Risk stratification of hemodynamically stable patients with acute pulmonary embolism: external validation and comparison of different strategies

Abstract Background Risk stratification for adverse outcome in hemodynamically stable patients with acute pulmonary embolism (PE) remains an issue. We aimed at validating and comparing currently available tools for risk stratification in this setting. Methods Hemodynamically stable patients (n= 5,036) from a prospective, multicentre study of adult patients with acute, symptomatic, objectively diagnosed PE (COPE) were included in this study. Study outcomes were in-hospital death, death at 30 days, PE-related death and the composite of clinical deterioration or death at 30 days. We assessed and compared the accuracy of ESC-2014, ESC-2019, PEITHO, Bova, TELOS, FAST and NEWS2 models to predict study outcome events by AUC statistics, negative and positive predictive values. Results Patient distribution in risk categories highly varied across different models (low-risk class: from 14.4% in ESC-2019 to 70.1% in FAST; highest risk class: from 16.4% in BOVA to 43.7% in ESC 2014). When the 3,544 patients with echocardiography performed and troponin assessed were considered, ESC-2014, PEITHO, TELOS and NEWS-2 had i) the best discriminatory powers for study outcome events, ii) the best performances for all study outcome events (ROC curves for death or clinical deterioration reported in the Figure) and iii) negative predictive values for PE-related death above 99%. Negative predictive values for death or clinical deterioration were above 98% for all the models except for FAST; positive predictive values were all below 10% except for TELOS. TELOS score had the best performance; NEWS2 score was very good by including clinical variables only (no instrumental/laboratory items). Conclusions The majority of available models for risk stratification of patients with acute PE have good accuracy for the identification of low-risk patients; their positive predictive values in hemodynamically stable patients remain modest, the best being that of TELOS score. These data are essential to inform management studies.