Comparative diagnostic accuracy of pre-test clinical probability scores for the risk stratification of patients with suspected pulmonary embolism: a systematic review and Bayesian network meta-analysis

Abstract

Background: The primary evaluation of pulmonary embolism (PE) is complicated by the presence of various pre-test clinical probability scores (pCPS) with different cut-offs, all equally recommended by guidelines. This lack of consensus has led to practice variability, unnecessary imaging, and worse patient outcomes. We aim to provide more definitive insights through a holistic comparison of available pCPS. Methods: PubMed, Embase and Web of Science, and Google Scholar were searched for studies evaluating pCPS in patients clinically suspected of PE until June 2023. Risk of bias was evaluated using QUADAS-2. Included pCPS were evaluated based on their diagnostic accuracy in: (1) Ruling-out PE (2) Utilization of imaging, and (3) Differentiating between patients needing d-dimer from imaging. Diagnostic test accuracy indices were synthesized using beta-binomial Bayesian methods. Results: Forty studies (37,027 patients) were included in the meta-analysis. Three-tier revised Geneva (RG) and three-tier Wells performed similarly in ruling-out PE (negative likelihood ratio (LR-) [95% credible interval (CI)]: 0·39[0·27–0·58] vs 0·34[0·25–0·45]). However, RG performed better in utilization of imaging (LR + : 6·65[3·75–10·56] vs 5·59[3·7–8·37], p < 0.001) and differentiating between patients needing d-dimer vs imaging (diagnostic odds ratio (DOR): 8·03[4·35–14·1] vs. 7·4[4·65–11·84], p < 0.001). The two-tier Wells score underperformed in all aspects (LR-: 0·56[0·45–0·68], LR + : 2·43[1·81–3·07], DOR: 4·41[2·81–6·43]). PERC demonstrated a reliable point estimate for ruling out PE, albeit with a wide CI (LR-: 0·36[0·17–0·78]). Conclusions: RG outperforms other pCPS for primary evaluation of suspected PE. While the difference is not large, RG's independence from subjective items supports its recommendation over three-tier Wells. Two-tier Wells underperforms significantly compared to the rest of pCPS. PERC shows considerable promise for minimizing unnecessary D-dimer testing in crowded emergency departments; however, more evidence is needed before its definitive recommendation. Protocol registration: PROSPERO (CRD42023464118).