Meeting Brochure and registration form      SMDM Homepage

Wednesday, October 24, 2007
P4-37

DECISION-MAKING WITH INSUFFICIENT DATA: THE CASE OF NVAF

Liana Fraenkel, MD, MPH, Joanne Foody, MD, and Terri Fried, MD. Yale University; VA CT Healthcare System, New Haven, CT

Background: Patients with NVAF have twice the mortality rate of age and gender matched controls, which is largely explained by an increased risk of stroke. While randomized controlled trials have demonstrated a reduction in risk of stroke with oral anticoagulants and antiplatelet agents, long-term anticoagulation with warfarin does not confer a mortality benefit over aspirin and is associated with a 45% increase in the odds of major bleeding compared with aspirin. The decision whether or not to initiate anticoagulation is therefore preference-sensitive. In order for NVAF patients to be able to consider the trade-offs involved in this decision, individualized risk information is required. Purpose: To review currently available risk information to inform decision-making in NVAF. Methods: We reviewed all the English randomized controlled trials, meta-analyses and decision support papers related to NVAF and the use of aspirin and/or warfarin in the MEDLINE database and searched the Internet for additional decision support tools and risk calculators. Results: Although numerous studies have been published detailing the risk and benefits of anticoagulation for patients with NVAF, we noted the following: 1. Decision tools and risk calculators employ varying risk gradients and time horizons resulting in inconsistent impressions of treatment effects. 2. Decision tools and risk calculators classify the magnitude of stroke risk using arbitrary cut-offs which do not take patient values into account. 3. Studies often describe outcomes in patient-years, which is problematic when communicating risk at the individual patient level. 4. There are insufficient data to enable accurate determination of baseline bleeding risks for patients with variable co-morbidities. Risk data from by patients randomized to placebo groups have limited generalizability, whereas selection bias limits the applicability of rates generated by observational studies. 5. Estimating long-term outcomes are based on assumptions which do not consider likely changes in co-morbidities over time. Conclusions: The data needed to promote informed individual decision-making in patients with NVAF are lacking. Consequently, there is significant variability in the content and risk estimates utilized to inform guidelines, recommendations, and decision support tools developed for both providers and patients. Further data are required in order to support decision making in NVAF. This is especially true for older adults, who will make up more than 50% of the patients with NVAF by the year 2050.