Adherence to study drug in a stroke prevention trial"?>

Objective: Standards for reporting and analyzing adherence to medical therapy have recently improved due to international consensus efforts. If applied to clinical trial research in patients with stroke, these improvements have the potential to identify when in the sequence of trial operations participants are at risk for non-adherence and opportunities to safeguard adherence.

Methods: We analyzed three phases of adherence according to the European Society for Patient Adherence, COMpliance, and Persistence (ESPACOMP) Medication Adherence Reporting Guideline (EMERGE) taxonomy in the Insulin Resistance Intervention after Stroke (IRIS) trial: initiation (did patient start drug), implementation (did patient take a drug holiday, defined as temporary cessation lasting ≥14 days), and persistence (did patient prematurely and permanently discontinue drug). IRIS was a randomized, placebo controlled, double-blind trial testing pioglitazone to prevent stroke or myocardial infarction in patients with a recent ischemic stroke or transient ischemic attack. Adherence was classified by self-report. Researchers used coaching algorithms to seek adherence recovery if participants went off drug.

Results: During 2005-2013, 3876 participants were enrolled from 179 sites in seven countries and followed for a mean of 4.8 years. Less than 1% of participants in each group did not initiate study drug. 20% of patients assigned to pioglitazone and 17% assigned to placebo took at least one drug holiday. 36% and 30%, respectively, discontinued the study drug prematurely with or without a prior holiday. The risk for stopping the study drug (temporarily or permanently) in the first year after randomization was twice the risk in each of the subsequent four years. This was true both for patients assigned to active therapy and placebo. More participants assigned to pioglitazone, compared to placebo, took a drug holiday or permanently stopped study drug, but the difference in rates of discontinuation was only evident in year one. In years two through five, rates of discontinuation were similar in the two treatment groups. The difference in rates during year one was the result of adverse effects related to the active study drug, pioglitazone. During the remainder of the trial, the attribution of discontinuations to adverse effects potentially related to pioglitazone was reduced but still higher in those assigned to active drug. Other reasons for discontinuation were similar between treatment groups and were largely unrelated to pharmacodynamic effects of the study drug. Rates of discontinuation varied widely among research sites.

Conclusion: Patients in a drug trial for stroke prevention are at greatest risk for premature drug discontinuation early after randomization. Reasons for discontinuation change over time. Variable discontinuation rates among sites suggests that adherence can be improved by using best practices from high-performing sites.