External validation of novel bleeding risk scores, ARC-HBR and PRECISE-HBR, in 5116 consecutive ACS patients in clinical practice

Abstract introduction Risk prediction scores for bleeding such as ARC-HBR (Academic Research Consortium, HBR=high bleeding risk) and PRECISE-HBR have been developed for patients undergoing coronary stenting with dual antiplatelet therapy (DAPT). Most validation studies are from clinical trials but their performance in real world populations of acute coronary syndrome treated with DAPT has seldom been explored. Methods We used a previously defined cohort of consecutive ACS patients treated with DAPT for 1 year in 5 English hospitals (ROBOT-ACS). Clinical outcomes were identified from the national Hospital Episode Statistics (HES) Admitted Patient Care database. All readmissions with a code for any form of bleeding underwent centrally coordinated adjudication using two experienced clinicians. Bleeding events were categorised using BARC (Bleeding Academic Research consortium) 3-5 criteria and PLATO (Platelet inhibition and Patient outcomes). ARC-HBR and precise-HBR bleeding risk scores were determined from covariates in the database and additional information from HES ICD-10 codes. Predicted versus actual major bleeding events and c-statistics (Reciever operator curves) for BARC 3-5 major bleeding event were determined for the total population and for ACS patients with potent DAPT (aspirin and ticagrelor) versus standard DAPT (aspirin and clopidogrel) Results 5116 patients formed the total study population. The median age was 68 and 34% were female. 52% of patients underwent coronary revascularisation (39% percutaneous coronary intervention, 13% coronary artery bypass surgery). Equal numbers of patients were prescribed ticagrelor versus clopidogrel (allied to aspirin) 1194 Bleeding events were tracked to 11 different English institutions. Clinical outcome data was 100% complete for bleeding. A total of 213 (4.2%) patients were adjudicated as suffering a BARC 3-5 major bleeding or PLATO major bleeding in the year following ACS. 26.6% (1302/4902) and 33.8% (1730/5116) of patients were categorised as high bleeding risk according to Precise-HBR and ARC-HBR scores respectively. Of those assigned very HBR by Precise-HBR, 6.3% had a BARC 3-5 bleed. The corresponding figure was 5.3% for ARC-HBR. Both scores predicted bleeding HBR better when potent antiplatelet (ticagrelor) was prescribed (allied to aspirin) (HBR as determined by Precise HBR: incidence of 1 year major bleed 8.5% vs 2.0% [non-HBR] p<0.001, HBR as determined by ARC-HBR: incidence of 1 year major bleed 6.6% versus 2.0%[non-HBR]. P<0.001 (table) The c-statistic for BARC 3-5 bleeding over 1 year, for all patients, was 0.62 and 0.61 for Precise-HBR and ARC-HBR respectively. Conclusion The performance of both precise-HBR and ARC-HBR bleeding risk scores was modest, but improved when assessing bleeding risk with potent antiplatelet prescription. Further work is required to improve risk tools to predict major bleeding, out with clinical trial populationsPerformance of bleeding risk scores