Parsimonious versus extensive score: can we simplify bleeding risk stratification after percutaneous coronary intervention without losing specificity?

Background/Introduction Prediction of bleeding risk in patients receiving dual antiplatelet therapy (DAPT) after percutaneous coronary intervention (PCI) has been repeatedly attempted and resulted in multiple risk models. To promote the adoption and use of risk stratification among clinicians, a consortium in the South East of the Netherlands, the "ZON-HR", has created a simple binary variant of the existing logistic models. This model includes commonly available factors which can be incorporated into the PCI report to generate stratification. In this "parsimonious" model, high bleeding risk (HBR) is stratified according to: history of intracranial haemorrhage, previous spontaneous bleeding, Hb < 11 g/dL, eGFR< 30. Purpose We aimed to perform an external validation of our model for bleeding risk and to demonstrate its discriminative value in comparison with the PRECISE DAPT score, an acknowledged risk score. Methods We performed a post hoc analysis of the GLOBAL LEADERS trial, a multicenter randomized trial, comparing ticagrelor monotherapy after one month of DAPT with standard DAPT after PCI. We stratified the patients for HBR according to the ZON-HR or a PRECISE-DAPT score ≥ 25. Bleeding events one year after PCI were compared in the entire population and in subgroups for chronic coronary syndrome (CCS), acute coronary syndrome (ACS), Ticagrelor monotherapy and DAPT. Results Of the 15968 patients included in GLOBAL LEADERS, 1059 patients were excluded due to missing variables required for either model. Among the 14909 patients, 17% had HBR according to the PRECISE DAPT score, 4% according to ZON-HR and 3% according to both. Bleeding events occurred in 8.5% and 13% of HBR defined by PRECISE DAPT or ZON-HR respectively. Patients with HBR according to both models showed 13% bleedings as opposed to 5% in non-HBR. Table 1 presents the sensitivity and specificity. Regression analysis (figure 1) demonstrated that patients with HBR according to either model have significantly more bleedings. This was also significant in all subgroups except for ACS patients receiving standard DAPT when the ZON-HR model was used. Conclusion The ZON-HR model is more conservative compared to the PRECISE DAPT score as depicted by the sensitivity and specificity. This can be explained by the larger number of risk factors used in the PRECISE DAPT score that cumulatively contributes to the sensitivity as opposed to the binary ZON-HR model which is more specific. However, the survival curves and hazard ratios showed good predictive value of both models for bleeding risk in the overall population.Table 1.Figure 1.