Investigating Chest Pain: An Evidence-Based Approach

Many of the established beliefs and procedures involved in the interpretation of chest pain do not always produce reliable results and, in fact, can hinder or even prevent proper diagnosis and therapy. A look at the latest research reveals the shortcomings of some of these techniques and points to a more informed approach.

By Mark A. Graber, MD, and Stephen J. Scheckel, MD

Drs. Graber and Scheckel are associate professors of emergency medicine at the University of Iowa College of Medicine in Iowa City, where Dr. Graber is also associate professor of family medicine.

The practice of medicine entails recognizing in a patient's symptoms and physical and historical findings the patterns that allow us to narrow the differential diagnosis and reach a proper conclusion. The manifestations of illness, however, do not always fit textbook descriptions. Many of the generalizations that we were taught about the clinical presentation of illness do not hold true. By bridging that gap, the published results of primary research can serve as a vital diagnostic resource. In this discussion, we focus on an evidence-based approach to the symptoms and laboratory findings associated with various causes of chest pain-cardiac disease, pulmonary embolism, pericarditis, and panic attack.

back to top

Cardiac Disease

To understand how this approach can help you, let us consider the hypothetical case of a 45-year-old man who has atypical, stabbing chest pain that radiates only to the right arm. He does not smoke or have diabetes, hypertension, or a family history of cardiac disease. On physical examination, the man appears anxious, and his oxygen saturation level measures 95%. The rest of the vital signs are normal and an electrocardiogram (ECG) shows only nonspecific ST-T changes. The initial creatine phosphokinase (CPK) and troponin I levels measured in the emergency department are normal.

Because the history and physical examination generate a lot of data, it is important to sort out what portions are relevant to the clinical question. In the case of a patient with possible cardiac disease, two questions have to be asked: Do these symptoms and findings point to cardiac disease or to another serious cause of chest pain?  And should this patient be admitted to the hospital?

Unfortunately, as Green and Yates have found, many physicians misjudge the importance of risk factors when making decisions about whether or not a patient with chest pain has cardiac disease (Annals of Emergency Medicine, vol. 25, p. 451, 1995). Without a complete understanding of how the ECG results, history, and enzyme levels contribute clinically to the diagnosis of cardiac disease in patients with chest pain, a rational decision about the necessity of hospital admission cannot be made.

Risk factors. It is second nature to ask patients about aspects of their history and lifestyle that could indicate an increased risk for cardiac disease, but will such information help us decide whether they should be admitted? Unfortunately, the answer is no. The presence or absence of risk factors does not change the likelihood of cardiac disease enough to be useful in discerning who might have the disease. Such historical and risk factors as diabetes, smoking, hypertension, increased serum cholesterol levels, and family history of cardiac disease are associated with a likelihood ratio (LR) of only 2-that is, risk factors are only two times as likely to be present in patients with cardiac disease as in those without (see sidebar, below). That information therefore does not help us differentiate between those patients with cardiac disease and those without.

For example, in a 40-year-old man who does not smoke and has anginal chest pain, the risk of significant coronary artery disease (CAD) is about 6%. If this same patient is a smoker, the risk increases, perhaps as high as 13%. However, since we cannot afford to miss 6% of patients who have significant CAD, we should ignore the absence of a smoking history and admit the patient anyway. This same approach is necessary for patients who have other risk factors as well.

Even though the presence of risk factors may change the prior probability of cardiac disease and are useful in predicting the risk among populations, their absence does not reduce the risk of cardiac disease enough in individual patients to allow us to send them home. Thus, these factors should not be relied upon when making a decision about whether or not a patient with chest pain should be admitted to rule out cardiac disease (see table, below).

History and physical examination. The second element to consider in the diagnosis is the nature of a patient's chest pain. The classic description of cardiac pain is substernal pressure with radiation to the left shoulder and arm. However, many patients with cardiac disease do not present so typically.

As Canto and colleagues have found, up to 33% of patients, especially the elderly and diabetic patients, who have an acute myocardial infarction (MI) do not have pain (JAMA, vol. 283, p. 3223, 2000). Others will present with atypical pain. Far from reassuring, an atypical pattern of pain radiation, as Panju has found, actually increases the likelihood that a patient has cardiac disease: left shoulder radiation is associated with an LR of 2.3; right shoulder radiation, 2.9; and radiation to both shoulders, 7.1 (JAMA, vol. 280, p. 1256, 1998).

Clearly, fewer patients have radiation to the right shoulder or both shoulders, but the presence of such pain is a cause for concern. Likewise, the nature of chest pain provides no definitive proof of its source, cardiac or otherwise. Pain described as a pressurelike or squeezing sensation, for example, is associated with an LR of less than 2 and is therefore not very useful in differentiating pain of cardiac origin from that of other causes. This LR does not mean that most patients with cardiac disease do not have substernal pressure or squeezing pain but, rather, that this description could also apply to other causes of chest pain, such as an esophageal disorder.

On the other hand, pain that is described as pleuritic, sharp, or stabbing or is reproducible on palpation of the chest wall reduces the likelihood that a patient's chest pain is of cardiac origin. However, even the presence of this type of chest pain-which is associated with an LR of 0.2 to 0.3-does not absolutely rule out the possibility of cardiac disease. To quote Rosen and Barkin, "No astute clinician is reassured by chest wall tenderness that reproduces a patient's pain, because 15% of patients with acute MI will have chest wall tenderness on palpation that reproduces their pain" (Emergency Medicine Concepts and Clinical Practice, St. Louis, Mosby, 1999).

To avoid the critical mistake of misdiagnosis, we should be prudent and not jump to the conclusion that a patient with sharp or reproducible chest pain does not have cardiac disease.

Several specific findings of the physical examination increase the probability that a patient has cardiac disease, among which are hypotension (LR, 3.1), an S3 sound (3.2), and rales (2.1). Diaphoresis also increases the risk that a patient's pain is of cardiac origin.

Electrocardiogram. The third aspect to consider in the diagnosis of chest pain is the ECG. Like a patient's medical history and risk profile, the ECG is not a perfect indicator of cardiac disease. At least two studies have shown, for example, that only 50% of patients with a proven acute myocardial infarction have an initial ECG indicating the disorder, although nondiagnostic ECG changes may be present.

In fact, up to 76% of patients with unstable angina will have an initial ECG that is normal, nonspecific, or unchanged from previous ECGs. Therefore, if the initial ECG is used as a criterion for patient admission, a significant percentage of patients who have an acute coronary event-unstable angina or acute MI-will be sent home.

That said, a normal ECG indicates a significantly reduced probability that a patient has an acute MI. Associated with an LR of 0.1 to 0.3, a normal ECG is only 10% to 30% as likely to be obtained from a patient who has an acute MI as is an abnormal ECG. However, since we can't afford to miss detecting any patients who have an acute MI, even a normal ECG is not reliable enough to enable us to determine whom to admit and whom to send home.

Serum chemical markers. The final clues to consider are the laboratory test results. In the case of the patient described earlier, the CPK and troponin I levels are normal. Unfortunately, an increase in the level of serum chemical markers is not a reliable enough indication of cardiac risk during the first 6 to 12 hours to be helpful in determining whether a patient should be admitted to the hospital. Often, relying on such a sign will lead a physician to send home a patient who has unstable angina.

Another limitation associated with relying on serum chemical marker levels is that physicians frequently see patients during a time when the serum markers are still expected to be absent. For example, at hours 4 to 8 and 8 to 12 the CPK level is more sensitive (84% and 94%, respectively) in indicating an acute MI than is the troponin level (74% and 88%, respectively). After 12 hours, the troponin level is essentially 100% sensitive, whereas the CPK level becomes less sensitive. Consequently, if these markers were the determining factor for admission, up to 12% of patients with acute MI would be sent home within the 8- to 12-hour period. Serum markers should be drawn as a baseline but should not enter into the decision to admit a patient with chest pain to the hospital.

How, then, are we to decide whom to admit? First, the probability of cardiac disease must be determined (see table below). If the probability is high enough to be of concern, the patient should be admitted. Any patient whose ECG shows new changes should also be admitted, as should any patient with a history of unstable angina. Do we overadmit patients to rule out MI? Absolutely. Only about 15% of patients admitted to the hospital to undergo testing to rule out MI will have had one. To maintain a high level of diagnostic sensitivity-that is, to admit all patients who have cardiac disease-we sacrifice specificity. However, in the current medicolegal climate, this is the expectation.

back to top

Pulmonary Embolism

The second case we will examine is that of a 23-year-old woman presenting to the ED with sharp chest pain that began earlier in the day but has not resolved. She reports that she is taking oral contraceptives and that she had had an episode of syncope and anxiety at approximately the same time the chest pain developed. On examination, the patient has mild tachycardia (110 bpm) and an oxygen saturation level of 95%. Results of the heart and lung examinations are normal, and no leg swelling or tenderness is present. A chest film is normal, as is a spiral CT scan for pulmonary embolism (PE). A latex agglutination D-dimer assay indicates no evidence of PE.

Signs and symptoms. Pulmonary embolism is another classic cause of chest pain that physicians cannot afford to miss. Unfortunately, diagnosing PE requires a high degree of clinical suspicion, because not all patients have the usual signs and symptoms. According to some studies, only 78% of patients will have dyspnea. Among other signs and symptoms, pleuritic chest pain is found in 59%, cough in 43%, tachycardia in 30%, tachypnea in 73% and syncope in 13% of patients. On close examination of the history and physical findings in this case, we will find a number of factors that may erroneously lead a physician to discharge the patient.

Oxygen saturation, PO2, and A/a gradient. One of the most misleading clues in this case is the oxygen saturation level of 95%. A normal oxygen saturation level and A/a gradient-a measure of blood shunting in the lung caused by pulmonary embolism-do not rule out pulmonary embolism. Stein and associates found that among patients without underlying pulmonary disease, approximately 12% with PE will have a PO2 level greater than 80 mm Hg (Chest, vol. 109, p. 78, 1996). Other studies suggest that 17% of those with angiographically proven PE will have a PO2 greater than 80 mm Hg and 5% will have a PO2 of greater than 100 mm Hg.

According to Jones and colleagues, the A/a gradient is no more accurate than PO2 in detecting pulmonary embolism (American Journal of Emergency Medicine, vol. 16, p. 333, 1998). A study by Meingnan and colleagues reinforces the belief that not all patients with PE will be hypoxic or have tachycardia or tachypnea. They found that up to 50% of patients with deep vein thrombosis (DVT) have a silent, or asymptomatic, PE (Archives of Internal Medicine, vol. 160, p. 159, 2000). The diagnosis of PE should therefore not be excluded because some of the classic findings are absent.

Spiral computed tomography. Because it is quick and easy to perform, spiral CT has become an increasingly popular diagnostic procedure for suspected PE. However, spiral CT is not sensitive enough to reliably rule out the disorder.

In two recent studies, the sensitivity of spiral CT was less than 70% for detecting segmental or larger pulmonary emboli. The accuracy of the technique is also limited by significant interobserver variability-one radiologist may read a particular study as positive, whereas another will read the same study as negative.

Pointing to these and other reasons, Rathburn and colleagues, in a systematic review of spiral CT in the detection of PE, concluded that the "use of helical CT for the diagnosis of pulmonary embolism has not been adequately evaluated. The safety of withholding anticoagulant treatment in a patient with suspected pulmonary embolism and negative results on helical CT is uncertain. On the basis of the current best evidence, clinicians should not use negative results on helical CT as the diagnostic end point for excluding pulmonary embolism" (Annals of Internal Medicine, vol. 132, p. 227, 2000).

What role spiral CT will play in the diagnosis of PE remains to be determined. If the result of a scan of a patient at medium or high risk is positive, it is likely a true-positive result. If the result is negative, further testing with V/Q scanning or angiography will be necessary.

D-dimer assay. The last consideration in this case is the interpretation of the D-dimer assay. D-dimer is a product of ongoing thrombolysis, which should accompany thrombosis. Thus, if thrombolysis is not present, meaning the D-dimer level is low, the presence of a clot is unlikely. Unfortunately, the sensitivity of the D-dimer as a marker is dependent on the test used to detect it. According to Farrell and colleagues, the false-negative rate of latex agglutination tests is too high (up to 35%) for that test to be useful in ruling out PE (Annals of Emergency Medicine, vol. 35, p. 121, 2000).

The enzyme-linked immunosorbent assay (ELISA) is associated with a false-negative rate of only 0.7% to 3.0%, equivalent to that of V/Q scanning in patients who have a high probability of cardiac disease. As the sole determinant of the presence of PE or DVT, however, ELISA is not yet the diagnostic standard.

Another problem with the D-dimer assay is its lack of specificity. Many patients with an elevated D-dimer level do not have DVT or PE but instead may have suffered a recent contusion, or the elevated D-dimer may just reflect the normal balance between ongoing thrombosis and thrombolysis that naturally occurs.

The test is particularly nonspecific in the evaluation of patients with cancer, one group that is especially prone to DVT and PE. The D-dimer level will be elevated in most patients with cancer and therefore will have little meaning in the diagnosis of PE. Consequently, although negative results of an ELISA D-dimer assay can be useful in ruling out thrombosis, an elevated D-dimer level is of no help in diagnosing the presence of PE or DVT.

A final factor that is frequently misinterpreted is the presence of a vena caval filter. A common error is the presumption that the presence of a vena caval filter rules out the possibility of PE. In one of the few studies to objectively address the topic, Decousus and colleagues found that patients who had vena caval filters in place did not have a reduced risk of subsequent PE: the number of emboli and the mortality rate at two years were the same among both the control subjects and the patients who had a vena caval filter in place. In fact, the vena caval filter group had almost twice the number of episodes of DVT (21% versus 12%) (New England Journal of Medicine, vol. 338, p. 409, 1998). The presence of a vena caval filter, therefore, should have no bearing on a physician's decision to investigate a possible presence of PE.

The best approach is still the classic one: perform a clinical examination to determine the source of DVT-a swollen leg, for example-and follow through with the appropriate imaging studies, such as Doppler ultrasonography. Performing Doppler ultrasonography on an asymptomatic leg is simply a waste of resources. If the results are negative, a V/Q scan should then be performed. If the results of that test are negative, angiography would be the next step for patients at high and intermediate risk (see table, below). Remember, to reliably rule out DVT in a symptomatic leg, the results from two venous Doppler examinations performed one week apart must be negative.

back to top

Pericarditis

Pericarditis is another potentially confusing source of chest pain. The classic symptoms of pericarditis include sharp, stabbing, or pleuritic pain that is often most severe when the patient is supine and is relieved when he or she sits up and leans forward. These symptoms may not always be present, however. Because the pain of pericarditis can be mistaken for ischemic cardiac pain, a thorough history, physical examination, and review of the ECG are particularly important in establishing the correct diagnosis.

Although the physical examination is important, a friction rub is not always present or is only transiently audible, and its absence does not rule out the diagnosis. The ST- and T-wave changes seen on the ECG can be confused with changes suggestive of MI or ischemia. The diagnosis of acute pericarditis is helped by the presence of PR-segment depression, which is very specific for the disorder, and the diffuse nature of the ST- and T-wave elevations without reciprocal changes. Finally, the ECG may not always be diagnostic. Changes do not appear on approximately 10% of ECGs of patients who have pericarditis.

back to top

Panic Attack

Finally, keep in mind that panic attacks can cause chest pain. Because some of the symptoms of a panic attack are similar to those of MI, differentiating patients who have ischemic heart disease from those undergoing panic attack can be difficult. Both panic attacks and cardiac disease can be marked by shortness of breath or a smothering sensation, tachycardia, sweating, chest pain, and a fear of dying. In up to 30% of patients who have noncardiac chest pain, the diagnosis will be panic attack. The converse is also true. Among patients with documented supraventricular paroxysmal tachycardia (PSVT), 67% meet the criteria for a panic attack; in one study, PSVT was correctly diagnosed at the first patient visit only 45% of the time.

The presence of symptoms that are not shared between panic attack and cardiac disease can be a clue to differentiating the two disorders. For example, patients undergoing a panic attack may report psychiatric symptoms, such as a history of depression or suicidal ideation, or a fear of losing control of themselves. They might also describe a feeling of depersonalization, the sensation that one's body is unreal, or derealization, the feeling that reality is altered. Of course, these features may also be absent but are certainly not characteristic of cardiac disease.

The possibility of a panic attack should therefore not be ruled out when patients present with symptoms of MI or PSVT. Likewise, physicians should not be quick to dismiss chest pain symptoms as a panic attack. It is still incumbent upon us physicians to ensure that a patient's chest pain is not of cardiac origin.

Suggested Reading Black ER, et al.: Diagnostic Strategies for Common Medical Problems, 2nd edition. Philadelphia, American College of Physicians, 1999. Canto JG, et al.: Prevalence, clinical characteristics, and mortality among patients with myocardial infarction presenting without chest pain. JAMA 283:3223, 2000. Decousus H, et al.: A clinical trial of vena caval filters in the prevention of pulmonary embolism in patients with proximal deep-vein thrombosis. N Engl J Med 338:409, 1998. Farrell S et al.: A negative SimpliRED D-dimer assay result does not exclude the diagnosis of deep vein thrombosis or pulmonary embolus in emergency department patients. Ann Emerg Med 35:121, 2000. Green LA and Yates JF: Influence of pseudodiagnostic information on the evaluation of ischemic heart disease. Ann Emerg Med 25:451, 1995. Jones JS, et al.: Use of the alveolar-arterial oxygen gradient in the assessment of acute pulmonary embolism. Am J Emerg Med 16:333, 1998. Meignan M, et al.: Systematic lung scans reveal a high frequency of silent pulmonary embolism in patients with proximal deep venous thrombosis. Arch Intern Med 160:159, 2000. Kearon C, et al.: The role of venous ultrasonography in the diagnosis of suspected deep venous thrombosis and pulmonary embolism. Ann Intern Med 129:1044, 1998. Panju A, et al.: Is this patient having a myocardial infarction? JAMA 280:1256, 1998. Pozen MW, et al.: A predictive instrument to improve coronary-care-unit admission practices in acute ischemic heart disease. A prospective multicenter clinical trial. N Engl J Med 310:1273, 1984. White RH, et al.: A population-based study of the effectiveness of inferior vena cava filter use among patients with venous thromboembolism. Arch Intern Med 160:2033, 2000. Zalenski RJ, et al.: Assessing the diagnostic value of an ECG containing leads V4R, V8, and V9: The 15-lead ECG. Ann Emerg Med 22:786, 1993.

Printable version of this article

E-Mail this Article