The incidence of major bleeding in patients with acute coronary syndromes (ACS) ranges from 3 to 5%.1–5 Major bleeding events are associated with worse outcomes in this clinical setting.6–8 Several bleeding risk scores9–11 have been developed in recent years in order to properly predict bleeding complications in patients with ACS, and their use is recommended in the current clinical guidelines.12 However, it has been suggested that these scores have lower predictive ability in the elderly13 and in patients with comorbidities.

Anemia is a common comorbidity among patients with ACS,14 and its prevalence is expected to increase due to the aging of the population. Anemia is associated with higher risk for both bleeding and ischemic complications in patients with ACS.15 Information on the predictive ability of bleeding risk scores in patients with anemia is scarce; no study has assessed the performance of bleeding risk scores according to hematocrit status in patients with ACS. The aim of this study was to assess the predictive ability of the most widely used bleeding risk scores according to anemia status in a series of consecutive patients with ACS from routine clinical practice.

During the study period 2255 patients were admitted with a diagnosis of ACS. Their mean age was 62.4 years, and almost 77% were male. Anemia was present in 550 patients (24.4%). Patients with anemia were significantly older, less often male and had a significantly higher prevalence of comorbidities (Table 1).

Patients with anemia also had poorer renal function at admission, a lower incidence of ST-segment elevation myocardial infarction and a higher percentage of signs of heart failure at presentation. Values of CRUSADE, Mehran and ACTION bleeding risk scores were all significantly higher in patients with anemia.

Significant differences were also observed regarding clinical management (Table 2). Patients with anemia were less often treated with clopidogrel and abciximab, and more often needed invasive procedures during hospitalization, such as intra-aortic balloon counterpulsation, renal replacement therapy, invasive mechanical ventilation and temporary pacing. Patients with anemia less often underwent an invasive strategy, and in patients undergoing angiography a radial approach was used less often than in patients without anemia.

These patients had also a higher incidence of in-hospital complications such as atrial fibrillation, atrioventricular block, contrast-induced nephropathy and infectious complications requiring antibiotics. Bleeding was significantly more common in patients with anemia regardless of the definition used. In-hospital mortality was almost three times higher in patients with anemia.

The three bleeding risk scores adequately predicted major bleeding according to their own definitions in the whole cohort (Figure 1). No significant differences were observed regarding the predictive ability of the bleeding risk scores in patients with and without anemia. While the predictive ability of the CRUSADE score was almost the same as for the whole cohort, the AUC of the ACTION score was slightly greater and that of the Mehran score was slightly smaller, without reaching statistical significance. Figure 2 shows the AUCs of each of the three bleeding risk scores in patients with and without anemia.

Figure 1.

Receiver operating characteristic curves of each of the bleeding risk scores for predicting major bleeding in the whole cohort. AUC: area under the curve.

(0.09MB).

Figure 2.

Receiver operating characteristic curves of each of the bleeding risk scores for predicting major bleeding in patients with and without anemia according to the World Health Organization criteria. AUC: area under the curve.

(0.19MB).

The assessment of patients with severe anemia showed different findings. The predictive ability of both the CRUSADE and ACTION scores decreased slightly in patients with severe anemia, but without significant difference. In contrast, the predictive ability of the Mehran score showed a significant decrease in patients with severe anemia. Table 3 shows the AUC values of each of the bleeding risk scores in patients without anemia, with anemia according to the WHO criteria, and with severe anemia.

The main findings from our study are: (a) a high prevalence of anemia was found in a series of unselected ACS patients from routine clinical practice; (b) no significant differences were observed regarding the predictive ability of the three bleeding risk scores in patients with or without anemia according to the WHO criteria; and (c) only the Mehran risk score showed a significantly poorer predictive ability in patients with severe anemia.

Anemia is a common comorbidity in ACS, and is strongly associated with higher mortality and morbidity in this setting.15,21,22 The reasons for this association have not been clearly elucidated. Anemia worsens myocardial ischemia by reducing oxygen delivery to the injured myocardium as well as by increasing myocardial oxygen demands due to a larger stroke volume and higher heart rate. At the same time, anemia potentially reflects occult disease, such as malignancy or kidney disease, and can have unfavorable effects on the clinical course of noncardiac disease. Recent data suggest that causes of mortality in patients with ACS and anemia may differ according to age subgroups.23 As reported above, patients with anemia from our series were significantly older, with a higher incidence of comorbidities and higher in-hospital mortality.

In addition, anemia is a well-known predictor of bleeding24 in patients with ACS. However, although currently available bleeding risk scores have been successfully validated in different subsets,25–27 there is little information on bleeding risk stratification in patients with anemia. Our group previously described13 a poorer performance of the three bleeding risk scores in predicting major in-hospital bleeding in patients with ACS aged 75 years or older. One possible explanation for these findings is that certain aging-related variables such as frailty, disability, or comorbidities, which are rarely assessed in trials and registries in the cardiovascular area, might hamper bleeding risk stratification in this clinical scenario. In addition, most registries show a strong association between anemia and other comorbidities in different clinical settings. In the previously mentioned paper,13 elderly patients with ACS had a significantly higher prevalence of anemia compared to younger patients. Therefore, poorer predictive ability could also be expected in patients with anemia. However, data from our series showed no significant differences regarding the predictive ability of the bleeding risk scores in patients with and without anemia.

An important point when analyzing our findings is the role of hematocrit status in the composition of the different bleeding risk scores. Hemoglobin level is part of the CRUSADE,9 Mehran10 and ACTION11 risk scores, but the contribution of hemoglobin values to these scores varies. While in CRUSADE and Mehran hemoglobin levels account for around 10% of the overall points, in ACTION they represent almost 20% of the score. Assessment of the predictive ability of these bleeding risk scores stratified by one of their components might in itself lead to a reduction in the AUC. In our opinion, the non-significant reduction in AUC values in patients with anemia may at least in part be due to this fact, especially for the ACTION score, given the greater role of hemoglobin status in the overall composition of the score.

The differences in therapeutic management according to anemia status in our series also deserves particular mention. Patients with anemia underwent more conservative antithrombotic management, thus possibly leading to a reduction in the rate of bleeding and potentially affecting the performance of bleeding risk scores. However, this management was performed according to current recommendations, since patients with anemia are usually considered at higher bleeding risk and a more conservative antithrombotic approach is encouraged in this scenario. In our opinion, one of the strengths of this work is its assessment of the performance of the current bleeding risk scores in patients from our routine clinical practice managed according to the clinical judgment of the attending physician, taking into account current recommendations and patients’ profile in terms of ischemic and bleeding risk.

The different characteristics of the populations from whom these scores were derived also deserve special comment. The CRUSADE and ACTION scores were based on large American registries (89134 patients for CRUSADE and 90273 for ACTION) with baseline characteristics that were probably similar to patients included in our study. In contrast, the Mehran score was based on 17421 patients included in two clinical trials (ACUITY and HORIZONS AMI). Populations from clinical trials usually have a low prevalence of comorbidities and patients at higher risk are usually under-represented. The prevalence of anemia in Mehran et al.’s series was clearly lower (14.5%) than in ours. The poorer predictive ability of the Mehran score in patients with severe anemia may be related to this fact. However, these data must be interpreted cautiously due to the small sample size of this subgroup.

Our study has several limitations. This is a single-center study and therefore our conclusions should be applied only to similar populations undergoing similar clinical management. Since this is an observational study we cannot rule out the possibility of selection bias and residual confounding. The cutoff point of <11.0 g/dl for defining severe anemia was selected by investigators after assessing the distribution of hemoglobin in our series. The number of bleeding events was relatively small. In addition, the use of novel antithrombotic drugs like prasugrel or ticagrelor was uncommon. On the other hand, excluding patients with missing bleeding scores might have led to a certain bias. However, analysis of the baseline characteristics of these patients showed no significant differences from the other patients. Finally, the Mehran bleeding risk score was designed to predict bleeding during the first 30 days and not only during hospitalization as in our study.

In spite of these limitations, we believe that our findings show reasonably acceptable predictive ability in the main available bleeding risk scores in a series of consecutive ACS patients with and without anemia from routine clinical practice.

The prevalence of anemia was high in our series of unselected ACS patients from routine clinical practice. The ability of the most important currently available bleeding risk scores to predict in-hospital major bleeding was acceptable in patients with anemia, especially in patients with mild anemia. The progressive aging of the population makes it particularly important to improve risk stratification in patients with comorbidities.

The authors have no conflicts of interest to declare.