CRUSADE Score for bleeding risk after myocardial infarction

Clinical background

Treatment of NSTEMI rests on potent antithrombotic medication and an early invasive strategy, both of which lower ischaemic risk but increase bleeding risk. Major bleeding during the admission is itself linked to increased mortality, a longer stay and higher costs [1]. The decision on how intensive the antithrombotic and invasive strategy should be therefore requires assessment of both ischaemic and bleeding risk, and the ischaemic risk scores (TIMI, GRACE) long lacked a bleeding counterpart. The CRUSADE score was constructed precisely to fill this gap and to give a quantitative estimate of the risk of major bleeding during the admission in NSTEMI patients, independently of the treatment strategy subsequently chosen [1].

Calculating the CRUSADE score

The score sums eight variables that were independent predictors of major in-hospital bleeding on multivariable analysis in the derivation cohort [1]:

$$\text{CRUSADE} = P_{\text{HCT}} + P_{\text{CrCl}} + P_{\text{HF}} + P_{\text{SBP}} + 8 \cdot \mathbb{1}$${\text{female}} + 7 \cdot \mathbb{1}{\text{heart failure}} + 6 \cdot \mathbb{1}{\text{vascular disease}} + 6 \cdot \mathbb{1}{\text{diabetes}}

The four continuous variables (baseline haematocrit, creatinine clearance by Cockcroft–Gault, heart rate and systolic blood pressure) are assigned points by interval, with a low haematocrit, a low creatinine clearance, a high heart rate and a low systolic blood pressure scoring higher. The dichotomous variables are assigned fixed points: female sex 8 points, signs of heart failure at presentation 7 points, previous vascular disease (peripheral arterial disease or previous stroke) 6 points and diabetes mellitus 6 points. The score ranges from 0 to 100 [1].

The derivation cohort consisted of 71,277 patients with NSTEMI from the CRUSADE Quality Improvement Initiative, enrolled at 485 hospitals in the USA from February 2003 to December 2006 [1]. The validation cohort comprised 17,857 patients. Patients with unstable angina, patients on warfarin at presentation, patients transferred to another hospital and those who died within 48 hours were excluded. Patients undergoing coronary bypass surgery were censored at the time of the procedure. The outcome modelled was major in-hospital bleeding, defined as intracranial bleeding, retroperitoneal bleeding, a fall in haematocrit $\geq$12 percentage points from baseline to nadir, red cell transfusion at a baseline haematocrit $\geq$28 percentage points, or red cell transfusion at a baseline haematocrit <28 percentage points combined with a demonstrated source of bleeding [1].

Interpretation in practice

The score is divided into five risk bands with rising bleeding risk. The table below shows the observed bleeding rate in the derivation cohort [1]:

For patients in the lower bands (score $\leq$30), a full antithrombotic and invasive strategy is supported without particular modification, provided the ischaemic risk warrants it. At a score of 31 to 50, bleeding risk should be weighed in when choosing the anticoagulation strategy, titrating antiplatelet doses and selecting the access route at catheterisation. At a score >50 the observed bleeding risk is close to 20 per cent, which warrants reduced antithrombotic intensity where clinically possible (for example a dose-adjusted anticoagulant or UFH instead of a GPIIb/IIIa inhibitor) and radial rather than femoral access. The score should never be the sole reason for withholding an indicated invasive investigation, but it identifies patients in whom extra caution is warranted [1].

Validation and performance

In the derivation cohort the c-statistic for the score was 0.71 and in the validation cohort 0.70 [1]. The score performed equivalently whatever the treatment strategy: the c-statistic was 0.72 among patients given $\geq$2 antithrombotic drugs and 0.73 among those given fewer, 0.73 with an invasive strategy and 0.68 with a conservative one [1].

External validation in a Chinese cohort of 4,939 ACS patients undergoing PCI showed a c-statistic of 0.80 for all patients and 0.73 for NSTE-ACS patients specifically, with good calibration on the Hosmer–Lemeshow test [2]. In a Spanish cohort of 1,391 STEMI patients undergoing primary PCI, the c-statistic for CRUSADE was 0.77, with no significant difference from the ACTION score but significantly better than ACUITY-HORIZONS [3].

One methodological limitation is that in the Chinese cohort CRUSADE systematically overestimated the absolute bleeding risk in all risk bands: at high risk the predicted rate was 11.9 per cent against an observed rate of 5.3 per cent [2]. The overall bleeding rate in the Chinese cohort was only 1.1 per cent, compared with just over 9 per cent in the American derivation cohort, which is explained by the higher proportion of radial access (90 per cent) and by differences in drug choice [2].

Limitations

The derivation cohort was collected between 2003 and 2006 in American hospitals, before radial access became standard and before ticagrelor and prasugrel were introduced [1]. The general bleeding risk in NSTEMI has fallen as radial access, radial compression and modern dose adjustment have become routine, which means that in many contemporary populations the score overestimates the absolute bleeding risk even though discrimination is preserved [2].

The score does not include the chosen antithrombotic intensity, the access route or dose adjustment, since it was designed as a tool for decisions before treatment begins. The score should therefore be read as an estimate of the patient's underlying propensity to bleed, not as a prediction of the actual outcome given a particular strategy [1].

Age was a univariate predictor in the derivation cohort but did not remain independent in the multivariable model, possibly because its effect is partly captured by the lower haematocrit and creatinine clearance of older patients [1]. The score has not been validated specifically for patients on warfarin or direct oral anticoagulants at presentation, since such patients were excluded from the derivation. Patients who underwent coronary bypass surgery were censored at the procedure, so the score does not estimate the bleeding risk of surgical revascularisation [1].

With a conservative strategy, discrimination was lower (c-statistic 0.68) than with an invasive one, which may reflect the different composition of underlying disease in patients managed conservatively [1].

References

1. Subherwal S et al. Baseline risk of major bleeding in non-ST-segment elevation myocardial infarction: the CRUSADE bleeding score. Circulation 2009. PMID: 19332461
2. Liu R et al. Comparison of the performance of the CRUSADE, ACUITY-HORIZONS, and ACTION bleeding scores in ACS patients undergoing PCI: insights from a cohort of 4939 patients in China. J Geriatr Cardiol 2017. PMID: 28491083
3. Flores-Ríos X et al. Comparison of the performance of the CRUSADE, ACUITY-HORIZONS, and ACTION bleeding risk scores in STEMI undergoing primary PCI: insights from a cohort of 1391 patients. Eur Heart J Acute Cardiovasc Care 2013. PMID: 24062930