ABC | Volume 110, Nº1, January 2018

Original Article Cerqueira Junior et al GRACE Score in Octogenarians Arq Bras Cardiol. 2018; 110(1):24-29 2. Ischemic electrocardiographic alteration, consisting of inversion of the T wave (≥ 0.1 mV) or alterations of the ST segment (≥ 0.05 mV); and 3. Previously documented coronary artery disease, defined by a history of myocardial infarction with Q wave or previous angiography demonstrating coronary obstruction ≥ 70%. The protocol is in compliancewith theDeclaration of Helsinki, released by the Research Ethic Committee of the institution and all patients evaluated signed the Informed Consent. GRACE Score The clinical data of each patient’s admission in the emergency unit, electrocardiograms performed within the first 6 hours of treatment, troponin T and troponin I dosage in the first 12 hours of treatment and the value of the first plasma creatinine were used to calculate the GRACE Score. The increased myocardial necrosis marker as a component of the scores was defined as troponin over 99 percent. The GRACE Score consists of eight variables: five semi-quantitative ones, i.e. , different weight for each age range (systolic blood pressure, heart rate, plasma creatinine and Killip class); and three dichotomic ones (ST segment depression, elevation of myocardial necrosis marker and cardiac arrest at the moment of admission). The final score can range from 0 to 372. 4 Data analysis The accuracy of the GRACE Score was evaluated by discrimination and calibration analyses, which were compared between two groups: one referred to as “very elderly” and the other as “not very elderly”; the first one defined by individuals ≥ 80 years old. The GRACE Score has its performance evaluated by the ability to predict death by any given cause during the hospitalization period. Statistical analysis Numerical variables were expressed as mean and standard deviation when presenting normal distribution or a small deviation from normality, whilst median and interquartile interval were preferable in the presence of at least a moderate deviation from normality. The analysis of normality was performed through combined visualization of the histogram and Q-Q plots, description of skewness and kurtosis with confidence intervals, and normality tests (Shapiro-Wilk and Kolmogorov-Smirnov). Continuous variables were compared by the Student’s t-test or Wilcoxon test when they presented normal and non-normal distribution, respectively. Categorical variables were expressed in proportion and compared through the χ 2 test. The discriminatory capacity of the GRACE Score for mortality was evaluated by the area below the curve of receiver operator characteristics – ROC (statistic-C), which was compared between the two groups by the unpaired Hanley-McNeil test. 8 The calibration of the scores had a hypothesis test carried out by the Hosmer-Lemeshow technique and was described by the comparison between mortality predicted by GRACE and the one observed in each prediction quartile. The influence of age in the performance of GRACE was tested by the p-value of the interaction by logistic regression analysis. The SPSS software, version 21, was used. The statistical significance was defined by two-tailed p-value lower than 0.05. Results Characteristics of the sample A total of 994 individuals were studied, of which 57% were male and 77% had non-ST elevation ACS. The mean age of the sample was 65 ± 13 years old, of which 173 (17%) were classified as very elderly for being 80 years old or older. The mean age of the very elderly was 85 ± 3.7 years old, compared to 61 ± 11 years of age in the rest of the sample (p < 0.001). The GRACE Score for very elderly patients was 162 ± 34, significantly higher than the one of other patients (115 ± 35; p < 0.001). This higher score in GRACE for very elderly people is due to the difference not only in age, but also in the variables troponin, non-ST elevation, Killip and blood pressure. Percutaneous revascularization during hospitalization was similar in both groups, while surgical revascularization was less frequent in the group of the very elderly. During hospitalization, in-hospital mortality was 5.8% of the total sample, being significantly higher in the group of very elderly people in relation to patients with less than 80 years of age (16% versus 3.7%; p < 0.001) (Table 1). Discriminatory ability of the GRACE Score In the total sample, the GRACE Score had statistic-C of 0.87 (95% CI = 0.82 – 0.92) in predicting hospital death. GRACE’s statistic-C among the very elderly was 0.86 (95% CI = 0.78 – 0.93), without difference in relation to the value found in patients aged less than 80 years old (statistic‑C = 0.83; 95% CI = 0.75 – 0.91), with p = 0.69 in the comparison of both curves (Figure 1). In the logistic regression in which GRACE and very elderly people were simultaneously inserted in the prediction model, there was no interaction between these two variables (p = 0.25). In addition, GRACE remained an independent predictor of age (p < 0.001). According to the ROC curve, the cutoff score in GRACE with best performance in the group of not very elderly was 134, with sensitivity of 83% and specificity of 76%. Among the very elderly, the cutoff point is displaced upward, with a value of 184, corresponding to the sensitivity of 77% and specificity of 87%. Calibration of the GRACE Score In the prediction of the incidence of death during hospitalization, the Hosmer-Lemeshow test showed satisfactory calibration in both groups, very elderly ( χ 2  = 2.2; p = 0.98) and not very elderly ( χ 2 = 9.0; p = 0.35). Figure 2 presents the stratified analysis per quartile of the probability predicted by GRACE for hospital death, comparing the predicted and the observed within both age groups. Only the fourth quartile had an underestimated predicted mortality compared to the one observed, in both groups. 25