Classification of outcome events in the Dutch TIA trial: prognostic value of accepted and rejected events

Abstract

Objective: In large multicentre clinical trials adjudication of outcome events most often is done centrally. Some of these events would eventually be judged not to meet the criteria and hence would be rejected. If the classification procedures work correctly one would expect a higher risk of future vascular events after an 'accepted' rather than after a 'rejected event. In the present study we aimed at testing the adequacy of a classification procedure in a trial of patients with a transient ischemic attack (TIA) or minor ischemic stroke by comparing the further outcome between patients in whom a possible event was either rejected or accepted for the final analysis in this trial.

Research design and methods: The vascular outcome events were analysed in 3150 patients with TIA or minor stroke who participated in the Dutch TIA trial. We identified the patients with a first 'accepted' or 'rejected' non-fatal stroke or myocardial infarction (MI). In these two groups of patients we determined the occurrence of the subsequent vascular events (vascular death, stroke or MI). The incidence was compared with survival analysis techniques.

Results: Among 308 patients with a first nonfatal 'accepted' event in 81 (26.3%) a new vascular event occurred; among the 51 patients with a 'rejected' event there were 12 (23.5%) such events. The hazard ratio for new vascular events was 1.22 (95% CI 0.67-2.22). After multivariate adjustment for age, type of qualifying event, history of smoking, angina pectoris and myocardial infarction, and Rankin score > 3, the hazard ratio was 1.49 (95% CI 0.78–2.84).

Conclusion: our study suggested that the adjudication process of outcome events in the Dutch TIA trial was done correctly because a trend towards a higher recurrence rate of vascular events among patients with 'accepted' outcome events was found.

Key Words::

• Classification
• Clinical trial
• Errors
• Measurement
• Outcome