Once suspicion of PE occurs based on the predisposing factors, symptoms, and signs, we
recommend an escalating approach, starting with a validated clinical model, followed by a D-dimer. We prefer the Wells model (13) (see Table 34.3), although others such as the Geneva score or empiric estimates may perform well (14,15). In a multicenter study of 930 patients in whom only 86 (15%) had PE, emergency department physicians using the Wells criteria reported a high pretest probability in 64 (7%), moderate in 339 (36%), and low in 527 (57%) of patients. PE was found in 24 (40.6%) of high, 55 (16.2%) of moderate, and seven (1.3%) of low-probability patients (13). More recently, this group reported on 1,126 outpatients and inpatients with a prevalence of VTE of 15.2%. By using a cut point of 4 (vs. 2, as in their prior study), 670 (60%) were categorized as low probability, and PE was diagnosed in 5% of them (16). Fifty percent of inpatients had a low probability versus 69% of outpatients. The prevalence of PE was 20% among inpatients vs. 11% among outpatients.

The quantitative rapid enzyme-linked immunosorbent assay (ELISA) generally provides the most satisfactory likelihood ratios: DVT, positive, or sensitivity -0.96, and negative,-0.12 and for PE, sensitivity of 0.96 and negative, 0.09 (17). Alternative D-dimer assays produce equivalent results. In isolation, however, the D-dimer may be misleading. In a study of 1,177 patients with a prevalence of PE of 17%, a negative D-dimer with a normal ventilation-perfusion (V/Q) scan had a posttest probability of PE of 0.4%. With a nondiagnostic V/Q scan, the posttest probability of PE was 2.8%, and if the V/Q scan was high probability, the posttest probability of PE was 65.4% (18). The D-dimer should be used in association with other testing. A positive D-dimer only indicates the need for additional testing. Unreliable positive D-dimer tests are often found in patients with cancer, atrial fibrillation, postoperative states, pregnancy, and sepsis or similar conditions.

The clinical model should be mated with a D-dimer assay as the initial paradigm for diagnosing PE. A low or intermediate clinical probability (see Table 34.3) with a negative D-dimer effectively excludes PE; the posttest probability of PE ranges from 0.7 to 2.0% (19). We believe further testing is not necessary. However, some would also obtain a venous ultrasound of the lower extremities. If the clinical probability is high, a D-dimer need not be done, because even if negative, the likelihood of PE is greater than 15% (17). If the D-dimer is positive, then imaging studies are necessary.

The most commonly used initial imaging study for the evaluation of possible PE is CT of the pulmonary arteries (CTA), frequently coupled with CT venography of the thigh veins (CTV). These studies are minimally invasive (injection of the dye) and are of additional value in identifying other lung lesions. A systematic review of single-slice CTA outcome studies reported sensitivities of 53% to 100% and specificities of 81% to 100% (20). The identification of subsegmental clots has been 30% or less; however, some 6% to 36% of PEs have been limited to subsegmental vessels (21). In a prospective study of 259 patients, the sensitivity of single-slice CT angiography was only 70% [95% confidence interval (CI), 62% to 78%] and specificity, 91% (95% CI, 86% to 95%). The likelihood ratio for a false-negative CT was 0.3, close to that of a low-probability scan in PIOPED. The false-negative rate was reduced from 30% to 20% if ultrasonography was negative and to 5% if the lung scan was also nondiagnostic. The false-positive rate was 15% in lobar arteries and 38% in segmental arteries (22).

The Prospective Investigation for Pulmonary Embolism Diagnosis II (PIOPED-II) was a multicenter prospective investigation of the accuracy of multidetector CTA in combination with CTV (additional venography) and the application of a validated clinical model (13) (see Table 34.3) (23). PIOPED-II used a composite reference test to diagnose or exclude PE. PE was found in 192 (23%) of 824 patients. The sensitivity of CTA was 83%, and the specificity was 96%; 51 CTA studies were not adequate for interpretation. The likelihood ratio for a positive test was 19.6 (95% CI, 13.3 to 29.0), and the likelihood ratio for a negative test was 0.18 (95% CI, 0.13 to 0.24). A positive likelihood ratio of greater than 10 and a negative likelihood ratio of less than 0.1 provide a highly definitive change from pretest to posttest (24). The sensitivity of the CTA-CTV was 90%, and the specificity was 95%; 87 CTA-CTV studies were inadequate for interpretation; the likelihood of a positive test was 16.5, and for a negative test, 0.11 (see Table 34.4). Location of the PE has significant implications. The positive predictive values for PE in the main PA and primary branches were 97% (116 of 120), 68% (32 of 47) in a segmental vessel, and 25% (two of eight) in subsegmental vessels. Table 34.4 emphasizes the effect of discordance between the test and the clinical impression. In the case of major discordance, both clinical and test impressions should be reevaluated with consideration of additional testing. The substantial number of inconclusive readings highlights the need to review the initial interpretation. Additionally, the study was conducted in centers with experience beyond that generally available. A negative CTA alone is insufficient to exclude the diagnosis of PE, but a low clinical score and a negative CTA confirm the exclusion of PE.

PIOPED II was an accuracy study. It compared its results with an independent gold standard. Such studies provide the understanding of test performance in both diagnosing and excluding disease (25). In contrast, outcome studies provide guidance for clinical management. The recent outcome studies of Perrier et al. (25,26) and Christopher Study Investigators (15) support the validity of CTA.

Ventilation/perfusion (V/Q) lung scans or perfusion lung scans were the usual initial imaging tests for PE for more than 20 years before the evolution of CTA/CTV. The only ventilation/perfusion (V/Q) results that permit definitive, rational clinical decision making are those indicating high probability (more than one segmental or larger perfusion
defect with normal ventilation—a mismatch) and those that are normal (no significant defects). In PIOPED, a high-probability V/Q scan had a positive predictive value of PE in 87% of patients, and the likelihood of PE was greater than 96% if the clinical suspicion was high. However, PE was found in only 74% of patients with this reading. If they had a history of PE, a PE was found in 4% of those with a normal or a near-normal V/Q scan. Only 27% of patients had high (13%) and normal (14%) readings. Intermediate- and low-probability V/Q scans should be reported and considered nondiagnostic. PE was found in 33% of patients with intermediate-probability scans and 14% of patients with low-probability scans, an incidence too high to abandon the diagnosis of PE and too low to initiate treatment (27). In patients with substantial chronic obstructive pulmonary disease, only 5% had high-probability scans, and they had PE; no normal scans were found (28); lung scans are also of little value in patients with acute respiratory failure (29).