ABC | Volume 110, Nº1, January 2018

Original Article Cerqueira Junior et al GRACE Score in Octogenarians Arq Bras Cardiol. 2018; 110(1):24-29 Table 1 – Comparison of clinical characteristics, laboratory characteristics, GRACE Score and mortality between very elderly versus not very elderly Age ≥ 80 Age < 80 p-value Sample size 173 (17%) 821 (83%) – Age (years) 85 ± 3.7 61 ± 11 < 0.001 ‡ Male 82 (47.0%) 487 (59.0%) 0.004 § Non-ST elevation ACS 23 (13.0%) 205 (25.0%) 0.001 § Diabetes 60 (35.0%) 300 (37.0%) 0.613 § Non-ST elevation 55 (32.0%) 308 (37.5%) 0.155 § Positive troponin 123 (71.0%) 557 (68.0%) 0.403 § Classification of Killip < 0.001 Killip I 127 (73.0%) 724 (88.0%) Killip II 21 (12.0%) 49 (6.0%) Killip III 23 (13.0%) 41 (5.0%) Killip IV 2 (1.2%) 7 (0.9%) Systolic pressure (mmHg) 151 ± 32 155 ± 30 0.098 ‡ Heart rate 80 ± 17 80 ± 18 0.519 ‡ Serum Creatinine (mg/dl) 1.1 ± 0.5 1.1 ± 0.9 0.669 ‡ Hemoglobin at admission 13 ± 1.8 14 ± 1.9 < 0.001 ‡ Triarterial or LMD* 38 (30.0%) 126 (18.0%) < 0.001 § Percutaneous Coronary Intervention † 66 (39.0%) 368 (45.0%) 0.129 § Revascularization surgery † 4 (2.0%) 92 (11.0%) < 0.001 § GRACE Score 162 ± 34 115 ± 35 < 0.001 ‡ In-hospital death 28 (16.0%) 30 (4.0%) < 0.001 § ACS: acute coronary syndrome; *Coronariography performed during hospitalization; LMD: left main disease; † Myocardial revascularization treatments during hospitalization; ‡ Compared through Student’s t-test; § Compared through χ 2 test. Discussion The present study demonstrates that the GRACE Score presents satisfactory accuracy in predicting hospital death of very elderly individuals with ACS (octogenarian and nonagenarian ones). The comparison with individuals aged less than 80 years old did not show loss of discriminatory capacity or GRACE’s calibration as the age progressed. Statistic-C values above 0.80 with narrow confidence intervals, in addition to linear growth of mortality observed in the different quartiles of mortality predicted by GRACE, are clear evidence of maintenance of the performance of this score in very elderly. Although the fourth quartile of predicted mortality has underestimated the risk in relation to what was observed, this difference did not compromise the categorization of the fourth larger groups of risk, once that both the observed and the predicted were in mortality ranges considered high for ACS. 4 There was no interaction between the adequacy of GRACE’s model and the age range group defined by the cutoff point of 80 years of age, confirming GRACE’s accuracy among elderly. Age is the marker of greater influence on the probability of hospital death in patients hospitalized with ACS, with exponential risk growth as the value of this variable increases. 6,9,10 The uncertainty of GRACE’s accuracy among very elderly individuals comes from the possibility that there could be less variability of important predicting values within a very advanced age range. For instance, the uniformity of advanced age in this sample may deprive this variable of its discriminatory power, which would not depict great contrast among the individuals. The inclination of this risk function may be lower when there are only very elderly patients. The same may occur with other variables which may be systematically altered in a very elderly sample. Also, the calibration of the score in estimating the numerical risk of death may be different for these patients, once the alpha constant (intercept) tends to be greater in samples with the highest risk. This could explain the need for recalibration of the score. This uncertainty becomes greater when realized that octogenarian patients were not well represented by the sample which derived and validated the GRACE Score as a hospital death predictor. 4,11,12 The median age of that sample was 66 years old, with upper limit of 75 years of age for the interquartile interval, indicating that 3/4 of patients were less than 75 years old, with no description as to who were the octogenarian or the nonagenarian ones. Due to the 26