Prediction of long-term net clinical outcomes using the TIMI-AF score: Comparison with CHA2DS2-VASc and HAS-BLED

Background The TIMI-AF score was described to predict net clinical outcomes (NCOs) in atrial fibrillation (AF) patients receiving warfarin. However, this score derived from the ENGAGE AF-TIMI 48 trial, and no external validation exists in real world clinical practice. We tested the long-term predictive performance of the TIMI-AF score in comparison with CHA(2)DS(2)-VASc and HAS-BLED in a 'real-world' cohort of anticoagulated AF patients. Methods We included 1156 consecutive AF patients stable on vitamin K antagonist (INR 2.0-3.0) during 6 months. The baseline risk of NCOs (composite of stroke, life-threatening bleeding, or all-cause mortality) was calculated using the novel TIMI-AF score. During follow-up, all NCOs were recorded and the predictive performance and clinical usefulness of TIMI-AF was compared with CHA(2)DS(2)-VASc and HAS-BLED. Results During 6.5 years (IQR 4.3-7.9), there were 563 NCOs (7.49%/year). 'Low-risk' (6.07%/year) and 'medium-risk' (9.49%/year) patients defined by the TIMI-AF suffered more endpoints that low-and medium-risk patients of CHA(2)DS(2)-VASc and HAS-BLED (2.37%/year and 4.40%/year for low risk; 3.48%/year and 6.39%/year for medium risk, respectively). The predictive performance of TIMI-AF was not different from CHA(2)DS(2)-VASc (0.678 vs 0.677, P=.963) or HAS-BLED (0.644 vs 0.671, P=.054). Discrimination and reclassification did not show improvement of prediction using the TIMI-AF score, and decision curves analysis did not demonstrate higher net benefit. Conclusions In VKA-experienced AF patients, the TIMI-AF score has limited usefulness predicting NCOs over a long-term period of follow-up. This novel score was not superior to CHA(2)DS(2)-VASc and HAS-BLED identifying low-risk AF patients.