Established Clinical Prediction Rules for Bleeding Had Mediocre Discrimination in Left Ventricular Assist Device Recipients

Abstract

Left ventricular assist devices (LVAD) reduce mortality in patients with end-stage heart failure, but LVAD management is frequently complicated by bleeding. Bleeding prediction post-LVAD implantation is challenging as prediction rules for hemorrhage have not been rigorously studied in this population. We aimed to validate clinical prediction rules for bleeding, derived in the atrial fibrillation and venous thromboembolism populations, in an LVAD cohort. This was a retrospective cohort study of LVAD recipients at an academic center. The primary end-point was time to gastrointestinal bleed or intracranial hemorrhage after implant; the secondary end-point was time to any major hemorrhage after hospital discharge. Four hundred and eighteen patients received an LVAD (135 HeartMate II, 125 HeartMate 3, 158 HVAD) between November 2009 and January 2019. The primary endpoint occurred in 169 (40.4%) patients with C-statistics ranging 0.55-0.58 (standard deviation [SD] 0.02 for all models). The secondary end-point occurred in 167 (40.0%) patients with C-statistics ranging 0.53-0.58 (SD 0.02 for all models). Modifying the age and liver function thresholds increased the C-statistic range to 0.56-0.60 for the primary and secondary end-points. In a sensitivity analysis of HeartMate 3 patients, prediction rules performed similarly. Existing prediction rules for major bleeding had mediocre discrimination in an LVAD cohort.

Performance of risk scores in predicting major bleeding in left ventricular assist device recipients: a comparative external validation

“…However, we could not establish an association between risk scores and MB assessed by ISTH, INTERMACS, and INTERMACS+ criteria. Our findings regarding the predictive performance of conventional risk scores align with those of previous studies, reporting poor discriminative abilities (C-statistics ≤ 0.62) in LVAD recipients for HAS-BLED, HEMORR 2 HAGES, ATRIA, OBRI, and VTE-BLEED [ 9 , 26 ]. Additionally, our calibration results highlight extreme underprediction of MB risk in this population.…”

“…Five previous studies explored the utility of bleeding risk scores in LVAD patients [ [7] , [8] , [9] , 11 , 26 ]. Two lacked formal statistical evaluation of model performance, ie, discrimination and calibration assessments [ 7 , 8 ].…”

“…Two lacked formal statistical evaluation of model performance, ie, discrimination and calibration assessments [ 7 , 8 ]. The remaining 3 studies assessed discrimination of 1 or more risk scores, yet none included calibration measurements [ 9 , 11 , 26 ].…”

“…These factors may have contributed to the poor predictive performance of URBS in our cohort, affecting both its discriminative ability and its calibration. Nevertheless, in previous external validation studies focusing on the ability to predict 3-year GIB, unsatisfactory discrimination was reported as well (C-statistics ≤ 0.59) [ 11 , 26 ].…”