Prospective Study of the Diagnosis of Pulmonary Embolism (PE) Using Pretest Probability (PTP), D-Dimer (DD) and Appropriate Imaging

Abstract number: P2263

Patel¹ RK, Bagot¹ C, Lambie¹ J, Desai² S, Arya¹ R

¹¹Department of Haematological Medicine, King's College Hospital, London, UK ¹¹Department of Haematological Medicine, King's College Hospital, London, UK ²²Department of Radiology, King's College Hospital, London, UK

The optimal practical approach for diagnosis of PE should be rapid, accurate and safe. At King's College Hospital we devised a diagnostic algorithm that relies on clinical PTP (Well's score), plasma DD measurement and chest X-ray (CXR) in the initial phase. PE was excluded in those with low PTP and negative DD (<200 mg/L). Remaining subjects with normal CXR underwent ventilation-perfusion (VQ) scan, requiring subsequent multi-slice detector CT-pulmonary angiography (CTPA) only if results were non-diagnostic. Subjects with an abnormal CXR underwent CTPA, requiring further investigation with pulmonary angiography only if CTPA results were equivocal. In this prospective management study, 651 consecutive patients (59.1% female) with suspected PE (mean age 57.9 years, SD 18.3) were assessed over a 2-year period. PTP scores were as follows: 16.1% high, 80.9% moderate, 3% low. DD level (range10–9288 mg/L, median 2610 mg/L) was >200 mg/L in 88% of subjects. CXR was abnormal in 63.5% subjects. VQ scan was performed in 25.7% (61.5% low probability, 27.7% intermediate probability, 10.8% high probability scans) and CTPA in 71.7% of subjects (73% normal, 2.8% equivocal, 21.8% positive, 2.4% failed). Overall prevalence of PE was 20.1%, with 69.5% diagnosed by CTPA, 17.6% by VQ and 12.9% both. DD levels of <200 mg/L had a 98.5% negative predictive value (NPV) for PE and low PTP had a 77.8% NPV for PE (3.1% low PTP, 74% moderate PTP, 22.9% high PTP), confirming the value of PTP and DD in safely excluding PE. CTPA, currently recommended as imaging modality of choice, provided the diagnosis in 94.8% cases. The utility of VQ was improved by limiting scans to those with normal CXR thus reducing the proportion of patients with intermediate probability scans. Combining clinical probability and diagnostic tests in robust algorithms streamlines diagnosis, increases accuracy and promotes efficient use of resources.