ABC | Volume 114, Nº3, March 2020

Original Article Pivatto Júnior et al. Risk scores for surgery in endocarditis Arq Bras Cardiol. 2020; 114(3):518-524 Table 2 – Patient characteristics and surgical details. VARIABLE n = 107 Age (years) 58.1±14.5 Female sex 26 (24.3) Hypertension 60 (56.1) NYHA III/IV 53 (49.5) Abscess 40 (37.4) Previous cardiac surgery 35 (32.7) Degenerative valve disease 31 (29.0) Severe PAH 31 (29.0) Prosthetic endocarditis 31 (29.0) Acute renal insufficiency 30 (28.0) Severe renal dysfunction* 25 (26.0) Dialysis 22 (20.6) Thrombocytopenia 20 (18.7) Critical preoperative state 19 (17.8) LVEF ≤ 50% 17 (15.9) IDDM 14 (13.1) Previous infective endocarditis 11 (10.3) Rheumatic valvulopathy 10 (9.3) Bicuspid aortic valve 8 (7.5) Extracardiac arteriopathy 8 (7.5) Previous MI 8 (7.5) Chronic lung disease 7 (6.5) Poor mobility 7 (6.5) Recent MI 3 (2.8) CCS class 4 angina 1 (0.9) Location of infective endocarditis Aortic 47 (43.9) Mitral 35 (32.7) Aortic + Mitral 20 (18.7) Tricuspid 4 (3.7) Tricuspid + Mitral 1 (0.9) Identified causative microorganism 72 (67.3) Streptococcus viridans 19 (17.8) Enterococcus sp. 10 (9.3) Staphylococcus aureus 9 (8.4) Magnitude of intervention Single, non-CABG 81 (75.7) Two procedures 25 (23.4) Three procedures 1 (0.9) Urgency Urgent 98 (91.6) Emergent 9 (8.4) Associated CABG 8 (7.5) Extracorporeal circulation time (min) 84.0 (65.0-110.0) Cross-clamp time (min) 65.0 (51.0-84.0) CABG: coronary artery bypass graft; CCS: Canadian Cardiovascular Society; IDDM: insulin-dependent diabetes mellitus; NYHA: New York Heart Association; PAH: pulmonary arterial hypertension; LVEF: left ventricular ejection fraction; MI: myocardial infarction. *We excluded patients on preoperative hemodialysis (n = 22; 20.6%) and those for whom body weight data were unavailable (n = 11; 10.3%), which makes it impossible to calculate the creatinine clearance. Data expressed as mean ± standard deviation, n (%), or median (interquartile range). mortality; thus, adjustments are required. In our cohort, the logistic EuroSCORE seemed to be the best predictor of mortality risk. The causative microorganism was identified in only 67.3% of cases in this cohort, unlike in the validation cohorts of the IE-specific scores, in which the detection rate was 81.0- 86.6%. 10-12 Similarly, Staphylococcus aureus , which causes an aggressive and often fatal infection, 21 was the causative microorganism in only 8.4% of cases, while in the validation cohorts this percentage ranged from 17.5 to 19.9%. 11,12 These two factors probably explain, at least partly, the low accuracy of IE-specific scores in our cohort. The same occurred with other items included in specific scores, such as NYHA class IV in the AEPEI score 10 (37.7 vs. 20.6%), LVEF ≤ 50% in the EndoSCORE 7 (35.9 vs. 15.9%), cardiogenic shock and thrombocytopenia in the RISK-E score 12 (17.9 and 29.2% in the original study vs. 11.2 and 18.7% in the present study, respectively); although strongly associated with mortality, these factors were not significantly prevalent in our cohort. EuroSCORE II, the most commonly used score for preoperative risk assessment in current clinical practice, underestimated the observed mortality 2.5-fold and had poor discriminatory power (AUC = 0.69). The original EuroSCORE II study cohort had a very low percentage of patients with active IE (2.2%); 9 therefore, it is difficult to generalize EuroSCORE II results for IE populations. In an analysis of 149 patients undergoing cardiac surgery for active IE at two French referral centers for cardiac surgery, Patrat-Delon et al. 6 observed that, although EuroSCORE II showed good power of discrimination (AUC = 0.78; 95%CI: 0.70-0.84), its results should be interpreted with caution during the acute phase of IE, because it also underestimated postoperative mortality by 5-10% in half of patients with predicted mortality >10%. In Brazil, Oliveira et al. 22 conducted the only other study to date to evaluate a prediction score in patients with active IE undergoing heart surgery. In this study, which included 88 patients, the EuroSCORE II significantly underestimated hospital mortality, with a mortality ratio O/E of 2.31 (95%CI: 1.41-3.58; p = 0.002). ROC curve analysis was not performed. Patients with active IE were already underrepresented in the EuroSCORE cohort, 8 in which active IE was present in only 3.6% of all valve surgery patients. Madeira et al. 23 in a study including 128 patients who underwent heart surgery for active IE, compared EuroSCORE and EuroSCORE II for perioperative mortality prediction. They observed that the pattern of calibration differed between the scores: EuroSCORE showed a progressive trend towards overprediction, whereas EuroSCORE II tended to underpredict mortality. On the other hand, as in the present study, Mestres et al. 24 in a study including 181 patients with IE (93.2% active), described good discriminatory power (AUC 0.84) and an expected mortality (27.1%) very similar to that observed (28.8%; O/E ratio: 1.1). The need for a dedicated stratification tool, useful both for preoperative patient information and for bedside decision- making, arises from the peculiarities of IE surgery compared with general cardiac surgery: postoperative outcomes may be influenced not only by cardiovascular anatomic and functional 521