The RENAMI (REgistry of New Antiplatelet therapy in patients with acute Myocardial Infarction) registry is a retrospective, observational, multicenter, international registry with voluntary participation of 11 sites in 6 European countries (Spain, Italy, Switzerland, Greece, Serbia, and the United Kingdom). This registry is an unfunded, investigator-dependent registry that arose from the need to learn more about the clinical benefit (ischemic-hemorrhagic risk) of DAPT with ticagrelor vs prasugrel in ACS. At the European Society of Cardiology congress held in 2016, the registry was suggested at a meeting of young investigators, and the following inclusion criteria were established: a) consecutive patients discharged with a diagnosis of ACS during any period between January 2012 and January 2016; b) with coronary stenosis ≥ 50% in the left coronary trunk or ≥ 70% in the rest of the coronary tree; c) treated by coronary stenting during the index hospitalization, and d) treated at hospital discharge with DAPT: aspirin (100 mg every 24 hours) plus prasugrel (10 mg every 24 hours) or aspirin plus ticagrelor (90 mg every 12 hours).

A database was designed for retrospective collection of information on clinical, laboratory, angiographic, and follow-up variables (death, ischemic and hemorrhagic events) (Table 1 of the supplementary material). The above databases from each of the 11 participating sites (Table 2 of the supplementary material) were sent in encrypted format to the coordinating site (Hospital Universitario Álvaro Cunqueiro in Vigo, Pontevedra, Spain) and then merged into a single registry. Two investigators at the coordinating site (E. Abu-Assi and S. Raposeiras-Roubín) were responsible for constructing the combined registry. The study was conducted in accordance with the Declaration of Helsinki, and approval was obtained from the local ethics committees.

The study classified ACS as ST-segment elevation acute myocardial infarction (AMI), non–ST-segment elevation AMI, and unstable angina.10,11

In this study, patients were excluded from the 4424 patients included in the RENAMI registry (based on the original design of the PARIS score, which focuses on events after discharge) in the following cases: a) BARC (Bleeding Academic Research Consortium) type 3 or 5 major bleeding during hospitalization (n = 52 [1.2%])15 and b) spontaneous in-hospital AMI16 or stent thrombosis (probable or confirmed)17 (n = 62 [1.4%]).

The final cohort for this study consisted of 4310 patients. All diagnostic and therapeutic procedures were performed according to local practices.

The aim of the study was to evaluate the clinical usefulness of the PARIS score in assessing the ischemic-hemorrhagic net benefit in patients receiving DAPT with ticagrelor or prasugrel, as well as the predictive capacity of each score (PARISischemic and PARIShemorrhagic) to assess the risk of reinfarction/stent thrombosis and major bleeding.

Events were defined in this study according to the definitions adopted in the PARIS study.12 Major bleeding was considered to be BARC type 3 or 5 bleeding, whereas ischemic events were considered to be spontaneous AMI (defined as elevated myocardial injury markers above the upper limit of normal in combination with anginal symptoms or electrocardiographic abnormalities consistent with myocardial ischemia16) and confirmed or probable stent thrombosis according to the Academic Research Consortium.17

Information on events during follow-up was obtained from hospital and administrative medical records. The study only considered the first ischemic and bleeding event. The follow-up time was considered to have ended when the patient experienced an ischemic (n = 80 [1.8%]) or bleeding (n = 66 [1.5%]) event, death (n = 97 [2.3%]), DAPT suspension/withdrawal (n = 2609 [60.5%]), or end of follow-up in the local clinical registry.

The PARISischemic and PARIShemorrhagic scores were calculated according to the definitions used in the developmental cohort12 (Table 3 of the supplementary material).

Continuous variables are shown as mean ± standard deviation. Discrete variables are expressed as percentages. Continuous variables were compared using the Student t test, and discrete variables were compared using the Pearson chi-square test. To analyze the incidence of events (ischemic and bleeding), cumulative incidence curves were plotted.

The individual components of the PARIS score were determined and found to be independent predictors in our series. To do this, these components were introduced in a Fine-Gray competing-risks regression model,18 with death as the competing event, to examine their association with the events studied (both ischemic and bleeding). The extent of this association is expressed by the subhazard ratio (sHR) and the respective 95% confidence interval (95%CI). The predictive capacity of the 2 PARIS scores (PARISischemic and PARIShemorrhagic) was also assessed by a Fine-Gray competing-risk regression model, using a similar approach to that described above. For comparisons between groups, the low-risk category was taken as a reference. To take into account the grouping of patients within each hospital, the Stata cluster option was used to perform a nonhierarchical cluster analysis.

The predictive capacity of the final model was calculated by the C statistic, using the c-index function of the pec extension for R. The calibration was assessed by comparing the observed and predicted probability in the 3 risk groups originally established in the referral cohort of the PARIS score,12 and by indicating the P values obtained by applying the Hosmer-Lemeshow test.

The ischemic-hemorrhagic net benefit was calculated as the absolute difference between the predicted probabilities for ischemic events and bleeding events, analogous to the approach taken in the referral cohort for the PARIS score.13 The expected risk of ischemic and bleeding events for both scores of the PARIS scale were modeled by polynomial fractions. Positive differences indicate excess ischemic risk, whereas negative differences indicate excess bleeding risk.

The predictive capacity of both scores for mortality was analyzed by a Cox regression model (Figure 1 of the supplementary material, Figure 2 of the supplementary material, Figure 3 of the supplementary material, and Figure 4 of the supplementary material). The linearity requirement for the 2 PARIS scores was confirmed by the Stata nlcheck command.

In the RENAMI registry, complete data were available for all variables except baseline hemoglobin and serum creatinine (3.3% of lost values, n = 141) and height and weight (16.6% of lost values, n = 717). Lost values were replaced with the medians of each variable according to sex. To assess the impact of lost data on the predictive capacity of the score, the sensitivity was analyzed after exclusion of patients with lost values (Table 4 of the supplementary material and Table 5 of the supplementary material).

Statistical analyses were performed using Stata/MP 13.1 and R 3.3.1. Results were considered statistically significant if P < .05.

The mean age of the 4310 patients studied was 60.9 ± 11.5, and 20.8% were women. In all, 58.0% had ST-segment elevation AMI; 32.9% had AMI without ST-segment elevation, and 9.0% had unstable angina; 11.1% were in Killip class ≥ II.

A total of 61.4% (n = 2647) were treated with DAPT with ticagrelor. Six-month GRACE score data were available for 1528 (35.4%) patients, with a value of 120.7 ± 32.1 points. The mean GRACE score was 118.3 ± 32.1 points in patients with non–ST-segment elevation ACS and 122.7 ± 31.9 in patients with ST-segment elevation AMI.

During 17.2 ± 8.3 months, 97 (2.3%) patients died; 80 experienced ischemic events (cumulative incidence of 1-year events, 1.9%; 95%CI, 1.5%-2.3%), of which 41 were stent thrombosis; 66 patients experienced BARC type 3 or 5 bleeding (cumulative incidence of 1-year events, 1.6%; 95%CI, 1.2-2.0%). Table 1 compares the baseline characteristics of patients with and without ischemic and bleeding events.

The PARISischemic score was significantly associated with ischemic risk, both continuously (sHR = 1.27; 95%CI, 1.16-1.39; P < .001) and categorically (Figure 1A). The PARIShemorrhagic score was also significantly associated with hemorrhagic risk, particularly continuously (sHR = 1.14; 95%CI, 1.01-1.30; P = .038). As a categorical variable, the moderate risk group showed no differences in the incidence of bleeding compared with the low-risk group (Figure 1B).

Figure 1.

Cumulative incidence curves by risk group according to the PARIS scale for ischemic (A) and hemorrhagic (B) events. 95%CI, 95% confidence interval; sHR, subhazard ratio.

(0.2MB).

Table 2 shows the association of variables that comprise the PARIS risk scores with ischemic and bleeding events.

The discriminatory capacity of the PARIS score for ischemic events was modest (C statistic = 0.64; 95%CI, 0.62-0.66), with good calibration (Figure 2A), but was suboptimal for bleeding events (C statistic = 0.56; 95%CI, 0.53-0.59), with acceptable calibration, particularly for low- and moderate-risk groups (Figure 2B).

Figure 2.

Cumulative incidences of events (observed and predicted) based on risk groups according to PARIS score, differentiated between ischemic (A) and hemorrhagic (B) events.

(0.16MB).

Figure 3 shows the distribution of both PARIS scores, with the predicted probability of the respective events. Most patients were classified as low-to-moderate ischemic risk (≤ 4 points) and low hemorrhagic risk (≤ 3 points). Only a few patients were classified as high hemorrhagic risk (n = 280 [6.5%]).

Figure 3.

Frequency histogram of thrombotic and ischemic events (estimated frequencies modeled by polynomial fractions).

(0.17MB).

Most (85.5%) patients at low ischemic risk also had low hemorrhagic risk according to the PARIS scale. However, only 11.3% of patients at high ischemic risk were at high hemorrhagic risk (Figure 4).

Figure 4.

Distribution of thrombotic and ischemic risk groups.

(0.12MB).

Figure 5 contains a plot of the ischemic-hemorrhagic net benefit according to the patient's ischemic and hemorrhagic risk group. The ischemic-hemorrhagic net benefit was negative (higher risk of bleeding vs ischemic events) for most patients at low ischemic risk (higher as hemorrhagic risk increases), as well as for patients at moderate ischemic risk and moderate-to-high hemorrhagic risk.

Figure 5.

Ischemic-hemorrhagic balance based on PARIS score. The expected risk of ischemic and hemorrhagic events is modeled by polynomial fractions. DAPT, dual antiplatelet therapy.

(0.49MB).

In patients at high ischemic risk, the ischemic-hemorrhagic net benefit tended to be positive, particularly for those at low-to-moderate hemorrhagic risk. These results are similar to those obtained for ticagrelor and for prasugrel (Figure 5).

The main limitation of the present study is its retrospective nature, with the inherent limitations of this type of registry. As mentioned in the discussion, there is considerable therapeutic selection bias in the population because all patients had been prescribed DAPT with prasugrel or ticagrelor instead of clopidogrel according to the guidelines and preferences at each site and at the attending physician's discretion, primarily limited by the subjective estimate of a low or moderate bleeding risk. Additionally, the study excluded patients with ischemic and hemorrhagic events during hospitalization. Moreover, the results of the present study should be interpreted in the context of patients at moderate risk according to the GRACE score used to predict 6-month mortality. Another limitation of this study is that there were no data on the vascular access used for inpatient coronary angiography. Likewise, the findings can be generalized only to ACS patients treated with percutaneous coronary intervention during hospitalization and at discharge with DAPT with prasugrel or ticagrelor plus aspirin. Despite these limitations, the study supports the hypothesis that the PARIS ischemic-hemorrhagic scale could be useful when deciding the type and duration of DAPT after ACS. A prospective study should be conducted to determine whether DAPT should be personalized to each patient according to the PARIS scale to prevent ischemic and hemorrhagic events.