The one generalization about venous thromboembolism (VTE) that is free from controversy is that many aspects of this disorder remain controversial. There are multiple reasons why VTE continues to engender lively debate. Perhaps the major reason, notwithstanding the substantial advances that have been made since the late 1990s, is that a number of fundamental questions continue to exist regarding the pathogenesis, clinical presentation, diagnosis, and therapy of the disease.
VTE represents a potentially fatal disease process with a clinical presentation that is often silent or nonspecific and for which a wide range of diagnostic techniques is available, many with technical and interpretive limitations. Although estimates vary widely, the best available information suggests that there are at least 5 million episodes of venous thrombosis annually in the United States. The annual incidence of acute pulmonary embolism (PE) is approximately 70 per 100,000, and PE accounts for more than 200,000 hospital admissions per year on the basis of discharge diagnosis (ICD-9) codes. These data, of course, represent only those patients in whom the diagnosis of acute PE was made correctly. Autopsy series suggest that the true number of deaths from acute PE (including those patients who died without the diagnosis ever being made) is at least threefold higher. For this reason, it is likely that the majority of PE-related deaths can be attributed to PE that was undiagnosed (and therefore untreated) during life.
Up to 10% of patients who suffer embolism may die from their disease. The overwhelming majority of these deaths do not appear to arise from therapeutic failure. With the exception of patients who initially present with hemodynamic impairment, in whom mortality rates approach 20% to 30%, embolic recurrence is rare and death is uncommon once the diagnosis of embolism is confirmed and appropriate therapy initiated. The majority of deaths related to embolism appear to arise from a failure to prevent the disease in patients at risk of it and from a failure to make the diagnosis in those afflicted. The incidence of fatal PE appears to have declined over the past several decades. For example, the Centers for Disease Control Compressed Mortality File (CMF) reported an age-adjusted death rate of 875/100,000 in 1999, which steadily decreased over the decade to 747/100,000 by 2010. It is noteworthy that these statistics do not account for fatal PE that was undiagnosed clinically, as described earlier. Balancing these facts are some data that suggest that the increasing reliance on chest computed tomography (CT) angiography may be leading to overdiagnosis of PE in some populations. Despite this controversy, mortality from acute PE remains a substantial public health care problem as a result of the demographics of an aging population.
Contributing to the debate surrounding VTE is the involvement of a wide range of medical disciplines in its prevention, diagnosis, and management. Thrombosis is not a discrete clinical subspecialty. The problem of VTE involves pulmonologists, cardiologists, hematologists, internists, specialists in vascular disease, radiologists, a range of surgical subspecialists, obstetricians, and others. Because specific prophylactic, diagnostic, and therapeutic strategies within one discipline may not necessarily be applicable to another, a perception of coherence in the clinical approach to this disease process often appears to be lacking.
Many of the long-standing controversies surrounding the natural history, diagnosis, and therapy of VTE have been partially or completely reconciled, resulting in substantial changes in the diagnostic and therapeutic approach to the disease. The persistence of a number of unresolved issues, as well as the emergence of still others, should not be a cause for cynicism. In the approach to the patient with suspected VTE, an understanding of what is unknown can prove invaluable to the decision-making process.
Pathogenesis and Risk Factors
The triad of venous stasis, alterations in coagulation, and vascular injury identified by Virchow in 1856 as primary factors in the pathogenesis of VTE has been supported by a considerable amount of clinical and experimental evidence. Over the past several decades, severe abnormalities of the coagulation and fibrinolytic system, including isolated deficiencies of antithrombin III, protein C, protein S, and plasminogen, as well as the presence of a lupus anticoagulant, have been described and their association with first-time and recurrent VTE has been confirmed ( Table 57-1 ). In addition, there has been increasing recognition of less severe, but more common, inherited “thrombophilic” conditions that are capable of shifting the hemostatic balance toward thrombosis and mildly increasing the risk of venous thromboembolic events. Although they also appear to increase the risk of recurrence slightly, the magnitude of the increase is unlikely to warrant changes in therapeutic strategies for patients with those thrombophilias.
|Protein C deficiency|
|Protein S deficiency|
|Antithrombin III deficiency|
|Factor V Leiden mutation|
|Prothrombin 20210 G → A variation|
|Familial plasminogen deficiency|
|ACQUIRED SURGICAL PREDISPOSITIONS|
|Major thoracic, abdominal, or neurosurgical procedures requiring general anesthesia and lasting >30 min|
|Spinal cord injury|
|ACQUIRED MEDICAL PREDISPOSITIONS|
|Prior venous thromboembolism|
|Advanced age (>60 yr)|
|Congestive heart failure|
|Pregnancy and the postpartum period|
|Antiphospholipid antibody syndrome|
|Inflammatory bowel disease|
|Paroxysmal nocturnal hemoglobinuria|
The most common of the inherited thrombophilic conditions, first described in 1993 by Dahlback and designated factor V Leiden mutation, is the consequence of a single point mutation in the factor V gene (adenine for guanine) resulting in activated factor V (factor Va) with diminished sensitivity to the natural anticoagulant effect of activated protein C. Approximately 5% of whites in Europe and North America are heterozygous for this genetic defect; lower rates of carrier frequency have been reported among Native American, African, and Asian populations. Although initially detected in up to 60% of selected patients with VTE, subsequent studies have detected the mutation in 10% to 20% of unselected patients. The heterozygous state carries a 5- to 10-fold increase in lifetime risk for VTE. Whereas the relative risk among patients homozygous for this mutation has been estimated to be as high as 80-fold, the estimates are based on a small number of patients and may be somewhat imprecise. Factor V Leiden mutation appears to be an important risk factor for VTE during pregnancy, in the postpartum period, and during oral contraceptive use. Compared with women who do not use oral contraceptives and are not carriers of the factor V mutation, the risk of thrombosis among those with both risk factors is increased approximately 30-fold. In view of the prevalence of thrombophilia and the low prevalence of VTE in nonusers of combined oral contraceptives, the absolute risk remains low. Because the absolute increase in VTE is modest, selective thrombophilia screening based on previous personal or family history of VTE is more cost-effective than universal screening.
A sequence variation in the prothrombin gene (G20210A) was described in 1996 and is estimated to exist in approximately 2% to 4% of the population. This mutation results in an overproduction of prothrombin, which is otherwise normal. It is associated with a threefold to fourfold increased risk of lower extremity venous thrombosis.
Hyperhomocysteinemia has also been identified as a potential independent predisposition to VTE. Elevation of plasma homocysteine levels may be the result of genetic abnormalities; nutritional deficiencies of vitamins (B 6 , B 12 , folate); clinical disorders (renal insufficiency, hypothyroidism, inflammatory bowel disease); or a combination of the three. Although retrospective, case-control studies demonstrate an association between hyperhomocysteinemia and VTE, the results of prospective studies have not been uniform.
The three “common genetic thrombophilias” (factor V Leiden, prothrombin G20210A, and hyperhomocysteinemia) appear to act as independent risk factors for increasing thrombosis risk. Thus, the relative risk for patients with multiple thrombophilic conditions is higher than for patients with only one of them.
The identification of these risk factors, and the likelihood that others exist, raises the possibility that screening to determine relative thromboembolic risk may be feasible in the future. However, a consensus for such an approach does not exist at the present time. Despite its prevalence in the general population, screening for factor V Leiden mutation is not advised because the overwhelming majority of patients with this abnormality will never suffer a thromboembolic event. Furthermore, the absence of this abnormality should not influence the decision to provide prophylaxis to patients at clinical risk. In addition, there is no evidence that prolonged anticoagulation would be more beneficial for VTE patients with either factor V Leiden or the prothrombin G20210A gene than in VTE patients without these mutations.
Patients with spontaneous (“unprovoked”) VTE who have both factor V Leiden and the prothrombin G20210A gene appear to be at somewhat higher risk for recurrence than similar patients without those mutations. This observation does not establish a risk-benefit relationship for prolonged anticoagulation in patients with the combination of thrombophilic mutations, which remains unknown.
For all the reasons stated previously, the benefit of screening for the common thrombophilias is controversial. Patient groups in whom screening is likely to yield positive results include those with a history of recurrent VTE or with a confirmed family history of thromboembolism, a first episode of thromboembolism at an early age, spontaneous venous thrombosis, thromboses in unusual anatomic sites, arterial thrombosis, and thromboembolism associated with pregnancy or estrogen use. Whether screening should be performed before the initiation of oral contraceptive agents remains a matter of debate. The relative risk of VTE is increased approximately fourfold in users of oral contraceptives, although the increase in absolute risk is modest. A policy of routine screening would deny effective contraception to a substantial number of women while preventing only a small number of PEs. A similar situation exists for pregnancy, in which the presence of factor V Leiden or the prothrombin G20210A mutation increases the relative risk of VTE but only slightly increases the absolute risk.
In most patients, even those with an identified thrombophilic state, clinical conditions associated with venous stasis, intimal injury, or both serve as the basis for the thromboembolic event. Furthermore, the risk of VTE in hospitalized patients is not limited to patients undergoing surgical procedures. Thromboembolic risk in patients admitted with a wide range of acute medical problems is comparable with that seen in surgical patients.
Major risk factors include pelvic or lower extremity fractures, hip and knee surgery, a past history of VTE, acute paralytic stroke or spinal injury, major traumatic injury, open prostatectomy, and abdominal or pelvic surgery for malignant disease. Other factors that enhance risk include prolonged general anesthesia, advancing age, cardiac disease, pregnancy, the postpartum state, the use of estrogen-containing hormone replacement therapy, malignancy, nephrotic syndrome, the presence of a lupus anticoagulant or antiphospholipid antibody, and prolonged immobilization. Air travel is a relatively modest risk factor, as is the presence of inflammatory bowel disease. It is important to recognize that these risk factors may be multiplicative. Thromboembolic risk in an otherwise healthy 45-year-old individual undergoing an elective cholecystectomy is considerably less than the risk experienced by an obese 75-year-old with a history of prior VTE undergoing the same procedure. Similarly, the patient with hip fracture or hip replacement has, by virtue of that condition alone, a 60% to 70% risk of deep venous thrombosis (DVT) and a 2% to 4% risk of experiencing a fatal thromboembolic event in the absence of preventive measures. Add other risk factors, as well as the incidence of DVT, and the likelihood of a fatal complication will be even higher. These considerations allow the development of a reasonable “risk profile” in an individual patient, a profile that should influence the use and intensity of prophylactic intervention.
Natural History: Deep Venous Thrombosis
Venous thrombi appear to originate either in the vicinity of a venous valve cusp, where eddy currents arise, or at the site of intimal injury. Platelet aggregation and release of mediators initiate the sequence. With local accumulation of such factors, the coagulation cascade is activated and thrombus develops, composed primarily of fibrin and erythrocytes. As the thrombus extends, local fibrinolytic activity is enhanced. Thus thrombus behavior becomes a dynamic process that may result in complete dissolution, partial resolution resulting in a variable degree of intimal narrowing and valvular damage, progressive proximal extension, or embolization. In more than half of patients, the venous wall suffers some degree of permanent scarring, which is visible on ultrasound ; if the venous wall scarring causes severe obstruction, collateral veins develop.
Extensive autopsy and clinical studies have established that some 90% of PEs that elicit clinical attention arise from venous thrombosis in the deep veins of the lower extremities. In fact, a conservative estimate is that at least one third of deep venous thrombosis is complicated by symptomatic or asymptomatic PE. Venous thrombi capable of embolization can also arise from other sites. Primary iliac or proximal femoral thrombi may develop in patients undergoing surgery involving the hip, and pelvic vein thrombosis may develop in patients undergoing pelvic or prostatic surgery. Axillosubclavian vein thrombosis may be spontaneous, resulting from congenital abnormalities of the thoracic outlet, or may be related to indwelling central venous catheters, pulmonary artery catheters, or transvenous pacing wires. Increasing use of central venous catheters has been implicated in a rise in the incidence of upper extremity deep venous thrombosis. In patients with dilated right heart chambers or pulmonary arteries, thrombi can form at those sites and embolize distally into the branches of the pulmonary artery.
The likelihood of embolism is influenced by the location of thrombi in the veins of the lower extremity. Although the majority of thrombi originate in the veins of the calf, it has been clearly demonstrated that thrombi that remain limited to the calf veins rarely result in PE. However, 15% to 25% of symptomatic, isolated calf thrombi when left untreated will extend to involve the proximal veins (popliteal, superficial femoral, and common femoral veins, or even more proximally). (It is worth emphasizing that the superficial femoral vein is actually not superficial and represents one of the deep veins. ) Proximal extension poses a risk of embolization that approaches 50%. For that reason, about one in eight patients with distal deep vein thromboses will develop PE.
This natural history of DVT has several important diagnostic and therapeutic implications. First, because the vast majority of emboli arise from thrombi in the veins in the lower extremity, diagnostic approaches to DVT can focus on techniques that detect lower extremity DVT. Second, techniques that detect above-knee thrombi are of particular value whether or not they can detect calf-limited thrombi. Finally, although it is true that calf-limited thrombi rarely embolize, many have incorrectly concluded from this information that symptomatic, calf-limited DVT represents a clinically irrelevant condition. Calf-limited thrombi may extend proximally. Furthermore, symptomatic calf vein thrombosis appears to be subject to recurrence, albeit at a lower risk than proximal vein thrombosis.
Although most above-knee thrombi represent extensions from calf thrombi, some do arise in the larger, proximal veins de novo. This appears to be restricted principally to patients with hip fracture or replacement, pelvic surgery (including prostatic resection), and other high inguinal pelvic trauma.
At any time during this process, a portion or all of the thrombus can detach as an embolus. This risk is highest early in thrombus development before there is significant fibrinolysis or organization. Beyond this acute phase, the long-term outlook is influenced principally by the extent of residual venous obstruction and valvular damage. If significant obstruction or valvular damage persists, downstream stasis will be present, leading to a risk of recurrent DVT and development of the postphlebitic syndrome.
Natural History: Pulmonary Embolism
PE causes a number of consequences to gas exchange and other pulmonary functions. Regional obstruction to pulmonary blood flow and diversion of the flow to unobstructed portions of the lung may alter the ventilation-perfusion balance in both the obstructed and unobstructed regions. In the regions with pulmonary vascular obstruction, alveolar dead space is created. If a region’s blood flow is severely obstructed, there may be bronchoconstriction in the lung distal to the area of obstruction as a result of alveolar hypocapnia. This is probably uncommon in patients because they are free to inhale carbon dioxide–rich dead-space air into the associated lung regions and because obstruction is rarely total. PE almost always leads to hyperventilation, the mechanism for which remains uncertain.
The characteristic gas exchange abnormality is hypoxemia, generally caused by the venous admixture due to areas of low ventilation/perfusion ratio or of shunt. Arterial hypoxemia may be worsened when acute increases in right ventricular afterload lower the cardiac output enough to widen the arteriovenous oxygen difference and decrease the oxygen saturation of mixed venous blood. This lowering of the mixed venous oxygen content magnifies the effects of the normal venous admixture, thereby further lowering the resultant arterial oxygen pressure (P o 2 ). Another potential mechanism for hypoxemia in patients with massive PE is right-to-left shunt, on either an intrapulmonary or intracardiac basis. With embolic occlusion sufficient to increase pulmonary artery pressure, hypoxic vasoconstrictive mechanisms can be overwhelmed and perfusion may increase in poorly ventilated or nonventilated lung regions. On occasion, embolic events massive enough to increase right atrial pressure may result in intracardiac right-to-left shunting through a patent foramen ovale. The final mechanism for hypoxemia relates to the loss of pulmonary surfactant. Surfactant is not lost immediately; it requires approximately 24 hours of total occlusion and lack of blood flow to develop. At that time or later, surfactant becomes depleted in the obstructed alveolar zones, resulting in atelectasis and edema. If the thrombus resolves and perfusion to this atelectatic region resumes, hypoxemia may result.
One uncommon local consequence of PE is pulmonary infarction. Infarction is uncommon because the pulmonary parenchyma has three potential sources of oxygen: the pulmonary arteries, the bronchial arteries, and the conducting airways. In patients with no coexisting cardiopulmonary disease, large infarctions (such as those visible by chest radiography) are rare. However, autopsy series suggest that infarctions of smaller pulmonary arteries are more common. Infarction develops in approximately 20% to 33% of patients with significant cardiac or pulmonary disease that compromises either bronchial arterial flow or airway patency. In patients with left ventricular failure, infarction may result if the increased pulmonary venous pressure compromises bronchial flow.
The cardiac and hemodynamic effects of embolism are related to three factors: the degree of reduction of the cross-sectional area of the pulmonary vascular bed, the preexisting status of the cardiopulmonary system, and the physiologic consequences of both hypoxic-mediated and neurohumorally mediated vasoconstriction. Mechanical obstruction of the pulmonary vascular bed by embolism accounts for most of the increase in pulmonary vascular resistance (PVR), although resistance is typically worsened by release of vasoconstrictive substances such as endothelin, thromboxane A 2 , and serotonin. The combination of factors acutely increases the workload on the right ventricle, a chamber ill-equipped to deal with an acute elevation in pressure load. In patients without preexisting cardiopulmonary disease, obstruction of less than 20% of the pulmonary vascular bed results in a number of compensatory events that minimize adverse hemodynamic consequences. Pulmonary vessels are recruited and become distended, resulting in a normal or near-normal PVR and pulmonary artery pressure; cardiac output is maintained by increases in the right ventricular stroke volume and increases in the heart rate. As the degree of pulmonary vascular obstruction exceeds 30% to 40%, there are increases in pulmonary artery pressure and modest increases in right atrial pressure. The Frank-Starling mechanism maintains right ventricular stroke work and cardiac output. When the degree of pulmonary artery obstruction exceeds 50% to 60%, compensatory mechanisms are overwhelmed, cardiac output begins to fall, and right atrial pressure increases dramatically. With acute obstruction beyond this amount, the right heart dilates, right ventricular wall tension increases, right ventricular ischemia may develop, the cardiac output falls, and systemic hypotension develops. Hypotension worsens the situation by decreasing the coronary perfusion pressure to the already tense right ventricle. In patients without prior cardiopulmonary disease, the maximal mean pulmonary artery pressure that can be generated by the right ventricle appears to be 40 mm Hg (representing a pulmonary artery systolic pressure of ≈70 mm Hg). The correlation between the extent of pulmonary vascular obstruction and PVR appears to be hyperbolic; with increasing vascular obstruction, PVR rises slowly as the remaining pulmonary vascular bed expands and recruits additional vessels and then rises rapidly when that reserve is exhausted.
The hemodynamic response to acute PE in patients with preexisting cardiopulmonary disease may be considerably different from that in patients without prior disease. In patients without prior cardiopulmonary disease, there is a general relationship between the rise in pulmonary artery pressure and the pulmonary vascular obstruction, whereas in patients with prior cardiopulmonary disease, pulmonary artery pressures may rise disproportionately. As a result, severe pulmonary hypertension may develop in response to a relatively small reduction in pulmonary artery cross-sectional area. In addition, evidence of right ventricular hypertrophy (rather than right ventricular dilation) associated with a mean pulmonary artery pressure in excess of 40 mm Hg (pulmonary artery systolic pressure in excess of ≈70 mm Hg) in a patient suspected of embolism should suggest an element of chronic pulmonary hypertension resulting from a potentially diverse group of etiologic possibilities (e.g., chronic thromboembolic pulmonary hypertension, left ventricular failure, valvular disease, right-to-left cardiac shunts).
Beyond the acute embolic event, the behavior of emboli parallels that previously described for venous thrombi; that is, they undergo resolution by fibrinolysis, by organization and recanalization, or both. Although there is a great deal of interpatient variability, resolution of PEs typically is substantial during the first week, somewhat more gradual for the next 4 to 8 weeks, and then slow thereafter. (The most rapid resolution of a large embolus that has been documented is 51 hours. ) The term resolution is used here because it is uncertain, in humans, to what degree lysis (versus organization) participates in embolic resolution. Most sequential data regarding resolution in humans are based on perfusion scan, not angiographic, data. However, these data suggest that residual anatomic defects are common following embolism and, contrary to prior opinion, that complete restoration of pulmonary blood flow represents the exception rather than the rule. In terms of hemodynamic resolution, it would appear that a stable pulmonary artery pressure is reached within 6 weeks. How often anatomic and mild hemodynamic residuals persist is not known. Nearly one third of patients with acute PE may have residual perfusion defects, which may be associated with a spectrum of symptoms including dyspnea, lower exercise tolerance, and higher pulmonary arterial pressures of various intensities. However, residual obstruction sufficient to cause clinically significant pulmonary hypertension is rare. For this small group of patients with significant residual obstruction, the clinical course and management are dealt with later in this chapter.
The most common symptoms and physical findings of venous thrombosis include swelling, pain, erythema, and warmth. “Classic” findings such as Homan sign (calf pain with flexion of the knee and dorsiflexion of the ankle), Moses sign (pain with calf compression against the tibia), or a palpable cord are infrequent and nonspecific.
As established by multiple investigations, the clinical diagnosis of venous thrombosis is imprecise. In patients with clinical signs and symptoms suggestive of venous thrombosis, 60% to 80% will not have the diagnosis by objective testing. Furthermore, and even more disquieting, the majority of high-risk patients who are monitored and who develop DVT will not have signs or symptoms suggesting the diagnosis. Algorithmic clinical models incorporating risk factors, symptoms, and physical signs have been demonstrated to have the ability to stratify symptomatic patients into risk categories, although not to a level in which clinical diagnosis, in the absence of objective testing, can be relied on either to confirm or exclude the diagnosis. The differential diagnosis of DVT is extensive and includes cellulitis, arthritis, muscular injury or tear, neuropathy, arterial insufficiency, lymphedema, ruptured Baker cyst, superficial thrombophlebitis, and chronic venous insufficiency.
Similarly, the diagnosis of PE cannot be confirmed or excluded solely on clinical grounds. However, recognition of the clinical signs and symptoms associated with PE is valuable because clinical findings and clinical suspicion represent an essential first step in the diagnostic pathway. Although a somewhat arbitrary classification because presenting symptoms and signs of embolism frequently overlap, the presentation of PE can be categorized into one of three clinical syndromes: (1) isolated dyspnea; (2) pleuritic pain or hemoptysis; and (3) circulatory collapse. Among patients without prior cardiopulmonary disease in the Prospective Investigation of Pulmonary Embolism Diagnosis (PIOPED) study, the syndrome of pleuritic pain or hemoptysis was the most common mode of presentation, seen in approximately 60% of patients; isolated dyspnea was noted in approximately 25%, and circulatory collapse in 10%.
The most common presenting symptom of acute embolism is the sudden onset of dyspnea. In various studies, dyspnea was a presenting symptom in the majority of patients. However, it must be emphasized that, in the PIOPED study, dyspnea was not present in 27% of patients ultimately proven to have embolism. Pleuritic chest pain was present in 66% of patients, whereas hemoptysis (15%) was uncommon. Less than 50% of patients had cough (37%), leg swelling (28%), and leg pain (26%). A sense of impending doom also is reported, particularly with massive embolism. Angina also can result from massive embolism representing, in this circumstance, right ventricular ischemia. Syncope also may be a presenting complaint in major embolic occlusion.
The most common physical finding is tachypnea (respiratory rate >20/min). In the PIOPED study, however, tachypnea was not present in approximately 30% of patients with embolism. Clinical findings noted less frequently include crackles (55%), tachycardia (30%), and an increased pulmonic component of the second heart sound (S2; 23%). Fever may develop some hours after the event and often reaches but rarely exceeds 38.3° C. As noted previously, hemoptysis may be observed; it usually is quite modest in extent, although it may persist for some days. Brisk hemoptysis is rare and is almost never the initial finding. With massive embolism, there may be evidence of right ventricular overload or failure, such as a right ventricular tap along the left sternal border and an accentuated pulmonary valve closure sound. If right ventricular failure develops, there may be narrowed or fixed splitting of an S2, an S3, and/or an S4, distended neck veins, and cyanosis. Careful examination of the legs may elicit evidence suggesting venous thrombosis. In the PIOPED study, clinically apparent venous thrombosis was found in only 15% of patients.
Obviously, these symptoms and signs are nonspecific. In the PIOPED study, none of the presenting symptoms was capable of discriminating between patients with positive and negative angiograms. Also, in terms of presenting signs, only the presence of crackles, an S4, and an increased pulmonic component of S2 could differentiate between those with positive and negative angiograms. Furthermore, in patients with underlying cardiopulmonary disease, the presenting symptoms and signs frequently may be obscured by elements of the underlying illness. It is also important to recognize that the clinical presentation of embolism has been characterized in trials composed of symptomatic patients, although it is known that many PE do not produce symptoms. In prospective studies of high-risk patients with proximal DVT, PE can be documented in 40% of patients who had no symptoms of PE. It is likely that the frequency and severity of symptoms are influenced by the extent of embolic occlusion and the prior cardiopulmonary status of the patient. Small- or moderate-sized emboli may induce few or no symptoms in an otherwise normal individual. In patients with preexisting cardiopulmonary disease, symptoms are more common and severe.
Owing to the nonspecific presentation of PE, the differential diagnosis is varied and extensive, especially in hospitalized patients with coexisting cardiac or pulmonary disease. Common considerations include congestive heart failure, exacerbation of chronic lung disease, postoperative atelectasis, and viral pleurisy. PE presenting with fever, dyspnea, and chest radiographic abnormalities easily can be confused with a bacterial pneumonia. The presence of fever and leukocytosis (rarely >15,000 cells/µL) are uncommon but well-described accompaniments of VTE.
These precautionary statements regarding clinical diagnosis are not meant to suggest that the clinical presentation of venous thrombosis or PE cannot be used as a basis for clinical decision making. However, they are meant as a reminder that the clinical presentation of VTE and PE may often be atypical or subtle and should serve only to generate a suspicion of that diagnosis. A reliance on symptoms and signs that are considered “classic” before making the decision to proceed to confirmatory testing may lead to underdiagnosis and unnecessary mortality.
Diagnosis of Venous Thrombosis
The proper diagnostic approach to VTE must take into account the central fact that venous thrombosis and PE are manifestations of the same disease process: venous thrombosis representing the source of PE and PE representing a complication of venous thrombosis.
In validating any test, there must be a “gold standard.” In the case of lower extremity venous thrombosis, that standard is contrast venography ( Fig. 57-1 ). In investigative contexts, it is a good gold standard (as indicated later, however, that often is not the case in clinical contexts). The venogram is performed according to a specific protocol described by Rabinov and Paulin in 1972. The most reliable criterion for the diagnosis of venous thrombosis is a constant intraluminal filling defect evident in two or more views. Other criteria such as nonvisualization of deep veins, presence of venous collaterals, or nonconstant filling defects are less reliable. Under circumstances in which the proper protocol and interpretative criteria are utilized, contrast venography has high sensitivity and specificity. However, the study is not without shortcomings. As those who have seen many venograms recognize, the study is not easy to interpret, especially in patients with a prior history of venous thrombosis. Venous cannulation may often be difficult, especially in the presence of edema; expert interpretation is essential for accurate diagnosis; injection of contrast material with its associated allergic and nephrotoxic risks is necessary; venous thrombosis may be induced by the procedure itself; and the cost, invasive nature, and discomfort of the study make sequential studies impractical.
Owing to these limitations, a number of noninvasive studies capable of being performed on a sequential basis were introduced into clinical practice. At present, duplex ultrasonography is the most commonly used noninvasive technique. Magnetic resonance imaging (MRI) and CT have proved capable of detecting thrombi, but their widespread utilization has been limited by cost, limited access, and, in the case of CT, the need for contrast administration.
Since the late 1990s, duplex ultrasonography, which refers to the combination of Doppler venous flow detection and real-time B-mode imaging, has assumed a central role in the noninvasive diagnosis of symptomatic lower extremity DVT. A number of criteria are used to diagnose venous thrombosis, the most reliable of which is the noncompressibility of a venous segment ( Fig. 57-2 ). Secondary, less reliable criteria include the presence of echogenic material within the venous lumen ( eFig. 57-1 ), venous distention, and loss of phasic change with respiration, lack of response at the common femoral vein (CFV) to Valsalva maneuver, diminished or absent color flow with color Doppler sonography ( eFig. 57-2 ), and lack of augmentation of flow at the CFV with compression of the calf. The two signs showing a lack of increase in flow, either with Valsalva or with calf compression, can indicate obstruction of veins between the site of increased pressure and the measured vein. The absence of an echogenic luminal mass cannot be considered useful in excluding the diagnosis of venous thrombosis because acute thrombus may not demonstrate echogenicity ( eFig. 57-3 ).
Multiple studies since the late 1990s have demonstrated sensitivities and specificities exceeding 95% in symptomatic patients with proximal venous thrombosis. Although simplified compression examinations limited to the symptomatic leg or to the common femoral and popliteal veins (rather than the entire lower extremity venous system) have been suggested, the time saved with such approaches is limited, and a number of isolated superficial femoral vein or calf-limited thrombi may be overlooked. Asymptomatic thrombi in the contralateral leg can be detected in approximately 5% to 10% of patients presenting with symptomatic acute venous thrombosis. Although the detection of asymptomatic, contralateral thrombi has little impact on the immediate management of the patient, it may have long-term consequences when recurrence is suspected. A more prudent approach appears to be a complete examination extending from the inguinal ligament to the popliteal vein and examination of the contralateral extremity if thrombus is detected in the symptomatic leg.
Duplex ultrasonography is less accurate in the detection of symptomatic calf-limited thrombi (sensitivity ≈70%) and in asymptomatic proximal vein thrombi (sensitivity ≈50%), thereby limiting its utility as a screening study in high-risk populations. When ultrasonography is negative in patients with suspected venous thrombosis, a strategy of serial testing, consisting of one or two additional tests over the subsequent week, has proved to be effective in detecting proximal extension.
Magnetic Resonance Imaging
MRI techniques for detecting venous thrombosis include spin-echo magnetic resonance, gradient-recalled-echo magnetic resonance, and magnetic resonance venography. Preliminary reports suggest that MRI is at least as sensitive and specific as duplex ultrasonography. A potential advantage of MRI is that the entire length of the venous system, including the pelvic veins, can be evaluated. Disadvantages associated with MRI include cost and limited access, as well as the expertise required to perform and interpret the studies properly.
The role of CT as a stand-alone test for venous thrombosis is limited. The sensitivity and specificity of CT venography are comparable with those of ultrasonography but mandate contrast injection with its associated risks and radiation exposure. Potential advantages of CT venography include the ability to visualize the pelvic veins ( eFig. 57-4 ) and vena cava ( eFig. 57-5 ). A diagnostic approach combining CT venography with CT pulmonary angiography (CTPA) may have a role in patients undergoing evaluation for PE (see later).
The development of a rapid and accurate blood test capable of diagnosing VTE has held special appeal and has been the subject of considerable investigative interest. A number of different serologic markers have been investigated, including D-dimer, fibrin monomer, prothrombin fragment, thrombin–antithrombin III complex, fibrinopeptide B, and fibronectin. Of these, D-dimer, alone and in combination with other noninvasive studies, has been subjected to the most rigorous clinical evaluation. D-dimer testing has proved to be highly sensitive but not specific; that is, elevated levels are present in nearly all patients with thromboembolism but also in a wide range of circumstances, including advancing age, pregnancy, trauma, infections, the postoperative period, inflammatory states, and malignancy. Therefore the role of D-dimer testing is limited to one of exclusion of VTE. Multiple assays have been developed with sensitivities that range from 80% to almost 100%. Highly sensitive assays, such as the enzyme-linked immunosorbent assay, are capable of excluding thromboembolism but are associated with such a high frequency of false-positive results that their clinical utility is limited.
Less sensitive assays (e.g., latex agglutination, red cell agglutination) lack the ability to exclude thromboembolism in isolation but have been used successfully in combination with either a clinical probability estimate or a noninvasive diagnostic study. Although potentially of substantial value in diagnostic pathways, the burgeoning selection of available assays, variations in sensitivity and specificity related to the type of assay, a range of discriminate values for positivity, and lack of standardization have limited generalized application of the technique due to uncertainty among clinicians regarding the predictive value of the particular test they are using. D-dimer testing has been used successfully as part of a number of different diagnostic strategies, and negative results of standardized, highly sensitive assays have proved capable of safely excluding venous thrombosis in outpatients presenting with a low or intermediate clinical likelihood of the disease.
Clinical Prediction Rules
A major advance in the diagnostic approach to both venous thrombosis and PE has been a transition from a technique-oriented approach to one that utilizes Bayesian analysis. In this strategy, the pretest probability of the disease, calculated independently of a particular test result either through empirical means or through a standardized prediction rule, is evaluated in combination with a test’s likelihood ratio (derived from the sensitivity and specificity of that test) to create a posttest probability of the disease. This posttest probability can then be utilized as a basis for clinical judgment, either excluding the disease with a certain level of probability, confirming the disease with a certain level of probability, or supporting the need for additional diagnostic testing. This approach to diagnosis has proved especially useful in an era of noninvasive testing in which results are often presented as probabilities rather than as discrete answers.
Several clinical prediction rules for venous thrombosis have been developed and validated. The Wells rule, initially described in 1995 and revised subsequently to include nine clinical features, proved capable of stratifying patients with suspected venous thrombosis into three probability categories—low, moderate, and high—in which the incidence of venous thrombosis approximates 3%, 17%, and 75%, respectively. By utilizing this prediction rule in combination with lower extremity ultrasonography, the diagnosis of venous thrombosis can be safely excluded in patients with a low clinical likelihood of venous thrombosis in combination with a negative lower extremity ultrasound and confirmed in patients with a high clinical probability and a positive lower extremity ultrasound. Such an approach dramatically reduces the need for contrast venography or serial lower extremity ultrasound studies.
The Wells prediction rule was again revised to include 10 clinical characteristics capable of stratifying outpatients into “clinically likely” and “clinically unlikely” categories ( Table 57-2 ). For outpatients falling into the clinically unlikely category, DVT was reliably excluded when the result of a sensitive D-dimer assay was negative, thereby limiting the need for ultrasound evaluation. The ability to exclude venous thrombosis in outpatients using a clinical prediction rule and negative results of a D-dimer assay has been confirmed in other studies. It should be emphasized that clinical prediction rules constructed and validated in outpatients should be viewed critically before they are applied to an inpatient population.
|Active cancer (patient receiving treatment for cancer within the previous 6 mo or currently receiving palliative treatment)||1|
|Paralysis, paresis, or recent plaster immobilization of the lower extremities||1|
|Recently bedridden for 3 days or more, or major surgery within the previous 12 wk requiring general or regional anesthesia||1|
|Localized tenderness along the distribution of the deep venous system||1|
|Entire leg swollen||1|
|Calf swelling at least 3 cm larger than that on the asymptomatic side (measured 10 cm below the tibial tuberosity)||1|
|Pitting edema confined to the symptomatic leg||1|
|Collateral superficial veins (nonvaricose)||1|
|Previously documented deep venous thrombosis||1|
|Alternate diagnosis at least as likely as deep venous thrombosis||–2|
|Score||Clinical Assessment Probability|
Diagnosis of Pulmonary Embolism
Certain parallels between the approaches to the diagnosis of DVT and PE exist. Perhaps the most important parallels are that clinical evidence in isolation, although capable of raising the suspicion of the disease, cannot be relied on to confirm or exclude the diagnosis and that the use of clinical prediction rules in combination with noninvasive testing can substantially decrease the need for invasive diagnostic testing.
Standard Laboratory Evaluation
Routine laboratory studies cannot make the diagnosis of PE. Although none have the discriminatory power to confirm the diagnosis of embolism, they do provide valuable adjunctive information and support for therapeutic interventions and may confirm the presence of an alternative diagnosis.
The majority of patients with PE have abnormal chest radiographs. However, these abnormalities are usually subtle, nonspecific, and therefore nondiagnostic ( Fig. 57-3 ). In the PIOPED study, the most common radiographic abnormalities were atelectasis and pulmonary opacities. There is some confusion about the diagnostic configuration of radiographic abnormalities due to embolism. Although usually abutting a pleural surface, the opacities can be of any shape, not necessarily wedge-shaped ( eFig. 57-6 ). Although pleural effusions are seen in almost half of the patients, the majority of effusions are small and involve only blunting of the costophrenic angle. Findings once considered specific for embolism, such as the Westermark sign (focal areas of avascularity, eFig. 57-7 ), the Hampton hump (pleural-based, wedge-shaped opacity, eFig. 57-8 ), and the Fleischner sign (prominence of the central pulmonary artery, eFig. 57-9 ), have not proved to have discriminatory value. In a patient with hypoxemia or pulmonary complaints, a normal chest radiograph may be quite useful in raising the index of suspicion of embolism and in excluding confounding diagnostic options. The major roles of chest radiography in suspected PE, therefore, are to exclude competing diagnoses and, if ventilation-perfusion (V/Q) scanning is anticipated, to evaluate the pulmonary parenchyma.
Likewise, electrocardiographic findings in PE, although common, are diverse and nonspecific. The most common abnormalities include nonspecific tachycardia, T wave inversion, and abnormalities of the ST segment. With more extensive occlusion, the electrocardiogram may reveal the more “classic” findings of right heart strain, including an “S 1 Q 3 T 3 “ pattern, a pseudoinfarction pattern (Qr in V 1 ), a complete or incomplete right bundle-branch block, or right axis deviation. Rhythm disturbances other than sinus tachycardia are uncommon and usually confined to patients with underlying cardiac disease.
Arterial blood gas analysis is helpful, although not definitive. Arterial hypoxemia may be present, and the more massive the obstruction, the more severe the hypoxemia is likely to be. However, many other conditions also cause hypoxemia, and embolism often does not cause hypoxemia or even a widening of the alveolar-arterial P o 2 difference. Hypocapnia usually is present with embolism; hypercapnia, conversely, is rare. Hypercapnia appears with embolism only in patients with marked antecedent ventilatory limitation or when such limitation has been imposed because the patient is on controlled mechanical ventilation.
Echocardiography may serve a valuable role in the diagnostic approach to PE. Under appropriate clinical circumstances, the detection of unexplained right ventricular volume ( eFig. 57-10A ) or pressure overload might suggest the possibility of embolism and lead to confirmatory testing ( eFig. 57-10B ). A distinct echocardiographic pattern involving akinesia of the mid-free right ventricular wall with apical sparing has been described, known as McConnell sign ( ) Direct visualization of right heart chamber emboli is uncommon but possible ( ). Properly performed transesophageal echocardiography has demonstrated excellent specificity for the detection of proximal emboli that involve the pulmonary trunk and the right and left main pulmonary arteries. Transesophageal echocardiography also has proved valuable in the evaluation of competing diagnostic possibilities such as right ventricular infarction, endocarditis, pericardial tamponade, and aortic dissection in patients with unexplained shock and evidence of elevated central venous pressure. The overall sensitivity of transthoracic echocardiography in PE approximates 50%. Therefore it cannot be considered a primary diagnostic technique. Consideration can be given to its use in that subset of patients with suspected massive PE who are too ill for transportation or who have an absolute contraindication to the administration of a contrast agent.
Despite significant limitations, V/Q lung scanning can be a valuable step in the diagnosis of PE if one of the two definitive results are found (i.e., either a negative or a high-probability scan).
First, a negative study excludes the diagnosis of PE with the same degree of certainty as a negative pulmonary angiogram ( Fig. 57-4 ) and with a higher degree of certainty than is achieved by a negative CT scan. This conclusion is illustrated by the results of two large prospective clinical trials comparing perfusion scanning to pulmonary angiography, the PIOPED trial (performed in the United States) and the Prospective Investigative Study of Acute Pulmonary Embolism Diagnosis (PISA-PED) trial (performed in Europe). In both trials, normal pulmonary perfusion scanning was a highly sensitive method for excluding the presence of PE. The value of a normal perfusion scan was not diminished even in a subset of patients who had a high pretest probability for PE or were critically ill.
The significance of a normal perfusion scan, as reported by the PIOPED and PISA-PED trials, is consistent with all published longitudinal studies. A meta-analysis of diagnostic studies for PE calculated the incidence of PE following a normal perfusion scan to be 0.3%. A subsequent case series of consecutive patients followed clinically after objective testing for PE disclosed no PE in 188 patients who had normal perfusion scans. These data support the clinical guidelines of the American Thoracic Society, British Thoracic Society, American Heart Association, and European Society of Cardiology, all of which recommend that a normal perfusion scan be accepted as reliably ruling out PE, with the same validity as a pulmonary angiogram.
Second, a “high-probability” study (one characterized by multiple, segmental-sized, mismatched defects) is associated with embolism in approximately 87% of patients, as shown in the PIOPED study ; when coupled with a high clinical probability of embolism, the positive predictive value increased to 96% ( Fig. 57-5 ).
Limitations of V/Q scanning are significant. For example, the PIOPED data provided several pieces of disquieting information: (1) the overwhelming majority of patients with suspected embolism did not have scan findings that fell into a high-probability or normal category, the only categories that can be considered definitive; (2) the majority of patients with embolism did not have a high-probability scan finding; (3) the overwhelming majority of patients without embolism did not have a normal scan; and (4) a substantial and clinically significant percentage of patients with scan findings interpreted as intermediate probability (33%) and low probability (16%) were subsequently demonstrated to have angiographic evidence of embolism. It is essential that clinicians recognize that the concept of a low-probability scan ( eFig. 57-11 ) is misleading and potentially dangerous because of the frequency of PE in patients exhibiting this scan pattern.
In order to improve perfusion scan specificity, traditional interpretive criteria, including the PIOPED criteria, rely on the number and size of the perfusion defects, as well as on the results of a concurrent ventilation image. The intended basis for doing so is to differentiate primary vascular obstruction (“mismatched” defects) from primary parenchymal disorders that result in compensatory pulmonary vasoconstriction (“matched” defects). The PISA-PED investigators utilized a fundamentally different interpretive scheme that relied on the shape of the perfusion defects regardless of their number or size or their association with ventilation findings. The results of this study suggest that embolism can be diagnosed accurately and the need for angiography limited by perfusion results combined with an assessment of clinical likelihood in the absence of ventilation imaging. An analysis of a subset of patients from the PIOPED study came to a similar conclusion.
The diagnostic approach to PE in patients with underlying chronic obstructive pulmonary disease (COPD) remains especially problematic because the presentation of PE in this population may closely mimic an exacerbation of their underlying disease. Unfortunately, the value of V/Q scanning in this population is even more limited than that in the general population because an even higher proportion of scans fall into an indeterminate category. However, among the more than half of COPD patients who had high-probability, normal, or near-normal scans, both the positive predictive value and the negative predictive value were equivalent to that in the general population.
Computed Tomography Pulmonary Angiography
CTPA has represented a major advance in the diagnosis of PE (see also Chapter 18 ). Unlike V/Q scanning, it provides the ability to visualize emboli directly, as well as to detect parenchymal abnormalities that may support the diagnosis of embolism or provide an alternative basis for the patient’s complaints ( Fig. 57-6 ; ). The reported sensitivity of chest CT scanning for embolism has ranged from 57% to 100%, with a specificity ranging from 78% to 100%. Factors responsible for this wide divergence relate to the proximal extent of vascular obstruction that can be detected and, in part, to advances in CT technology that allow higher resolution, dramatically faster scanning times, more peripheral visualization, and less motion artifact than that provided by earlier-generation scanners. Sensitivity and specificity of CT scanning for emboli involving the main and lobar pulmonary arteries exceeds 95%. Vascular involvement confined to segmental or subsegmental pulmonary vessels is associated with a decline in both sensitivity and specificity. In one series, the sensitivity of CT scanning for subsegmental arteries reported by two readers ranged between 71% and 84% even after nonevaluable scans were excluded. Isolated involvement of the subsegmental pulmonary arteries is not unusual and, in various series, may be found in up to 30% of patients. These findings suggest that filling defects consistent with embolism involving the main or lobar pulmonary arteries can be considered diagnostic of embolism. By contrast, defects involving the segmental and subsegmental arteries can be considered suggestive of embolism but should be supported by additional objective data. The absence of detectable filling defects reduces the likelihood of embolism but appears incapable of excluding the possibility with the same degree of certainty as a negative V/Q scan. The importance of treating patients with CTPA findings suggestive of exclusively subsegmental emboli has been questioned recently, especially in patients with good cardiopulmonary reserve and no coexisting DVT or persistent risk factors. However, no large trials have yet demonstrated the safety of withholding anticoagulants in this patient population.
Although CTPA scanning has increased the number of cases of acute PE that are diagnosed, it has inherent limitations. The technique requires infusion of intravascular iodinated contrast agents, and the most common serious complications of testing arise from their use. During CT scanning, the peripherally infused dye fills the lumen of the pulmonary arteries, hopefully at the exact time that the chest is imaged. Emboli are detected as focal defects in pulmonary artery filling. When performed and interpreted expertly, CTPA scans are capable of identifying emboli in the segmental or larger pulmonary arteries (as does V/Q scanning ). However, certain areas, such as the hila, are prone to false positives. Reading emboli in these areas should be done with special care. Perhaps more importantly, CT scans have difficulty imaging emboli in subsegmental pulmonary arteries. In selected populations, smaller emboli may account for up to 20% to 30% of PEs and represent the very cases in which V/Q scans are the most limited.
The National Institutes of Health–funded Prospective Investigation of Pulmonary Embolism Diagnosis–2 (PIOPED-2) study highlighted the strengths and limitations of CTPA. Before enrolling the study population of 1090 patients, the investigators excluded 1350 patients because their abnormal creatinine levels reflected some degree of renal dysfunction, which would have increased the risk involved with the administration of contrast dye. An additional 272 patients were excluded because of a history of contrast dye allergy. During the performance of the trial, 6% of the scans were excluded because the images were of poor quality. Even after the inconclusive scans were disregarded, the sensitivity of the chest CT scans was only 83%, although it is difficult to be confident of the gold standard used to compare to CTPA.
The PIOPED-2 investigators also excluded another 976 patients from the study because they had histories of long-term anticoagulation therapy. This highlights another weakness of CTPA in that luminal filling defects remain long after an acute pulmonary embolic event, so the test cannot easily differentiate between chronic and acute VTE. This is clinically relevant because the rate of recurrence is about 7% during the half-year after an acute PE and about 3% per year for the subsequent 5 years.
Finally, CTPA can expose patients to clinically significant doses of radiation. Current clinical protocols deliver a radiation dose to the female breast ranging from 4 to 6 cGy per scan. This dose is especially concerning because most CTPA evaluations are negative, even when the criteria for scanning are rigorously followed. In addition, because PE has a relatively high rate of recurrence, patients are commonly reimaged after their first embolism. Younger women, who have a known elevated incidence of PE, are particularly at risk from radiation damage to breast and lungs.
From a clinical (outcomes-based) perspective, CT scanning and V/Q scanning have many similarities, and either can be used to exclude clinically significant PE in stable patients under many circumstances. Outcome studies have demonstrated that withholding anticoagulant therapy in patients with a negative CT scan coupled with a negative lower extremity ultrasound study appears to be a safe strategy except in those patients who present with a high clinical likelihood of embolism. Similarly, withholding anticoagulation from patients with “non–high-probability” findings on V/Q scan coupled with negative lower extremity studies is safe, except in those with inadequate cardiopulmonary reserve. A recent randomized, controlled trial comparing CT and V/Q for the management of suspected PE found the two studies to be comparable for excluding clinically significant PE. Specifically, the study disclosed no difference in outcome among patients in whom anticoagulation was withheld on the basis of the combination of a negative CT and negative leg studies compared with patients in whom anticoagulation was withheld based on the combination of a “non–high-probability” V/Q scan and negative leg studies or on a negative V/Q scan (without the need for leg studies). However, it should be emphasized that the outcome studies were done on relatively stable patients. Those with instability or poor cardiopulmonary reserve may require a higher degree of diagnostic certainty in order to rule out PE.
Single-Photon Emission Computed Tomography Ventilation/Perfusion Scans
Single-photon emission computed tomography (SPECT) is a nuclear medicine technique that constructs three-dimensional images from scintigraphic data, in a similar way that CT constructs three-dimensional images from transmitted radiographs. SPECT ventilation and perfusion imaging (SPECT-V/Q) is a promising tool for diagnosing acute PE. The tomographic images can identify perfusion defects in areas of the lung that are hard to visualize with planar V/Q scans because of interposing lung tissue (e.g., the medial basilar segments of the lower lobes ). As a result, SPECT-V/Q has far fewer nondiagnostic test results than planar V/Q, which removes one of the major limitations of planar V/Q testing. In fact, nondiagnostic results were reported in only 0.5% to 3% of SPECT-V/Q studies. Another advantage of SPECT-V/Q is that it delivers only about one fourth of the radiation to the breast tissue than what is typical for CTPA scans.
Although the lack of a diagnostic gold standard limits our knowledge of its true accuracy (a phenomenon common to many tests for PE), several different analyses support the accuracy of SPECT-V/Q for PE. SPECT-V/Q had a high degree of agreement with the results of CTPA in patients suspected of having PE. When compared with consensus diagnoses of PE (which, admittedly, included the SPECT-V/Q results themselves) SPECT-V/Q had sensitivities and specificities for PE that were typically in the 95% to 100% range. Similar to studies with CTPA scans, the clinical outcomes of patients with negative SPECT-V/Q, in whom anticoagulation was not given, were excellent. These findings suggest that SPECT-V/Q is highly sensitive for clinically important PE in the populations tested.
SPECT-V/Q is a promising technique but has not undergone extensive testing sufficient to merit replacement of CTPA as the primary diagnostic tool for PE. It may be especially useful for patients with nondiagnostic CTPA results or in those in whom the lower radiation dose to the chest would be especially advantageous. It may be useful as well for the follow-up of PE patients in order to detect and quantify residual perfusion defects.
Lower Extremity Venous Evaluation
Because the majority of PEs arise from the deep veins of the lower extremities, the detection of lower extremity proximal vein thrombosis in a patient suspected of embolism, although not confirming PE, is strongly suggestive of that diagnosis and has an equivalent therapeutic implication. Positive ultrasound findings without symptoms or signs referable to the lower extremities should be interpreted judiciously, especially in patients with low pretest probability of PE, because even a highly specific test can yield false-positive results in some circumstances. Lower extremity ultrasonography has a low yield in patients without leg symptoms or risk factors strongly suggestive of thromboembolism. Conversely, it is typically positive in only about 10% to 20% of patients with suspected PE and in 50% of patients with proven PE. Therefore a negative ultrasound finding cannot exclude the diagnosis. CT venography (see eFig. 57-4 ) as an adjunct to chest CTPA scanning appears to be capable of detecting femoropopliteal thrombosis with the same accuracy as duplex ultrasonography while also detecting pelvic and abdominal thrombosis. However, the combined CT technique is technically complicated and significantly increases the amount of pelvic radiation exposure to the patient.
The utility of D-dimer testing in PE diagnostic pathways is limited by the same shortfalls as those encountered in venous thrombosis pathways (i.e., a low specificity, which makes it most useful as an exclusionary technique in outpatients, and a lack of standardization). However, studies have demonstrated that a normal D-dimer result can safely exclude embolism in patients with a low clinical probability of disease. Although preliminary data suggest that a highly sensitive assay is capable of excluding embolism at all levels of clinical probability, these results require confirmation.
The studies reviewed to date are capable of excluding or confirming the diagnosis of embolism in the majority of patients with suspected embolism. Angiography should be considered in patients in whom the diagnosis has not been confirmed or excluded using noninvasive techniques and when it is considered unsafe to withhold anticoagulation, when cardiopulmonary instability is present, and when the results of diagnostic testing are at such odds with the clinical impression as to warrant the risk of the procedure. Like contrast venography, however, pulmonary angiography has a number of limitations as a gold standard. First, the procedure is invasive and not without risk, especially in patients with acute right ventricular failure. However, experience has demonstrated that the perception of risk associated with angiography outweighs the actual risk. Pulmonary angiography can be performed quite safely if certain safeguards are observed and experienced personnel are involved.
Even though the risk of angiography should be nominal, the procedure has other limitations. One is accessibility: Angiography is performed in a special facility to which the patient must be transported. In some institutions, the logistical problems involved are modest; in others, they are substantial. The other limitation is interpretation. The interpretation of pulmonary angiograms is heavily influenced by three factors: location of the thromboembolic obstruction, quality of the images, and experience of the interpreters. Only two angiographic findings are diagnostic of acute embolism: the filling defect and abrupt cutoff of a vessel ( Fig. 57-7 ). Technical adequacy of the angiogram is critical to accurate identification of both. Flow artifacts can falsely suggest a filling defect. It is essential that good vessel opacification be obtained and that the filling defects be identified as present on a sequence of images.