299 – Meta-Analysis and Other Innovative Comparison
A Statistical Methodology for Adjustment of Stroke Event Rates for CHADS2 Score for Indirect Comparison Across Trials - a Case Study
Vijay Chauhan
Alpha Stats Inc.
Stroke event rates in patients with atrial fibrillation vary according to patients' baseline characteristics reflected by a CHADS2 score based on prior heart failure, hypertension, diabetes mellitus, prior stroke, and age 75 year or older. Adjustment for CHADS2 score would help in indirect comparison of drugs across studies. We have illustrated a methodology to adjust stroke event rates for CHADS2 score, using published results of 3 new anticoagulants trials (dabigatran, apixaban, rivaroxaban) where warfarin was the common comparator. Mean CHADS2 score was ~2.1 in both dabigatran and apixaban trials, and 3.5 in rivaroxaban trial. Adjustment factors were derived using regression of natural log of event rates on CHADS2 scores from warfarin treatment group only. Estimates were transformed back to the original scale. Event rates adjusted to a CHADS2 score of 3.0 for dabigatran, apixaban and rivaroxaban all compared with warfarin were 1.35 vs 1.93, 1.46 vs 1.87, and 1.36 vs 1.82, respectively. Whereas, the originally reported event rates were 1.11 vs 1.69, 1.27 vs 1.60, and 1.70 vs 2.20, respectively. Adjustment to a CHADS2 score of 3.0 gave the minimum mean square error. The increase or decrease in adjusted event rates was relative to the baseline CHADS2 score. Adjusted event rates were meaningful and realistic.
