Eighteen studies (n=5,997 patients) were included in the review. The median sample size was 171 (range 37 to 1,100). The majority of studies reported independent application of the reference standard (except where clinical probability testing was added to the D-dimer test); most clinical assessments were blinded, but reports of blinding of the reference standard were lacking.
Although all likelihood ratios were close to the value of one (indicating marginal effect), the most prominent clinical features for diagnosing the presence of acute pulmonary embolism were: syncope, likelihood ratio (LR) 2.38 (95% confidence interval (CI):1.54 to 3.69); shock, LR 4.07 (95% CI:1.84 to 8.96); thrombophlebitis, LR 2.20 (95% CI: 0.435 to 3.29); current deep vein thrombosis, LR 2.05 (95% CI:1.12 to 3.73); leg swelling, LR 2.11 (95% CI:1.59 to 2.79); sudden dyspnoea, LR 1.83 (95% CI: 1.07 to 3.13); active cancer, LR 1.74 (95% CI: 1.17 to 2.59); surgery, LR 1.63 (95% CI:1.23 to 2.12); haemoptysis, LR 1.62 (95% CI:1.23 to 2.15); and leg pain, LR 1.60 (95% CI: 0.936 to 2.74).
Clinical features for ruling out acute pulmonary embolism were reported as the absence of sudden dyspnoea, LR 0.430 (95% CI: 0.254, 0.730); any dyspnoea, LR 0.521 (95% CI: 0.372, 0.729), and tachypnoea, LR 0.561 (95% CI: 0.404, 0.780).
Many of the analyses were carried out in the presence of statistically significant heterogeneity (p<0.001).