Background: We describe innovations in the study design and the efficient data coordination of a randomized multicenter trial of Argatroban in Combination with Recombinant Tissue Plasminogen Activator for Acute Stroke (ARTSS-2).
Methods: ARTSS-2 is a 3-arm, multisite/multiregional randomized controlled trials (RCTs) of two doses of Argatroban injection (low, high) in combination with recombinant tissue plasminogen activator (rt-PA) in acute ischemic stroke patients and rt-PA alone. We developed a covariate adaptive randomization program that balanced the study arms with respect to study site as well as hemorrhage after thrombolysis (HAT) score and presence of distal internal carotid artery occlusion (DICAO). We used simulation studies to validate performance of the randomization program before making any adaptations during the trial. For the first 90 patients enrolled in ARTSS-2, we evaluated performance of our randomization program using chi-square tests of homogeneity or extended Fisher's exact test. We also designed a four-step partly Bayesian safety stopping rule for low and high dose Argatroban arms.
Results: Homogeneity of the study arms was confirmed with respect to distribution of study site (UK sites vs. US sites, P=0.98), HAT score (0-2 vs. 3-5, P=1.0), and DICAO (N/A vs. No vs. Yes, P=0.97). Our stopping thresholds for safety of low and high dose Argatroban were not crossed. Despite challenges, data quality was assured.
Conclusions: We recommend adaptive designs for randomization and Bayesian safety stopping rules for multisite Phase I/II RCTs for maintaining additional flexibility. Efficient data coordination could lead to improved data quality.
Keywords: Acute Stroke; Adaptive design; Argatroban; Bayesian methods; Data quality assurance; Multisite randomized clinical trials.
Copyright © 2015. Published by Elsevier Inc.