Background-Limited research has compared the measures of summarizing international normalized ratio (INR) control over time. Measures that are more predictive of patient outcomes would be preferred as would those that are easier to calculate and understand. Methods and Results-We examined 676 patients who received long-term warfarin therapy to treat atrial fibrillation: 125 patients who experienced major hemorrhage and 551 matched controls who did not. Patient INR control was characterized using various measures, from simple (proportion of INR values in range) to complex (eg, area under the curve above target range, squared) measures. Conditional logistic regression was used to examine the ability of each measure to predict the outcome of clinically relevant bleeding across quintiles of control. All measures were associated with clinically relevant bleeding to some extent: patients with the poorest control had significantly more bleeding events compared with patients with the best control. The measure most strongly associated with bleeding was a combination of percent time in therapeutic range and INR variability (odds ratio of 4.34, comparing the lowest to the highest quintiles of control). The strongest single predictor was INR variability, followed closely by time in therapeutic range. More computationally complex measures, which had been expected to perform better, were not so strongly associated with bleeding. Conclusions-INR variability was the most strongly associated predictor of clinically relevant bleeding followed closely by time in therapeutic range. Using both measures together had an even stronger association. These findings support continued use of INR variability, time in therapeutic range, or both for research and quality assurance efforts.
Bibliographical noteFunding Information:
This study was funded by US Department of Veterans Affairs HSR&D Investigator-Initiated Research 10-374 (principal investigator [PI]: A.J. Rose) and the Kaiser Permanente Colorado Pharmacy Department (PI: T. Delate).
© 2015 American Heart Association, Inc.
- case-control studies
- quality of health care
- research design