Stroke in Patients With Peripheral Artery Disease

Insights From the EUCLID Study

Brad J. Kolls, MD, PhD, MMCi; Shelly Sapp, MS; Frank W. Rockhold, PhD; J. Dedrick Jordan, MD, PhD; Keith E. Dombrowski, MD; F. Gerry R. Fowkes, MB ChB, PhD; Kenneth W. Mahaffey, MD; Jeffrey S. Berger, MD; Brian G. Katona, PharmD; Juuso I. Blomster, MD, PhD; Lars Norgren, MD, PhD; Beth L. Abramson, MD; Jose L. Leiva-Pons, MD; Juan Carlos Prieto, MD; German Sokurenko, MD; William R. Hiatt, MD; W. Schuyler Jones, MD; Manesh R. Patel, MD


Stroke. 2019;50(6):1356-1363. 

In This Article


The data, methods used in the analysis, and materials used to conduct the research will not be made available for purposes of reproducing the results or replicating the procedure.

Trial Design

EUCLID was an international, double-blind, randomized controlled trial that included 13 885 patients with symptomatic PAD randomized to receive antiplatelet monotherapy with ticagrelor or clopidogrel. The design of the study and the primary results have been published previously.[11–13] Patients were followed closely at 3-month intervals using in-person and telephone visits. Median follow-up was 30 months.

Study Population

Eligible patients were required to be ≥50 years of age with lower extremity PAD as defined by (1) an abnormal ankle-brachial index (ABI) ≤0.80 at screening, or (2) prior revascularization of the lower extremity >30 days before randomization. Key exclusion criteria included planned use of dual antiplatelet therapy or the use of aspirin, treatment with anticoagulation, poor metabolizer status for CYP2C19, planned revascularization (any territory), or major amputation within 3 months. All patients provided written informed consent, and institutional review boards at participating institutions approved the protocol.

Study end Points

The primary efficacy end point for the EUCLID trial was a composite of cardiovascular death, MI, or ischemic stroke. The primary safety end point was major bleeding. Key secondary end points included analysis of the components of the primary end point including the occurrence of ischemic stroke, hemorrhagic stroke, and TIA. End points were site-reported in an electronic web-based capture system with submission of supporting source documentation where applicable.

All suspected stroke and TIA events were systematically adjudicated by 2 trained neurologists as part of the clinical events classification process. Stroke was defined as a focal neurological deficit that could be attributed to a vascular territory and lasted >24 hours or was associated with a new lesion on computed tomography scan or magnetic resonance imaging. If the reviewers disagreed about an event, the event was then rereviewed by the clinical events classification committee (including at least 1 neurologist) to obtain a final classification.

Statistical Methods

Baseline characteristics for patients who had stroke events after randomization and those without stroke events are summarized using descriptive statistics. Categorical data are presented as frequencies and percentages and continuous data presented as the medians with 25th and 75th percentiles. Event rates are reported as events per 100 patient-years. Risk difference, defined as the ticagrelor event percentage minus the clopidogrel event percentage, and the corresponding 95% CI of the risk difference are also provided.

There is a competing risk of death when performing nonfatal event analyses during clinical trials (eg, death without the occurrence of stroke is a competing risk for having a future stroke during study conduct). Therefore, competing risk survival analyses were performed, where the stroke end points were the events of interest and deaths without prior strokes were specified as competing risks. Competing risk models based on Fine and Gray subdistribution hazards models evaluating the association of baseline characteristics with adjudicated all-cause stroke were used to develop a multivariable model.[14–16] Univariate competing risk models were first used to assess the association of each baseline characteristic with all-cause stroke. Given the number of baseline characteristics evaluated, only characteristics univariately significant at a 0.1 level were considered as potential covariates in the multivariable model. The final multivariable model of all-cause stroke was developed by evaluating the potential baseline covariates in a stepwise fashion using a significance level of 0.05. After the significant baseline characteristics were selected for the multivariable model, treatment group was included in the final model. Linearity assumptions for the continuous baseline characteristics were examined using spline functions, and there was no evidence indicating violations of this assumption. The proportional hazards assumption was evaluated using Schoenfeld residuals; variables in the final model met this assumption. Competing risk models were also used to assess the association of treatment with all-cause stroke and TIA end points. A multivariable competing risk model of ischemic stroke including the same baseline characteristics in the final multivariable model of all-cause stroke was generated to evaluate the adjusted treatment effect. Multivariable models of hemorrhagic stroke and TIA were not produced due to the small number of events. In addition to the competing risk analyses, standard Cox proportional hazards models not taking into account a competing risk were generated, and the standard results were similar to the competing risk results. All reported hazard ratios (HRs; 95% CIs) and P values based on the Wald statistic are from competing risk models. Plots of cumulative incidence functions from competing risk analyses of all-cause stroke and ischemic stroke are provided. All statistical analyses were performed using SAS software, version 9.4 (SAS Institute, Inc, Cary, NC).