A number of scoring methods are available for bleeding risk prediction in adult cardiac surgery [1, 12,13,14,15,16]. However, all but the WILL-BLEED risk score [1] are nonspecific addressing general adult cardiac surgery cases.
CABG remains one of the most commonly performed major surgeries, with well-established symptomatic and prognostic benefits in patients with multivessel and left main coronary artery disease [17]. CABG is on average performed at a rate of 44 per 100,000 individuals [17]. Number of CABG procedures per 100.000 inhabitants varies widely from 4 per 100,000 in Mexico to 79 per 100,000 in the United States and 91 per 100,000 in Hungary, respectively [17]. Reported transfusion rates for isolated CABG (rising over 90% with a huge proportion of those transfusions being unnecessary) call for a user-friendly screening tool to stratify bleeding risk and patients identified with low risk define the subgroup of patients to whom transfusion treatment might be avoided.
Several important considerations should be made when it comes to the development of the bleeding risk score:
1) Homogeneity of the study population makes it more reliable to create a scoring system. We know without a scoring system how complex cardiac surgery procedures carry a markedly higher risk of bleeding than isolated CABG.
2) The parameters we consider when thinking of bleeding risk constantly evolve. Some of those parameters are persistent, though. Our understanding, however, of bleeding risk evolves and when thinking of the inextricable association between bleeding and transfusion requirements, our focus switched from personalized point-of-care guided transfusion management in high-risk patients towards complete avoidance of transfusion in patients previously considered to have a low predicted risk of bleeding. Our SHOULD-NOT-BLEED bleeding risk score identifies patients at low risk of bleeding with a specificity as high as 94% and sensitivity of 24%.
Having in mind how transfusion rates climb up to over 90% in CABG patients, it becomes apparent that our target was to identify low bleeding risk CABG patients for whom transfusion treatment might be avoided.
SHOULD-NOT-BLEED score compared to other bleeding risk scores
r The ROC analysis of the SHOULD-NOT-BLEED risk score calculator showed an adequate discriminatory ability (AUC 0.723 95% CI (0.694–0.753), p > 0.001). This discriminatory ability is comparable to the WILL-BLEED score (AUC 0.725, 95% CI 0.686–0.763, p = 0.033) [1]. It is important to stress how the SHOULD-NOT-BLEED score is designed to recognize patients at low risk for bleeding, whereas the WILL-BLEED, as well as other predictor scores including: the ACTION score [18], CRUSADE score [19], Papworth score [12], TRUST (Transfusion Risk Understanding Scoring Tool) score [13], and TRACK (Transfusion Risk And Clinical Knowledge) bleeding score [14], each of which were designed to identify patients at high risk of bleeding.
The WILL-BLEED bleeding risk score [1] was, to the best of our knowledge, the only score based on the isolated CABG population. The SHOULD-NOT-BLEED score is now the second one based on isolated CABG patients. Because patients undergoing off-pump CABG were excluded from the data analysis [8], The SHOULD-NOT-BLEED score may not be applicable for patients undergoing this surgical procedure. Such an approach sounds reasonable given the fact that on-pump CABG substantially differs to off-pump CABG (the use of cardiopulmonary bypass alters haemostatic mechanisms) as well as the priority given to study cohort homogeneity.
The problem with scores being developed on the general adult cardiac surgery population is that the scoring system inevitably carries non-specific parameters such as “complex cardiac surgery”. Moreover, it seems less feasible to use the same score for off-pump cases, on-pump CABG and complex aortic surgery cases. Therefore, our concept presents a kind of shift towards a precise, more focused, and personalized approach. Such an approach sets a priority to a homogeneity of the study cohort. The discriminatory ability of the SHOULD-NOT-BLEED score (AUC 0.723) is comparable to the WILL-BLEED score (AUC 0.725) [1]. In the WILL-BLEED score, few baseline characteristics and information on “potent” antiplatelet drugs use allows an accurate stratification of bleeding risk [1]. Our score presents a more accurate assessment of preoperative haemostatic properties. We have a parameter on recent (less then 5 days) clopidogrel use, which is more or less the case for the WILL-BLEED score. In contrast to our score, WILL-BLEED accounts for potent antiplatelet drugs use within 5 days. Our database of isolated elective CABG patients allowed only for the assessment of recent clopidogrel use. This is more specific to the drug evaluated in the context of bleeding risk. On the other hand, the chances to include elective patients exposed to ticagrelor for further studies are small as elective patients strictly adhere to the current guidelines on dual antiplatelet therapy. The SHOULD-NOT-BLEED score provides a more detailed insight into haemostatic properties. Our application is based on a single center database where all patients were exposed to Aspirin preoperatively and Aspirin was continued throughout the procedure. Accordingly, there was no need to specify whether or not someone was using Aspirin preoperatively into the app. However, our research group previously confirmed the presence of a subset of patients who have a prolonged and pronounced platelet inhibitory response to Aspirin [20], which in turn reflects bleeding tendency. We recently showed that patients with an adequate platelet inhibitory response to Aspirin are prone to excessive bleeding [20]. A Multiplate ASPI test value of AUC < 25 U was found to be predictive of excessive bleeding (OR 2.82 [95% CI 1.43–5.55], p = 0.003) which generates the idea of a subset of patients who have pronounced platelet inhibition on Aspirin therapy and who could benefit from preoperative Aspirin cessation [20]. The mathematical risk modelling platform used for the SHOULD-NOT-BLEED risk score has also proved that the Multiplate ASPI test value is as an independent predictor for bleeding [8]. The SHOULD-NOT-BLEED risk score is the first bleeding risk score that accounts for drug specific platelet reactivity in calculating bleeding risk. Current guidelines on dual antiplatelet therapy suggest continuation of Aspirin peri-procedurally [21]. Our approach adds to current knowledge and will hopefully contribute to the change of this paradigm. It is apparent that some patients under Aspirin treatment have a higher risk of bleeding (OR 2.82), therefore, inclusion of an Aspirin sensitive platelet function test into the bleeding risk score calculator sets a new moment.
The homogeneity of the study cohort rules out some of the confounding variables and leaves space for some new predictors. The WILL-BLEED score was designed to detect patients undergoing CABG who are at high risk of bleeding and to modify antithrombotic treatment if possible [1]. In other words, the WILL-BLEED score was mainly driven by the idea that a proportion of patients undergoing CABG are at high risk of bleeding and as such may be subject to possible haemostatic interventions, be it pre and/or intraoperative intervention. In contrast, our SHOULD-NOT-BLEED score is driven by the idea that a huge amount of transfusions (up to 93% according to the literature) in patients undergoing cardiosurgical procedures such as CABG which carry the lowest risk of bleeding are, in fact, unnecessary. The first step in addressing unnecessary transfusions in patients undergoing CABG is to identify patients primarily considered to be at low risk of bleeding. Our paradigm is that all patients with a high risk of bleeding should be treated in the same/similar way using point-of-care (POC) -guided transfusion algorithms to optimize haemostasis. The major clinical and economic burden arises from unnecessary transfusions, and when it comes to unnecessary transfusions, we should start with patients undergoing low risk procedures such as CABG who are at the same time at a low risk of bleeding.
Another advantage of the SHOULD-NOT-BLEED score is that use of the application is user friendly and self-explanatory. The idea of this application is not to guide and/or alter the clinical decision-making process. Since the development of this application is based on and validated by data from a primary source, it may be a useful tool for clinicians involved in preoperative screening and risk assessment.
Being based on the data from elective patients and primarily focused on elective patients makes this application not just a useful tool for preoperative risk assessment, but it also allows for modifiable parameter modification prior to surgery. For example, if someone is using clopidogrel in close proximity to surgery and at the same time has a low ASPI test value, postponing surgery with temporary discontinuation of Aspirin could modify the risk of bleeding.
The same holds for anemic patients with low a red blood cell count before surgery. The optimization of the red blood cell mass is the first pillar of patient blood management and may easily be considered modifiable if it contributes to high bleeding risk before surgery, as per the SHOULD-NOT-BLEED application. On the other hand, a low fibrinogen level contributing to high risk of bleeding on the calculator may prompt early fibrinogen supplementation if bleeding occurs after surgery.
When the statistical model used to design the SHOULD-NOT-BLEED-SCORE was applied to our existing database, an astonishing reduction of 39.1% in transfusion requirements could theoretically be reached. The cost savings reach 48.2% for PRBCs, 38.9% for fresh frozen plasma (FFP), 10.9% for platelet concentrate and 17.9% for fibrinogen, respectively [8]. Aforementioned cost savings pertain solely to blood product manufacturing costs [8]. Having in mind the additional cost of product administration as well as overhead expenses and the costs associated with treating complications secondary to transfusion therapy itself, it becomes apparent that real-life cost savings could potentially be much higher. Notably, indirect costs of transfusion treatment may reach over 65% of all expenditures related to transfusion therapy [22, 23].
The SHOULD-NOT-BLEED bleeding risk score may serve as an impetus for further refinements in haemostatic management. For this reason, we call for multicentric collaboration in developing and refining haemostatic management. Firstly, we propose validation of this score through multicenter collaboration. Secondly, cost-effectiveness of such a score may be calculated in a stepped wedge design prospective interventional multicentric trial [23]. Multicenter collaboration would yield a huge database allowing for more complex statistics. More patients recruited would make it possible to add some new parameters in considerations. We know from our practice several factors that could be implemented into considerations and this could be achieved in new study with bigger study cohort. We need to count on more parameters contributing to the bleeding risk. On the other hand, the major limitation of the current scoring system is that all parameters need to be available so to calculate the bleeding risk. The next generation of the bleeding risk score should be able to calculate the risk even in cases where missing values for some parameters. For this new generation risk scoring app, we need multicentric collaboration in order to recruit more patients and to achieve study sample size that would allow for such analyses.