PE-SARD Score for bleeding risk in acute pulmonary embolism

Clinical background

In acute pulmonary embolism, anticoagulation must be started promptly, but the treatment itself is the commonest cause of early major bleeding, which in turn is linked to increased mortality. The choice of agent, dose and level of care would be easier with an instrument that identifies patients at raised bleeding risk early. Several bleeding scores have been proposed for venous thromboembolism (VTE-BLEED, RIETE, BACS), but they have shown weak discrimination precisely for the early phase of pulmonary embolism. The PE-SARD bleeding score was developed specifically to fill this gap: to predict major bleeding up to discharge in patients with acute pulmonary embolism.

The score is intended to complement, not replace, assessment of the severity of the embolism. It weighs bleeding risk, not mortality, and should be combined with a severity measure (for example sPESI) when the level of care and treatment strategy are decided.

Calculating the PE-SARD bleeding score

The PE-SARD bleeding score rests on three clinical variables available on arrival:

$$\text{PE-SARD} = 1{.}5 \times \text{Syncope} + 2{.}5 \times \text{Anaemia} + 1 \times \text{Renal dysfunction}$$

where syncope at presentation scores 1.5 points (yes/no), anaemia is defined as a haemoglobin <12 g/dL and scores 2.5 points, and renal dysfunction is defined as an eGFR <60 mL/min/1.73 m² and scores 1 point. The score ranges from 0 to 5.

The score was derived from a multicentre prospective cohort of 2,754 patients with acute pulmonary embolism [1]. Multivariable logistic regression identified the three predictors among 82 major bleeding events (3.0%; 95% CI 2.39 to 3.72%). The model was internally validated by bootstrapping (500 iterations). In the derivation cohort, 52.2% were classified as low risk (0 points), 35.2% as intermediate risk (1 to 2.5 points) and 12.6% as high risk (>2.5 points). Observed bleeding rates rose from 0.97% in the low-risk group to 8.93% in the high-risk group. The c-index was 0.74 (95% CI 0.73 to 0.76) and the Brier score 0.028 [1]. PE-SARD performed better than VTE-BLEED, RIETE and BACS in the same cohort and led to a high proportion of reclassification, both of patients who bled and of those who did not [1].

Interpretation in practice

The calculator defines three risk bands:

A patient with a low score can as a rule be managed according to standard protocol without specific measures for bleeding risk. At an intermediate score, the bleeding risk should be taken into account when choosing an anticoagulant, particularly if the patient also has a high embolism-related mortality risk. At a high score, the early bleeding risk is appreciable and there is a strong case for individually tailored treatment, close monitoring and a documented balancing of embolism and bleeding risk.

It is important to note that the score does not indicate whether anticoagulation should be withheld. In acute pulmonary embolism, anticoagulation is indicated regardless of the bleeding score. The score informs rather the intensity of monitoring and the degree of caution in the choice of agent and dose.

Validation and performance

Three external validations have been published, with varying results.

The RIETE registry (50,686 patients). Chopard et al. validated PE-SARD in the large RIETE registry [2]. Over 30 days there were 640 major bleeds (1.3%). The bleeding rate was 0.6% in the low-risk group, 1.5% in the intermediate group and 2.5% in the high-risk group. The odds ratios were 2.22 (95% CI 2.02 to 2.43) for intermediate versus low risk and 3.94 for high versus low risk. The c-index was 0.654 and calibration was described as excellent. PE-SARD performed better than BACS and PE-CH. The results were consistent across clinical subgroups and over shorter follow-up periods (3 and 7 days) [2].

COMMAND VTE Registry-2 (2,781 patients, Japan). Nishimoto et al. validated the score in a Japanese cohort with acute pulmonary embolism [3]. The distribution was 29% low, 51% intermediate and 20% high risk. At 30 days the bleeding rate was 1.8% in the low-risk group, 4.6% in the intermediate group and 8.2% in the high-risk group. The c-statistic was 0.65 (95% CI 0.61 to 0.70). Calibration was good for scores below 4, except in patients with active cancer, in whom the score did not calibrate well [3].

Swiss cohort (687 patients, aged ≥65 years). Villiger et al. carried out an independent validation in a prospective Swiss multicentre cohort of patients aged 65 years or older [4]. Here the outcome was major bleeding within 7 days (primary) and over longer follow-up (median 30 months, secondary). The bleeding prevalence was 2.0% at 7 days. The distribution was 40.2% low, 42.2% intermediate and 17.6% high risk. Observed bleeding at 7 days was 1.8% in the low-risk group, 2.1% in the intermediate group and 2.5% in the high-risk group, that is, an essentially flat gradient. The AUC was 0.52 (95% CI 0.48 to 0.56) at 7 days and rose to 0.60 by the end of follow-up. Calibration was adequate (p > 0.05) throughout follow-up. The authors concluded that PE-SARD did not predict early major bleeding in this population and may transport poorly to older patients with pulmonary embolism [4].

In summary, discrimination fell from a c-index of 0.74 in the derivation cohort to 0.65 in both RIETE and COMMAND VTE, and to 0.52 in the Swiss cohort of older patients. The score performed best in large, broad cohorts and less well in a selected older population.

Limitations

PE-SARD was derived from a post hoc analysis of a prospective registry, not from a study primarily designed as a prediction study. The number of events in the derivation cohort was low (82 bleeds), which limits the statistical precision of the point weights.

The score comprises only three variables and does not take into account several well-recognised bleeding risk factors in pulmonary embolism, including active cancer, thrombocytopenia, a previous bleeding history, concomitant antiplatelet treatment and advanced age as such. In the COMMAND VTE validation, the score calibrated poorly in patients with active cancer [3], which is to be expected since cancer is not part of the model.

The Swiss validation shows that the score may be unreliable in older patients (≥65 years) [4], a group over-represented in clinical practice. The flat risk gradient and an AUC close to 0.50 in that cohort suggest that the score should not be used alone to decide treatment intensity in the elderly.

The score was developed for patients with acute pulmonary embolism and does not apply to decisions about prophylaxis, to deep vein thrombosis without pulmonary embolism, or to patients already on anticoagulation at presentation. It does not weigh the severity of the embolism and should not be used alone to determine the level of care.

References

1. Chopard R, Piazza G, Falvo N et al. An Original Risk Score to Predict Early Major Bleeding in Acute Pulmonary Embolism: The Syncope, Anemia, Renal Dysfunction (PE-SARD) Bleeding Score. Chest 2021;160(5):1832-1843. PMID: 34217683
2. Chopard R, Bertoletti L, Piazza G et al. External validation of the PE-SARD risk score for predicting early bleeding in acute pulmonary embolism in the RIETE Registry. Thromb Res 2024;235:22-31. PMID: 38295598
3. Nishimoto Y, Yamashita Y, Morimoto T et al. External validation of the Pulmonary Embolism-Syncope, Anemia, and Renal Dysfunction bleeding score for early major bleeding in patients with acute pulmonary embolism: from the COMMAND VTE Registry-2. J Thromb Haemost 2024;22(10):2784-2796. PMID: 38944241
4. Villiger R, Méan M, Stalder O et al. Prediction of very early major bleeding risk in acute pulmonary embolism: an independent external validation of the Pulmonary Embolism-Syncope, Anemia, and Renal Dysfunction (PE-SARD) bleeding score. J Thromb Haemost 2023;21(10):2884-2893. PMID: 37149148