Author + information
- Received April 25, 2002
- Revision received May 31, 2002
- Accepted June 7, 2002
- Published online October 16, 2002.
- Kelly L Dunn, BA*,
- Jonathan P Wolf, BS*,
- David M Dorfman, MD, PhD†,
- Patricia Fitzpatrick, BS†,
- James L Baker, MD, MPH‡ and
- Samuel Z Goldhaber, MD*,* ()
- ↵*Reprint requests and correspondence:
Dr. Samuel Z. Goldhaber, Cardiovascular Division, Department of Medicine, Brigham and Women’s Hospital, 75 Francis Street, Boston, Massachusetts, USA 02115.
Objectives We sought to determine: 1) whether normal d-dimer enzyme-linked immunosorbent assay (ELISA) assays predicted the absence of pulmonary embolism (PE) in the high-volume emergency department (ED) of the Brigham and Women’s Hospital, and 2) whether ED physicians accepted normal d-dimer levels as confirmation of no PE without further diagnostic testing such as lung scanning, chest computed tomography (CT) scanning, or pulmonary angiography.
Background Although the plasma d-dimer ELISA is a sensitive screening test for excluding acute PE, this laboratory marker has not been widely integrated into clinical algorithms such as creatine kinase-MB fraction or troponin testing for acute myocardial infarction.
Methods We mandated that ED physicians order d-dimer ELISA tests on all patients suspected of acute PE. We reviewed the clinical record of each ED patient initially evaluated for suspected PE during the year 2000. We determined whether additional imaging tests for PE were obtained and whether the final diagnosis was PE.
Results Of 1,106 d-dimer assays, 559 were elevated and 547 were normal. Only 2 of 547 had PE despite a normal d-dimer. The sensitivity of the d-dimer ELISA for acute PE was 96.4% (95% confidence interval [CI]: 87.5% to 99.6%), and the negative predictive value was 99.6% (95% CI: 98.7% to >99.9%). Nevertheless, 24% of patients with normal d-dimers had additional imaging tests for PE.
Conclusions The d-dimer ELISA has a high negative predictive value for excluding PE. By paying more attention to normal d-dimer results, fewer chest CT scans and lung scans will be required, and improvements may be realized in diagnostic efficiency and cost reduction.
Although the plasma d-dimer enzyme-linked immunosorbent assay (ELISA) is a sensitive screening test for excluding acute pulmonary embolism (PE) (1), this laboratory marker has not been widely integrated into clinical algorithms such as creatine kinase-MB fraction or troponin testing for acute myocardial infarction. Accordingly, we sought to determine: 1) whether normal d-dimer ELISA assays predicted the absence of PE in the high-volume emergency department (ED) of the Brigham and Women’s Hospital, and 2) whether ED physicians accepted normal d-dimer levels as confirmation of no PE without further diagnostic testing such as lung scanning, chest computed tomography (CT) scanning, or pulmonary angiography.
We mandated that ED physicians order d-dimer ELISA tests on all patients suspected of acute PE. We utilized the VIDAS d-dimer assay (bioMérieux; Marcy l’Etoile, France) (2,3), and our laboratory provided turnaround within 2 h on a round-the-clock basis. A normal d-dimer to exclude acute PE was defined as ≤500 ng/ml. After approval by our Human Research Committee, we reviewed the clinical record of each ED patient initially evaluated for suspected PE during the year 2000. We determined whether additional imaging tests for PE were obtained and whether the final diagnosis was PE, as determined by high-probability lung scan, positive chest CT, or positive pulmonary angiogram. Furthermore, we followed up each patient in whom PE was excluded for the next six months to determine whether any further imaging tests were ordered and whether PE was subsequently diagnosed by lung scan, chest CT, or pulmonary angiography. Our results and recommendations for patients whom we defined as having no PE were formulated on the basis of this six-month observation period.
Overall, 1,106 d-dimer assays were obtained in our ED for initial evaluation of PE, with 559 abnormally elevated levels and 547 normal results (Fig. 1). The normal d-dimer group and the elevated d-dimer group had similar demographic characteristics (Table 1). Of the 55 patients diagnosed with acute PE during the initial ED evaluation or six months of follow-up, 53 presented with elevated d-dimer levels. Only 2 of 547 had PE despite a normal d-dimer (Table 2).
During six months of follow-up, no additional patients in the normal d-dimer group (≤500 ng/ml) were subsequently diagnosed with PE. Thus, the sensitivity of the d-dimer ELISA for acute PE was 96.4% (95% confidence interval [CI]: 87.5% to 99.6%), and the negative predictive value was 99.6% (95% CI: 98.7% to >99.9%). The specificity was 52.0% (95% CI: 48.8% to 54.9%), and the positive predictive value was 9.5% (95% CI: 7.2% to 12.2%).
One of the two patients with false negative d-dimer results (461 ng/ml) was a 75-year-old woman who was hospitalized with an exacerbation of chronic obstructive pulmonary disease. On the second inpatient day, she underwent CT scanning, which demonstrated acute PE. The other patient was a 26-year-old woman with a history of recurrent PE who had been noncompliant with anticoagulant therapy prescribed before the ED visit. Her initial d-dimer was 376 ng/ml, and she was discharged home. She returned to the ED within 24 h, had a repeat d-dimer of >1,000 ng/ml, and was diagnosed with acute PE by chest CT scan.
Our results indicate that the plasma d-dimer ELISA confers a high negative predictive value for excluding acute PE in a large ED cohort of consecutive patients in whom the diagnosis of PE is suspected. However, despite these results, our ED physicians pursued the work-up of acute PE and disregarded the normal d-dimer values in one of every four patients. This clinical practice resulted in additional costly testing, radiation exposure, and delay of the disposition of patients presenting to the ED.
With a high sensitivity, high negative predictive value, and rapid turnaround time, the d-dimer ELISA functioned as an excellent screening test for suspected PE. However, as others have found (4), the d-dimer ELISA was often elevated in the absence of PE. It had a low specificity and poor positive predictive value.
Some investigators have had results that differed from ours. In a study of 198 ED patients, a whole blood assay had a negative predictive value of only 83% (5). This finding calls into question whether rapid whole blood d-dimer assays have adequate sensitivity compared with ELISA when utilized for suspected PE. When a normal whole blood d-dimer assay result was combined with a normal bedside alveolar dead-space test, a low prevalence of PE was accurately predicted (6).
Our findings do not pertain to inpatients suspected of PE. When d-dimer assays were tested as part of the diagnostic algorithm among inpatients suspected of PE, costs increased with d-dimer testing because more patients with elevated d-dimer results underwent expensive imaging studies (7). Unfortunately, there was no improvement in clinical outcomes among patients in the d-dimer group. The investigators questioned whether d-dimer testing on inpatient services was cost-effective. We note that almost all studies of plasma d-dimers in the diagnosis of suspected venous thromboembolism, including ours, have been undertaken in outpatients, not inpatients (8). The available evidence indicates that in contrast to outpatients, the diagnostic strategy of d-dimer measurement appears to be useless in hospitalized patients (9).
We studied a population of outpatients, many of whom had no concomitant acute medical problem (which might elevate d-dimer levels) other than possible PE. Patients for whom a diagnosis of PE was initially excluded were followed for six months to determine if PE was subsequently diagnosed. Our findings of high sensitivity and high negative predictive value are strengthened by the large sample size of 1,106 consecutive patients. Our results are consistent with those of Perrier et al. (10) who studied 444 consecutive outpatients with suspected PE. However, our results cannot necessarily be generalized to other ELISA assays of d-dimer because each assay has a different sensitivity and negative predictive value for acute PE (11).
A cautionary note is appropriate because almost all diagnoses of PE in this study were established or excluded on the basis of noninvasive testing. It is important to recognize and acknowledge the vagaries and limitations of noninvasive testing for suspected PE. Criteria for interpreting lung scans are complicated (12), and lung scanning, unlike chest CT, does not provide adequate confirmation or exclusion of PE in the majority of patients who undergo this test (13). Among unselected outpatients with suspected PE, as many as 30% of the emboli may escape detection on first-generation single-slice CT scanners (14). New, multislice scanners increase imaging speed and allow improvement in resolution from 5 mm to 1 mm (15). In the short time that has elapsed since completion of our ED study in which lung scanning was used almost twice as frequently as chest CT scanning, CT has now overtaken lung scanning as the primary noninvasive imaging modality for suspected PE (16,17). However, despite the limitations of noninvasive testing, an integrated diagnostic strategy has been validated in which very few patients ultimately require pulmonary angiography (18). Therefore, we believe that our sensitivity and specificity analyses for the VIDAS d-dimer ELISA are valid in the ED population that we describe.
Our findings indicate that a negative VIDAS d-dimer ELISA can almost always exclude acute PE in the ED setting. Furthermore, additional opportunities exist for educational initiatives on the role of the d-dimer ELISA in the evaluation of possible PE. Dual intervention with intensified teaching and more rigidly enforced diagnostic algorithms may yield improvements in efficiency and cost reduction.
- emergency department
- enzyme-linked immunosorbent assay
- confidence interval
- computed tomography
- pulmonary embolism
- Received April 25, 2002.
- Revision received May 31, 2002.
- Accepted June 7, 2002.
- American College of Cardiology Foundation
- Miron M.J.,
- Perrier A.,
- Bounameaux H.,
- et al.
- van der Graaf F.,
- van den Borne H.,
- van der Kolk M.,
- de Wild P.J.,
- Janssen G.W.,
- van Uum S.H.
- Gottschalk A.,
- Sostman H.D.,
- Coleman R.E.,
- et al.
- ↵Herold CJ. Spiral computed tomography of pulmonary embolism. Eur Respir J 2002;Suppl 35:13S–21S
- Bounameaux H.