IJCS | Volume 31, Nº6, November / December 2018

563 Lacerda et al. Paradoxical choice for radial access in ACS Int J Cardiovasc Sci. 2018;31(6)562-568 Original Article the effect of medical decision making. Effectiveness represents the performance of the therapy in the real world, in which allocation depends on the medical decision making. Effectiveness is then optimized when allocation of treatment prioritizes patients at high risk for the outcome expected to be prevented by the intervention in question. Radial access would be more effective for patients at higher risk of bleeding who are allocated to this intervention. In a recent study, Wimmer et al., 7 reported a risk-treatment paradox, in which the radial approachwas less frequent in patients at higher bleeding risk than in those at lower risk. The present study aimed to explore this phenomenon. Using the Prospective Registry of ACS, we testedwhether the radial access was the first choice for PCI in patients at high risk of bleeding, which was evaluated by the CRUSADE 8 and the ACUITY 9 scores. Also, we identified predictors of radial access and developed a propensity score of representative, predicting factors of medical decision making. Methods Sample selection We included in the study patients consecutively admitted to the coronary unit of a tertiary hospital betweenDecember 2011 and January 2016with diagnosis of ACS (unstable angina or myocardial infarction) with previous diagnostic or therapeutic invasive cardiac procedures. ACS was defined as precordial discomfort in the 48 hours prior to admission, associated with at least one of the following criteria: 1) myocardial necrosis markers, defined as troponin T ≥ 0.01 ug/L or troponin I > 0.034 g/L, corresponding to values above the 99 th percentile; 10 2) ischemic electrocardiographic changes, consisting of T-wave inversion (≥ 0.1 mV) or ST-segment changes (≥ 0.05 mV); 3) previous coronary artery disease, defined as previous Q-wave myocardial infarction or coronary obstruction ≥ 70% confirmed by angiography. Patients who declined to participate were excluded from the study. The study protocol was in accordance with the Helsinki declaration and approved by the local ethics committee. All patients signed the informed consent form. Study protocol This is a registry of ACS, composed by collection of prospective data. Variables of these data were used for calculation of bleeding scores. Access site for the first arterial puncture in the first (diagnostic or therapeutic) coronary procedure was systematically registered on data collection form. Major bleeding was defined as BARC (Bleeding Academic Research Consortium) type 3 or type 5. 11 The criteria for type 3 bleeding were as follow – decrease in hemoglobin of 3-5 g/dL or need for transfusion (type 3a); hemoglobin drop ≥ 5 g/dL, cardiac tamponade, requirement of surgical intervention or hemodynamic instability for control (type 3b); and intracranial or intraocular bleed (type 3c). Type 5 bleeding is a definite fatal bleeding (direct causal relationship, type 5a) or a probable fatal bleeding (indirect causal relationship, type 5b). Minor bleeding (type 1 or type 2) or cardiac surgery-related bleeding (type 4) were not included in the analysis. Bleeding risk scores The CRUSADE score was used to evaluate the baseline risk of bleeding. This instrument is composed of eight variables – four categorical variables (female sex, signs of heart failure, diabetes and peripheral arterial disease) and four numerical variables (baseline hematocrit, creatinine clearance, heart rate, and systolic blood pressure). The point scores were calculated based on the value of each variable; the sum of all variables indicated pre- determined levels (low, intermediate and high). 8 Bleeding risk was also confirmed by the ACUITY score, composed of seven variables – four categorical variables (female sex, anemia, bivalirudin therapy and type of ACS) and three numerical variables (age, creatinine clearance, white blood cell count). 9 Statistical analysis Although the collection of the variables included in the primary analysis was predetermined, the association between bleeding score and the access route was a posteriori exploratory analysis. Nevertheless, we estimated that a minimum of 100 patients with radial or femoral access would allow the insertion of 10 covariables into the propensity model, based on the logistic regression principle, which establishes the need of at least 10 patients with the outcome in question for each covariable. 12 Numerical variables were described as mean and standard deviation or median and interquartile range, as appropriate. Normality of numerical variables was verified by the Shapiro-Wilk test. Categorical variables