cardiac arrhythmias. There are major differences between our study and that of Calkins and coworkers in the definition of cardiac arrhythmias and the population studied. Calkins et al restricted their analysis to patients already known to have neurocardiogenic syncope (32 patients) and syncope due to atrioventricular block (n = 32) or ventricular tachycardia (n = 16). Their study design highlights the differences in clinical characteristics between these 2 diagnoses. In our study, on the other hand, all patients presenting to our medical center in whom the cause of syncope was not clear after history and physical examination were analyzed. This is the usual clinical scenario faced by clinicians evaluating these patients, where the need for risk stratification is the greatest to plan further diagnostic testing.
We acknowledge that there are several limitations to our study. First, our cohort consisted of patients with syncope presenting to a tertiary academic medical center. Although most of our patients (374 of 497) presented themselves to our emergency department, some patients may have been referred to our institution specifically for syncope evaluations. The generalizability of our findings may need further testing in other clinical settings. Second, electrophysiologic studies were performed based on usual clinical practice. Although more arrhythmias may have been diagnosed if electrophysiologic studies were performed in all patients, the interpretation of these tests is still highly controversial. Problems with sensitivity and specificity of many findings led us to use this test as deemed necessary by the physician caring for the patient. Last, our follow-up period was only 1 year. Our results may have been different with a longer period of follow-up.
In conclusion, our study shows that symptoms associated with syncope, although important in assigning many noncardiac causes, are not useful in risk-stratifying patients. Thus, triage decision and management plans should be based on the presence of cardiac disease or ECG abnormalities that are important predictors of arrhythmic syncope and mortality.15
Accepted for publication May 26, 1998.
This study was supported in part by grant RO1 HL36735 from the National Heart, Lung, and Blood Institute, Bethesda, Md. Dr Kapoor is a recipient of Research Career Development Award from the National Heart, Lung, and Blood Institute (K04L 01899).
Reprints: Wishwa N. Kapoor, MD, MPH, Montefiore University Hospital, Suite W933, 200 Lothrop St, University of Pittsburgh Medical Center, Pittsburgh, PA 15213-2582 (e-mail: Wnk@med.pitt.edu).