Does This Patient With Palpitations Have a Cardiac Arrhythmia?

A 58-year-old woman presents to the emergency department with intermittent episodes of palpitations. She describes “heart fluttering” that usually lasts less than 5 minutes, which is associated with a sense of “impending doom,” sweating, and paresthesia in both hands. She is unable to tell whether the rhythm is regular or irregular and denies a regular rapid-pounding sensation in the neck. There is no associated presyncope or syncope. She has a history of panic disorder but is otherwise healthy and takes no medications. Her pulse rate and rhythm are palpably normal and the rest of her physical examination along with a 12-lead electrocardiogram is normal.

Palpitations are a common, unpleasant, and often alarming awareness of heartbeats,¹ with a prevalence as high as 16% in medical outpatients.² They often pose a clinical challenge³ because of the wide differential diagnosis (Box). Quiz Ref IDPalpitations may occur due to a change or abnormality in heart rhythm, such as an arrhythmia (ie, an abnormal, disordered, or disorganized heartbeat), due to an appropriate increase in normal sinus rate, or with a normal sinus rate and rhythm due to heightened sensitivity and perception of one's heartbeats (eFigure 1).

In one study, primary cardiac disease (43%) and anxiety or panic disorder (31%) were the most common causes in patients presenting with palpitations to the emergency department, admitted to the hospital, or attending a medical clinic.⁴ Among patients with cardiac disease, palpitations were attributable to arrhythmia in 91% of cases. Thus, the pretest probability of cardiac arrhythmia in a similar patient population would be 39%. In 2 other studies, 19% of patients presenting with palpitations were found to have a clinically significant arrhythmia.^{5 - 6}

Because a minority of patients have palpitations while being examined by their physician, the challenge is to capture a recording of the cardiac rhythm during symptoms. While event monitors have been designed to facilitate this process, the diagnostic yield varies with the frequency of symptoms and duration of the monitored period. Arrhythmias also may occur in individuals who have no symptoms at all.⁷ Therefore, the presence of an arrhythmia on diagnostic testing does not confirm that it is the cause of a patient's symptoms.⁷ To be certain, their symptoms must be correlated with an electrocardiographically documented rhythm disturbance. Similarly, if the patient repeatedly has a normal cardiac rate and rhythm during typical symptoms, one can reassure the patient that the cause is likely nonarrhythmic.

While palpitations are usually benign, they may be a manifestation of life-threatening conditions. More importantly, recurrent palpitations can be associated with significant disability, including impaired work performance and the inability to perform household duties.^{4 ,8} However, using diagnostic tests such as event monitors and echocardiograms for every person with palpitations can be costly and of low diagnostic yield. Therefore, we reviewed the utility of clinical history, physical examination, and resting routine electrocardiography as screening tests for identifying patients with palpitations whose symptoms are likely or unlikely to be due to a cardiac arrhythmia.

Most demographic and historical features do not significantly influence the likelihood of clinically significant arrhythmias.^{5 - 6} Patient age may be important because supraventricular tachycardias, particularly ones that use a bypass tract (atrioventricular reentry tachycardia) (eFigure 1D), may be first experienced earlier in life.^{9 - 10} In young athletes with palpitations, it is important to consider clinically significant arrhythmias associated with sudden cardiac death. Atrial fibrillation, flutter, atrial tachycardia, and ventricular tachycardia (eFigure 1A, B, I) tend to occur later in life and are often associated with structural heart disease. Some arrhythmias such as atrioventricular node reentry tachycardia (eFigure 1C) may be more common in women than men.^{9 - 11}

A history of panic disorder should be explored.¹² The details of a family history of palpitations should be recorded, especially if family members have established diagnoses such as arrhythmogenic right ventricular cardiomyopathy¹³ or atrial fibrillation.¹⁴ Any history of previous cardiac disease may predispose patients to more clinically significant arrhythmias^{4 ,6} and suggest the need for a more aggressive search for a cardiac cause.

Patients should be asked to tap out the rhythm of their palpitations, or to choose from cadences tapped by the physician, to identify the regularity and speed of the palpitations. Single skipped beats or a sensation of the heart stopping and then starting with a pounding, flipping, or jumping sensation, especially while sitting quietly or lying in bed and lasting only for brief periods, have traditionally been attributed to premature atrial or ventricular extra systoles.^{15 - 16} An irregular heartbeat, both in rhythm and strength, that begins and terminates abruptly suggests atrial fibrillation.

The association of polyuria and palpitations may indicate supraventricular tachycardia because increased atrial pressures stimulate production of natriuretic peptides.¹⁷ Quiz Ref IDA regular rapid-pounding sensation in the neck may signify atrioventricular node reentry tachycardia when the contraction of the atria against closed atrioventricular valves produces increased right atrial pressures and reflux of blood into the superior vena cava.¹⁸ Associated shirt flapping, defined as visible movement of patient's clothes during the episode, also has been described with both atrioventricular node reentry tachycardia and atrioventricular reentry tachycardia.¹⁹ Presyncope or syncope may represent more clinically significant arrhythmias such as ventricular tachycardia (eFigure 1I).¹⁶ However, syncope can occasionally result from acute vasodilatation and/or rapid heart rate with low cardiac output that occurs at the beginning of a supraventricular tachycardia^{20 - 21} or due to conversion pauses occurring at the end, especially in patients with underlying sinus node disease. Conditions such as hyperthyroidism could be associated with sinus tachycardia or atrial fibrillation (eFigure 1G, I). Similarly, palpitations associated with a psychiatric diagnosis such as panic disorder could suggest sinus tachycardia. However, it is essential to rule out clinically significant arrhythmias before attributing palpitations to the patient's psychiatric condition.^{4 ,8 ,22}

Onset during catecholamine excess, such as during exercise, may suggest ventricular tachycardia or sinus tachycardia (more commonly).²³ Palpitations starting during sleep or states of increased vagal tone (eg, at termination of exercise) can be associated with vagal-mediated atrial fibrillation or, less likely, certain subtypes of long-QT syndromes.²⁴ Other triggers for tachycardias include alcohol or caffeine consumption.^{25 - 26}

While patients with QT prolongation and associated arrhythmias usually present with syncope, a medication review is warranted. Drugs that prolong QT and predispose patients to torsades de pointes and other ventricular arrhythmias include antiarrhythmics, antimicrobials, antihistamines, psychotropic drugs, and other miscellaneous drugs such as motility drugs, diuretics via electrolyte depletion, and protease inhibitors for human immunodeficiency virus.^{27 - 30}

Most patients with episodic palpitations are examined when asymptomatic. Typically, the purpose of the physical examination in this setting is to identify structural heart abnormalities that may give rise to an arrhythmia. When a patient is examined while having palpitations or the examiner detects an asymptomatic arrhythmia, certain physical examination features may be useful. Quiz Ref IDAtrial fibrillation is suggested by a pulse that is not regular and has no repeating pattern (ie, irregularly irregular), the presence of a pulse deficit (ie, obtaining a lower pulse rate at the wrist than at the apex), or the auscultation of variable first heart sound intensity. These findings are due to beat-to-beat variation in stroke volume that occurs during atrial fibrillation. The presence of cannon A waves on the jugular venous pressure suggests an arrhythmia associated with atrioventricular dissociation such as ventricular tachycardia.³¹ A cannon A wave is a prominent wave in the jugular venous pressure that occurs due to the contraction of the right atrium against a closed tricuspid valve.

Standard 12-lead electrocardiography is the initial test in patients with palpitations and may identify the arrhythmia or provide insight into underlying structural and electrical abnormality that may be a precipitant for arrhythmias. Patients with electrical or structural abnormalities on 12-lead electrocardiography may warrant a more aggressive search for a cardiac cause of palpitations.

The prototypical clinical event monitor is the Holter monitor that continuously and simultaneously records 2 or 3 electrocardiographic leads. At the end of the monitoring period (typically 24 or 48 hours), the data are analyzed for arrhythmias (eFigure 1) and are correlated with symptoms recorded by the patient. The Holter monitor detects asymptomatic arrhythmias and may capture arrhythmias in patients who are unable to trigger the device (eg, during syncope). Frequently, patients may not experience their usual symptoms during monitoring and the test is nondiagnostic.

Intermittent event recorders can be worn continuously (loop recorders) or applied at the time of symptoms (event recorders).³² Traditionally, intermittent event recorders store electrocardiographic monitoring for several minutes once activated by the patient and hence cannot capture asymptomatic arrhythmias or those associated with loss of consciousness.³² Newer loop recorders provide continuous, real-time outpatient electrocardiographic monitoring and can automatically detect asymptomatic arrhythmias in addition to being activated by the patient.^{33 - 34} Intermittent event recorders allow for prolonged monitoring (weeks to months) in patients who have infrequent symptoms. These devices may have a higher specificity because the patient activates the recording during symptoms. Specifically, loop monitors save information for a predetermined period prior to the patient trigger, and hence, can help identify the initiation sequence for arrhythmias. These stored events can be transmitted through a telephone for physician review.

An electrophysiologic study is an invasive test of the electrical conduction system of the heart. Although often performed for diagnostic and therapeutic purposes in patients with a known arrhythmia or who have presented with syncope or resuscitated sudden cardiac death, it is occasionally performed as a diagnostic test in patients with palpitations in whom there is high suspicion for cardiac origin.³⁵

Exercise treadmill testing with a standard Bruce protocol may be useful in patients whose palpitations typically occur during exercise or are provoked by cardiac ischemia.^{15 ,36}

When palpitations occur infrequently or are associated with serious events such as syncope that cannot be identified using intermittent event recorders, implantable loop recorders (implanted under the skin in the left parasternal region) can record the patient's electrocardiogram continuously for prolonged periods (several months to years).^{37 - 38} Patients keep a diary of their symptoms for symptom-rhythm correlation. The device can also be triggered with an external activator.

Echocardiography may identify structural heart diseases that may be a precipitant for arrhythmias.^{15 ,39} While the presence of structural heart disease increases the likelihood of a clinically significant arrhythmia and suggests the need for a more aggressive search for an arrhythmic substrate, it does not prove that the patient's palpitations are secondary to an arrhythmia.

Structured MEDLINE (1950 to August 25, 2009) and EMBASE (1947 to August 2009) literature searches were performed to identify English-language articles relevant to the precision or accuracy of the clinical examination for patients with palpitations. Search terms included palpitations, heart racing, heart pounding, physical examination, medical history taking, professional competence, “sensitivity and specificity,” reproducibility of results, observer variation, “diagnostic tests, routine,”decision support techniques, Bayes theorem, and mass screening. Two authors independently reviewed the abstracts of the search and retrieved potentially relevant articles and a third author resolved disagreements. Additional articles were identified by reviewing the reference lists of retrieved articles and expert suggestions.^{15 - 16 ,40}

Articles reporting original empirical studies evaluating historical features, physical examination, or diagnostic tests against a reference standard for the diagnosis of palpitations secondary to an arrhythmia were included. Acceptable reference standards included clinical event monitors, intermittent event recorders, implantable loop recorders, in-hospital telemetry, 12-lead electrocardiographic monitoring during symptoms, or electrophysiological study. Excluded studies (1) focused primarily on nonarrhythmic diagnoses in patients with palpitations; (2) enrolled patients with several presenting complaints but did not provide separate data for the subgroup with palpitations; (3) focused only on comparison between specific arrhythmias or used the presence of arrhythmias as opposed to palpitations as inclusion criteria; or (4) did not require symptom rhythm correlation for the diagnosis of arrhythmia. From the results of the same literature search, studies were identified providing data on the diagnostic yield of the various tests (eg, electrocardiography and loop monitoring).

The data extracted were the number of patients enrolled, symptoms, signs or tests assessed, the number of patients with and without arrhythmia for each clinical parameter, and the frequency of typical symptoms and clinically significant arrhythmias (when present). From this, the likelihood ratios (LRs) were calculated for the individual findings described, along with the 95% confidence intervals (CIs).

Where possible, the LRs were separately calculated for detecting any arrhythmias and clinically significant arrhythmias. An arrhythmia was defined as any rhythm with a heart rate of 60/min or less, or 100/min or greater, and/or that was not normal sinus rhythm. Clinically significant arrhythmias were those that likely require specific management including ventricular tachycardia, atrioventricular node reentry tachycardia, atrioventricular reentry tachycardia, atrial fibrillation, atrial flutter, atrial tachycardia, junctional tachycardia, or ventricular ectopic beats occurring in salvos.

The yield of the various diagnostic tests was calculated and defined as the number of patients who had any arrhythmia or clinically significant arrhythmia during monitoring. Whenever available, separate data were provided for the subgroup of patients who had their typical symptoms during the monitoring period.

Articles were graded for methodological quality using standard methods with a threshold of more than 100 patients distinguishing level 1 from level 2 studies.⁴¹

Only 7 studies met inclusion criteria for the assessment of diagnostic accuracy (Table 1; eFigure 2 is a flow diagram illustrating the identification of articles).^{4 - 6 ,11 - 12 ,18 ,42} Palpitations were the predominant presenting complaint in these studies (99.4% of the included patients). The majority of the data was extracted from the 2 level 1 studies.^{5 - 6} Only 1 study assessed a limited number of physical examination signs in patients with palpitations.⁶ No study evaluated a combination of historical and physical examination features or the precision of any historical or physical examination feature. The reference standards in the included studies were electrophysiological study,¹⁸ 24-hour Holter monitor,^{12 ,42} intermittent event recorders,^{5 - 6} and in 2 studies a combination of methods.^{4 ,11} Among studies that used loop recorders (both for diagnostic accuracy and yield data), only 1 study³⁴ had the automatic trigger feature to record asymptomatic arrhythmias.

Only 2 of the 7 diagnostic accuracy studies distinguished clinically insignificant and significant arrhythmias,^{5 - 6} although only 1 allowed for the calculation of LRs for both types of rhythm disturbances.⁵ Other studies looked only at clinically significant arrhythmias^{11 ,18 ,42} or did not differentiate between the 2.^{4 ,12}

Seven studies examined the utility of the features on history for diagnosing an arrhythmia as the cause of palpitations (Table 2). Most of the data are obtained from studies with small sample sizes. Although several features increase the likelihood that a patient's palpitations were secondary to an arrhythmia, most have 95% CIs crossing unity and thus may not be clinically useful.

The only findings with an LR of 2.00 or greater for any arrhythmia were a history of cardiac disease (LR, 2.03; 95% CI, 1.33-3.11) and palpitations affected by sleeping (LR, 2.29; 95% CI, 1.33-3.94; which are presumably palpitations that are severe enough to wake patients up from sleep) or while the patient was at work (LR, 2.17; 95% CI, 1.19-3.96). Although description of palpitations as either regular (LR, 1.66; 95% CI, 1.20-2.29) or irregular (LR, 1.65; 95% CI, 1.22-2.22) had little value in the likelihood of cardiac arrhythmia, this information may be helpful in the right context because certain arrhythmias are typically regular while others are irregular.

The 2 factors with an LR of 0.50 or less for any arrhythmia were an underlying history of panic disorder (LR, 0.26; 95% CI, 0.07-1.01) and duration of palpitation less than 5 minutes (LR, 0.38; 95% CI, 0.22-0.63) (Table 2). However, these observations are based on single studies and the upper bound of the 95% CI for a history of panic disorder included 1.00.

The presence of an associated regular rapid-pounding sensation in the neck (LR, 177; 95% CI, 25-1251) increased the likelihood that the patient's symptoms of palpitations are due to atrioventricular node reentry tachycardia.¹⁸ The absence of an associated regular rapid-pounding sensation in the neck significantly decreased the likelihood of atrioventricular node reentry tachycardia (LR, 0.07; 95% CI, 0.03-0.19). However, a second study¹¹ found that neck fullness is not useful for distinguishing atrioventricular node reentry tachycardia from other arrhythmias (LR, 0.85; 95% CI, 0.44-1.64), but the presence of visible neck pulsations may be useful (LR, 2.68; 95% CI, 1.25-5.78). More recently, the description of palpitations in the neck in patients with documented narrow complex tachycardia was shown to distinguish atrioventricular node reentry tachycardia from atrioventricular reentry tachycardia with an LR of 2.41 (95% CI, 1.54-3.76).¹⁰ From the reviewed studies, no other features appear to be useful for ruling in or ruling out a clinically significant arrhythmia.

In these 7 studies,^{4 - 6 ,11 - 12 ,18 ,42} all patients including those who had no symptoms during the monitored period were included in the accuracy analysis. Only 1 study provided data on a subgroup of 81 patients who had their typical symptoms during the monitoring period.⁵ For these patients, the most useful feature for detecting an arrhythmia was the occurrence of palpitations at work (LR, 2.38; 95% CI, 1.03-5.50). Potentially useful features for detecting clinically significant arrhythmias include palpitations described as regular (LR, 1.52; 95% CI, 1.04-2.24) or those that were affected by sleeping (LR, 1.83; 95% CI, 1.03-3.27).

Although no study specifically assessed the accuracy of the associated symptom of shirt flapping in patients with palpitations, 1 study¹⁹ found that in 326 patients with documented arrhythmias, the proportion of patients with atrioventricular node reentry tachycardia (58%) who reported shirt flapping was greater than that reported by patients with other arrhythmias such as atrioventricular reentry tachycardia (44%), ventricular tachycardia (32%), atrial flutter (17%), and atrial fibrillation (13%).

Based on a single study⁶ (Table 3), the presence of resting bradycardia (<60/min) during the examination increases the likelihood of a clinically significant arrhythmia (LR, 3.00; 95% CI, 1.27-7.08). No other physical examination findings, including the presence of murmurs, have been evaluated in patients presenting with palpitations

No studies reported on the sensitivity and specificity of baseline 12-lead electrocardiographic abnormalities in predicting a cardiac arrhythmia as a cause of symptoms. Nevertheless, baseline electrocardiography is typically performed prior to other diagnostic tests.^{39 ,43}

A total of 16 studies^{4 - 6 ,12 ,34 ,37 ,42 ,44 - 52} provided diagnostic yield data (eTable 1). Diagnostic yield refers to the occurrence of a cardiac arrhythmia detected during the monitoring period in either all of the patients included in a study or only in the subgroup of patients who had symptoms during monitoring.

The diagnostic yield of 12-lead electrocardiography performed during symptoms ranged from 3% to 26%^{4 ,6} for any arrhythmias and 2%⁶ for clinically significant arrhythmias. The yield of the 24-hour Holter monitoring was 34%⁴⁴ for any arrhythmia and ranged between 3% and 24%^{44 - 45} for clinically significant arrhythmias. One study⁴⁶ used a 48-hour Holter monitor as a reference standard and had a diagnostic yield of 21% for any arrhythmia and 0% for clinically significant arrhythmias. The diagnostic yield for loop monitors ranged from 34%⁶ to 84%⁴⁷ for any arrhythmia and from 8%⁴⁸ to 36%⁴⁷ for clinically significant arrhythmias. A 2-week loop recorder had a greater yield than a 1-week recorder, however, using it for 3 weeks had minimal⁴⁹ or no additional yield.⁴⁷ The use of loop recorders with an automatic trigger function for asymptomatic arrhythmias had a slightly higher yield for clinically significant arrhythmias.³⁴ The yield for event recorders ranged from 30%⁵ to 60%⁵⁰ for any arrhythmia and from 17%⁵⁰ to 19%⁵ for a clinically significant arrhythmia (eTable 1). Among patients with typical symptoms during monitoring (eTable 2), the yield of intermittent event recorders was higher for any arrhythmia and for clinically significant arrhythmias.^{5 - 6 ,51} Based on 1 study,³⁷ the yield of implantable loop recorders was 73% for clinically significant arrhythmias during a mean (SD) monitoring period of 279 (228) days.

Only 7 studies provided data on diagnostic accuracy and only 2 of these^{5 - 6} were of high methodological quality (level 1). Many of the studies had small sample sizes and in some cases data from subgroups in single studies were relied on to evaluate the accuracy of clinical parameters. Therefore, caution should be taken in interpreting our results.

Our review focused on studies of patients presenting with palpitations rather than patients with conditions, such as hyperthyroidism, who had palpitations as part of their symptom complex. As such, our results are most relevant for patients without an obvious underlying medical problem or structural heart disease that might cause their palpitations. Also, the data does not provide information about the specific kind of arrhythmia experienced.

A significant proportion of the patients in the included studies did not have symptoms during the monitored period and only a few features have been evaluated in the 81 patients with symptoms during monitoring. Therefore, the majority of the LRs were calculated using all included patients rather than only those with symptoms during monitoring. Similarly, except in 3 studies (eTable 2),^{5 - 6 ,51} the yield data was extracted from studies in which some of the patients did not experience their symptoms during monitoring. Because some of these patients may actually have an arrhythmia, we could have underestimated the true yield of the diagnostic tests.

Finally, there are no published data evaluating combinations of features or the precision of the clinical examination. As a result, multiplying together individual LRs or applying them sequentially may substantially overestimate posttest probability.

The pretest probability of any cardiac arrhythmia in this patient based on a study⁴ that would have enrolled this patient is 40%. While the nonspecific description of symptoms as “heart racing” and the inability to tell the cadence of the rhythm are not helpful, the previous diagnosis of panic disorder (LR, 0.26) and the duration of less than 5 minutes (LR, 0.38) decreases the likelihood of any arrhythmia. The posttest probability of any arrhythmia based only on the patient's previous history of panic disorder would be 15%. The absence of a regular rapid-pounding sensation in the neck decreases the likelihood of atrioventricular node reentry tachycardia (LR, 0.07).

While many physicians would not pursue further testing initially because of the relatively low posttest probability, it is essential to recognize that patients with panic disorders may also have clinically significant arrhythmias^{22 ,42} ; and it may be prudent to perform long-term electrocardiographic monitoring to rule out clinically significant arrhythmias before attributing the symptoms to panic disorder. Structural cardiac assessment could be deferred unless the patient has persistent palpitations, develops more alarming symptoms such as syncope, or a clinically significant arrhythmia is identified on electrocardiographic monitoring.

When evaluating patients with palpitations, the presence of underlying medical conditions should be carefully considered. In the emergency department, primary cardiac diagnoses are the most common reason for palpitations (43%), but anxiety or panic disorders are also frequent (31%).⁴

Quiz Ref IDA known history of cardiac disease (LR, 2.03; 95% CI, 1.33-3.11) and palpitations affected by sleeping (LR, 2.29; 95% CI, 1.33-3.94) or while at work (LR, 2.17; 95% CI, 1.19-3.96) slightly increase the likelihood of a cardiac arrhythmia while palpitations lasting less than 5 minutes (LR, 0.38; 95% CI, 0.22-0.63) and a known history of panic disorder make the diagnosis less likely (LR, 0.26; 95% CI, 0.07-1.01). The presence of a regular rapid-pounding sensation in the neck as opposed to neck fullness in association with palpitations increases the likelihood that the patient has atrioventricular node reentry tachycardia (LR, 177; 95% CI, 25-1251), whereas its absence makes atrioventricular node reentry tachycardia less likely (LR, 0.07; 95% CI, 0.03-0.19). The presence of visible neck pulsations also increases the likelihood of atrioventricular node reentry tachycardia (LR, 2.68; 95% CI, 1.25-5.78).

Because of the limitations of the literature and the consequences of missing an important rhythm disturbance, no clinical examination features appear to be sufficiently accurate to exclude other clinically significant arrhythmias, especially in high-risk patients. Quiz Ref IDTherefore, when a clinically significant arrhythmia is suspected, further testing including evaluating cardiac structure with transthoracic echocardiography and attempting to establish symptom-rhythm correlation with prolonged electrocardiographic monitoring should be undertaken. The selection of the monitoring type depends on the frequency of the symptoms. If the symptoms occur daily, a Holter monitor may be of reasonable diagnostic yield, with emphasis on the importance of accurate diary recording. If symptoms occur more infrequently, an intermittent event recorder such as a loop monitor is a more appropriate test.