External Validation of the WILL-BLEED Risk Score

Perioperative bleeding is recognized as a major complication affecting the early and late outcome of patients undergoing cardiac surgery.[1] Such a risk is of particular relevance in patients undergoing coronary artery bypass grafting (CABG) because they are often exposed preoperatively to potent antithrombotics and not infrequently require urgent or emergency operation, which prevents the discontinuation of these drugs. Recently, we developed a risk scoring method, the WILL-BLEED score, for the prediction of severe and massive perioperative bleeding in patients undergoing CABG.[2] The value of this risk score was questioned in the absence of an external validation.[1] [3] Therefore, we sought to validate its predictive ability in an institutional series.

The present study includes 2,761 patients who underwent isolated CABG from June 2006 to December 2013 at the Oulu University Hospital, Finland. Patients undergoing any other major cardiac procedure were excluded from this analysis. Data on pre-, intra- and postoperative variables were retrospectively collected into an Institutional Access database. Data on the types and amount of blood products transfused in these patients were retrieved from a prospective electronic hospital registry. Preoperative anaemia was defined a haemoglobin level <12.0 g/dL in women and <13.0 g/dL in men. The outcome measures of this analysis were the amount of units of transfused red blood cells (RBC), transfusion of > 4 units of RBC and/or resternotomy for bleeding, and 30-day mortality. The predictive ability of this risk score was compared with that of the TRUST[4] and the Papworth[5] bleeding risk scores.

Statistical analysis was performed using an SPSS statistical software (version 24.0; IBM Corporation, Armonk, New York, United States) and the freely available software easyROC (ver. 1.3; http://www.biosoft.hacettepe.edu.tr/easyROC/, accessed 17 November 2017). Continuous variables were reported as the mean and standard deviation as well as median and interquartile range where appropriate. Nominal variables were reported as counts and percentages. Pearson's chi-square and Mann–Whitney tests were used for univariate analysis. The Spearman's test was used to evaluate the correlation between the WILL-BLEED risk score and the amount of transfused units of RBC. Discriminatory ability and calibration of the WILL-BLEED risk score in predicting the study outcomes were assessed by evaluating the area under (AUC) the receiver operating characteristic (ROC) curve and by the Hosmer–Lemeshow (H-L) test, respectively. The WILL-BLEED risk score was also dichotomized according to the Youden test. All tests were two-sided with the α level set at 0.05 for statistical significance.

Details on this series are reported elsewhere.[6] In brief, the mean age of these patients was 67.0 ± 9.0 years, 581 patients (21.0%) were females, 662 (24.0%) had anaemia, 118 (4.3%) had an estimated glomerular filtration rate <45 mL/min/1.73 m², 282 (10.2%) had atrial fibrillation and 177 of them (6.4%) were on warfarin treatment preoperatively, 1,611 (58.3%) had acute coronary syndrome, 1,502 (54.4%) required urgent or emergency operation and 217 patients (7.9) were in critical preoperative state. Furthermore, 2,579 patients received aspirin (93.4%), 829 patients received clopidogrel (30.0%), 73 patients received ticagrelor (2.6%), 2 patients received prasugrel (0.1%) and 233 patients received warfarin (8.4%) preoperatively. A total of 471 patients (17.3%) were on dual-antiplatelet treatment within 5 days before surgery.

Thirty-day mortality rate was 3.2%. RBC transfusion was required in 63.5% of patients and resternotomy for bleeding in 6.5% of patients. Transfusion of > 4 units of RBC was observed in 18.5% of patients and transfusion of >4 units of RBC and/or resternotomy for bleeding was necessary in 20.7% of patients and increased along with the WILL-BLEED risk score (p < 0.0001, [Fig. 1]).

Fig. 1 Proportion of transfusion of >4 units of red blood cells (RBC) and/or resternotomy for bleeding after coronary artery bypass grafting according to the WILL-BLEED risk score.

The WILL-BLEED risk score significantly correlated with the amount of transfused RBC (rho: 0.486, p < 0.0001). The WILL-BLEED risk score showed good discriminatory ability for prediction of transfusion of >4 units of RBC and/or resternotomy for bleeding (AUC of ROC, 0.721, 95% confidence interval [CI]: 0.697–0.745, H-L test: p = 0.082), transfusion of >4 units of RBC (AUC of ROC, 0.756, 95% CI: 0.733–0.780, H-L test: p = 0.016) and 30-day mortality (AUC of ROC, 0.754, 95% CI: 0.697–0.811, H-L test: p = 0.106). However, calibration was satisfactory only for prediction of transfusion of >4 units of RBC and/or resternotomy for bleeding and of 30-day mortality. The best cutoff value of the WILL-BLEED risk score for prediction of transfusion of >4 units of RBC and/or resternotomy for bleeding was 5 (>5: 42.8% vs. ≤5: 13.5%, p < 0.0001, sensitivity 0.651, specificity 0.690). Patients with a WILL-BLEED score >5 had a significantly increased amount of transfused RBC units (median 4.0 ± 4.0 vs. 1.0 ± 3.0 units, mean 4.7 ± 4.5 vs. 1.8 ± 2.5 units, p < 0.0001). The best cutoff of the WILL-BLEED risk score for prediction of 30-day mortality was 6 (>6: 10.1% vs. ≤6: 1.6%, p < 0.0001, sensitivity 0.667, specificity 0.771).

In this series, the TRUST bleeding risk score showed good discriminatory ability for prediction of transfusion of >4 units of RBC and/or resternotomy for bleeding (AUC of ROC, 0.705, 95% CI: 0.681–0.729, H-L test: p = 0.659), which was similar to that of the WILL-BLEED score (p = 1.00). The Papworth bleeding risk score showed a satisfactory discriminatory ability for prediction of transfusion of >4 units of RBC and/or resternotomy for bleeding (AUC of ROC, 0.659, 95% CI: 0.635–0.684, H-L test: p = 0.017), but this was inferior to that of the WILL-BLEED score (p = 0.001).

This analysis showed that the WILL-BLEED risk score correlated with the amount of transfused RBC and has a good calibration for prediction of transfusion of >4 units of RBC and/or resternotomy for bleeding (H-L test: p = 0.082), and 30-day mortality (H-L test: p = 0.106), but suboptimal calibration for prediction transfusion of >4 units of RBC (H-L test: p = 0.016). These findings suggest that the WILL-BLEED risk score seems to be an easy and valuable tool to stratify the risk of perioperative bleeding in patients undergoing isolated CABG. The predictive ability of this bleeding risk score should be evaluated in patients undergoing heart valve surgery and in those undergoing percutaneous coronary intervention.

• References

• 1 Pagano D, Milojevic M, Meesters MI. , et al; Authors/Task Force Members; Document Reviewers. 2017 EACTS/EACTA Guidelines on patient blood management for adult cardiac surgery: The Task Force on Patient Blood Management for Adult Cardiac Surgery of the European Association for Cardio-Thoracic Surgery (EACTS) and the European Association of Cardiothoracic Anaesthesiology (EACTA). Eur J Cardiothorac Surg 2018; 53 (01) 79-111
• 2 Biancari F, Brascia D, Onorati F. , et al. Prediction of severe bleeding after coronary surgery: the WILL-BLEED Risk Score. Thromb Haemost 2017; 117 (03) 445-456
• 3 Roldán V, Marín F. Predicting bleeding risk after coronary surgery: Let's focus on modifiable risk factors and simple, practical decision making. Thromb Haemost 2017; 117 (04) 647-649
• 4 Alghamdi AA, Davis A, Brister S, Corey P, Logan A. Development and validation of Transfusion Risk Understanding Scoring Tool (TRUST) to stratify cardiac surgery patients according to their blood transfusion needs. Transfusion 2006; 46 (07) 1120-1129
• 5 Vuylsteke A, Pagel C, Gerrard C. , et al. The Papworth Bleeding Risk Score: a stratification scheme for identifying cardiac surgery patients at risk of excessive early postoperative bleeding. Eur J Cardiothorac Surg 2011; 39 (06) 924-930
• 6 Tauriainen T, Koski-Vähälä J, Kinnunen EM, Biancari F. The effect of preoperative anemia on the outcome after coronary surgery. World J Surg 2017; 41 (07) 1910-1918