Pulmonary embolism (PE) is a relatively common vascular disease. PE and deep vein thrombosis (DVT) comprise two clinical presentations of the same pathologic process: venous thromboembolism (VTE) [1, 2]. The reported prevalence of PE is 0.4%, and according to data collected between 1979 and 1999, the incidence is 600,000 cases annually in the United States [1]. Acute PE is the third most common acute cardiovascular disease, after myocardial infarction and stroke [3, 4]. Untreated PE is fatal in up to 30% of patients, but its mortality rate is 2% to 10% with timely diagnosis and treatment [3].

Risk factors include older age, a history of previous VTE, active cancer, neurologic disease with extremity paresis, surgery, prolonged bed rest (especially if due to a medical disorder), and congenital or acquired thrombophilia [1]. However, in up to 30% of patients presenting with PE, there are no predisposing factors, so called unprovoked or idiopathic PE [1]. Greater that 90% of patients diagnosed with PE also have lower or upper extremity DVT [5]. There are many different tests to investigate a patient with suspected PE or DVT, and each of the diagnostic tests has strengths and limitations. This review article discusses the different imaging modalities available with an emphasis on the role of pulmonary CT angiography (CTA) in the workup of PE. The diagnostic role of pulmonary CTA in specific populations is addressed, including patients with impaired renal function or allergy to iodinated contrast material, pregnant patients, patients in extremis, and those of reproductive age. We further discuss issues such as pulmonary CTA combined with indirect CT venography (CTV), when to use CTV versus lower limb sonography, radiation dose considerations, the management of isolated subsegmental PE, the role of ECG-gated CTA, assessment of PE severity (i.e., clot burden and right heart overload), new technologic developments such as dual-source/dual-energy pulmonary CTA, and unanswered questions and research issues.

In general, clinical signs, symptoms, and routine laboratory tests cannot reliably exclude or confirm the diagnosis in a patient with suspected acute PE. However, a combination of these variables in addition to one of the available clinical prediction rules can be used to more accurately determine the clinical probability of PE [6]. The most frequently used prediction rules in clinics or emergency settings are the Wells score [7] (Tables 1 and 2), the revised Geneva score [8], and the Miniati or Pisa score [1, 9, 10]. The first prediction rule for PE was reported by Hoellerich and Wigton [11] in 1986. Within the past two decades, many different rules have been described. The most widely used clinical prediction rule is the Wells rule. However, the Wells rule has been criticized for its lack of objectivity because it includes the physician's judgment of whether an alternative diagnosis is more likely than PE. This criterion, which carries a major weight in the score (3 points), is subjective and cannot be standardized. Answering yes to this one question puts the patient at least in the moderate risk category. Therefore, several efforts were made to develop entirely objective scores, such as the Geneva, Charlotte, and Miniati rules. In the past decade, the most used rules have been modified to increase their usefulness and acceptability for clinicians. All of these prediction rules are simple to use and are based on easily collectable information or clinical variables [1]. The Pisa score may be more useful in inpatient settings [10]. Recent guidelines recommend first assessing the clinical probability of PE before deciding on the use of any diagnostic test (pretest probability) [10]. The posttest probability can then be determined according to both the characteristics of the diagnostic test and the pretest (clinical) probability [1, 10]. When the pre- and posttest probabilities of PE are discordant, further objective testing is advised [10].

Several different laboratory and radiologic tests are used to make a diagnosis of PE and DVT, including serum d-dimer, ventilation-perfusion (V/Q) scintigraphy, SPECT, pulmonary MR angiography (MRA), MR venography (MRV), pulmonary CTA indirect CTV, catheter pulmonary angiography, and transesophageal echocardiography. The decisive criterion for diagnosis of PE differs among nuclear medicine modalities (V/Q scan and SPECT), pulmonary CTA, and pulmonary MRA. With V/Q scintigraphy, PE is diagnosed by visualization of functional parameter changes, with reduced perfusion in areas of the lung that are normally ventilated (V/Q mismatches), whereas with pulmonary CTA and MRA, PE is diagnosed on the basis of embolus detection in the pulmonary arterial vasculature (morphologic changes) [3].

Although catheter pulmonary angiography was the reference standard for diagnosis of PE, its use is limited secondary to its invasiveness [1, 12]. The choice of diagnostic test depends on several factors, such as the clinical probability of PE, the patient's clinical condition, the availability of diagnostic testing, the risks associated with exposure to iodinated contrast agent or ionizing radiation, and the costs of each modality [13]. Table 3 shows the diagnostic accuracy for some of the tests.

Plasma d-dimer, a degradation product of cross-linked fibrin, is a test with a high negative predictive value (NPV) and a low positive predictive value (PPV) [1]. It is the first-choice test in patients with a low to moderate probability of clinical assessment (pretest probability) [13]. Plasma d-dimer concentrations above 0.5 mg/L have a sensitivity of 95% and specificity of 55% for VTE [6].

A negative d-dimer test result, when measured with a highly sensitive assay (over 95% sensitivity) (e.g., quantitative enzyme-linked immunoabsorbent assay), can be used to exclude PE in patients with a low or moderate pretest probability without the need for further testing [1, 6, 10, 13]. The resulting 3-month thromboembolic risk would be less than 1% in patients with a negative test who are left untreated [1]. On the other hand, when d-dimer is measured with moderately sensitive tests (e.g., quantitative latex-derived assay or whole blood agglutination assay), a negative result can exclude PE only in patients with a low clinical probability, resulting in a 3-month thromboembolic risk of less than 1% if the patient is not treated [1, 10].

For patients with high-probability clinical assessment, the d-dimer test is not helpful [6, 10, 13]. Moreover, normal d-dimer levels are uncommon in patients with a history of DVT or PE, older age (greater than 80 years), pregnancy (especially after 20 weeks), cancer, and hospitalized inpatients [10, 14]. An abnormal d-dimer result in any clinical setting or with a high clinical probability indicates the need for further diagnostic testing [6, 13].

V/Q scintigraphy was the main imaging test used for the diagnosis of PE before the introduction of contrast-enhanced pulmonary CTA. It is a safe modality, with few allergic reactions and a lower radiation exposure than helical CT [1]. Various studies have shown that V/Q scintigraphy can gave a definitive diagnosis in 28–74% of cases [10]. A recent analysis of data from the Prospective Investigation of Pulmonary Embolism Diagnosis (PIOPED) II study showed a sensitivity of 77.4% in patients with high-probability V/Q results (PE present) and specificity of 97.7% in patients with very low probability or normal V/Q results (PE absent) [15]. Combining the results of V/Q scintigraphy and the Wells score [16] resulted in a PPV and NPV greater than 90% [15].

Recent studies have suggested that the ventilation part of the examination can be omitted without affecting diagnostic accuracy [17]. This is particularly true in patients with normal chest radiographs, such as young and pregnant patients [17], and can also reduce both cost and radiation exposure [17]. Perfusion scintigraphy combined with chest radiography has the same diagnostic accuracy as pulmonary CTA and V/Q scintigraphy [17].

The most important limitation of V/Q scintigraphy is more nondiagnostic test results when compared with pulmonary CTA [15]. This results in lower diagnostic accuracy of V/Q scintigraphy [3] and the need for additional testing to confirm a diagnosis [16]. Sostman et al. [15] showed that nondiagnostic pulmonary CTA results were most often due to technically inadequate studies, whereas nondiagnostic V/Q scintigraphy results were most often due to inconclusive interpretations. However, when nondiagnostic studies were removed from the data analysis, the sensitivity and specificity of V/Q scintigraphy (77% and 98%, respectively) were similar to those of pulmonary CTA (83% and 96%, respectively) [15].

There is little evidence about the role of V/Q scintigraphy in different patient populations, but most nuclear medicine physicians would agree that V/Q scintigraphy results for inpatients and critically ill patients are more difficult to interpret than those for outpatients [15] (Fig. 1).

SPECT has higher diagnostic accuracy compared with V/Q scintigraphy. It has higher contrast resolution than planar V/Q imaging and therefore it can detect abnormalities at the subsegmental level and in the lung bases [18]. The higher sensitivity of SPECT for the detection of peripheral PE compared with pulmonary CTA makes it an important alternative when pulmonary CTA is not diagnostic or when it is contraindicated [3, 18]. This is mainly due to the visualization of peripheral perfusion defects resulting from small emboli at the subsegmental arterial level [3].

In previous studies investigating the diagnostic accuracy of SPECT, its sensitivity varied from 80% to 100% and its specificity varied from 64% to 100%, with a PPV ranging from 50% to 62% [18] (Fig. 2).

Since the late 1960s, catheter pulmonary angiography has been considered the reference standard and the most accurate test for the evaluation of PE and the reference test to which new diagnostic techniques are compared [19, 20]. Consequently, the major articles that evaluated the accuracy of catheter pulmonary angiography have used clinical outcome as a reference standard [19, 20]. Catheter pulmonary angiography is invasive, with a 2% morbidity and small risk of mortality, which may have contributed to underuse [21, 22]. However, another important point is that pulmonary CTA has also shown that catheter pulmonary angiography is not as accurate as once thought, is particularly poor for evaluation of the small pulmonary arteries, and is an imperfect standard or reference test.

Several studies have shown deterioration in angiographic interobserver agreement as the branch order of arteries being visualized increases, with poor interobserver agreement for subsegmental arteries [23–25]. When looking at the smaller pulmonary arteries, particularly the subsegmental vessels, there is considerable interobserver disagreement for the presence or absence of an embolism at pulmonary angiography [26]. Baile et al. [27] showed that there was no difference between pulmonary CTA and catheter pulmonary angiography for the detection of subsegmental-sized PE, concluding that pulmonary CTA is comparable to catheter pulmonary angiography for detection of pulmonary emboli, but catheter angiography was only 87% sensitive (not 100%). In studies that evaluate the accuracy of pulmonary CTA (or other technologies) using catheter pulmonary angiography as the reference test, this should be kept in mind.

More recently, a retrospective evaluation of the causes of discordant pulmonary CTA and catheter pulmonary angiographic readings from the PIOPED II study found that on catheter pulmonary angiography there were one false-positive examination and 13 false-negative examinations [28]. With pulmonary CTA there were two false-negative examinations. Four studies that were true-negative at pulmonary CTA became positive for thrombus by the time of catheter pulmonary angiography. This gave sensitivities for detection of PE of 87% for pulmonary CTA and 32% for catheter pulmonary angiography (p = 0.007). The largest missed thrombus at angiography was subsegmental in eight patients, segmental in two patients, and lobar in three patients; at CT it was subsegmental in two patients [28].

The risk of recurrent PE after negative catheter pulmonary angiography is low, with NPVs in the range of 99–100% reported in original studies and in a systematic review [29–32].

MRA is one of the new imaging techniques that can be used when there is contraindication to ionizing radiation or iodinated contrast media. Its estimated sensitivity and specificity were 78% and 99% respectively in the PIOPED III study, when patients with technically inadequate images (constituting 25% of the study patients) were excluded [33]. With an assumed prevalence of 11–50% for PE, the estimated PPV ranged from 91% to 98%, and the estimated NPV ranged from 97% to 82% [33]. The same study showed that the sensitivity of pulmonary MRA decreased when detecting smaller emboli, at 79% in the main or lobar pulmonary arteries versus 50% in the segmental arteries and 0% in the subsegmental arteries [33]. Other studies of gadolinium-enhanced pulmonary MRA have shown a sensitivity of 100% for PE in the central and lobar arteries, 84% in the segmental arteries, and 40% in the segmental branches [6].

In the PIOPED III study, when pulmonary MRA was combined with MRV, the number of patients with technically inadequate studies increased from 25% to 52% [33]. However, with exclusion of the technically inadequate group, the sensitivity and specificity rose to 92% and 96%, respectively. The PPV and NPV for combined pulmonary MRA–MRV ranged from 74% to 98% and 99–92%, respectively [33].

A positive pulmonary MRA or pulmonary MRA–MRV result with a high or moderate Wells score is reported to be predictive of a 91–99% probability of pulmonary embolism [33]. However, when the Wells score is of low probability, a positive pulmonary MRA or a positive pulmonary MRA–MRV was associated with an 84% and 62% probability of PE, respectively [33].

Thus, due to the higher number of technically inadequate studies with pulmonary MRA, more complex and less robust examinations with pulmonary MRA compared with pulmonary CTA, longer examination times with pulmonary MRA, limited patient access to MRI, limited ability of pulmonary MRA to detect cardiopulmonary disorders other than PE, and contraindication to MRI in patients with pacemakers or other implanted devices, the PIOPED III investigators and Medicare suggest that pulmonary MRA should be considered only for patients with contraindications to standard tests and in centers that routinely perform it [6, 33] (Fig. 3).

Pulmonary CTA alone or in combination with indirect CTV is now the standard diagnostic test for the evaluation of patients with a high clinical probability of PE as well as for patients with low to moderate probabilities of PE with a positive d-dimer test [3, 13]. The rapid and widespread availability of pulmonary CTA, its noninvasiveness and accuracy, the speed of diagnosis, the ability to concurrently perform lower extremity venous imaging with CTV, the higher proportion of definitive diagnostic results compared with V/Q scintigraphy, and the ability to diagnose alternative or additional causes of chest pain make pulmonary CTA the first-line diagnostic imaging modality in most patients with suspected acute PE [6, 15] (Fig. 4).

Given the improved diagnostic accuracy and performance, MDCT scanners have now largely replaced single-detector CT scanners [3, 6, 34]. Improvements in CT technology allow routine visualization of pulmonary arteries to the subsegmental level, with less motion-related artifact and fewer indeterminate results [6].

In the PIOPED II study, pulmonary MDCTA had a sensitivity of 83% and specificity of 96% for diagnosis of PE, with a positive likelihood ratio of 19.6 (95% CI, 13.3–29.0) and a negative likelihood ratio of 0.18 (95% CI, 0.13–0.24) [2]. The PPV and NPV were 86% and 95%, respectively. The PPV for PE in a main or lobar artery was 97%, whereas for segmental and subsegmental vessels, the PPVs were 68% and 25%, respectively [2]. Outcome studies have shown that the diagnosis of PE can be safely excluded by a negative pulmonary MDCTA [10]. Although PIOPED II may currently be the best study on pulmonary CTA published, it nevertheless suffers from limitations. The main limitation of the study is use of noninvasive diagnostic tests as part of the reference standard. Other limitations of the study include the restriction of recruitment to patients who could safely undergo the extra testing within 36 hours after the reference test. The reported values of diagnostic accuracy may not apply to pregnant women; patients with renal failure; and patients who are critically ill, in shock, or receiving ventilatory support because these patient groups were not included in the study. Among the 1,090 recruited patients, a significant proportion (238 patients) did not complete the diagnostic reference testing, primarily because the diagnosis was inconclusive on noninvasive testing and patients or their clinical team declined catheter angiography. It is uncertain whether this affected the results. It is also not apparent whether a lack of screening and recruiting of patients during the night and on weekends affected the results. Patients who presented on weekdays may have had generally milder symptoms than those who presented at night or on weekends.

In a recent meta-analysis, the clinical validity of using a pulmonary CTA (either a single-detector or MDCT) to rule out PE was reported to be similar to that of a pulmonary catheter angiography [6, 35].

There are a few reasons to use ECG-gating during pulmonary CTA. First, an objective assessment of right ventricular (RV) function could help stratify patients with RV dysfunction, who have a much more guarded prognosis, and guide certain therapeutic decisions, such as intensive therapy with thrombolytic agents or mechanical embolectomy, either at catheter angiography or surgery [6]. Second, the clinical presentation of patients suspected of having acute PE is nonspecific, and it is well established that clinical signs and symptoms of PE and myocardial infarction overlap. Therefore, the possibility of using ECG-gated CTA for assessment of coronary artery disease as a potential cause for chest pain or dyspnea could improve patient evaluation and triage, especially in the emergency department [6].

In general, the use of ECG gating adds radiation exposure. Other limiting factors for ECG-gated acquisitions include longer scanning times compared with those for non–ECG-gated acquisitions and image quality degradation due to irregular rhythms or high heart rates. These limitations are expected to be lessened with the introduction of dual-source CT technology, which should enable radiologists to provide clinicians with cardiac functional information in routine clinical practice [6].

Dual-source/dual-energy CT was first introduced in 2006 and is a subgroup of MDCT [3]. The scanner consists of two tubes and two detectors mounted orthogonally to each other [3]. It has the advantage of providing CT data at the same energy (dual-source scanning), or at two different tube voltages simultaneously (dual-energy scanning) [3, 36].

The application of dual-energy pulmonary CTA in the diagnosis of PE mainly relies on the ability to identify contrast medium or specific tissue infiltration, which cannot be precisely detected with single-energy CT [36]. There are several advantages of dual-energy pulmonary CTA compared with other diagnostic modalities for detection of PE because with this technique, blood volume compared with blood flow can be visualized [36] (Figs. 5A, 5B, 5C, and 5D). It can provide high-quality morphologic analysis and functional information on the pulmonary circulation from the same dataset, comparable to pulmonary MRA or V/Q scintigraphy [36]. It allows comparison of pulmonary CTA images acquired at different energies in the same patient at the same time point after contrast medium injection [36]. In cases of suboptimal arterial concentration of contrast medium, the low-energy acquisition enables one to generate images with increased vascular enhancement [36]. A disadvantage of dual-energy pulmonary CTA is the higher radiation exposure compared with single-source CT [36]. In a study of 13 patients, the sensitivity, specificity, and NPV of dual-energy pulmonary CTA compared with pulmonary perfusion scintigraphy were estimated as 75%, 80%, and 66%, respectively [3].

A manufacturer has developed a scanner that can perform dual-energy imaging with a single (source) tube using rapid energy switching in the tube. This is made possible by the use of a new detector material produced by changing the molecular structure of garnets. This scintillator is capable of delivering images 100 times faster, with up to 33% greater detail within the body and up to 47% greater detail within the heart. This imaging uses up to 2,496 views per rotation, a 2.5 times increase on previous technology, to deliver improved spatial resolution and improved image quality across the entire field of view. Dual-energy fast kilovoltage switching registers energies at least 165 times faster than dual-source CT at a 330-millisecond rotating speed. It offers 128 sections of data per rotation and 101 user-selectable energy levels for viewing. This also allows both better imaging and radiation dose reduction, reducing dose by up to 50% within the body and by as much as 83% for cardiac examinations.

Combining pulmonary CTA with CTV in the PIOPED II study resulted in an increase in inconclusive interpretations (from 6% with pulmonary CTA alone to 10.5% with pulmonary CTA–CTV) but with sensitivity increased from 83% to 90% without a change in specificity (95%) [2]. Although the increase in sensitivity was not statistically significant [6, 10], the PIOPED II investigators concluded that pulmonary CTA alone with a false-negative rate of 17% was not adequate to exclude PE, and its combination with CTV was necessary [2, 6]. In addition, 8% of subjects in the PIOPED II study with VTE had DVT only [2]. In another study of patients with proven PE and a negative pulmonary CTA, 16% had a positive diagnostic CTV [10]. Several other studies have shown that the use of CTV in addition to pulmonary CTA increases the percentage of patients requiring anticoagulation by 5–27% [6]. Pulmonary CTA with CTV allows “one-stop shopping” with assessment for both PE and DVT; the greatest benefit from CTV combined with pulmonary CTA has been shown in sicker patients, in centers with less experience, and with older equipment [6] (Fig. 6).

Despite this, the role of CTV is currently less established than that of pulmonary CTA [10]. Questions remain about the utility of CTV: Does lower extremity imaging fit in the diagnostic workup for PE? Is sonography or CTV the preferred modality? Should CTV be performed routinely?

The literature is inconclusive regarding whether the additional diagnostic yield from CTV justifies the additional time, expense, and patient irradiation [6]. Numerous clinical outcome studies have shown similar NPVs for a negative pulmonary CTA alone and a negative catheter pulmonary angiography, suggesting that CTV may not be required [6]. The PIOPED II study showed that the use of the Wells criteria [7] along with pulmonary CTA alone has sensitivity similar to that of pulmonary CTA–CTV [6].

As would be expected, the predictive values of pulmonary CTA and pulmonary CTA–CTV change when assessment of the clinical probability is taken into account, with higher PPVs among patients with a higher pretest probability of PE and lower NPVs among patients with lower pretest probabilities of PE [2]. The sensitivity, specificity, and proportion of inconclusive interpretations with pulmonary CTA as well as pulmonary CTA–CTV are not affected by age or sex according to the analysis of the PIOPED II study results [34].

The PIOPED II investigators calculated radiation doses from pulmonary CTA to the chest at 3.8 mSv, from indirect CTV to the pelvis at 6.0 mSv, and for the thighs at 3.2 mSv [2]. Kalva et al. [37] calculated effective radiation doses from CTV to the pelvis at 5.2 mSv ± 0.5 (SD) and 0.6 mSv ± 0.2 for the lower extremities. Those authors concluded that CTV of the pelvis after pulmonary CTA did not significantly improve the detection of VTE, indicating that CTV may be limited to the lower extremities, thus reducing the radiation dose [37]. The increased gonadal radiation dose is significant for pulmonary CTA–CVT compared with pulmonary CTA alone, with the gonadal radiation dose increasing 1,500-fold for women and 2,000-fold for men [2].

With continuous versus discontinuous imaging during CTV, using data from PIOPED II, Goodman et al. [38] showed that there was agreement for the presence of DVT in at least one leg (same side) or for the absence of DVT in both legs in 89% of cases. There was substantial agreement between the techniques with a kappa statistic of 0.75 between the consensus interpretations and the test interpretations. The authors concluded that there was good agreement between continuous helical and discontinuous axial imaging for the detection of DVT, and there appeared to be little difference between the two approaches. However, adopting discontinuous imaging and other dose-reduction strategies could reduce pelvic radiation by more than 75% [38].

The PIOPED II study showed that CTV detects thrombosis in the inferior vena cava (IVC) or iliac vein alone in 3% of cases [2]. Therefore, pelvic radiation can be decreased considerably by scanning from the acetabulum to the knees (i.e., scanning the femoral veins) rather than from the iliac crest to the knees (i.e., scanning the femoral and iliac veins) [39].

When comparing CTV to lower-limb sonography for the diagnosis or exclusion of DVT, using PIOPED II data, Goodman et al. [38] showed that there was 96% concordance between CTV and sonography for the diagnosis or exclusion of DVT. There was near perfect agreement between the two techniques with a kappa statistic of 0.81. The sensitivity and specificity of combined pulmonary CTA and CTV were equivalent to those of combined pulmonary CTA and sonography. Diagnostic results in subgroups, including patients with signs or symptoms of DVT, asymptomatic patients, and patients with a history of DVT, were similar whether CTV or sonography was used. The authors concluded that CTV and sonography showed similar accuracy in confirming or excluding DVT [38]. The incidence of positive studies in patients without signs, symptoms, or history of DVT was low. In terms of clinical significance, CTV and lower-extremity sonography yield equivalent diagnostic results. The incidence of positive studies in patients without signs, symptoms, or history of DVT is low, thus the choice of imaging technique can be made on the basis of safety, expense, and time constraints [40]. Moreover, numerous studies have shown an approximate 97% agreement between CTV and sonography [6]. In contrast to CTV, sonography is well accepted, has low cost, and does not expose the patient to ionizing radiation [6].

Therefore, reasonable strategies for reducing pelvic radiation dose would include not routinely performing CTV on younger patients, in particular women of child-bearing age or in patients with no leg symptoms, i.e., no clinical evidence of DVT. CT venography is contraindicated in pregnant patients. CT venography could be reserved for older patients who are less sensitive to the effects of radiation, for sick or ICU patients to prevent these patients from being imaged by both CT and sonography, and in patients with leg symptoms (i.e., clinical evidence of DVT). Pelvic radiation could further be reduced by performing discontinuous rather than continuous imaging and commencing scanning from the acetabulum to the knees (i.e., scanning the femoral veins) rather than from the iliac crest (i.e., scanning the femoral and iliac veins) when performing CTV. Pelvic and lower-extremity radiation could be completely eliminated by performing sonography rather than CTV for the diagnosis or exclusion of DVT.

In hemodynamically stable patients with a low or intermediate pretest probability for PE determined by a clinical prediction rule many authorities recommend performing a serum d-dimer assay as the next step [1, 13] (Fig. 7). If the serum d-dimer assay is negative, given its high NPV of 98%, it is safe to withhold further investigation for PE or treatment [13, 41].

If the d-dimer assay is positive or the pretest probability for PE is high, it is recommended to proceed with further testing (Fig. 7). Most of the PIOPED II investigators and the European Society of Cardiology (ESC) Task Force recommend pulmonary MDCTA as the first-line test in patients with a positive d-dimer or high pretest probability patients [1, 13]. The Fleischner Society recommends either pulmonary MDCTA or V/Q scintigraphy in these patients [6]. The PIOPED II investigators recommended performing indirect CTV at the same time as the pulmonary CTA [13]. Since then, many articles have been published showing that CTV and lower-limb venous sonography are equivalent in terms of accuracy [40]. Therefore, the choice of either test will depend on the individual patient's circumstances, with pulmonary CTA and CTV advocated as a one-stop-shop in sick and intensive-care patients [6].

In hemodynamically unstable patients, pulmonary MDCTA is recommended as the first-line test, with echocardiography as an alternative if pulmonary MDCTA is not available [1].

In low pretest probability patients with a negative pulmonary MDCTA, it is safe to withhold therapy due to the very high NPV of pulmonary MDCTA of more than 99% [35] (Fig. 8). In low pretest probability patients with a positive pulmonary MDCTA, the ESC Task Force recommends treatment with anticoagulation [1] (Fig. 8). The PIOPED II investigators stratify management on the basis of the size of the vessels involved with emboli [13]. For emboli in the main, right, left, or lobar pulmonary arteries, treatment is recommended [13]. For emboli in the segmental or subsegmental vessels, further evaluation with imaging tests is recommended because of the low PPV of pulmonary MDCTA in these patients [13], who represent a discordant group—that is, a low clinical probability with a positive test. Further evaluation could be performed with repeat pulmonary MDCTA if there was a technical issue with the initial pulmonary MDCTA or, alternatively, pulmonary catheter angiography or V/Q scintigraphy could be performed [13]. If the lower-limb veins had not been assessed, this could be done with sonography or MRV.

In moderate pretest probability patients with a negative pulmonary MDCTA, it is safe to withhold therapy because of the relatively high NPV of pulmonary MDCTA [2] (Fig. 9). The PIOPED II investigators mentioned lower limb evaluation with MRV or sonography as an option at this point if this has not been performed [13]. In moderate pretest probability patients with a positive pulmonary MDCTA, the PIOPED II investigators and the ESC Task Force recommend treatment [1, 13] (Fig. 9). Further evaluation could be performed with repeat pulmonary MDCTA if there are technical issues with the initial pulmonary MDCTA or, alternatively, pulmonary catheter angiography or V/Q scintigraphy could be performed [13]. In addition, if the lower-limb veins have not been assessed, this could be done with sonography or MRV.

In high pretest probability patients with a negative pulmonary MDCTA, the PIOPED II investigators and ESC Task Force both recommend further evaluation [1, 13] (Fig. 10). These patients represent a discordant group—that is, a high clinical probability with a negative test. The options again include repeat pulmonary MDCTA if there were technical problems with the initial study, an alternative test such as pulmonary catheter angiography or V/Q scintigraphy, or evaluation of the lower-limb veins with sonography or MRV if the lower limbs were not evaluated. In high pretest probability patients with a positive pulmonary MDCTA, the PIOPED II investigators and the ESC Task Force recommend treatment [1, 13] (Fig. 10).

The choice of diagnostic tests depends on several factors, such as the clinical probability of PE, characteristics of the patient, availability of diagnostic tests, the costs, and the risks associated with each test (e.g., exposure to iodinated contrast agent or radiation) for each individual patient [13]. During the past decade, the role of pulmonary CTA in the diagnosis of PE has dramatically increased [6]. This raises concerns regarding the potential carcinogenic effects of ionizing radiation on radiosensitive tissues, such as the thyroid gland; gonads; and, in particular, the female breast [14]. There are also issues regarding the use of iodinated contrast agents in patients with allergy to contrast material or those with impaired renal function. Current concerns are no longer centered on showing the clinical value of pulmonary CTA but are more about how to optimize its application in various categories of patients [6]. The PIOPED II investigators [13], the Fleischner Society [6], and the ESC [1] have published recommendations for special patient populations on the basis of previous studies.

Because anticoagulation was rapidly accepted into clinical practice at a time when diagnostic tests were basic and isolated small emboli rarely diagnosed, there have been very few controlled studies of anticoagulation in the modern era. Most of the evidence is indirect, and a true understanding of the consequences of small PE is difficult. The majority of patients with small PE are never suspected clinically and are never evaluated [68, 69]. In fact, autopsy studies show evidence of old or recent PE in 51–90% of patients when there is careful examination of the pulmonary vessels [68, 69]. The majority of PEs that are thought to be fatal are not suspected clinically and are not treated [68].

Swensen et al. [70] assessed the outcome of withholding anticoagulation from patients with suspected acute PE in whom CT findings were interpreted as negative for PE. They found the incidence of VTE or fatal PE among patients with suspected acute PE and negative CT results with no other evidence of VTE to be low. Withholding anticoagulation in these patients appears to be safe. In a meta-analysis of 23 studies reported on 4,657 patients with negative pulmonary CTA who did not receive anticoagulation, the 3-month rate of subsequent VTE events was 1.4% (95% CI, 1.1–1.8%), and the 3-month rate of fatal PE was 0.51% (CI, 0.33–0.76%) [71]. However, studies withholding anticoagulation from patients with suspected acute PE in whom pulmonary CTA findings are negative for central, lobar, or segmental pulmonary embolism but positive for isolated subsegmental pulmonary embolism in the absence of DVT have not been performed.

In fact, in the modern era there has only been one randomized controlled trial in which the recurrence and mortality rates of patients treated with anticoagulation therapy for proven VTE were compared with those of patients with proven VTE who did not receive anticoagulation. In 1994, Nielsen et al. [72] examined 87 ambulatory patients with venography-proven DVT but no symptoms of PE, although 49% of these patients had proven occult PE. One half (44 patients) were treated with anticoagulation, and one half (43 control patients) were treated with an antiinflammatory agent. At 3 months, 19 patients in each group had developed progressive VTE, either DVT or PE. Thus, anticoagulation did not appear to alter disease progression. There were no deaths among the 43 control patients despite progressive VTE in 19 of them, but one patient who was undergoing anticoagulation therapy died [72].

There is further evidence that withholding anticoagulation therapy in patients with isolated subsegmental PE in the absence of DVT may not be harmful. In the PIOPED I study, 20 patients who had negative pulmonary catheter angiography results at their local hospital and did not receive anticoagulation therapy were subsequently found to have PE on the original pulmonary catheter angiograms as decided by an expert panel of angiographers [73]. Of these 20 nontreated patients, the PE-associated fatality rate was 5% and the PE-associated nonfatality rate was 5%, which is comparable to the PE-associated fatality rate of 2.5% and the recurrence rate of 3.5% for the patients in PIOPED I who received anticoagulation therapy. The 20 nontreated patients had a limited clot burden. All of them had fewer than three mismatched subsegmental equivalent perfusion defects compared with 60% of all patients with PE [73]. The angiographic results were positive only at the segmental or subsegmental level in 84% of these patients versus in 36% of all the patients with PE. Stein et al. [73] concluded that mild untreated PE carries a lower immediate mortality from recurrent PE than PE described in prior decades [73].

Schultz et al. [74] prospectively performed pulmonary MDCTA for the assessment of possible PE or DVT in moderately to severely injured inpatients 3 to 7 days after admission, who had no signs or symptoms of VTE. Of these 90 patients, 24% had PE. Four patients with major occult PE and one patient with isolated subsegmental PE and DVT were treated with anticoagulation therapy. The remaining 17 patients, who had subsegmental emboli only, were not treated with anticoagulation therapy. No patients had clinical evidence of PE during their hospitalization. The 10 patients who were available for 3-month follow-up were symptom free [74].

Eyer et al. [75] showed that for 37% of patients with isolated subsegmental PE and for 85% of patients with inconclusive pulmonary MDCTA results, the primary physician chose not to administer anticoagulation treatment. Two patients in each subgroup returned with signs or symptoms of PE, but all of the patients had negative repeat imaging results. Thirty-two percent of patients were treated with anticoagulation, and five had recurrent signs and symptoms, but no VTE was found in any patient [75]. In another study involving patients who were not treated with anticoagulation therapy, 2% of patients with an uncertain diagnosis of PE were subsequently found to have PE, whereas 0.8% of patients with a diagnosis of “PE excluded” were subsequently found to have PE [76].

In 1994, Hull et al. [77] proposed that in patients with adequate cardiopulmonary reserve and nondiagnostic V/Q scintigraphy, anticoagulation was not required if serial studies of the lower extremities showed normal findings. Hull et al. defined adequate cardiopulmonary reserve as the absence of any of the following conditions during the first 10 days after presentation: pulmonary edema, RV failure, hypotension (systolic blood pressure less than 90 mm Hg), syncope, acute tachyarrhythmia, or respiratory failure indicated by severely abnormal spirometric results (forced expiratory volume in 1 second less than 1.0 or vital capacity less than 1.5 L) or blood gas measurements (partial pressure of oxygen less than 50 mm Hg or partial pressure of carbon dioxide greater than 45 mm Hg while breathing room air) [77]. Wells et al. [78] proposed a similar strategy involving the use of serial lower-extremity sonography.

There are three scenarios in which most would agree that even small PE requires treatment: patients with small PE and inadequate cardiopulmonary reserve, patients who have a small embolus and coexisting acute DVT, and patients who have recurrent small PE possibly due to thrombophilia to prevent chronic PE and pulmonary arterial hypertension [79]. However, there appear to be subsets of patients with small or questionable PE in whom the risks associated with anticoagulation may outweigh the benefits: symptomatic patients who have clots limited to the subsegmental vessels, no DVT, and adequate cardiopulmonary reserve; patients with indeterminate pulmonary MDCTA or V/Q scintigraphy results, no DVT, and adequate cardiopulmonary reserve; asymptomatic patients with incidentally discovered small emboli, no DVT, and adequate cardiopulmonary reserve; patients with contraindications to anticoagulation (e.g., intracranial hemorrhage, recent surgery, or trauma), isolated subsegmental PE, and no DVT; and patients with contraindications to anticoagulation, indeterminate pulmonary MDCTA results, and no DVT [75, 77, 78]. All of these scenarios have central to them isolated subsegmental PE or indeterminate pulmonary MDCTA results and no DVT plus other features.

There are also several secondary factors that might strengthen the decision to with-hold anticoagulation: no or few risk factors for VTE, transient (e.g., recent surgery or recent injury) rather than persistent (e.g., coagulopathy, cancer) risk factors for VTE, other cardiopulmonary disease that might explain the patient's signs and symptoms, and negative d-dimer test results [41, 80, 81].

It is not clear whether small PE in the absence of demonstrable DVT justifies the expense, mortality, and serious morbidity associated with anticoagulation. In well-controlled studies (international normalized ratio, 2.0–3.0), the rate of major bleeding caused by warfarin therapy has been below 3.0% and the mortality rate has been less than 0.5% at 3 months [82]. Under less stringent control of anticoagulation, the mortality and morbidity at 1 year are 1% and 7%, respectively [83].

No clear consensus exists in the literature as to whether patients with isolated subsegmental PE in the absence of DVT should be treated with anticoagulation, with the decision to treat often based on physician preference, clinical suspicion, and other test results. Therefore, it is extremely important to know if the risk of developing a life-threatening PE in a patient with isolated subsegmental PE in the absence of DVT is greater than the risk of major complication from treating an isolated subsegmental PE with anticoagulation (Fig. 15).

Risk stratification is important because the workup, treatment, and monitoring of PE will vary depending on the overall prognosis. With fatal PE, patients usually die as a result of RV failure and circulatory collapse, which often occur within a few hours of admission [3]. Therefore, RV dysfunction needs to be diagnosed rapidly to identify patients who may benefit from thrombolytic therapy. Acute right-sided heart failure can be diagnosed on pulmonary CTA by assessing the dimensions of the right-sided cardiac chambers, superior vena cava, or azygos vein. Measurements include RV to left ventricular (LV) short-axis ratios and main pulmonary artery diameter to aorta diameter ratios, which are usually not greater than 1. Other features to indicate RV dysfunction include a dilated IVC, reflux of contrast medium into the IVC, and leftward bowing or sigmoid configuration of the interventricular septum (Fig. 16). With the use of ECG-gating, it is also possible to measure LV and RV ejection fractions [3].

Thrombus burden (clot load) scores can also be used to assess the magnitude or severity of PE, such as those of Miller et al. [84] and Walsh et al. [85]. The Miller Index for pulmonary angiography uses an objective score for the vessel level of arterial obstruction (with 1 for segmental vessels and larger arteries scored on all the segments that arise distally, with a maximum score of 16) and a subjective score for peripheral perfusion in the lungs (with three zones on each side, each scored 0 for normal perfusion, 1 if moderately reduced, 2 if severely reduced, and 3 if absent, with a maximum score of 18) [84]. Since the advent of CT and MDCT, the Miller and Walsh scores have been modified for CT [86]. In addition, newer scoring systems have been developed that include qualitative assessments of the degree of vascular obstruction (with 0 for no occlusion, 1 for partial occlusion, and 2 for complete occlusion) [87]. These scores correlate well with the Miller Index and echographic assessment of right heart overload and dilatation [87]. Newer scoring systems that use quantitative assessments of the percentage of obstruction in vessels (with 1 for < 25%, 2 for 25–49%, 3 for 50–74%, 4 for 75–99%, and 5 for 100%) also correlate well with echographic assessment of cor pulmonale and pulmonary hypertension [88].

The clinical utility of the thrombus burden (clot load) scoring systems in daily practice is uncertain. The decision to withhold anticoagulation therapy is based on the absence of PE, and the decision to treat with anticoagulation therapy is usually based on the presence of emboli rather than on clot burden. However, patients with abnormal RV function on echocardiography, pulmonary CTA, or pulmonary MRA may be treated with thrombolysis or with mechanical or surgical embolectomy, and a CT obstruction index greater than 40% has been shown in more than 90% of patients with RV dilatation [87].

The severity of PE should be understood as an individual estimate of PE-related early mortality risk rather than the anatomic burden and the shape and distribution of intrapulmonary emboli [1]. Therefore, current guidelines suggest replacing potentially misleading terms such as “massive,” “submassive,” and “nonmassive” with the estimated level of the risk of PE-related early death [1]. High-risk PE is a life-threatening emergency requiring specific diagnostic and therapeutic strategy (short-term mortality > 15%) [1]. Non–high-risk PE can be further stratified according to the presence of markers of RV dysfunction or myocardial injury into intermediate- and low-risk PE. Intermediate-risk PE is diagnosed if at least one RV dysfunction or one myocardial injury marker is positive. Low-risk PE is diagnosed when all checked RV dysfunction and myocardial injury markers are found negative (short-term PE-related mortality < 1%) [1].

There are an increasing number of reports of incidental, asymptomatic PE that are detected in patients undergoing chest CT for reasons other than suspected PE. These incidental diagnoses are due to the widespread use of MDCT scanners and to their high acquisition speed, which results in increased spatial resolution and improved visualization of peripheral pulmonary arteries. With the increasing use of MDCT, incidental diagnoses of PE are becoming a common problem in clinical practice. However, information on the prevalence and on the natural history of unsuspected PE is extremely limited [89]. Pulmonary emboli are detected incidentally in asymptomatic patients at a prevalence of approximately 1.0–1.5% in the general patient population but the clinical importance of these unsuspected emboli is unknown [90]. Inpatients have a higher prevalence, around 4–5%. Many patients found to have incidental pulmonary emboli have known risk factors for thromboembolic disease, such as clotting disorders, recent surgery, or malignancy [90].

In a series that looked at incidentally detected PE in oncology patients, the authors found 4.0% of the 403 patients had pulmonary emboli [90]. However, only 25% of the patients with emboli were identified at initial clinical CT image interpretation. All had multiple emboli involving at least the lobar arteries [90]. Missed emboli were typically solitary and involved smaller arteries. Sixty percent of patients with emboli who underwent any lower extremity imaging had DVT [90].

In a recent meta-analysis of incidental asymp tomatic pulmonary embolism, the mean prevalence of incidental PE was 2.6% (95% CI, 1.9–3.4%) based on 12 studies with a total of more than 10,000 patients included [89]. The authors concluded that the prevalence of incidental PE is clinically relevant and that future studies are necessary to properly evaluate the clinical history of these patients [89].

Overall, the diagnosis of unsuspected PE is associated with increased morbidity and mortality rates, especially in certain groups of patients, such as those with cancer. However, the optimal therapeutic strategies when asymptomatic PE is incidentally diagnosed are uncertain. In the absence of evidence on the risk-to-benefit ratio of an active treatment, it is currently recommended that the same initial and long-term anticoagulation as for comparable patients with symptomatic PE be prescribed [89].

Pulmonary CTA has become the first-line imaging test for the diagnosis of PE during recent years. The technologic advances in CT have had a dramatic impact on its accuracy to diagnose PE [14]. The role of pulmonary CTA will continue to grow with the emergence of MDCT and dual-energy CT and their improved capabilities [14]. However, the need for any given CT examination should always be justified on the basis of the individual patient's benefits and risks [14]. Accurate risk stratification for selection of diagnostic method of PE is crucial, and both radiologists and referring physicians need to work together to develop better selection criteria of patients for CT [14].

Supported in part by the General Electric-Radiological Society of North America Radiology Educational Scholarship Award.

Address correspondence to P. Cronin ([email protected]).