Diagnosing Pulmonary Embolism With Improved Certainty and Simplicity

Pulmonary embolism (PE) is both a common and ubiquitous disorder which, if incorrectly diagnosed and untreated, is associated with high patient morbidity and mortality due to nonfatal and fatal recurrent venous thromboembolism. Pulmonary embolism is one of the “great masqueraders” of medicine. The presenting symptoms and signs are nonspecific; therefore, diagnostic testing is necessary to establish the presence or absence of PE. The current diagnostic pathways available for the clinician are a profound improvement compared with those available historically.

Over the past 25 years, substantive improvement in clinical trials methods and major advances in diagnostic technology have resulted in more rigorous approaches to the diagnosis of deep vein thrombosis and PE. Discharge data from the National Hospital Discharge Survey from 1979 to 1999 indicate sequential introduction of improved diagnostic strategies for deep vein thrombosis and PE.¹ Diagnostic approaches to deep vein thrombosis showed an initial marked increase in use of venography followed by a rapid decline that coincided with increased use of Doppler ultrasonography.¹ Diagnosis of PE was characterized by initial marked increases in lung scanning followed by a rapid decline as use of ultrasonography increased considerably and pulmonary angiography modestly increased.¹ By 2000, the use of spiral computed tomography (CT) for hospitalized patients sharply increased and exceeded the use of ventilation-perfusion lung scanning.²

The literature on the diagnosis of PE reflects a classic pathway of initial enthusiasm subsequently tempered by the limitations of the evidence and the need for complex pathways. Divergent and contradictory approaches were recommended for the use of lung scanning in the diagnosis of patients with clinically suspected PE during the initial years of its use.³ Several rigorous studies^{3 - 5} including PIOPED⁵ showed that lung scanning alone or combined with clinical probability assessment established or ruled out the diagnosis of PE only for a minority of patients–those with concordant clinical and lung scan findings. In particular, the use of the low- or intermediate-probability lung scan findings alone resulted in diagnostic uncertainty, placing the patient with suspected PE at an unacceptable risk.^{3 - 6}

This uncertainty with using clinical assessment combined with lung scanning or lung scanning alone for the majority of patients with suspected PE, as well as the strong association of PE with proximal vein thrombosis,³ led to studies using serial noninvasive leg testing. The hypothesis was that if serial leg testing was negative for deep vein thrombosis, anticoagulant therapy could be safely withheld in patients with suspected PE and nondiagnostic lung scans. This hypothesis was confirmed for patients with suspected PE who had good cardiorespiratory reserve⁷ . This strategy was further refined in subsequent outcome studies^{8 - 10} using clinical probability assessment, ventilation-perfusion lung scanning, D-dimer testing, and serial Doppler ultrasonography of the lower extremities. Although complex and somewhat impractical due to the need for serial leg testing, this strategy had the advantage of reducing the need for pulmonary angiography. These trials improved understanding of the management of PE both for the accuracy of diagnosis and for prognosis, and importantly, for the safety of withholding anticoagulant therapy.

The introduction of spiral CT has had a profound influence on the questions addressed in recent clinical trials in patients with suspected PE. Spiral CT of the chest is now more widely available than ventilation-perfusion lung scanning² and has become a pivotal test in patients with suspected PE. Although some argue that single-detector spiral CT could be used alone,¹¹ current evidence^{12 - 13} indicates the need for caution and argues strongly against using single-detector spiral CT alone due to its inadequate sensitivity. For an acceptable patient outcome, single-detector spiral CT should be used with clinical probability assessment combined with a D-dimer assay or a Doppler ultrasound evaluation.^{14 - 16}

Multidetector-row spiral CT with simultaneous acquisition of multiple sections per scanner rotation is now widely available,¹⁷ and offers improved detection of PE compared with single-detector CT.¹⁷ Images of the entire chest can be acquired with 1 millimeter or less resolution with a breath hold of less than 10 seconds for 16-slice CT. Spiral CT venography can be readily performed in conjunction with a spiral CT of the chest, demonstrating the presence or absence of proximal vein thrombosis.¹⁷ Multidetector (slice) CT provides rapid acquisition of diagnostic images. From the time of dye injection, chest imaging is acquired in less than 30 seconds. After a 3- to 3.5-minute wait, spiral CT venographic imaging is acquired within 10 seconds.

Several recent studies^{18 - 20} have evaluated multidetector spiral CT of the chest. Perrier et al¹⁹ studied multidetector row CT in 756 emergency department patients with suspected PE using a multibranch diagnostic algorithm that included clinical probability assessment, D-dimer measurement, and lower limb ultrasonography. The authors showed that for patients who have a low or intermediate clinical probability and a negative D-dimer result, it is safe to withhold anticoagulant therapy without the need for chest imaging; this applied to one third of the patients presenting with PE. In addition, use of multidetector spiral CT obviated the need for lower extremity ultrasonography. However, limitations were that this was not a “true” management study, since it incorporated use of ultrasound and pulmonary angiography and there was a high rate (25%) of exclusions. Ghanima et al¹⁸ evaluated 432 outpatients with suspected PE using a strategy based on clinical probability, D-dimer testing, and multislice CT. The authors reported that their algorithm yielded a definite diagnosis in 96.5% of patients, and suggested that multislice CT was unnecessary in approximately 20% of the patients presenting with suspected PE (ie, those who had a low to intermediate clinical probability assessment and negative D-dimer result).

In this issue of JAMA, van Belle et al provide new, important data on management of suspected PE. The authors report the findings of the Christopher Study,²⁰ a clinical outcome study of 3306 consecutive patients with clinically suspected PE who were evaluated using a diagnostic algorithm with a dichotomized clinical decision rule, D-dimer testing, and CT. The dichotomized clinical decision rule⁹ allowed a clinical probability assessment for PE of unlikely or likely, thereby simplifying the diagnostic algorithm. The findings indicated that it was safe to withhold anticoagulant therapy for patients who had an unlikely clinical probability of PE and a negative D-dimer result; these patients represented one third of patients presenting with suspected PE.

The Christopher Study²⁰ is one of the largest studies of management of patients with suspected PE, and is a “true” management study because all clinical decisions were based on the decision rule and the D-dimer and CT results. Dichotomizing the decision rule made it more practical and no diagnostic tests other than D-dimer and CT were needed to guide management. In addition, in this large cohort the exclusion rate was low (5.6%).

The diagnostic algorithm reported by van Belle et al and supported by other studies^{18 - 19} offers the clinician a practical and safe pathway for evaluating patients with suspected PE, avoids ionizing radiation exposure in as many as one third of patients, and provides high likelihood of a good outcome. However, in the absence of rigorous evidence confirming the safety of a negative finding by multidetector spiral CT imaging of the chest, anticoagulant therapy should not be withheld using spiral CT alone. The clinician must combine spiral CT findings with clinical probability assessment and the D-dimer assay.

It is possible that combining CT venography with spiral CT of the chest may provide diagnostic pathways; the recently completed PIOPED II study²¹ addresses this question. Alternatively, lung scanning and Doppler ultrasonography combined with a clinical probability assessment and D-dimer assay offers the advantage of less ionizing radiation and avoids exposure to radiographic contrast. Although the diagnostic algorithms are much more complex, they offer a diagnostic strategy for patients in whom contrast exposure should be avoided or ionizing radiation should be limited, such as those with renal impairment or who are pregnant.

The importance of using clinical probability assessments in patients with deep vein thrombosis is addressed in another article in this issue by Wells et al.²² Based on their review of data from 14 studies with more than 8000 outpatients with suspected deep vein thrombosis, the authors found that the prevalence of deep vein thrombosis based on probability estimates from clinical prediction rules was 5%, 17%, and 53% respectively, in the low, moderate, and high clinical probability groups.

Based on their estimates of sensitivity, specificity, and likelihood ratios for D-dimer testing across the clinical probability groups, the authors suggest that in patients with low probability of deep vein thrombosis and a negative D-dimer result, the diagnosis of deep vein thrombosis can be excluded without use of ultrasonography. However, the authors emphasize that for patients with high clinical suspicion of deep vein thrombosis, D-dimer results should not affect clinical decisions. These conclusions are supported by the findings of a recent systematic review examining D-dimer for the exclusion of acute venous thrombosis and PE.²³

As the direct result of rigorous studies and innovative advances in diagnostic testing over the past 15 years, firm recommendations can now be made concerning practical and fairly simple diagnostic algorithms for evaluating patients with suspected PE or deep vein thrombosis. Future data from high-quality prospective management studies will continue to provide evidence-based refinement of these diagnostic pathways and ultimately will serve to improve outcomes for patients with suspected thromboembolic disease.