Comparative Effectiveness of Aspirin Dosing in Cardiovascular Disease

Trial Design

The trial design and processes have been described previously.14 In brief, the trial was conducted in 40 centers and in one health plan participating in PCORnet (Table S1 in the Supplementary Appendix, available with the full text of this article at Patients with established atherosclerotic cardiovascular disease were identified with the use of electronic health record data at each institution through a cohort identification query (termed “computable phenotype”).18.19 The trial protocol and statistical analysis plan are available at The protocol and all patient-facing materials were designed with a group of nine patient-partners (who called themselves “Adaptors”), and patient-engagement activities were coordinated by the Health eHeart Alliance (San Francisco). An independent data and safety monitoring committee also approved the trial protocol and monitored patient safety throughout the trial. All the patients provided electronic informed consent before enrollment.

Trial Population and Recruitment Strategies

Eligible patients had established atherosclerotic cardiovascular disease as defined by the inclusion and exclusion criteria in the Supplementary Appendix. Details of the recruitment strategies have been reported previously and are described in the Supplementary Appendix.14

Baseline demographic characteristics were reported by the patients, including age, sex, race, ethnic group, current tobacco use, and medication use before randomization. Baseline clinical characteristics and medical history were retrieved by means of a trial-specific query of the electronic health record (with the use of the PCORnet Common Data Model format) at enrolling health centers. All baseline medical history was recorded with a look-back period of 5 years from the date of enrollment.

Randomization and Trial Treatment

After providing informed consent, patients were randomly assigned through the patient portal in a 1: 1 ratio to take 81 mg or 325 mg of daily aspirin, and they purchased the assigned dose over the counter. Patients received $ 25 remuneration for trial participation. Patients were also randomly assigned in a 1: 1 ratio to follow-up visits every 3 months or every 6 months to better understand the effect of the frequency of clinical trial assessments on patient engagement and follow-up. There were no in-person visits at the trial centers during follow-up. Patients were asked to return to the patient portal for an Early Study Encounter (conducted through telephone contact by the call center for non-Internet participants) between 1 and 3 weeks after randomization to confirm adherence to the appropriate dosage and answer questions about secondary contact information .

Routine follow-up encounters then occurred every 3 or 6 months, depending on the randomized group. Internet participants were sent email reminders to complete trial visits, and non-Internet participants received telephone calls from the call center. At each encounter, patients were asked about adherence to the trial medication, the use of concomitant medications, recent hospitalizations (and primary diagnoses of hospitalizations), and patient-reported outcomes. Internet participants who had not completed a trial follow-up encounter for 6 months were converted to non-Internet participation and contacted by the call center in order to complete follow-up. When patients missed an encounter and returned for a subsequent encounter, they were asked to complete information on hospitalizations (trial outcomes) that had occurred since the last complete encounter.

Clinical Outcomes

The primary effectiveness outcome was the time to a first occurrence of any event in the composite of death from any cause, hospitalization for myocardial infarction, or hospitalization for stroke. Prespecified secondary outcomes included coronary revascularization (percutaneous coronary intervention or coronary-artery bypass grafting), the individual components of the primary outcome, and hospitalization for transient ischemic attack. The primary safety outcome was hospitalization for major bleeding with an associated blood-product transfusion. The PCORI Patient-Reported Outcomes Common Measures short form was administered through the patient portal (or call center) at baseline and every 6 months.

Data Sources

Outcomes were ascertained from multiple data sources, including patient report at scheduled trial encounters, queries of electronic health record data organized according to the PCORnet Common Data Model format, linkage with data sources from PCORnet private health plan partners (Aetna, Anthem, and Humana) , and linkage with fee-for-service Centers for Medicare and Medicaid Services claims data (Fig. S1). Details on the programming algorithms for outcomes, ascertainment of death, outcome validation plan, outcome reconciliation and confirmation, and censoring rules are described in the Supplementary Appendix.

Statistical Analysis

The planned sample size was 20,000 patients. In 2017, the sample size was reevaluated by the coordinating center (with PCORI oversight) with the use of data on trial recruitment rates and blinded, aggregate event rates, and the trial size was reduced to 15,000 patients. These changes are described in the Supplementary Appendix. Comparisons based on randomized treatment assignments were performed according to the intention-to-treat principle with the use of time-to-first-event analyzes. Descriptive summaries of baseline demographic and clinical variables were generated for each randomized treatment group. Continuous baseline variables were presented as medians with interquartile ranges, and discrete variables were summarized with the use of frequencies and percentages.

Cumulative event rates (estimated percentages) were estimated at median follow-up in each treatment group with the use of Kalbfleisch and Prentice’s nonparametric estimator of the cumulative incidence function. Event-free survival rates for the primary effectiveness outcome and death from any cause were compared with the use of Cox proportional-hazards models; censoring rules are described in the Supplementary Appendix. The secondary effectiveness outcomes were compared with the use of the Fine and Gray method to take into account the competing risk of death from any cause. The proportional-hazards assumption was checked for the randomized treatment assignment with the use of weighted Schoenfeld residuals. Hazard ratios with 95% confidence intervals are presented comparing the 81-mg group with the 325-mg group, such that hazard ratios greater than 1 indicate a higher risk of events in the 81-mg group. There was no prespecified plan to adjust for multiple comparisons. Results of analyzes of secondary outcomes and subgroup analyzes are reported with 95% confidence intervals, without P values. The confidence intervals have not been adjusted for multiple comparisons and should not be used to infer definitive treatment effects.

Event-free survival rates for the primary safety outcome were compared with the use of the Fine and Gray method to account for the competing risk of death from any cause; the same censoring rules that were used for the analyzes of the primary effectiveness outcome and death from any cause were applied. The test was two-tailed and was performed at an overall alpha level of 0.05.

Sensitivity analyzes were performed to assess robustness of primary results to potential misclassification of events based on electronic phenotypes and underreporting of events that occurred outside PCORnet health systems. A landmark analysis was conducted to allow for washout of previous aspirin use. A modified per-protocol analysis was also performed with reported aspirin dose modeled as a time-dependent covariate. (See the Supplementary Appendix for details.) All analyzes were performed with the use of SAS software, version 9.4 (SAS Institute).

Leave a Reply

Your email address will not be published.