PE-SARD – Big Fail Predicting Bleeding Risk with PE
June 9, 2023
Written by Millie Cosse
Spoon Feed
PE-SARD failed to accurately predict risk of major or clinically relevant bleeding in patients ≥65 years with acute PE at 7 days or on longer follow-up.
Why does this matter?
Major bleeding occurs in 3% of patients with acute pulmonary embolism and carries a high mortality. Accurately identifying the patients who are most susceptible to such bleeding is crucial for determining appropriate anticoagulation and reperfusion strategies. The PE-SARD bleeding score has not previously been externally validated in patients aged ≥65 years.
PE-SARD: Predictive excellence or coin-toss coincidence?
This prospective multicenter study of 687 Swiss patients aimed to externally validate the accuracy of PE-SARD in patients ≥65 years. The primary outcome was major bleeding (MB) at 7 days, and secondary outcomes were MB and clinically relevant bleeding at various time points over the entire follow-up period.
The accuracy of PE-SARD for MB when shown as area under the receiver operating characteristic (ROC) curve was 0.52 (95%CI 0.48-0.56) at 7 days and increased to a peak value of 0.60 (95%CI 0.56-0.64) over time. For the secondary outcome of clinically relevant bleeding at 7 days after acute PE, PE-SARD had an area under the ROC curve of 0.57.
As a reminder, the area under the ROC curve of a perfect classifier will approach 1, and a random classifier would have a value of 0.5 (basically, a coin-toss).
My take: The PE-SARD uses syncope, anemia, and renal dysfunction to predict risk of major bleeding. Unfortunately, all three of these findings may be coincidentally seen in patients ≥65 years, and the result is that PE-SARD is barely more accurate in predicting this patient population’s risk of major bleeding 7 days after acute PE than flipping a coin.
Source
Prediction of very early major bleeding risk in acute pulmonary embolism: an independent external validation of the PE-SARD Bleeding Score. J Thromb Haemost. 2023 May 4;S1538-7836(23)00347-1. doi: 10.1016/j.jtha.2023.04.025. Online ahead of print.