Clinical Decision Rules for Excluding Pulmonary Embolism: A Meta-Analysis

Study Questions:

What are the relative merits of physician gestalt or clinical decision rules, in combination with D-dimer testing, to exclude a suspected diagnosis of pulmonary embolism (PE)?


The authors performed a systematic literature review and meta-analysis using a bivariate analysis approach. They searched MEDLINE and EMBASE for articles published between 1966 and June 2011 reporting prospective studies of various clinical decision rules or clinician gestalt in patients suspected of having PE. Included were only those studies in which the estimated probability of PE was verified by an appropriate reference standard. They used a bivariate model for diagnostic meta-analysis to obtain summary estimates of sensitivity and specificity, an approach that incorporates any correlation which may exist between sensitivity and specificity, and also uses a random-effects approach.


The authors identified 52 studies including 55,268 subjects that met inclusion criteria. There were 15 studies that used gestalt (sensitivity 0.85; specificity 0.51), 19 studies that used the Wells rule with the cutoff value less than 2 (sensitivity 0.84; specificity 0.58), 11 studies that used the Wells rule with cutoff value less than 4 (sensitivity 0.60; specificity 0.80), five studies that used the Geneva rule (sensitivity 0.91; specificity 0.37), and four studies that used the revised Geneva rule (sensitivity 0.91; specificity 0.37). A greater prevalence of PE was associated with higher sensitivity and lower specificity. The authors observed significant heterogeneity among studies in terms of PE prevalence and differences in threshold. The only combination of decision rule and testing that did not safely exclude PE was the less-sensitive Wells rule (cutoff value ≤4) combined with less-sensitive qualitative D-dimer testing.


The authors concluded that clinical decision rules and gestalt can safely exclude PE when combined with sensitive D-dimer testing. The authors recommended standardized rules over gestalt because of lower specificity, but believed that the particular rule and D-dimer test performance depends on both PE prevalence and the clinical setting.


Due to the proliferation of validated clinical decision rules for establishing the risk of PE in patients presenting with suspicion of PE, as well as the multiplicity of quantitative and qualitative D-dimer tests used, there may be increasing confusion as to which clinical algorithm performs best, and whether any are better than empirical clinical assessment (gestalt). Although individual studies of specific clinical prediction rules have demonstrated safety for most of them, this very well conducted meta-analysis provides validation for the continued clinical use of these decision tools. The bottom line is that a validated clinical prediction rule combined with D-dimer testing is a safe and effective way to exclude the diagnosis of PE, and that neither clinician gestalt nor clinical decision rules are sensitive enough without D-dimer testing. Specific clinical decision rules may work better in a high PE prevalent situation (such as Wells 4), whereas evaluation of patients in a low prevalence situation may be more effectively done with higher sensitivity rules (such as Wells 2 or Geneva). But for the average clinician, any established protocol that utilizes a validated decision rule and D-dimer testing is well supported.

Clinical Topics: Vascular Medicine

Keywords: Prevalence, Fibrin Fibrinogen Degradation Products, Cardiology, Pulmonary Embolism, Cardiovascular Diseases

< Back to Listings