AMERICAN THORACIC SOCIETY
The Diagnostic Approach to Acute Venous Thromboembolism
Clinical Practice Guideline

	INTRODUCTION

TOP INTRODUCTION CONTENTS INTRODUCTION THE DIAGNOSTIC APPROACH TO... THE DIAGNOSTIC APPROACH TO... THE DIAGNOSTIC APPROACH TO... THE DIAGNOSTIC APPROACH TO... THE FUTURE REFERENCES

THIS OFFICIAL STATEMENT OF THE AMERICAN THORACIC SOCIETY WAS ADOPTED BY THE ATS BOARD OF DIRECTORS, FEBRUARY 1999

	CONTENTS

TOP INTRODUCTION CONTENTS INTRODUCTION THE DIAGNOSTIC APPROACH TO... THE DIAGNOSTIC APPROACH TO... THE DIAGNOSTIC APPROACH TO... THE DIAGNOSTIC APPROACH TO... THE FUTURE REFERENCES

Introduction

The Diagnostic Approach to Acute Deep Venous Thrombosis

Background

Symptoms and Signs

Contrast Venography

Impedance Plethysmography

  Background

  Physiology and Technique

  Limitations of Impedance Plethysmography

  Early Clinical Trials: Establishing Accuracy in

    Symptomatic Acute Proximal DVT

  Management Studies: Early Success and Later Questions

  Impedance Plethysmography in Asymptomatic Patients

  Recurrent and Chronic Deep Venous Thrombosis

Compression Ultrasound with Venous Imaging

  Background

  Technique

  Limitations of Compression Ultrasound with

    Venous Imaging

  Symptomatic Acute Proximal Deep Venous Thrombosis

  Asymptomatic Acute Proximal Deep Venous Thrombosis

  Acute Calf Deep Venous Thrombosis

  Recurrent and Chronic Deep Venous Thrombosis

  Upper Extremity Deep Venous Thrombosis

Magnetic Resonance Imaging

The Diagnostic Approach to Acute Pulmonary Embolism

Background

Symptoms and Signs

Electrocardiography

Arterial Blood Gas Analysis

Chest Radiography

D-Dimer

The Ventilation-Perfusion Scan

  The Effect of Prior Cardiopulmonary Disease

  The Perfusion Scan Alone

  The Nondiagnostic Ventilation-Perfusion Scan: Use of

    Lower Extremity Studies

Pulmonary Angiography

Spiral (Helical) Computed Tomography

Magnetic Resonance Imaging

Echocardiography

The Diagnostic Approach to Acute Venous Thromboembo-

    lism: Final Summary and Recommendations

The Diagnostic Approach to Acute Pulmonary Embolism:

    Final Summary and Recommendations

The Future

References

	INTRODUCTION

TOP INTRODUCTION CONTENTS INTRODUCTION THE DIAGNOSTIC APPROACH TO... THE DIAGNOSTIC APPROACH TO... THE DIAGNOSTIC APPROACH TO... THE DIAGNOSTIC APPROACH TO... THE FUTURE REFERENCES

Venous thromboembolism (VTE) represents a spectrum of disease that includes both deep venous thrombosis (DVT) and pulmonary embolism (PE). Pulmonary embolism most commonly results from DVT occurring in the deep veins of the lower extremities, proximal to and including the popliteal veins. Both DVT and PE are frequently clinically unsuspected, leading to significant diagnostic and therapeutic delays and accounting for substantial morbidity and mortality. While there are as many as 260,000 patients in the United States in whom VTE is diagnosed and treated each year, more than half of the cases that actually occur are never diagnosed and as many as 600,000 cases may therefore occur (1). Because of the magnitude of the problem, and the variable diagnostic approaches that are feasible, this official statement outlining acceptable diagnostic approaches to VTE is presented. The treatment of acute VTE will not be addressed.

To present a coherent position on the diagnostic approach to VTE, clinical trials evaluating the diagnostic approach to DVT and PE have been reviewed and are categorized as level 1 or level 2 (2). Level 1 studies are those that incorporate the following three criteria: (1) previous establishment of objective diagnostic criteria for normal and abnormal diagnostic studies, (2) independent comparison of the diagnostic result with contrast venography (CV) for DVT or with pulmonary angiography for PE, with readers blinded to the other test result, and (3) the prospective evaluation of patients who were enrolled consecutively. A clinical trial was accepted as enrolling consecutive patients only if this was explicitly stated or if the study stated that patients were excluded only if they refused consent or could not tolerate the diagnostic procedure. Other clinical trials were considered to be level 2. It should be emphasized that CV and pulmonary angiography have been established as gold standard diagnostic tests by default, so that when other modalities are evaluated, this a priori assumption exists (3). Relatively more data are presented for impedance plethysmography (IPG) and for ultrasound (US) imaging than for other diagnostic modalities because the data involving these technologies are more extensive and complex. More level 1 data exist for these techniques than for newer technology such as spiral computed tomography (CT) scanning or magnetic resonance imaging (MRI). For DVT and PE, background information is presented, followed by a discussion of the clinical diagnosis. Subsequently, each diagnostic technique is addressed. Final guidelines are ultimately presented for the diagnostic approach to both DVT and PE. The recommendations of the American Thoracic Society Clinical Practice Committee (4) were reviewed as this statement was developed and our goal was to adhere to these guidelines. The committee preparing this document was multidisciplinary, as recommended. Because different medical centers have different resources, clinical flexibility was built into the recommendations. The latter concept is of particular importance in the diagnostic approach to venous thromboembolism because although level 1 studies have been performed at some medical centers, validated protocols or the specific technology is not available everywhere and the resulting data may not be applicable at other centers.

	THE DIAGNOSTIC APPROACH TO ACUTE DEEP VENOUS THROMBOSIS

TOP INTRODUCTION CONTENTS INTRODUCTION THE DIAGNOSTIC APPROACH TO... THE DIAGNOSTIC APPROACH TO... THE DIAGNOSTIC APPROACH TO... THE DIAGNOSTIC APPROACH TO... THE FUTURE REFERENCES

Background

The clinical diagnosis of DVT of the lower extremities cannot be established with certainty without objective testing. Contrast venography is invasive, requires contrast media, and is no longer appropriate as the initial diagnostic test for the evaluation of symptoms that suggest acute DVT. The proven utility of noninvasive technology, including IPG and compression US, as well as increasing experience with MRI, have rendered CV much less popular. Nonetheless, venography remains the gold standard test. The availability of and familiarity with certain technology may influence the diagnostic approach. The specific clinical scenario impacts on the diagnostic algorithm that is chosen. For example, while IPG and US are reliable for the diagnosis of symptomatic proximal DVT (involving the popliteal and/or more proximal veins), they are much less reliable for recognizing asymptomatic DVT. The sensitivity of certain diagnostic tests is influenced by thrombus location. Thrombi located between and including the popliteal and the iliac veins are the easiest to locate, and those above the iliac veins and in the calf veins are more elusive. The diagnosis of recurrent DVT remains a challenge. The D-dimer test has been evaluated in the setting of both acute DVT and acute PE and is discussed in the section, THE DIAGNOSTIC APPROACH TO ACUTE PULMONARY EMBOLISM, below. Currently available diagnostic modalities are reviewed, followed by recommendations for their use. Diagnostic algorithms are then presented. The following clinical scenarios are considered in the context of each diagnostic test: (1) symptomatic proximal DVT, (2) asymptomatic proximal DVT, (3) calf DVT, (4) recurrent and chronic lower extremity DVT, and (5) upper extremity venous thrombosis.

Symptoms and Signs

Innumerable clinical investigations have established that DVT cannot be reliably diagnosed on the basis of the history and physical examination, even in high-risk patients (5). Patients with lower extremity DVT often do not exhibit erythema, warmth, pain, swelling, or tenderness. When five clinical studies were compared, for example, the sensitivity of calf pain for acute DVT varied from 66 to 91% and the specificity varied from 3 to 87% (6). In six studies that included evaluation for calf tenderness, the range for sensitivity was 56 to 82%, and the range for specificity was 26 to 74%. For Homans' sign, the sensitivity varied from 13 to 48%, and the specificity from 39 to 84% (6). Swelling of the calf or leg as a marker was also inconsistent, with the sensitivity ranging from 35 to 97% and the specificity from 8 to 88% (6). When present, however, these findings merit further evaluation despite their lack of specificity. Thus, the clinical evaluation may imply the need for further evaluation but cannot, by itself, be relied on to confirm or exclude the diagnosis of DVT. The presence of risk factors for DVT should always be rigorously scrutinized. The clinical examination and laboratory testing have been reviewed elsewhere (5). Our focus is on the diagnostic approach once DVT is clinically suspected but also includes the asymptomatic high-risk patient. Objective testing is also necessary to diagnose recurrent DVT.

Contrast Venography

While CV remains the gold standard technique for the diagnosis of symptomatic DVT, it is rarely performed because of the accuracy of noninvasive testing. Venography should be performed whenever noninvasive testing is nondiagnostic or impossible to perform. The technique of Rabinov and Paulin has been used consistently (8). Contrast venography has been considered nearly 100% sensitive and specific provided it is technically adequate and that strict diagnostic criteria are adhered to. Level 1 studies have not been performed because CV has been established, by default, as the gold standard test. Adequate CV requires complete visualization of the deep venous system, from the calf to the pelvic veins and inferior vena cava. The most reliable criterion for the diagnosis of acute DVT is a constant intralumenal filling defect evident in two or more views (8). An abrupt cutoff of a deep vein is another reliable criterion but requires cautious interpretation in patients with previous DVT. Other criteria such as nonvisualization of deep veins (may be clarified with injection of more contrast material), venous collaterals, or nonconstant intralumenal filling defects are less reliable and should not be used to confirm the diagnosis of acute DVT.

For symptomatic proximal DVT, CV is extremely sensitive and specific but noninvasive tests are more appropriate for first-line testing. Although CV is also sensitive for asymptomatic proximal DVT, it is generally not utilized as a screening test except in clinical trials. Venography appears to be the most sensitive test for calf DVT. The diagnosis of recurrent lower extremity venous thrombosis has proven challenging. It can be difficult to visualize a constant intralumenal defect with CV when veins have been thrombosed previously. Venography has been considered the gold standard technique for upper extremity thrombosis, but other modalities, such as US, are generally attempted first.

Disadvantages of CV include invasiveness, which may result in phlebitis or hypersensitivity reactions; however, it is generally safe and accurate. It may be painful, and poor venous access may make the test difficult or impossible to perform. Deep venous thrombosis may occasionally result from the procedure. Direct toxicity of the contrast agent may result in nausea and vomiting, flushing, nephrotoxicity, or cardiotoxicity. Nephrotoxicity is generally manifested by transient renal failure. Idiosyncratic reactions are not dose related and include urticaria, angioedema, bronchospasm, and cardiovascular collapse. Venography is more expensive than IPG or US but the cost varies among different institutions. Thus, CV has its limitations (3).

Relative contraindications to CV include acute renal failure, and chronic renal insufficiency with a creatinine level greater than 2 to 3 mg/dl. Idiosyncratic reactions may be minimized with antihistamines and corticosteroids. Arterial insufficiency is a relative contraindication in view of the possibility of extravasation of contrast with resultant cellulitis and the potential for tissue necrosis. Advantages and disadvantages of CV are outlined in Table 1.

Impedance Plethysmography

Background. Impedance plethysmography was developed in 1969 and has been extensively investigated in a number of prospective clinical trials, mainly from Canada and Europe (9). Compared with other diagnostic tests for DVT, it takes less technical training, is less expensive, and is portable. This technique detects increased venous outflow resistance in the deep veins of the proximal lower extremities. Impedance plethysmography has been compared with CV in consecutive symptomatic patients with suspected proximal DVT in a number of clinical investigations. Clinical trials have been conducted to establish the sensitivity of serial IPG and outcome in patients in whom the initial IPG was negative. Despite extensive outcome data, this diagnostic modality is less commonly used today, with US being more widely employed for evaluating suspected acute lower extremity DVT. An overview of the technique of IPG including its limitations, as well as results of clinical trials in both symptomatic and asymptomatic patients conducted, are presented.

Physiology and technique. Impedance plethysmography is a sensitive method for evaluating the rate of venous return from the lower extremity. The test relies on the principle that the volume of blood in the leg affects its ability to conduct an applied electrical current, which is inversely proportional to the impedance between two electrodes placed along the calf. To conduct the test, a small electrical current is passed between one set of electrodes, while the second measures changes in voltage. A cuff is inflated around the thigh to obstruct venous outflow but not arterial inflow. As blood accumulates in the leg below the cuff, impedance between the calf electrodes falls. When venous pressure builds to the point that it equals that of the cuff, venous outflow is reestablished, and the tracing plateaus. The sudden release of cuff pressure results in a sudden surge of blood flow proximally (the blood volume of the leg decreases), resulting in a rapid increase in impedance. If DVT is present in any major vein draining the lower extremity (from the popliteal to the iliac veins) the rate of venous emptying (and the increase in impedance) is significantly slower, and the tracing reveals a slower than normal return toward baseline. This technique is insensitive to thrombi that do not decrease the rate of venous outflow, such as most calf thrombi and small, nonobstructing thrombi in the proximal veins. Other causes of slow venous outflow, such as elevated central venous pressure, may yield bilateral false-positive results on IPG.

It has been demonstrated that the sensitivity and specificity of IPG for detecting proximal DVT are both dependent on adhering to the validated protocol (13). This includes careful leg positioning to avoid compression of the popliteal or femoral veins. The occlusive cuff is inflated to 45 cm of water for 45 s and is rapidly released. The impedance rise at the end of the occlusion is plotted against the fall as recorded 3 s after cuff release. If an equivocal or abnormal result is obtained, the procedure is repeated for 45 s of occlusion, then 120 s of occlusion, then 45 s and again 120 s. A result in the normal range terminates the sequence. A validated graph for plotting the results should be used. Results using computerized IPG devices have not been validated. It is important to emphasize that if different protocols are utilized, a degree of institution specificity will be imparted that may contribute to differences in results. The McMaster investigators, for example, have validated criteria for the technique based on their early results (13). Impedance plethysmography should be performed with standardized and commercially available equipment by trained technicians. Studies not utilizing such a protocol or those employing nonvalidated devices should not be used.

Limitations of impedance plethysmography. Certain limitations of IPG must be considered. The technique does not distinguish between venous obstruction due to DVT and that caused by nonthrombotic entities. Correct positioning of the legs is important to avoid obstructing venous outflow. Potential causes of false-positive results include increased intrathoracic pressure, increased intraabdominal pressure, and decreased venous return from the lower extremities, such as might occur owing to obstruction of blood flow by tumor. False-positive results can also result from poor arterial inflow such as low cardiac output states or severe peripheral vascular disease. These conditions are outlined in Table 2. Advantages and limitations of IPG are outlined in Table 3. Potential limitations of the sensitivity of serial IPG in symptomatic patients are discussed below.

Early clinical trials: Establishing accuracy in symptomatic acute proximal DVT. Impedance plethysmography has the distinct advantage of outcome data that are available from large, prospective clinical investigations. Early studies (1976 to 1982) revealed sensitivities of 92 to 98% for symptomatic proximal DVT with confirmation using CV (9, 14), although one early study revealed a sensitivity of only 81% (10). The poor sensitivity for calf DVT was soon established (sensitivity approximately 20%) (9, 16) although it appears that embolization from calf DVT is unlikely unless proximal extension occurs (21). Sensitivity and specificity values from clinical trials comparing IPG with CV in consecutive patients with symptomatic, suspected DVT and with independent interpretation of each study are included in Table 4 (9, 14, 16, 20). Subsequent developments included the use of a computerized IPG device (22) and clinical trials comparing IPG with US in outpatients and hospitalized patients (23, 24). These are discussed below. A major advance was the realization that serial IPG determinations over a 10- to 14-d period could detect extension of calf vein thrombi into the proximal veins, which necessitates treatment (25). While such extension may be the reason that IPG studies become positive during serial testing, it has also been suggested that some degree of undetectable extension into the proximal veins may have already occurred at the time of the initial test.

Management studies: Early success and later questions. Determining clinical outcome without anticoagulation in patients with negative serial IPG studies has been crucial in evaluating the validity of the technique. These management trials were conducted to prove that it was safe to withhold treatment in patients with suspected DVT if serial IPG studies remained negative during the 10- to 14-d study period. The importance of serial testing after an initially negative test is emphasized by five clinical trials revealing a conversion rate to positive of 1.4 to 19% (25) with the combined rate of conversion being 89 of 1,637 patients (5.4%) (Table 5). Although most such conversions occur during the first 3 d, some patients will take up to 2 wk to develop a positive test. The precise sensitivity of serial testing could not be determined because patients with persistently normal IPG studies did not undergo CV. However, in the five studies noted above, subsequent DVT or PE (no fatal PE) was documented in a maximum of only 2.5% of patients with normal serial IPG. Unfortunately, in another clinical trial, four patients died of PE after having normal serial IPG (30). In this trial, serial IPG testing was performed in 311 patients with clinically suspected DVT in whom initial IPG testing was normal. Four patients (1.3%) developed fatal PE despite the normal serial tests. There are several possible explanations for the poor outcome (31). These investigators used a protocol and a computerized device that differed from that validated by the McMaster group. It is conceivable that equipment or other technical factors may have played a role. Unvalidated protocols are not acceptable and computerized IPG devices cannot be considered appropriate at the present time. It is also possible that the four deaths were chance occurrences. The sensitivity of IPG in this study was 86%. Serial IPG studies have been compared with serial US in patients with suspected, symptomatic acute DVT and an initially negative study (see COMPRESSION ULTRASOUND WITH VENOUS IMAGING, below).

Additional concerns arose from the results of a retrospective clinical trial conducted by one of the McMaster groups (Henderson General Hospital), a group experienced with IPG. Anderson and associates (19) performed CV (or compression ultrasound in a minority of patients) in patients with abnormal IPG results, in those with normal IPG testing in whom DVT was highly suspected, and in those in whom serial IPG testing would be difficult. Impedance plethysmography was abnormal in only 37 of 56 patients with confirmed DVT (sensitivity, 66%). Of the 19 proximal DVT not detected by IPG, 12 (63%) were occlusive and 11 (58%) involved at least the popliteal and superficial femoral veins. Thus, these investigators reported a lower sensitivity for IPG at their center than had been previously reported in symptomatic outpatients. Although consecutive patients had been enrolled, the study was retrospective. Further studies were indicated.

The same investigators, together with the group from Padua, Italy, then prospectively compared IPG and US in 495 symptomatic outpatients with suspected DVT, using CV as the definitive answer (24). The prevalence of DVT was 130 of 495 (27%). Of these, 109 of 130 (84%) were proximal. Overall, the sensitivity of IPG was 77% and the specificity was 93%, compared with 90 and 98%, respectively, for US. There were significant differences in sensitivity and specificity between the two centers as a consequence of differences in size and location of thrombi. The majority of proximal thrombi not detected by IPG and US involved less than 5 cm of the distal half of the popliteal vein and most of these thrombi occurred at one center (Hamilton). Exclusion of these thrombi from the analysis increased the sensitivity of US for proximal thrombi to 86 of 87 (99%) and improved the sensitivity of IPG to 72 of 79 (91%). The positive predictive value of US was strongly influenced by the number of abnormal venous segments. A higher prevalence of patients with less extensive, less occlusive thrombi at the Hamilton center appeared to be a factor in the difference in sensitivity.

Ginsberg and colleagues (20), another McMaster group (Chedoke-McMaster Hospital), also elected to reevaluate prospectively the sensitivity of IPG for proximal DVT as well as to relate the location and size of thrombi to the IPG result. Clinically suspected DVT in 132 consecutive patients was evaluated with IPG and 118 of these patients underwent CV. The other 14 patients underwent US and were felt to be definitively diagnosed with proximal DVT. Of the 132 patients, 40 (30%) had proximal DVT, 7 (5%) had calf DVT, and 85 (64%) did not have DVT. The sensitivity of IPG for proximal DVT was 65% and the specificity was 93%. Of the proximal vein thrombi, IPG detected 3 of 13 popliteal thrombi (23%) not involving the superficial femoral vein and 23 of 27 thrombi (85%) involving the superficial femoral vein. Changes in referral patterns may have resulted in more patients with less severe symptoms and smaller, less occlusive thrombi being referred (20). Potential explanations for the lower sensitivity include biases that may have inflated sensitivities from earlier studies such as repeated IPG testing prior to CV, and inclusion of patients with a known abnormal IPG in the study population (31). It has been suggested that modern CV techniques may detect early thrombi that would have been previously overlooked (32, 33). A shift in the referral pattern to patients with less extensive, less occlusive thrombi as well as heightened awareness of DVT and improved availability of testing facilities are explanations that have been given substantial credence (20, 24, 31, 32). Ginsberg and colleagues (20) recommended that, on the basis of their results, patients with a high clinical likelihood of DVT but a normal initial IPG should undergo US or CV instead of serial IPG. It has been argued that, on the basis of clinical outcome trials (34), the latter approach has not been proved necessary.

Impedance plethysmography in asymptomatic patients. The diagnostic accuracy of IPG in asymptomatic patients has been evaluated in a number of clinical trials, predominantly in patients undergoing total hip replacement or surgery for hip fracture (35). The sensitivity for proximal DVT has ranged from 12 to 64% and was less than 30% in three of these studies. In 106 asymptomatic patients undergoing IPG, US, and CV after total hip or total knee replacement, the sensitivity for IPG was 41.2% for proximal thrombi compared with 64.7% for US (42). Impedance plethysmography was also insensitive to calf vein thrombi in these patients. The low sensitivity of IPG in these asymptomatic patients may be attributed to the fact that the thrombi are often smaller and less likely to be occlusive (43). Agnelli and associates (22) utilized serial computerized IPG in 246 asymptomatic patients with a negative initial IPG undergoing elective total hip replacement or surgery for hip fracture. The sensitivity and specificity for DVT were 22 and 87% in the operated leg and 14 and 95% in the nonoperated leg, respectively. The same investigators (41) subsequently determined that there was a significantly higher proportion of proximal DVT (p = 0.001), a significantly higher Marder score (index of thrombus size) (p = 0.0001), and a significantly higher proportion of occlusive DVT (p = 0.001) in symptomatic patients than in asymptomatic patients. Screening for DVT in asymptomatic high-risk patients has not proved useful (see COMPRESSION ULTRASOUND WITH VENOUS IMAGING, below).

Recurrent and chronic deep venous thrombosis. The clinical diagnosis of recurrent DVT is nonspecific (44, 45). Impedance plethysmography may be especially useful to diagnose recurrence, since positive findings revert to normal as the DVT resolves and/or collateral circulation develops. Resolution rates for IPG-documented acute proximal DVT at 3, 6, 9, and 12 mo have been found to be 67, 85, 92, and 95%, respectively (46). Thus, IPG appears to be reliable in diagnosing recurrent DVT when the previous episode is more remote. Impedance plethysmography is not useful for early recurrences unless IPG normalization has been documented.

Compression Ultrasound with Venous Imaging

Background. Ultrasound has been studied extensively in the setting of suspected acute DVT as well as for screening asymptomatic patients deemed at high risk for acute DVT. Compression ultrasound with venous imaging (real-time B-mode imaging) is noninvasive, widely available, and has been proved accurate for diagnosing acute, symptomatic proximal DVT. In contrast to Doppler venous flow detection, which only offers information regarding blood flow, real-time sonography permits a two-dimensional cross-sectional representation of the lower extremity veins. The combination of these two techniques is termed duplex ultrasound. Ultrasound technology has been advanced by the development of color duplex instrumentation that displays Doppler frequency shifts as color superimposed on a gray-scale image. Color duplex images display both mean blood flow velocity, expressed as a change in hue or saturation, and direction of blood flow as displayed as red or blue. Among the useful features of US imaging techniques are the ability to identify pathology other than DVT. Baker's cysts, superficial or intramuscular hematomas, lymphadenopathy, femoral artery aneurysm, superficial thrombophlebitis, and abscesses may be suggested or diagnosed (47). Advantages and disadvantages of US imaging are listed in Table 6.

Technique. Most medical centers utilize a combination of gray-scale, duplex, and color Doppler imaging. The technique requires a 3- to 7.5-MHz real-time transducer. The patient is positioned supine with the leg slightly externally rotated. The reverse Trendelenburg position may facilitate the examination by increasing venous distention. The compression technique is used, beginning at the inguinal ligament, and the common femoral vein and greater saphenous vein are evaluated. Radiologists frequently identify the vein below the bifurcation of the common femoral vein as the superficial femoral vein. This may be confusing because the superficial femoral vein is actually a component of the deep venous system. (The term "femoral vein" has replaced "superficial femoral vein," emphasizing its importance.) The deep femoral vein is evaluated at the bifurcation of the common femoral vein but cannot generally be visualized along its entire length. The prone or lateral position may aid in evaluating the calf and popliteal veins and the popliteal should be scanned at least to the level of the venous trifurcation, or 10 cm below the midpatellar point. Compression is applied with the transducer at short intervals over the entire length of the vessels. The pressure applied should be enough to indent the skin but not enough to compress arterial flow. This will allow complete compression of the normal opposing venous walls. Certain areas of incomplete compressibility (greater saphenous vein and common femoral vein juncture and superficial femoral vein at the adductor canal) may exist in the absence of DVT. Doppler studies can be used to confirm the presence of spontaneous venous flow. Respiratory phasicity and cessation of flow with the Valsalva maneuver offer indirect evidence of abdominal and pelvic venous patency. Color imaging appears to offer a superior evaluation of flow than can be achieved with duplex scanning. Nonocclusive thrombi may be more easily documented with color flow imaging, and calf vein evaluation and studies in obese patients are generally more easily achieved with this technique. Compression US with venous imaging (real-time B-mode imaging), duplex US, and color Doppler all rely on compression, at least to some degree, for the diagnosis of DVT. While there are differences between the techniques, a clear advantage of one over another has not been demonstrated in prospective clinical trials as long as compression is used. Color Doppler energy (power Doppler) has been utilized in the evaluation of thrombotic disorders. The color map in the power Doppler display shows the integrated power of the Doppler signal, which is related to the number of red blood cells producing the Doppler shift. Power Doppler imaging is more sensitive for the detection of low-amplitude, low-velocity flow than color Doppler and is relatively Doppler angle independent. However, power Doppler provides no velocity or directional information and is motion sensitive. This technique has proved valuable in other vascular US imaging applications and could prove useful for imaging patients with DVT to assess early recanalization or nonocclusive thrombus. However, no clinical trials have been performed to assess power Doppler in patients with DVT. Criteria for the diagnosis of acute DVT using US imaging are listed in Table 7.

There has been controversy over the necessary extent of the US examination. When the (superficial) femoral vein is not evaluated, diagnostic efficacy may be reduced, perhaps to a clinically significant extent (48). In a study by Frederick and coworkers (48), six cases of isolated superficial femoral venous thrombosis were missed with an abbreviated protocol, amounting to 4.6% of the DVT diagnosed. It has been suggested by others that US evaluation from the inguinal ligament to the calf veins is not necessary. Pezzullo and colleagues (49), retrospectively evaluated 160 US examinations in 155 symptomatic patients and found 146 cases of proximal thrombosis. In 145 cases (99%), either the common femoral or popliteal vein was involved. In the other 14 of 160 cases (9%), isolated calf vein thrombosis was diagnosed. The limited examination decreased the examination time by 9.7 min, or 54%. More recent data have suggested an excellent outcome with US performed serially over 1 wk for suspected DVT, using a limited examination (also see SYMPTOMATIC ACUTE PROXIMAL DEEP VENOUS THROMBOSIS, below). In addition to controversy over the limited examination, the issue of unilateral versus bilateral studies in the setting of unilateral symptoms is debated (50). The unilateral examination has been reported to decrease scanning time and cost, without a decrease in diagnostic yield (51). Naidich and associates (52) evaluated 245 patients with unilateral symptoms and determined that 180 had no DVT, 44 had ipsilateral DVT, 18 had bilateral DVT, and 3 had contralateral DVT. While it was argued that this supported the bilateral examination, the incidence of contralateral DVT was low (1%) and the presence of bilateral DVT has not been proved to have more impact on outcome than unilateral DVT.

Limitations of compression ultrasound with venous imaging. Venous compressibility may be limited by patient characteristics such as obesity, edema, and tenderness as well as by casts or immobilization devices that limit access to the extremity. While there may be areas of focal noncompressibility in these situations, these areas are generally bilaterally symmetric and color flow imaging will usually reveal venous filling. Other potential causes of false-positive results include extrinsic compression of a vein by a pelvic mass or other perivascular pathology (47) and thrombosis in the distal popliteal vein. False-negative studies may occur in the presence of calf DVT, with proximal DVT in asymptomatic (even high-risk) patients (53) or in the presence of a thrombosed duplicated venous segment. Ultrasound techniques are unreliable in detecting DVT in the iliac veins; CV and MRI are much more reliable in this setting (54). Finally, because US may not return to normal after acute DVT has been diagnosed, it must be interpreted with caution when attempting to diagnose recurrent DVT (Table 8). These limitations are discussed further in the following two sections.

Symptomatic acute proximal deep venous thrombosis. A number of clinical trials have suggested the accuracy of US in diagnosing suspected, acute DVT. A large retrospective outcome trial in which anticoagulation was withheld in the setting of negative US revealed only five episodes of VTE in 1,022 symptomatic patients (57). Two patients developed fatal PE more than 3 mo after the initial event. Prospective clinical trials in which consecutive patients have been evaluated and in which real-time B-mode compression US has been compared with CV in an independent, blinded manner have been useful in confirming the accuracy of the technique in patients with suspected acute DVT (42, 58). Other clinical trials utilizing the same technique have been conducted less rigorously, in that consecutive enrollment of patients is not documented (63). These level 1 and level 2 studies are shown in Table 9. The duplex and color-flow techniques have been evaluated in similar studies and the results are similar to the above trials. Level 1 (69) and level 2 (73) studies for duplex US are shown in Table 10, with studies for color-flow Doppler in Table 11 (77). Studies in which independent, blinded readings of the diagnostic studies were not performed or not specified were not evaluated (83). When US is negative in patients with suspected DVT, serial US has proved to be a sensitive means by which to detect proximal extension of calf DVT in symptomatic outpatients. Heijboer and colleagues (29) found that when serial compression US remained negative (Days 1, 2, and 8), the incidence of VTE during the 6-mo follow-up period was only 1.5%, compared with 2.5% for serial IPG. These investigators examined only the common femoral and popliteal veins.

In a more recent clinical trial, Birdwell and associates (86) evaluated 405 consecutive outpatients with a suspected first episode of acute DVT. If the simple compression US (common femoral from inguinal line to bifurcation, and popliteal vein from proximal popliteal fossa to a point 10 cm distal to the midpatella) was normal, anticoagulation was withheld regardless of symptoms and the test was repeated 5 to 7 d later. The initial US was normal in 342 patients and 7 of these patients developed an abnormal study during the serial follow-up. The initial US was abnormal in 63 patients. Over the 3-mo follow-up period, 2 of the 335 patients (0.6%) with normal serial US studies, from whom treatment had been withheld, developed VTE while 4 of the 70 patients (5.4%) with either an initially abnormal or subsequently abnormal study (treated) developed recurrent VTE. None of the patients in the study died from acute PE.

Similarly, Cogo and colleagues (87) evaluated the safety of withholding anticoagulation in patients with suspected DVT when compression US was initially negative and remained negative at repeat testing 1 wk later. A simplified compression US procedure limited to the common femoral vein in the groin and the popliteal vein down to the trifurcation of the calf veins was also performed in this study. Of the 1,702 patients included, US was abnormal in 400 patients initially and in 12 patients at 1 wk. Venous thromboembolic complications occurred in only one patient during the week of serial testing and in eight patients during the 6-mo follow-up period. It is important to note that although the extended popliteal examination did allow for the earlier identification of patients with proximal DVT, the procedure resulted in more false-positive results. The positive predictive value for the assessment of the common femoral vein and the popliteal vein in the popliteal fossa was 98.5%, but decreased to 79% for the distal popliteal region. Thus, it appears safe to withhold anticoagulation in patients in whom one or two serial US (including distal popliteal scanning) are negative over 5 to 7 d. The studies described above (86, 87) suggest that a single repeat study at 5 to 7 d is adequate if the initial study includes the femoral vein, the popliteal fossa, and scanning to 10 cm below the midpatella or to the trifurcation of the calf veins. When patient follow-up cannot be guaranteed or in centers in which US has not proved sufficiently reliable, these serial US protocols should not be utilized.

Asymptomatic acute proximal deep venous thrombosis. As is the case with IPG, real-time B-mode US, duplex US, and color-flow Doppler US have been used as surveillance techniques to evaluate asymptomatic patients at high risk for DVT. They have proved insufficiently sensitive in this setting. Without prophylaxis, the risk of DVT is approximately 50% after total hip replacement and as high as 65% after total knee replacement (89). Prospective clinical trials enrolling consecutive patients and using previously established objective criteria for CV and US with independent, blinded comparisons of the two techniques were assessed (level 1 trials) (43, 68, 90). Other studies were deemed level 2 (76, 99). Of the 11 level 1 studies, 5 used real-time B-mode US, 4 utilized duplex US, and 2 were color Doppler studies. When level 1 studies were considered, US had a sensitivity of 62% (95 of 144 patients), a specificity of 97%, and a positive predictive value of 66% for detecting proximal DVT. For level 2 studies, the sensitivity was 95%, the specificity was 100%, and the positive predictive value was 100%. Asymptomatic patients undergoing orthopedic surgery have been scrutinized by metaanalysis and although duplex and color Doppler imaging may have theoretical advantages compared with B-mode imaging, this has not been clearly demonstrated (53). It is likely that the lower sensitivity of US in asymptomatic high-risk orthopedic patients occurs because thrombi in asymptomatic patients are smaller, fresh, and more easily compressible and nonocclusive. Outcome data evaluating US screening in these asymptomatic patients are now available. In one double-blind, randomized, controlled trial involving 1,024 elective total hip or knee arthroplasty patients receiving warfarin prophylaxis (and asymptomatic for DVT), screening US was performed at discharge (103). In patients in whom DVT was detected, warfarin was continued at a therapeutic dose, while in those with negative studies, it was discontinued. The total outcome event rate (venous thromboembolism plus bleeding) at 90 d was 1% for each group. In a large, prospective, Canadian clinical trial of 1,984 consecutive hip or knee arthroplasty patients receiving enoxaparin prophylaxis, predischarge compression US revealed only 3 patients (0.15%) with DVT (104). These results suggest that a screening US at discharge in high-risk orthopedic patients receiving enoxaparin or warfarin prophylaxis is unnecessary.

Acute calf deep venous thrombosis. When acute DVT is suspected, one or both lower extremities are evaluated. A search specifically for isolated calf DVT is not generally undertaken since the proximal lower extremity is also evaluated in the setting of suspected calf DVT. However, it is useful to discuss the sensitivity and specificity of US for calf DVT since this entity is either treated or followed with serial noninvasive studies. Contrast venography has been considered the most accurate diagnostic test for acute calf DVT. As is the case with IPG, US cannot be relied on to exclude calf vein thrombosis. As noted above, serial US (or IPG) is appropriate in patients with symptoms of acute DVT and a negative initial study (86, 87), and some of these patients may have undetected calf DVT, which can be assessed (for possible extension) at follow-up. If, in a particular patient with suspected DVT, the initial US (or IPG) is negative and follow-up with serial studies cannot be guaranteed, then CV would be appropriate. Ultrasonography is specific for symptomatic acute calf vein DVT, and a positive test in this setting can usually be relied on. These recommendations are based on level 1 studies. Because the calf veins are smaller and characterized by slower flow, and because they are more anatomically variable than the proximal lower extremity veins, US assessment is more difficult. Technically inadequate studies result more commonly than when the proximal veins are examined. In symptomatic patients with isolated calf DVT, the sensitivity of US has been shown to be 73% for compression US (65), 81% for duplex US (71), and 87% with color-flow Doppler (78). In each of these prospective studies, independent, blinded readings were performed for both US and CV. Except for the unclear question of consecutive enrollment in the compression US study (65), level 1 methodology was employed in these investigations. When the calf veins can be adequately visualized, the sensitivity and specificity are improved and range from 88 to 100% and from 83 to 100%, respectively (63, 81, 105, 106). Because of the above-described technical considerations, however, the sensitivity is frequently much lower (67). The sensitivity of US for detecting isolated calf DVT in asymptomatic high-risk patients is even lower, ranging from 33 to 58% (91, 93, 97, 107). Clinical investigations evaluating US techniques for calf DVT are both level 1 and level 2. While many of the level 2 studies are otherwise well designed, frequently it is not explicit that consecutive patients were enrolled.

Recurrent and chronic deep venous thrombosis. Distinguishing between acute and chronic DVT is crucial because after several weeks thrombi become adherent to the wall of the vein and are not likely to embolize. When patients present with recurrent symptoms, some will have recurrent DVT and others will have postphlebitic syndrome. Ultrasound techniques should not be considered reliable for recurrent DVT unless the test has been shown to normalize prior to the suspected recurrence. However, the rate of normalization of an abnormal US test after a first episode of acute DVT has been determined to be only 44 to 52% after 6 mo and 55% after 12 mo in two prospective follow-up clinical investigations (108, 109). Clot echogenicity does not accurately discriminate between acute and chronic DVT, but there does appear to be a positive correlation between venous distention and the age of the thrombus (109). However, a study of 975 legs of patients with suspected DVT evaluated vein diameter in normal veins and in those with acute and chronic thrombosis (110). It was concluded that although veins involved by acute DVT tend to be larger than normal veins, and veins with chronic changes tend to be smaller than normal vessels, the mean differences are small. The differences appear to be most useful at the extremes of size. Thus, when evaluating a patient with suspected acute DVT, vein size should be interpreted in the context of other sonographic findings. Because previous DVT is a risk factor for recurrence, it may be appropriate to perform a follow-up US between 3 and 6 mo after anticoagulation is initiated, to serve as a baseline in the event that symptoms recur. There is, however, no uniformly accepted standard of care for repeating US after DVT is diagnosed.

Upper extremity deep venous thrombosis. Axillary-subclavian vein thrombosis commonly results from indwelling venous catheters but may be spontaneous, including the syndrome of "effort thrombosis." The diagnosis may be made by US, CV, or MRI. When US is utilized, criteria for the diagnosis are the same as in the lower extremities. Although compression techniques are employed, portions of the subclavian vein behind the clavicle cannot be compressed and greater reliance on Doppler evaluation is required. The internal jugular, subclavian, axillary, and brachial veins are generally evaluated. The superior vena cava and brachiocephalic vein are inaccessible or only partially accessible to US. While the sensitivity of US for symptomatic upper extremity thrombosis may range from 78 to 100% (111, 112), it has been shown to be as low as 31% in asymptomatic individuals after subclavian catheter removal (113). Most of the false-negative studies appeared to be due to short, nonocclusive thrombi. In a prospective study of 58 consecutive patients with suspected upper extremity DVT, central venous catheters, thrombophilic states, and previous leg DVT were significantly associated with upper extremity thrombosis (114). All patients were evaluated by objective testing for PE. Pulmonary embolism was detected in 36%. Thus, it appears that PE occurs in a substantial proportion of these patients.

Magnetic Resonance Imaging

Preliminary reports using MRI to detect DVT suggested that MRI was at least 90% sensitive and specific for acute symptomatic proximal DVT (115). A number of prospective clinical trials have evaluated MRI, using CV as the gold standard (Table 12), with several revealing sensitivity and/or specificity values as high as 100%. Less information is available for MRI as a screening modality in asymptomatic patients. It has been suggested that silent lower extremity DVT may be demonstrated with MRI (118). This is logical since MRI directly images thrombi, and can image nonocclusive clots. It does not rely on compression or other adjunctive techniques. However, in calf DVT, a change or lack of change on the images during compression from above and below may be useful (54). Magnetic resonance imaging appears less sensitive than CV for calf DVT (54, 119) but no level 1 studies with large numbers of calf DVT have been performed. Magnetic resonance imaging has also been compared with compression US for the evaluation of acute DVT (120).

There are a number of potential advantages of MRI (Table 13) and the technique is evolving. Preliminary studies suggest excellent sensitivity and specificity not only for thigh DVT, but also for acute pelvic vein thrombosis (54, 118, 121, 122). Pelvic DVT may be difficult to evaluate by US and even by CV. Although CV and IPG are accurate for iliac vein DVT, MRI may prove to be the superior test for noniliac pelvic vein thrombosis. Studies have validated the use of gradient echo "white blood" (blood imaged as a brighter intensity against a relatively darker background) MRI for the detection of DVT. Such images may be supplemented by spin echo or fast spin echo "black blood" images, but the latter are not recommended for primary diagnosis. Imaging should be performed in the axial plane and interpretations should be based on review of source images rather than reprojections. As the MRI study is performed, the attending radiologist can carefully scrutinize areas of suspected abnormality by using different techniques, and the success of the technology depends on the active involvement of an experienced radiologist. Distinguishing acute from chronic DVT is a potentially advantageous feature of MRI. Criteria that may suggest chronic DVT have also been used for CV and include irregular wall thickening in the presence of collateral veins, and a diminutive lumen (121). Erdman and colleagues (118) have suggested that inflammation surrounding a thrombosed vessel indicates acute DVT, while the absence of edema suggests more chronic DVT. Such criteria require validation.

Magnetic resonance imaging appears useful in evaluating upper extremity venous thrombosis (118, 123), although large comparative trials with CV have not been performed. The opportunity to diagnose nonthrombotic conditions by MRI is attractive. Diseases that have been diagnosed by MRI in patients with suspected DVT include cellulitis, edema, varices, hematomas, superficial phlebitis, joint effusions, myositis, and adenopathy (118). Other advantages of MRI include noninvasiveness, lack of operator dependence (although reader experience is necessary), and the ability to scan patients without intravenous access or contrast. Finally, preliminary (level 2) studies suggest that MRI is promising for the diagnosis of PE, so that it may be the first technique enabling both the lungs and the lower extremities to be evaluated for clot at the same time (124, 125).

There are disadvantages of MRI. Patients must be carefully screened for contraindications to MRI, particularly with regard to metallic devices from injury or surgery. Other potential contraindications include significant claustrophobia, the inability to cooperate, and massive obesity. Although MRI is available at all large hospitals, it may not be at smaller instititutions, and reader expertise is crucial. It is relatively expensive, although cost-benefit analyses need to be performed. Multicenter, randomized clinical trials would be useful. An algorithm for the diagnostic approach to symptomatic, suspected acute DVT is presented in Figure 1.

View larger version (13K): [in this window] [in a new window]   Figure 1.   Diagnostic algorithm for patients with symptoms suggesting acute deep venous thrombosis. The recommended diagnostic approach allows for some flexibility depending upon the resources at a particular institution. First episode of suspected (symptomatic) acute lower extremity DVT. Duplex or color-flow US may be utilized, although there is no proven superiority over compression US alone. Scenarios associated with false-positive IPG studies must be realized. Because the sensitivity of US or IPG for calf DVT is lower than for proximal DVT, three studies over 7 to 14 d (IPG) or one to two studies over 5 to 7 d (US) are needed to detect proximal extension (see text). Outcome study results from large clinical trials may not necessarily be applicable at all medical facilities. If a patient cannot return for a repeat study, if iliac thrombosis is suspected, or if an urgent diagnosis is deemed necessary, a more rapid evaluation using venography (or MRI) should be undertaken. § Although MRI appears accurate for lower extremity DVT (level 2 studies), it is more expensive than US or IPG and is generally not indicated as the initial diagnostic test.

	THE DIAGNOSTIC APPROACH TO ACUTE PULMONARY EMBOLISM

TOP INTRODUCTION CONTENTS INTRODUCTION THE DIAGNOSTIC APPROACH TO... THE DIAGNOSTIC APPROACH TO... THE DIAGNOSTIC APPROACH TO... THE DIAGNOSTIC APPROACH TO... THE FUTURE REFERENCES

Background

As with DVT, the diagnosis of PE cannot be established with certainty without objective testing. Suspicion of the diagnosis, based on the presence of risk factors and frequent, but nonspecific, clinical findings should lead to a thorough diagnostic evaluation that leads to either confirmation or exclusion of PE.

Symptoms and Signs

It is well established that PE cannot be unequivocally diagnosed solely from the history and physical examination and this is underscored by the frequent failure to make the diagnosis antemortem (126, 127). While certain symptoms are common, and may serve as important clues, the lack of specificity mandates additional testing when the clinical presentation is consistent with PE. Pulmonary embolism should be considered whenever unexplained dyspnea occurs. Dyspnea with or without associated anxiety, as well as pleuritic chest pain and hemoptysis, are common in PE, but are nonspecific, and one or more of these symptoms may develop with pneumothorax, pneumonia, pleuritis, exacerbations of chronic obstructive lung disease, congestive heart failure, or lung cancer. Tachypnea and tachycardia are the most common signs of PE but are nonspecific. Lightheadedness and syncope may be caused by PE but may also result from a number of other entities that result in hypoxemia or hypotension. Pulmonary embolism should always be suspected in the setting of syncope or sudden hypotension and these often indicate a large clot burden. The cardiac and pulmonary physical examinations are both nonspecific for PE. The index of clinical suspicion does, however, become a more useful parameter when considered in conjunction with ventilation-perfusion (/) scanning (128). Diagnostic efforts directed at possible PE may be appropriate despite alternative explanations if risk factors and the clinical setting are suggestive. Dyspnea, tachypnea, clear lung fields, and hypoxemia may be attributed to a flare of chronic obstructive disease or asthma when underlying PE may in fact, be present.

Electrocardiography

While electrocardiographic abnormalities may develop in the setting of acute PE, they are generally nonspecific and include T-wave changes, ST segment abnormalities, and left or right axis deviation. In the Urokinase Pulmonary Embolism Trial (UPET) electrocardiographic abnormalities were demonstrated in 87% of patients with proven PE and who were without underlying cardiopulmonary disease (129). These findings were not specific for PE, however. In this large clinical trial, 26% of patients with massive or submassive PE and 32% of those with massive PE had manifestations of acute cor pulmonale (S1 Q3 T3 pattern, right bundle branch block, P-wave pulmonale, or right axis deviation). The low frequency of specific electrocardiogram (ECG) changes associated with PE was confirmed in the PIOPED (Prospective Investigation of Pulmonary Embolism Diagnosis) study (128).

Arterial Blood Gas Analysis

Hypoxemia is common in acute PE, but is not universally present. Young patients without underlying lung disease may have a normal Pa_O2. In a retrospective analysis of hospitalized patients with proven PE, the Pa_O2 was greater than 80 mm Hg in 29% of patients less than 40 yr old, compared with 3% in the older group (130). The alveolar-arterial oxygen tension difference was abnormal in all patients, however. A subset of patients participating in the PIOPED study and suspected of PE with no history or evidence of preexisting cardiac or pulmonary disease was evaluated, and the Pa_O2 and alveolar- arterial difference values were compared (131). Patients with and without PE could not be distinguished on the basis of either of these values. The alveolar-arterial difference was elevated by more than 20 mm Hg in 76 of 88 (86%) patients with PE, however. The diagnosis of acute PE cannot be excluded on the basis of a normal Pa_O2 and although the alveolar-arterial difference is usually elevated, it may be normal in patients without preexisting cardiopulmonary disease.

Chest Radiography

The majority of patients with PE have an abnormal but nonspecific chest radiograph. Common radiographic findings include atelectasis, pleural effusion, pulmonary infiltrates, and elevation of a hemidiaphragm (131). Classic suggestions of pulmonary infarction such as Hampton's hump or decreased vascularity (Westermark's sign) are suggestive but infrequent. A normal chest radiograph in the setting of severe dyspnea and hypoxemia without evidence of bronchospasm or anatomic cardiac shunt is strongly suggestive of PE. The presence of a pleural effusion increases the likelihood of PE in young patients who present with acute pleuritic chest pain (132) In general, however, the chest radiograph cannot be used to prove or exclude PE conclusively. Diagnosing other processes such as pneumonia, pneumothorax, or rib fracture, which may cause symptoms similar to acute PE, is important, but PE may coexist with other cardiopulmonary processes.

D-Dimer

Noninvasive blood tests have been evaluated in hopes of identifying a specific marker of VTE. D-dimer is a specific degradation product released into the circulation when cross-linked fibrin undergoes endogenous fibrinolysis (133). A number of clinical trials have been undertaken to determine the utility of this test. Strategies have included the combination of / scanning and D-dimer testing. Different assays have been evaluated with different cutoff values utilized. Generally, either an enzyme-linked immunosorbent assay (ELISA) or a latex agglutination test has been performed. In patients with suspected PE, a low plasma D-dimer concentration (< 500 ng/ ml), measured by ELISA, has a 95% negative predictive power. However, low D-dimer levels have been found in only about 25% of patients without PE (134, 135). A latex agglutination test indicating a normal D-dimer level does not appear to be reliable in excluding PE (136, 137).

When the medical literature is systematically reviewed for publications that compare D-dimer results with the results of other diagnostic tests for venous thromboembolism, there appears to be substantial variability in assay performance, heterogeneity among the patient population, and inconsistent use of definitive diagnostic criteria for venous thromboembolism (138, 139). Becker and colleagues (138) performed a thorough review of the available literature and evaluated publications that compared D-dimer results with those of objective diagnostic tests for DVT or PE. Each study was evaluated independently by three reviewers. Articles meeting appropriate standards were designated level 1. The following conclusions were reached: (1) results of clinical studies utilizing one manufacturer D-dimer assay cannot be extrapolated to another; (2) no one test has been established as the best. The ELISAs are sensitive but cannot be performed rapidly. The latex tests, while rapid, have not been proved to be sufficiently sensitive. There are insufficient data available regarding the newer immunofiltration techniques; (3) future studies should be more rigorous regarding the definitive presence or absence of DVT and PE, and should as well address issues such as the extent of thrombosis, clinical setting, and comorbidity; and (4) additional outcome studies are needed.

Since the publication of above-described review, both DVT and PE management studies have been performed with therapeutic decisions based, in part, on D-dimer results. Ginsberg and colleagues (140) evaluated the results of a bedside whole blood agglutination D-dimer assay together with IPG in patients with suspected DVT. When both studies were negative, anticoagulation was withheld and the patients were monitored for 3 mo. In this group of patients, the negative predictive value was 98.5% (95% confidence interval, 96.3-99.6). For the D-dimer test alone, the negative predictive value was 97.2%. Perrier and colleagues (141) evaluated 308 consecutive patients presenting to the emergency room with suspected PE. Each patient was managed according to a diagnostic protocol including an assessment of clinical probability, / scan, ELISA plasma D-dimer, and lower extremity US. Of the 308 patients, 106 (34%) had diagnostic / scans (high probability in 63 and normal in 43). The noninvasive evaluation was diagnostic in 125 patients (62%). In 48 patients, PE was ruled out by a nondiagnostic lung scan together with low clinical probability. In 53 cases, it was ruled out by a quantitative D-dimer of less than 500 µg/L. Only 77 of the 202 patients with nondiagnostic / scans required pulmonary angiography. At 6-mo follow-up, only 2 of the 199 patients in whom the diagnostic protocol had ruled out PE had a VTE event. Using the same cutoff value for the quantitative D-dimer test, these investigators subsequently reported that of 198 patients with suspected PE and a D-dimer level, < 500 µg/L, 196 were free of PE, 1 had PE, and one was lost to follow-up (142). Thus, the negative predictive value of the D-dimer test was approximately 196 of 198 (99%). These data, although from one group of investigators, are encouraging. Rapid "bedside assays" are becoming increasingly available and additional outcome studies will further define their role. However, the D-dimer test cannot be recommended as a standard part of the PE or DVT diagnostic algorithm at the present time.

The Ventilation-Perfusion Scan

The ventilation-perfusion (/) scan has long been considered the pivotal diagnostic test in acute PE. Unfortunately, the / scan is diagnostic in a minority of cases; that is, it is rarely interpreted as normal or high probability. Most lung diseases affect pulmonary blood flow to some extent as well as ventilation, decreasing the specificity of the / scan (143- 149). Pulmonary embolism frequently occurs in the setting of concomitant lung disease such as chronic obstructive pulmonary disease (COPD) or pneumonia, further complicating the diagnostic evaluation (127, 150, 151).

The Prospective Investigation of Pulmonary Embolism Diagnosis (PIOPED) was a multicenter, collaborative effort designed to determine the sensitivity and specificity of the / scan in patients with suspected acute PE (128). The importance of clinical suspicion (made without knowledge of the scan results) combined with the / scan was a crucial aspect of the investigation. In this clinical trial, PE was proven or excluded by pulmonary angiography or by autopsy. In patients in whom the pulmonary angiogram was nondiagnostic, PE was excluded by the absence of an adverse event over the course of 1 yr, without therapy. Criteria for the interpretation of / scans from the PIOPED subsequently became widely adopted. The most important information derived from the study was the concept that PE is often present in patients with nondiagnostic lung scans when associated with a high clinical suspicion of PE. In this setting, a high-probability lung scan is associated with proved PE in 96% of cases, but a low-probability scan is also associated with PE in 40% of patients. When a high-probability / scan is associated with a low or uncertain clinical suspicion for PE likelihood, the likelihood of PE is only 56 and 88%, respectively (Table 14). A treatise on PE based on the vast amount of data accrued by the PIOPED has been published (152).

The effect of prior cardiopulmonary disease. As the extent of cardiopulmonary disease increases, it becomes increasingly likely that the lung scan will be nondiagnostic. On the basis of the original PIOPED criteria, patients with normal chest radiographs had intermediate-probability / scans in only 13% of cases, while low- and near normal/normal-probability scans occurred in 35 and 45% of these patients, respectively (128). Intermediate-probability scans were seen in 33% of patients with no prior cardiopulmonary disease and in 43% of those with any form of cardiopulmonary disease. With even more complex underlying disease, i.e., COPD, 60% of patients had intermediate-probability scans. Fewer patients with COPD had high-probability scans or nearly normal scans than did patients without cardiopulmonary disease. However, among each of the above-described patient groups, including those with COPD, the positive predictive value of high, intermediate, low, and nearly normal scans was similar (151, 153)

The perfusion scan alone. The value of the perfusion scan without a ventilation scan has been examined (128, 154). A randomly selected subset of patients from the PIOPED with suspected acute PE had perfusion scans interpreted in a blinded manner, independent of, as well as in combination with, the ventilation scan. Pulmonary embolism was proved by pulmonary angiography in 29 of these 98 patients, and excluded by angiography in 33 patients or by outcome analysis in 5 patients. Neither outcome analysis nor angiography could be performed in the remaining 31 patients. In the 67 patients in whom the presence or absence of PE was certain, the positive predictive value of a high-probability perfusion scan (93%) did not differ from the / scan group (94%). Similarly, an intermediate-probability perfusion scan was no less predictive of PE than an intermediate probability / scan, and a low probability perfusion scan was no less predictive than a low-probability / scan. There were not enough near normal/normal perfusion scans to make a useful comparison with / scans. The available data would suggest that if a ventilation scan cannot be performed, an isolated perfusion scan is useful if the scan is high probability, low probability, near normal, or normal (154). Unfortunately, there has been controversy over whether a ventilation scan should ever be performed after a perfusion scan. Certain individuals believe that a ¹³³Xe ventilation scan can be effectively performed after a perfusion scan (155). Others suggest that the scattered radiation from the previously administered ^99mTc perfusion particles substantially decreases the accuracy of the washout phase of the ventilation scan, particularly if ^99mTc is used for the ventilation scan (156). Because ¹²⁷Xe has a higher inherent photon energy than ^99mTc, ¹²⁷Xe ventilation studies can be performed after the perfusion scan. However, ¹²⁷Xe is expensive and not readily available. Details regarding appropriate techniques for / scanning are available elsewhere (157).

In the PISA-PED study, only perfusion scans were utilized (158) and one or more segmental perfusion defects were considered diagnostic of PE. This is important because a single perfusion defect has not been found to be a consistent predictor of PE. A positive perfusion scan had a positive predictive value of 95% and a negative scan had a negative predictive value of 81%. Only 21% of patients had clinical and perfusion scan results that were contradictory. These results appear superior to the PIOPED results, but the study populations differed. In the PISA-PED, 24% of patients had normal perfusion scans compared with 2% in the PIOPED. The interpretive criteria differed as well. On the basis of the PIOPED results, probability estimates for PE have been correlated with / scan results and some of the original / scan diagnostic criteria have been revised. It has been suggested that these revised criteria be applied (159). It is important to emphasize that probability estimates based on different interpretive schemes have varied considerably (152).

The nondiagnostic ventilation-perfusion scan: Use of lower extremity studies. When the lung scan is nondiagnostic, evaluation of the lower extremities is an alternative means by which to establish the need for anticoagulatio