This is a prespecified subanalysis (ClinicalTrials.govNCT04630288, Safety and Efficacy of Ticagrelor vs Clopidogrel in Patients With Acute Coronary Syndrome) of the CREA-ARIAM registry (ClinicalTrials.gov Identifier NCT02500290, Antiplatelet Therapy in Acute Coronary Syndrome (ACS). Safety and Efficacy of Switching Antiplatelet), a prospective, multicenter investigator-initiated branch of the ARIAM-Andalucía registry (Analysis of Delay in Acute Myocardial Infarction in Andalucía). Details of design, definitions, and primary results of the ARIAM-Andalucía registry and CREA-ARIAM registry have already been reported elsewhere.13,14 The quality control of data in ARIAM-Andalucía registry is carried out at regular intervals through independent external audits by the Agency for Healthcare Quality of Andalucía. Patients with ACS admitted to cardiac care units between March 2015 and April 2019, were prospectively screened for eligibility if they were intended to receive at least 12 months of DAPT with clopidogrel or ticagrelor. Major exclusions were a history of previous intracranial hemorrhage or recent major bleeding, patients discharged on prasugrel or oral anticoagulation, and patients lost to follow-up or with missing data (methods 1 of the supplementary data). The study was approved by the Regional Research Ethics Committee of Andalusia and the Institutional Review Boards at each participating center. All eligible patients were required to give written informed consent.

The primary endpoint was the first occurrence of major adverse cardiac events (MACE), a composite of all-cause mortality, myocardial infarction,15 stroke, unplanned target lesion revascularization, or definite stent thrombosis at 1 year.16 Secondary endpoints included each individual component of the primary endpoint, and a more restrictive definition of MACE (MACE-2) including cardiovascular death, myocardial infarction, unplanned target lesion revascularization, and definite stent thrombosis. Major bleeding was defined as Bleeding Academic Research Consortium type 3 or 5 bleeding17 (methods 2 of the supplementary data).

Clinical endpoints and DAPT exposure status were systematically and prospectively tracked throughout 1 year after discharge during scheduled postdischarge outpatient follow-up visits, and by trained research coordinators through structured telephone interviews with patients or relatives planned at 1, 6 and 12 months after discharge (methods 2 of the supplementary data). These included a dedicated questionnaire in which patients were asked to provide detailed information on hospital readmissions, outpatient visits, drug-related adverse effects, or any change in antiplatelet therapy from the last contact. Specifically, we collected dates of stopping and restarting medication, reasons for drug cessation, and the replacement therapy, when applicable. Self-reported information from patients was systematically validated with data manually extracted from electronic medical records. All potential endpoints identified during the follow-up period underwent formal adjudication by consensus between 2 experienced investigators who were blinded to calendar year and DAPT status, using previously anonymized original source data. Additionally, medication adherence was measured at regular intervals irrespective of clinical status by means of the medication possession ratio (methods 2 of the supplementary data). For this study, the term cessation refers to any unplanned DAPT discontinuation before 12 months, defined as temporary interruption (< 14 days), or permanent discontinuation of ticagrelor or clopidogrel for more than 3 days, with or without aspirin cessation. In keeping with the PARIS registry,5 we considered 2 modes of cessation: physician-guided discontinuation and disruption due to nonadherence, adverse effects, or bleeding. Treatment interruption due to surgery or the need for invasive procedures was categorized as physician-recommended discontinuation. We further categorized nonadherence according to the timing of drug discontinuation in relation to hospital discharge into early (< 90 days) and late (>90 days) cessation, and by the P2Y12-i choice at the time of cessation. Additionally, specific reasons for discontinuation were systematically captured and categorized according to the predefined cessation modes.

Baseline and procedural characteristics according to DAPT cessation status are described using frequencies and percentages for categorical variables, and mean±standard deviation, or median [interquartile range] for continuous variables, as appropriate. Patient characteristics of patients in each cessation group were compared using chi-square tests for categorical variables, and independent samples Student t tests, Mann-Whitney U-tests, or Kruskal-Wallis tests for continuous variables, as appropriate. The inverse probability of censoring weighting (IPCW) approach was used to account for dependent censoring resulting from time-varying confounders18 (methods 2 of the supplementary data). The cumulative incidence of different nonadherence patterns was summarized using weighted Kaplan-Meier estimators. Predictors of premature cessation were assessed using multivariate time-updated Cox regression models using the multivariate fractional polynomials approach.19 The adjusted risk of MACE associated with DAPT cessation was analyzed taking death as a competing event, fitting multivariate time-updated Cox regression models with each cessation mode entered as a time-varying covariate. Models included doubly robust IPCW estimators with participating hospitals entered as cluster random-effect variable using robust estimators to account for the lack of independence induced by the weighted data, and interhospital variability in clinical performance (methods 2 of the supplementary data). Results are expressed as adjusted hazard ratios (aHR) with 95% confidence intervals (95%CI). We performed subgroup analysis across nonadherence categories stratified by P2Y12-i status using the formal (Wald) test for interactions with uninterrupted DAPT as the reference group. In addition, interactions were assessed on the additive scale using the relative excess risk due to interaction (RERI) and the attributable proportion, with 95%CI calculations by the delta method.20 All subgroup analyses were deemed exploratory and no adjustments for multiplicity were applied. To ensure the robustness of the main results, a set of sensitivity analyses were performed: a) for different datasets to address reverse causality, excluding patients more likely to discontinue DAPT; b) fitting flexible parametric survival models using restricted cubic spline functions to model DAPT duration as a continuous covariate; and c) according to PARIS criteria, considering DAPT interruption due to surgery separately. All analysis tests were 2-tailed with alpha set at 5% and performed using Stata 14.2 (Stata Corp 2016. Stata Statistical Software: College Station, Stata Corp LP, United States).

Out of 2828 patients screened, 2180 were included in this analysis. Of these, 174 (8.3%) patients discontinued DAPT, 126 (6.0%) were categorized as physician-guided discontinuation, and 48 (2.4%) as disruption (figure 1). The annual rate of DAPT cessation significantly decreased over the study period (Pfor trend=.035) (Figure 1 of the supplementary data). Table 1 and table 2 summarize patient demographics, clinical and procedural characteristics, and medication use according to DAPT cessation status. Compared with patients on DAPT, those who prematurely discontinued treatment were older, more frequently had prior bleeding, impaired renal function, chronic obstructive pulmonary disease or anemia, and were more likely to have myocardial infarction with nonobstructive coronary arteries and receive medical treatment only for the index ACS. In contrast, they were less likely to receive drug-eluting stents and guideline-directed medical therapies at hospital discharge, including potent P2Y12-i. There were no between-group differences in clinical presentation upon admission, the prevalence of multivessel disease, or completeness of revascularization, nor were there major differences when we analyzed modes of cessation (tables 1-2 of the supplementary data).

Figure 1.

Patient flow chart. ACS, acute coronary syndrome; DAPT, dual antiplatelet therapy; ICH, intracranial hemorrhage; OAC, oral anticoagulation.

(0.3MB).

Table 3 depicts the cumulative incidence and timing of DAPT cessation according to the predefined nonadherence categories. Out of 174 episodes of unplanned cessation, most (88.5%) were categorized as permanent, whereas temporary interruptions occurred in a minority for a median of 5.5 [5-9.5] days. Most discontinuations (90%) were for P2Y12-i only, while a minority included both aspirin and P2Y12-i. In the overall cohort, DAPT cessation was most commonly driven by physician decision (figure 2). The median time until any DAPT cessation occurrence was 204 [96-321] days. Compared with physician-guided discontinuation, DAPT disruption, in particular that due to nonadherence, resulted in shorter courses of treatment (table 3 of the supplementary data). The cumulative incidence of cessation for clopidogrel doubled that of ticagrelor, mainly driven by more frequent physician-guided discontinuation (table 3, figure 2 of the supplementary data). By contrast, disruption rates were proportionally higher in ticagrelor-treated patients than in those with clopidogrel, even though, differences were only evident within 90 days after discharge (table 4 of the supplementary data). Further considering the reasons for cessation, differences between ticagrelor- and clopidogrel-treated patients were largely driven by the higher observed rates of noncompliant disruption in the former (figure 3 of the supplementary data). Overall, the duration of DAPT did not differ significantly by type of P2Y12-I (table 3), or the mode and timing of cessation (table 4 of the supplementary data). However, the median time to DAPT disruption within the first 90 days after discharge was significantly shorter with ticagrelor than with clopidogrel (P=.035) (figure 4 of the supplementary data).

Figure 2.

Relative rates of reasons for dual antiplatelet therapy cessation. DAPT, dual antiplatelet therapy; HBR, high bleeding risk; LIR, low ischemic risk.

(0.23MB).

There was a notable overlap between most of the predictors of DAPT cessation and high bleeding risk factors, whereas neither the occurrence of dyspnea nor PY12-i class were predictors of drug cessation (table 4).

At 1 year, 218 (10%) patients experienced a MACE, of whom 188 (86%) were on DAPT, and 30 (14%) had discontinued treatment before experiencing the event (table 5). Specifically, cardiac events clustered within the first 90 days after discharge (n=106, 48.5%) and declined thereafter. In multivariate models accounting for fixed and time-varying confounders, compared with adherent patients, premature cessation of DAPT was associated with an increased risk of MACE. Differences were largely driven by disruption, whereas physician-guided discontinuation was not associated with a greater risk of MACE (table 5, figure 3). The results remained consistent in direction and magnitude for individual components of the primary endpoint, and with the use of a more restrictive definition of MACE.

Figure 3.

Central illustration. DAPT cessation in ACS patients treated with ticagrelor vs clopidogrel. 95%CI, 95% confidence interval; aHR, adjusted hazard ratio; DAPT: dual antiplatelet therapy; IPCW, inverse probability of censoring weighting; MACE, major adverse cardiac events; MI, myocardial infarction; uTLR, urgent target lesion revascularization; ST, stent thrombosis.

(0.76MB).

In subgroup analysis, the adjusted risk of MACE after DAPT cessation was significantly higher with ticagrelor than with clopidogrel regardless of cessation mode (Pinteraction<.001) (figure 5 of the supplementary data), particularly during the early postdischarge period (figure 6 of the supplementary data). However, when jointly considering the mode and timing of cessation, this interaction was only evident for DAPT disruption within the first 90 days after discharge (Pinteraction<.001) (figure 7 of the supplementary data). There was a stepwise increase in MACE risk according to the underlying reasons for cessation, with the highest risk noted after noncompliant disruption, particularly within the first 30 days after discharge (figure 8 of the supplementary data, table 5 of the supplementary data). In spline analysis, MACE risk after cessation increased exponentially as DAPT duration shortened, to gradually decrease with prolonged courses of treatment irrespective of cessation mode and the choice of P2Y12-i (figure 4); there were no signals for a time-varying interaction between DAPT cessation and P2Y12-i therapy on MACE risk at any DAPT duration strata (table 5 of the supplementary data, figure 9 of the supplementary data). Likewise, RERI analysis revealed that, compared with clopidogrel, while ticagrelor disruption contributed to a nonsignificant 12% increase in MACE risk, the shorter DAPT regimens resulting from early disruption contributed to a significant increase of 64% in MACE risk associated with cessation (table 6 of the supplementary data). Last, the results of a sensitivity analysis according to PARIS criteria (table 7 of the supplementary data), and in selected datasets (table 8 of the supplementary data) remained consistent with the overall study results.

Figure 4.

Duration-response curves depict the aHRs (solid lines in red), and 95%CIs (grey shaded areas) for MACE as a continuous function of DAPT duration. In each panel, the dashed and solid vertical lines plotted over the x-axis denote the duration of continuous DAPT exposure at which the smoothed curves for the aHR and corresponding 95%CI cross the horizontal dashed line of “null effect” (aHR=1). 95%CI, 95% confidence interval; aHR, adjusted hazard ratio; DAPT: dual antiplatelet therapy; IPCW, inverse probability of censoring weighting; MACE, major adverse cardiac events; MI, myocardial infarction; uTLR, urgent target lesion revascularization; ST, stent thrombosis.

(0.51MB).

Some limitations should be also acknowledged. First, relationships between cessation patterns and outcomes cannot be inferred, as we cannot definitely exclude the presence of unmeasured/hidden confounders and potential detection bias. Nonetheless, the results of this study are strengthened by the high level of data granularity collected, which is difficult to achieve in practice in the overwhelming majority of studies, and by the fact that data were prospectively collected independently of treatment and outcomes. Second, nonadherence status was self-reported from patients, which may lead to underreporting because of recall bias. However, to minimize this type of bias, our registry required adjudication of all cessation events by use of any available source documentation. Third, we lacked information on socioeconomic status and medication copayment, which may contribute to medication nonadherence. Nevertheless, this limitation did not seem to significantly affect the study results, given that only a small proportion of patients reported noncompliant disruption because of affordability issues. Last, sensitivity and subgroups analyses were not powered to demonstrate differences across groups and should therefore be considered hypothesis-generating.