In this multicenter observational study of patients prospectively enrolled in the second phase of the ProACS13 with a diagnosis of NSTE-ACS and aged ≥80 years, predictors of IHM and in-hospital MB were analyzed and prediction models (PM) were created for these endpoints. The accuracy of the GRACE and CRUSADE scores and the prediction models for IHM (IHM-PM) and MB (MB-PM) were then compared.

We analyzed 544 consecutive patients aged ≥80 years diagnosed with NSTE-ACS in 14 Portuguese hospitals and included in the ProACS between October 1, 2010 and October 25, 2012.

NSTE-ACS was defined on the basis of an admission diagnosis of non-ST-elevation myocardial infarction (NSTEMI), defined as lack of persistent ST elevation (<30 min) associated with elevated cardiac biomarkers (troponin or CK-MB) in a clinical context suggestive of myocardial ischemia, or unstable angina (UA), defined as angina or anginal equivalent, with or without ECG changes and without elevated cardiac biomarkers.

The ProACS12 is an observational study with continuous and prospective inclusion of all adult patients (≥18 years) diagnosed with acute coronary syndrome (ACS) within 48 hours of symptom onset. MI types 2, 4 and 5 according to the Joint ESC/ACCF/AHA/WHF Task Force for the Redefinition of Myocardial Infarction (2007)15 were excluded. Further information on the registry protocol is available on the website of the Portuguese Society of Cardiology.16

The following variables from the ProACS were included in the univariate analysis: demographic data (age, gender, body mass index [BMI]); cardiovascular risk factors (documented history of hypertension, diabetes, dyslipidemia, current smoking in the 30 days before admission, family history of CAD); personal cardiovascular (MI, heart failure [HF], stroke/transient ischemic attack) and non-cardiovascular history (cancer, clinically significant or life-threatening bleeding, dementia, chronic obstructive pulmonary disease [COPD], chronic renal disease [defined as previous history or presence of at least one of the following criteria: creatinine prior to admission >2 mg/dl, dialysis or kidney transplantation]); medication prior to admission (antiplatelets: aspirin, clopidogrel, both, or others; anticoagulants: vitamin K antagonists, others); data on hospital admission (location, mode of transport to the hospital, symptoms, blood pressure, heart rate, Killip class,17 ST-segment depression on ECG); and laboratory test results (isolated troponin elevation, hemoglobin, HbA1c, hematocrit, platelets, blood glucose, and creatinine). Hematocrit was calculated on the basis of hemoglobin level in accordance with the formula of Bain and Bates18 (hematocrit=hemoglobin [g/dl]×3). Creatinine clearance was estimated using the Cockcroft-Gault formula.19 Data on hospital stay were also recorded, including duration, laboratory values (minimum hemoglobin, peak creatinine), medication (antiplatelets: aspirin, clopidogrel, others; anticoagulants: unfractionated heparin, enoxaparin, fondaparinux, vitamin K antagonists, others; glycoprotein IIb/IIIa inhibitors, beta-blockers, angiotensin-converting enzyme [ACE] inhibitors or angiotensin receptor blockers [ARBs], statins, nitrates, calcium channel blockers, ivabradine, aldosterone antagonists, diuretics, antiarrhythmics, insulin, oral antidiabetic agents, inotropic agents, levosimendan), whether coronary angiography was performed, vascular access, use of vascular closure devices, culprit vessel, number of vessels with ≥50% stenosis, percutaneous coronary intervention (PCI), type of stent used, coronary artery bypass grafting (CABG), left ventricular (LV) function, temporary pacing, intra-aortic balloon pump, and invasive and non-invasive mechanical ventilation, as well as in-hospital complications (reinfarction, mechanical complications, arrhythmias, resuscitation following cardiac arrest, stroke, HF, major bleeding and death).

The GRACE5 and CRUSADE6 scores were calculated for all patients.

Further information on the definition of these variables is available on the website of the Portuguese Society of Cardiology.20

The study endpoints were IHM and MB.

IHM was defined as death from any cause during hospitalization for NSTE-ACS. MB was defined as intracranial bleeding or bleeding resulting in hemodynamic compromise requiring treatment, the criteria in the Global Use of Strategies to Open Occluded Arteries (GUSTO) trial.21 The GUSTO criteria were chosen because this classification is used in the ProACS; the authors also considered it the most appropriate for the population and the study objectives, since it was important to identify bleeding that was sufficiently severe to justify not using therapies which reduce ischemic complications but increase bleeding risk in ACS.

An overall descriptive analysis of the study population was performed based on all the parameters assessed. Categorical variables were expressed as absolute and relative frequencies, and continuous variables as means and standard deviation.

Independent predictors of each of the study endpoints, IHM and MB, were identified, using univariate analysis to test the association between the endpoint and each potential predictor. The chi-square test was used for categorical variables and the t test for continuous variables. The Mann-Whitney test was used when the assumptions for the t test were not met. The normality of the distribution of continuous variables and their equality of variances were determined by the Kolmogorov-Smirnov test and the Levene test, respectively.

Logistic regression techniques were applied to identify independent predictors of IHM and MB considering all variables with statistical significance, using the forward stepwise method with likelihood ratios. The model included variables used in the CRUSADE score and, separately, variables related to medication during hospitalization: dual antiplatelet therapy, dual antiplatelet therapy plus anticoagulants (triple therapy), or dual antiplatelet therapy plus glycoprotein IIb/IIIa inhibitors (triple antiplatelet therapy). The entry and exit values were set at 0.10 and 0.15, respectively. The risk for each endpoint associated with each predictor was estimated using odds ratios and 95% confidence intervals (CI).

On the basis of the logistic regression analysis (“Enter” method), prediction models for IHM and MB were created on the basis of variables at admission with statistical significance for each endpoint or clinical significance: IHM-PM (LV ejection fraction <30%, Killip class >I, ST-segment depression on ECG, systolic blood pressure, heart rate, blood glucose, creatinine, and isolated troponin elevation) and MB-PM (history of bleeding, history of COPD, Killip class at admission >I, and admission creatinine). The fit of the models was assessed using the Hosmer-Lemeshow test.

On the basis of the regression coefficients, the predictive value of each model was determined and its discriminatory capacity was estimated by calculating the area under the receiver operating characteristic (ROC) curve, together with its sensitivity and specificity. ROC curves evaluate the ability of a model to attribute a higher probability of events to patients who suffered the study endpoints. In this way the diagnostic accuracy of the GRACE score and the IHM-PM for IHM, and the CRUSADE score and the MB-PM for MB, were assessed, using MedCalc for Windows version 9.2.0.1. The rest of the statistical analysis was performed with SPSS version 19, using the Hanley-McNeil test. The level of significance was taken to be 5%.

The characteristics of the 544 patients studied are presented in Table 1. Ages ranged from 80 to 92 years (mean 84). Diagnosis at admission was NSTEMI in 91% (n=495) and UA in 9% (n=49). Figure 1 shows the distribution of the total population and of patients with IHM and MB according to the GRACE and CRUSADE scores, respectively. In the study population, 93.6% of patients who suffered IHM were classified as high risk by the GRACE score, while the CRUSADE score classified only 53.9% of those with MB as high or very high risk.

IHM: in-hospital mortality; MB: major bleeding.'> IHM: in-hospital mortality; MB: major bleeding.' title='Distribution of the total population and of patients with events according to GRACE and CRUSADE scores. IHM: in-hospital mortality; MB: major bleeding.'/>

Figure 1.

Distribution of the total population and of patients with events according to GRACE and CRUSADE scores. IHM: in-hospital mortality; MB: major bleeding.

(0.21MB).

Around 98% of patients were medicated with aspirin, 86.9% with clopidogrel and 5.9% with glycoprotein IIb/IIIa inhibitors. Enoxaparin was the most commonly prescribed anticoagulant. Coronary angiography was performed in 59.4% of patients, PCI in 34.7% and CABG in 14.7%. The mean GRACE score in this population was 161±27 and the mean CRUSADE score was 44±14. Median hospital stay was five days (25th percentile: 3; 75th percentile: 8); the IHM endpoint was observed in 6.6% and the MB endpoint in 2.9% (Table 1).

In univariate analysis, IHM was associated with higher heart rate and creatinine on admission, peak creatinine, and Killip class >I, ST-segment depression, isolated troponin elevation, and lower systolic blood pressure on admission. In-hospital survival was associated with medication with beta-blockers, calcium channel blockers and ACE inhibitors/ARBs, coronary angiography and PCI, and absence of left main disease. Patients who died during hospitalization more frequently received levosimendan, inotropic agents, mechanical ventilation and temporary pacing, and had a higher prevalence of ischemic and bleeding complications (Table 2). No association was seen between IHM and age, cardiovascular risk factors, previous medication, vascular access or closure devices used in coronary angiography, type of stent used for PCI, or CABG.

In multivariate analysis, independent predictors of IHM were Killip class >1 on admission, LV ejection fraction <30%, and MB (which increased the risk of IHM almost 11-fold). Use of coronary angiography had a protective effect (Table 3).

The GRACE score showed good diagnostic accuracy in identifying patients who died during hospitalization, with an area under the curve (AUC) of 0.75 (95% CI 0.63–0.87, p<0.001) (Figure 1). The Youden index for the GRACE score was 170 (sensitivity 77%, specificity 67%).

The IHM-PM presented a good fit for IHM according to the Hosmer-Lemeshow test (p=0.533). ROC curve analysis showed excellent discriminatory capacity for IHM, with an AUC of 0.81 (95% CI 0.72–0.90, p<0.001), although no significant difference was seen between the AUCs of the two models (0.06, p=0.239) (Figure 2).

IHM-PM for assessment of in-hospital mortality. AUC: area under the curve; CI: confidence interval; Difference: difference between the AUCs of the two ROC curves; IHM-PM: in-hospital mortality prediction model.'> IHM-PM for assessment of in-hospital mortality. AUC: area under the curve; CI: confidence interval; Difference: difference between the AUCs of the two ROC curves; IHM-PM: in-hospital mortality prediction model.' title='ROC curve analysis of the GRACE score and IHM-PM for assessment of in-hospital mortality. AUC: area under the curve; CI: confidence interval; Difference: difference between the AUCs of the two ROC curves; IHM-PM: in-hospital mortality prediction model.'/>

Figure 2.

ROC curve analysis of the GRACE score and IHM-PM for assessment of in-hospital mortality. AUC: area under the curve; CI: confidence interval; Difference: difference between the AUCs of the two ROC curves; IHM-PM: in-hospital mortality prediction model.

(0.19MB).

MB was associated with a history of bleeding or COPD, Killip class >I on admission, higher peak creatinine and lower minimum hemoglobin, treatment with unfractionated heparin, antiplatelets other than aspirin or clopidogrel, and aldosterone antagonists, longer hospital stay and higher IHM (Table 4). On univariate analysis no association was seen between MB and age, gender, other cardiovascular risk factors, previous medication, BMI, admission hemoglobin, creatinine or platelets, medication with glycoprotein IIb/IIIa inhibitors, or coronary angiography, PCI, CABG, type of stent, or vascular access or closure devices.

On multivariate analysis, of all the associations with statistical significance on univariate analysis, only medical therapy during hospitalization (unfractionated heparin, antiplatelets other than aspirin or clopidogrel, and aldosterone antagonists) were independent predictors of MB. The same analysis including these predictors and the parameters of the CRUSADE score identified history of COPD and medication with unfractionated heparin and aldosterone antagonists as predictors of MB. In multivariate analysis including the same predictors and dual antiplatelet or triple therapy separately, marginally significant p values were obtained for COPD and unfractionated heparin, while antiplatelets other than aspirin or clopidogrel and aldosterone antagonists were predictors of MB. In a similar analysis including triple antiplatelet therapy, this was shown to be an independent predictor of MB, as were COPD and medication with aldosterone antagonists (Table 5).

ROC curve analysis showed that the CRUSADE score had poor diagnostic accuracy in predicting MB in this population, with an AUC of 0.51 (95% CI 0.30–0.63, p=0.942; Figure 3).

AUC: area under the curve; CI: confidence interval; MB-PM: major bleeding prediction model.'> AUC: area under the curve; CI: confidence interval; MB-PM: major bleeding prediction model.' title='ROC curve analysis of the CRUSADE score and MB-PM for assessment of major bleeding. AUC: area under the curve; CI: confidence interval; MB-PM: major bleeding prediction model.'/>

Figure 3.

ROC curve analysis of the CRUSADE score and MB-PM for assessment of major bleeding. AUC: area under the curve; CI: confidence interval; MB-PM: major bleeding prediction model.

(0.19MB).

The MB-PM presented a good fit on the Hosmer-Lemeshow test (p=0.111) and on ROC curve analysis showed good discriminatory capacity for MB in the study population (AUC 0.68, 95% CI 0.52–0.84, p=0.032; Figure 3).

The authors declare that no experiments were performed on humans or animals for this study.

The authors declare that they have followed the protocols of their work center on the publication of patient data.

The authors have obtained the written informed consent of the patients or subjects mentioned in the article. The corresponding author is in possession of this document.