Pulmonary Embolism:
Diagnostic Evaluation

Jason Fleming, M.D.; Jeff Hersh, Ph.D., M.D., FAAP, FACP

Abstract
Pulmonary embolism (PE) is a fairly common disease, with very significant morbidity and mortality. However, since the presenting signs and symptoms of PE are nonspecific, it is also a very under-diagnosed disease. We review many of the blood tests, imaging modalities, clinical scoring systems and other studies (including electrocardiograms) that can be used in the evaluation of this difficult to diagnose entity. Stemming from this presentation, we discuss an algorithm that can be used to guide a systematic workup in the evaluation of a patient with a possible PE. This algorithm is a modification of existing recommendations and incorporates recent outcome results on the use of spiral-computed tomography. The algorithm uses a clinical scoring system to risk stratify patients, and then guides the clinician through a systematic evaluation of their patient utilizing widely available laboratory and radiological imaging studies.

The Problem
Pulmonary embolism (PE) occurs when some substance (usually thrombus, fat, air, or amniotic fluid) lodges in the pulmonary vasculature resulting in obstruction of the blood flow to that portion of the lung. PE is a relatively common disease, with an annual incidence of 23 cases per 100,000 persons in the United States.

Patients suffering from PE often present with nonspecific signs and symptoms (Table 1) that may mimic a long list of benign and life-threatening conditions. The differential diagnosis of PE is therefore an understandably long list (see Table 2).

The presenting signs and symptoms of acute PE are determined by the extent of pulmonary vascular occlusion, as well as the patient's baseline cardiopulmonary status. Dyspnea is the most common symptom of PE, and tachypnea is the most common sign. If neither dyspnea, pleuritic chestpain, nor tachypnea are present, PE is unlikely in the absence of risk factors (see Table 3). However, the entire classic triad of dyspnea, pleuritic chest pain and hemoptysis are present in less than 25 percent of patients with PE.

All known risk factors for deep venous thrombosis (DVT) and PE stem from Virchow's Triad of venous stasis, hypercoagulability, and vessel wall injury or abnormality. However, it is important to recognize that 15 percent of patients with PE have no demonstrable risk factors at the time of presentation. The single most important risk factor is a history of prior DVT or PE, which increases the incidence of PE up to 30-fold. A family history of early PE may suggest an inherited disorder of coagulation or thrombolysis. All three trimesters of pregnancy place the patient at increased risk for PE, and it is important to remember that this risk extends for a 3-month period postpartum.

The consequences of missing the diagnosis of PE can be severe, and so the treating physician must include PE in the differential diagnosis of any patient presenting with acute shortness of breath, chest pain, syncope, or unexplained tachycardia or hypoxemia. PE is the third most common cause of death in the United States. Approximately 50 percent of patients with PE are undiagnosed. Ten percent of patients with PE die within the first hour. The mortality of undiagnosed PE is estimated at 30 percent.

Diagnosing patients with pulmonary embolism (PE) remains one of the greatest challenges to the treating physician. Timely diagnostic testing is necessary to make an appropriate diagnosis and initiate treatment; however, PE is difficult to diagnose because a single definitive test to exclude PE does not exist.

The Solution
Baye's theorem underlies all that we do as clinicians on a day-to-day basis. Simply put, Baye's theorem states that the predictive value of any given test depends not only on the sensitivity and specificity of the test, but also on the prevalence of the disease in the population tested. A simple example will illustrate this point.

Suppose we have a test that is 99 percent sensitive and 99 percent specific for a particular disease. Let's apply this test to a sample population where the incidence of disease is about 1 percent, with 100 cases per 10,100 people (see Table 4). If 100 people have the disease, this 99 percent sensitive test is positive in 99 and falsely negative in 1. If 10,000 people do NOT have the disease, this 99 percent specific test is negative in 9,900 but falsely positive in 100. The positive predictive value of this test on this population (true positives divided by all positives) is the probability that a patient with a positive test has the disease and in this example is less than 50 percent (that is 99/(100+99) < 50%). This remarkably sensitive (99 percent) and specific (99 percent) test is not a definitive test for this population.

As clinicians, we take histories from and perform physical exams on our patients so that we can gain a pretest probability of what is wrong with them. In essence, we are creating a subset of patients to order specific tests on, so that the results of these tests will help us diagnose and subsequently treat our patients.

In 1990 the PIOPED study applied this concept. In this study a random 933 of 1493 patients with suspicion of PE were studied prospectively. The treating clinician was asked to assign a pretest probability (low 0-19%, intermediate 20-79%, or high 80-100%) that their patient had a PE based on their clinical assessment. A ventilation-perfusion (V/Q) scan was then obtained in 931 of these 933 patients and the results interpreted within the context of the assigned pretest probability. Pulmonary angiography was performed on 755 of these 931 patients, and the results of the angiography were taken as the gold standard for the final diagnosis or exclusion of PE.

There are several weaknesses of the PIOPED approach. The PIOPED investigators did not offer a clinical scoring system to bring the phrase "pretest probability" out of the subjective realm of the individual physician's clinical judgment. Furthermore, 28 percent of patients with a low clinical probability of PE were subsequently diagnosed with PE. The variability in the interpretation of a V/Q scan is another important consideration (see Table 5), and also limited the conclusions of this landmark study.

More recent research has addressed these shortcomings, and incorporated new diagnostic tests into clinical algorithms. These will be discussed below.

Evaluation and Diagnosis

There are many tests available that can aid in the diagnostic workup of PE. Many of these tests will help to rule in or rule out other conditions that could be the cause of the patient's symptoms. Others will help in the diagnosis of PE.

Evaluation of the patient with possible PE begins with a careful history and exploration of potential risk factors, as well as a focused physical exam. This initial step is necessary to gather information about the patient's complaints, to begin a determination of the likelihood of PE and to estimate a pretest probability before ordering any further evaluations.

The workup proceeds to include an electrocardiogram, complete blood count (CBC), electrolytes, cardiac enzymes and chest radiography, as well as consideration of arterial blood gases, liver function tests and pancreatic tests. These tests may uncover an alternative diagnosis, or they may provide additional evidence to support the diagnosis of PE. Furthermore, the chest radiograph is necessary to correctly interpret a V/Q scan. If these screening tests do not uncover another diagnosis (e.g. myocardial infarction, pneumothorax, pneumonia, etc.) that explains the patient's signs and symptoms, a pretest probability for PE will need to be determined. The subsequent selection of appropriate diagnostic modalities will depend on the clinical pretest probability of PE.

Before we discuss this approach in more detail, it will be useful to review specific information about various tests and how they relate to the diagnosis of PE.

Blood Tests
The measurement of D-dimer (a degradation product after plasmin lysis of cross-linked fib-rin) can be an extremely useful screening test in the workup of suspected PE. The quantitative ELISA assay is the D-dimer test we will discuss here, since it has proved to be the most reliable evaluation of D-dimer. Available rapid quantitative D-dimer assays have a sensitivity and negative predictive value of 95-99 percent. Therefore, when coupled with a low pretest probability, a normal D-dimer allows the clinician to rule out PE early in the workup and focus on other competing diagnoses. Unfortunately, D-dimer assays are not very specific. In fact, the specificity has been reported as low as 9 percent. Therefore, if the result is positive further testing is required to establish the diagnosis of PE.

Arterial blood gas (ABG) analysis can confirm hypoxemia, which is one of the possible clinical indicators for the evaluation of the pretest probability of PE. It is important to note that PaO2 can be greater than 80 mmHg in as many as 12 percent of patients with PE. However, 86 percent of patients with proven PE were found to have an alveolar-arterial oxygen gradient of greater than 20 mmHg. Of course many other disease processes can give abnormal ABG values as well.

Other blood tests such as CBC, serum chemistries, liver and pancreatic function tests are not particularly helpful in diagnosing PE, however they may lend evidence for other diagnoses (see Table 2) that may have signs and symptoms that overlap those of PE.

Electrocardiogram
The electrocardiogram (EKG) is not a specific test for PE. It may be useful in suggesting alternative diagnoses in the patient with signs and symptoms suggestive of PE such as cardiac ischemia, pericarditis, or rhythm disturbances. The EKG is abnormal in the great majority of patients presenting with PE.

According to the Urokinase-Pulmonary Embolism Trial(UPET), 87 percent of patients with PE have an abnormal EKG. The most frequent findings were sinus tachycardia or non-specific T-wave inversions, each occurring in as many as 40 percent of patients with some patients having both. Other EKG findings include changes reflecting right sided heart strain such as peaked P waves in lead II suggesting right atrial enlargement, right axis deviation or right bundle branch block. Atrial fibrillation or an S₁Q₃T₃ (S wave in lead I and Q wave with T wave inversion in lead III) pattern can also occur. The 'characteristic' S₁Q₃T₃ abnormality occurs in only 11 percent of patients with PE.

Chest Radiography

The chest x-ray is helpful for two important reasons: it is necessary to interpret a subsequent ventilation-perfusion scan, and it could suggest other disease processes that might be causing a patient's symptoms such as pneumothorax, pneumonia, or lung mass. The CXR is rarely normal in PE. In the PIOPED study, the CXR was abnormal 88 percent of the time, and this finding is fairly consistent with most other studies, although the percentage of a given abnormality varies. The timing of the CXR might be a factor in the particular abnormality noted, and this might explain some of these variations. For example, it has been shown that infiltrates usually occur several days after the PE. Results of CXR findings from several studies are presented in Table 6. It should be noted that Westermark's sign and Hampton's hump are very uncommon findings on the CXR.

Ventilation/Perfusion Scans


Data from the Prospective Investigation of Pulmonary Embolism Diagnosis (PIOPED) study shows that almost all patients with PE had abnormal ventilation/perfusion (V/Q) scans of high, intermediate, or low probability, but so did most without PE. If the V/Q scan was interpreted as high probability, it carried a sensitivity of 41 and a specificity of 97 percent. If all scans interpreted as high or intermediate probabilities are included, the sensitivity is 82 percent and the specificity drops to 52 percent. If all non-normal V/Q scans (high, intermediate, and low probability) are included, the sensitivity goes to 98 percent but the specificity drops further to 10 percent. Table 7 summarizes results of the V/Q scans done in the PIOPED study.

Spiral Volumetric Computed Tomography
Contrast-enhanced thin collimation spiral computed tomography of the thorax with 1 to 2 mm image reconstruction (also known as CT angiography or CTA) has been recognized as a potential test in the diagnosis of PE since 1992, where the sensitivity and specificity was initially reported as 99 percent and 96 percent, respectively. A clinical policy statement from the American College of Emergency Physicians has reviewed the more recent literature, and when subsegmental PE are included, the overall sensitivity of CTA drops to 77 percent, with 89 percent specificity. However, outcome data from CTA studies are much more encouraging, with reported incidence of only 1 percent of subsequent PE in patients with a negative CTA in one trial and a three month incidence of fatal PE of only 0.3 percent reported in a retrospective study of 1510 patients. These outcomes compare favorably with the 0 percent subsequent PE for a normal V/Q and 3 percent subsequent PE for a low probability V/Q reported by Goodman et al. The addition of venous studies of the legs improves the overall sensitivity, because diagnosis of venous thrombosis in a patient with signs and symptoms of PE strongly supports the diagnosis. Since IVcontrast is utilized in CTA, patients with an allergy to IV contrast dye or those with impaired renal function may not be candidates for this test.

The outcome studies noted above indicate that a negative CTA can be considered essentially equivalent to a normal V/Q scan in the evaluation of suspected PE. An abnormal CXR limits the interpretability of a V/Q scan, and so CTA may be the preferred test in these patients. Furthermore, CTA is more likely than V/Q scan to identify an alternative diagnosis.

Pretest Probability and Predictive Value
The first step in the evaluation of any patient presenting with possible pulmonary embolism is to determine the pretest probability of a PE. Ideally, this would be a quantifiable score using objective criteria. As per the discussion of Baye's theorem above, the predictive value of a test depends not only on the sensitivity and specificity of that test to identify a certain disease, but also on the prevalence of the disease in the tested population. Therefore, we want to identify patients who are likely to have the disease on a clinical basis to help in the interpretation of subsequent diagnostic testing.

Recently, both Wells, et al. and Wicki, et al. have proposed scoring methods to asses clinical probability of PE at the bedside.

The Wells system is summarized in Table 8. This system is based only on factors that can be obtained from the history and physical examination; no laboratory results are needed. As would be expected, it gives weight to risk factors for PE such as malignancy, immobilization or a history of prior DVT or PE. It also gives weight to typical findings from physical exam, such as evidence of DVT or tachycardia, and to findings from the patient history of hemoptysis. The weakness of this system is that it also weights the subjective factor of whether the treating physician feels that PE is the likely diagnosis, reminiscent of the PIOPED study.

The Wicki criteria (outlined in Table 9) overcome the weakness in the Wells criteria of a subjective rating from the treating physician, making it a more objective scoring system overall. This system also weights factors of tachycardia, recent surgery (rather than just immobilization) and a history of DVT or PE. However, CXR and blood gas results are required for this system, and there will still be some subjectivity in the reading of the CXR.

Despite the subjective criteria in the Wells scoring system, its ease of use and immediate results (since it is based only on history and physical examination) may make it the preferred system at the present time. A diagnostic algorithm has been proposed by Wells et al. and further developed by Wolfe and Hartsell and we present a slightly modified version in the discussion below.

Diagnostic Algorithm
Wells et al. as well as Wolfe and Hartsell have outlined elegant and feasible diagnostic algorithms to guide the workup of a patient with suspected PE. The strength of these approaches is that they articulate which patients require diagnostic imaging, and have the potential to save valuable healthcare resources. We present a modified flow algorithm in Table 10 that summarizes these concepts.

The first step in the evaluation of a patient with suspected PE is to take a careful history and examine the patient. The next step is to determine the likelihood that the patient has a DVT. A scoring system to help quantify the likelihood of DVT has been developed by Anderson et al. (Table 11). If there is a high clinical likelihood of DVT, an immediate evaluation of the lower extremities by duplex ultrasound (US) or CT of the lower extremities/pelvis should be obtained, and if positive treatment should be started. Otherwise, the pretest probability of PE should be determined using a clinical scoring system as described above.

Low Pretest Probability:
If the clinical pretest probability is low, the next step is to obtain a plasma D-dimer. If this is negative, PE has been excluded. If the plasma D-dimer is positive, evaluation of the chest radiograph is necessary to determine the potential utili- ty of a ventilation/perfusion (V/Q) scan. If the chest x-ray is normal a V/Q scan should be obtained. If the CXR is abnormal, the utility of a V/Q scan is diminished (although is still of value and may still be utilized), and so consideration of CTA of the chest should be considered. If the V/Q scan is interpreted as normal or the CTA is negative, PE has been excluded. If the V/Q scan is interpreted as high probability or if the CTA is positive for PE, treatment for PE should be initiated. If the V/Q scan has been interpreted as low or intermediate probability or if the CTA is equivocal, duplex US or lower extremity/pelvic CT should be done. Positive evidence of DVT will require treatment. Negative evaluation of the lower extremities should be followed up with repeat evaluation in 5-7 days.

Moderate Pretest Probability:
If the clinical pretest probability is moderate, the D-dimer should still be obtained. A negative D-dimer result does not negate the need for further imaging studies in a patient who has a moderate pretest probability. As in the case of a low pretest probability, the chest x-ray should be used to help select between V/Q scan and CTA. A normal V/Q scan or negative CTA excludes PE. If the V/Q scan is interpreted as high probability or the CTA is positive for PE, treatment should be initiated. If the V/Q scan shows intermediate or low probability, the D-dimer test once again becomes helpful. If the D-dimer is negative, PE has been excluded. If the D- dimer is positive, duplex US of the legs or lower extremity/pelvic CT should be obtained. If these tests are positive, the diagnosis of PE should be made and treatment for PE should be initiated. If they are negative, they will need to be repeated in 5-7 days. An indeterminate CTA or a patient with very poor cardio-pulmonary status will need further evaluation. Evaluation of the lower extremities with duplex US of the legs or lower extremity/pelvic CT should be obtained, and if positive treatment should be started. If these tests are negative, a pulmonary angiogram will need to be considered, and treatment based on this result.

High Pretest Probability:
If the clinical pretest probability is high, a chest x-ray should be used to help select between V/Q and CTA. A normal V/Q scan or negative CTA excludes PE. If the V/Q scan is interpreted as high probability or the CTA is positive for PE, treatment should be initiated. If the V/Q scan shows intermediate or low probability, or CTA is equivocal, duplex US of the legs or lower extremity/pelvic CT should be obtained. If these tests are positive, the diagnosis of PE should be made and treatment for PE should be initiated. If they are negative, an angiogram will need to be obtained to rule out PE. The results of a pulmonary angiogram are felt to be definitive.

Unstable Patient: Transesophageal echocardiography (TEE) is useful in the diagnosis of massive PE and can be performed at the bedside, and should be considered in hemodynamically unstable patients. Findings suggestive of massive PE include right ventricular dilatation, right ventricular dysfunction, and possibly visualization of the thrombus. If PE is felt to be the likely diagnosis in an unstable patient, treatment may need to be begun empirically. Consideration of pulmonary angiogram and possible embolectomy may also be considered.

Treatment
All patients with chest pain and any patient suspected of having a PE should have cardiac monitoring, intravenous access and supplemental oxygen on arrival to the emergency department. Endotrachial intubation may be required if hypoxemia cannot be corrected with supplemental oxygen, or if ventilation is compromised. Shock in the absence of pulmonary edema may require aggressive fluid resuscitation with crystalloid fluids and then dopamine if necessary. All patients with known PE should be admitted for monitoring and anticoagulation or thrombolytic therapy.3

Anticoagulation is the mainstay of therapy for PE. Rapid anticoagulation is essential. Initial anticoagulation is usually done with intravenous heparin or low molecular weight heparin (LMWH). Heparin does not dissolve thrombus; it simply interferes with the coagulation cascade to inhibit the action of thrombosis and slow thrombus formation. This allows the natural thrombolytic mechanisms of the body to reduce the extent of existing thrombosis. Heparin acts by activating antithrombin III and so a deficiency of antithrombin III renders heparin ineffective as an anticoagulant. In these cases alternate anticoagulants such as hirudins must be employed. Heparin is felt to be safe in pregnancy and so is the preferred anticoagulation method in pregnant patients.

Long-term anticoagulation is usually maintained with warfarin. Warfarin acts by interfering with the action of vitamin K dependent clotting factor II, VII, IX and X, as well as the action of anticoagulant factors protein C and protein S. Warfarin should not be started in patients with PE without prior heparin therapy because warfarin may reduce the levels of anticoagulants protein C and protein S before it reduces the levels of procoagulant factors, producing a transient hypercoagulable state in the first several days of therapy. Warfarin anticoagulation is considered effective for PE when the international normalized ratio(INR) is above 2.5.3 Warfarin is teratogenic and is con- traindicated in pregnancy. It is secreted in breast milk in minute quantities only and is considered safe for nursing mothers.

Hemorrhagic complications may occur with any of the anticoagulants discussed above. If this occurs, appropriate reversal of the anticoagulation utilized should be considered. In these cases, or in cases where anticoagulation is contraindicated, an intravenous filter placed in the vena cava (sometimes referred to as an umbrella) should be considered.

Thrombocytopenia is the other major complication of heparin therapy. The more common type of thrombocytopenia from IV heparin therapy occurs in up to 25 percent of patients and is a benign transient decrease in platelet count usually not below 100,000/microL. Heparin-induced throm- bocytopenia occurs in 1-5 percent of patients treated with IV heparin, and may be life threatening. It is due to an antibody reaction to platelets. It usually occurs after 5-10 days of IV heparin therapy, but may occur sooner in patients who have been treated with IV heparin in the past. Very low platelet levels, under 50,000/microL, along with arterial thrombosis may occur. Treatment is discontinuation of the IV heparin and use of a different anticoagulation agent.

Most patients with an initial DVT or PE should be anticoagulated for six months or longer. A subset of patients who have a reversible cause of venous thrombosis may be candidates for a three month course of oral anticoagulation. Long-term anticoagulation is indicated for patients with a second episode of thrombosis or with a known irreversible risk factor.

Fibrinolytic agents are the treatment of choice in patients with hemodynamic compromise from PE. Unstable patients treated with lytic agents are more likely to survive than those treated only with anticoagulants. However, fibri- nolysis may cause intracranial bleeding in up to 1.2 percent of patients. If a serious non-compressible bleeding complication results from lytic therapy, the infusion of the lytic agent should be stopped immediately and further treatment for the bleeding as well as alternative treatment for the PE (embolectomy) should be considered.

In the pre-fibrinolytic era, immediate pulmonary embolectomy was the only effective therapy for patients with a massive PE who were hypotensive and did not respond to fluid and pressor agents. Presently, embolectomy is reserved for the patient with severe pulmonary or cardiac compromise who is not a candidate for fibrinolysis or in whom fibrinolytic therapy has failed.

Conclusion
Pulmonary embolism continues to present a diagnostic challenge. Atypical presentations of PE are common, especially in elderly patients, and the clinician must maintain a high index of suspicion to make the proper diagnosis.

The use of a clinical scoring system to assign a pretest probability to a patient presenting with signs and symptoms of PE can effectively triage them into the appropriate probability category of low, intermediate, or high. We prefer the Wells scoring system discussed above because of its simplicity. However, it is limited in that this scoring system contains the subjective criterion, "Is PE as likely, or more likely than an alternative diagnosis?" This introduces variability into the decision-making algorithm. However, its ease of use, and the fact that it can be calculated from the history and physical exam without relying on any laboratory evaluation is extremely valuable to the physician who is expected to routinely handle multiple critically ill patients simultaneously.

We have also presented a slightly modified version of a treatment algorithm of Wells et al. and Wolfe and Hartsell that is feasible for treating physicians to follow. This algorithm or one similar to it can help the treating physician utilize the diagnostic modalities available to them in a logical and efficient way, while insuring that their patients receive the very best care possible.

Treating physicians continue to confront the diagnostic challenge presented by pulmonary embolism. Guidelines and algorithms designed to assist the clinician in the selection of appropriate diagnostic tests continue to be developed and refined, and increasingly are supported by evidence-based medicine. The advent of more accurate diagnostic modalities, as well as clear guidelines for their use, has made it possible for physicians to more effectively diagnose and treat their patients.

© 2004 American Association of Physician Specialists, Inc.

Click here for references.

Return to the Home Page