A Grey Area: warfarin for atrial fibrillation in dialysis patients

I recently wrote a short post detailing safety concerns surrounding warfarin use for atrial fibrillation (AF) in dialysis patients. I focussed mostly on the emerging evidence that it may promote vascular and valvular calcification by the inhibition of vitamin-K dependent gamma-carboxylation, as well as recent observational data linking warfarin use to higher mortality rates in ESRD. I finished the piece with this line: "adopting a restrictive policy towards warfarin use in ESRD seems wise, such as reserving it for patients with a CHADS score greater than 2". The use of the CHADS score to stratify dialysis patients with AF has been addressed in a study in this month's Kidney International, and the results suggest that even this restrictive approach may not be safe.
Briefly, CHADS2 is a risk score that stratifies members of the general population with AF by stroke risk, in order to permit informed clinical decision-making regarding the risks/benefits of warfarin anticoagulation. Its name is an acronym derived from the component independent stroke risk factors, namely Congestive heart failure, Hypertension, Age over 75 years, Diabetic status, and history of Stroke (or TIA). This study highlights several caveats regarding the use of CHADS2 (and warfarin use in general) in the dialysis population, information you’ll need next time you’re thrashing out this thorny issue with your Cardiology colleagues.
The authors examined 17,500 dialysis patients from the international DOPPS study, among whom the prevalence of AF was 12% and the incidence of new-onset AF was 1 case per 100 patient years.

1. CHADS2 divides this representative sample of dialysis patients 50:50 into high ( greater than 4 per 100 pt-yrs) and low (less than 2) risk for stroke. 
2. In the low-risk group, the risk of stroke equals the rate below which anticoagulation is not recommended for AF in the general population and, as such, would support a decision not to anticoagulate a dialysis patient with AF and a low CHADS2 score. 
3. Two key components of the CHADS2 score, HTN and heart failure, did not independently predict stroke in this study. As a result, CHADS2 may overcall stroke risk in some low-risk HD patients, and caution should be exercised when calling a patient high risk based on these 2 conditions. A dialysis-specific CHADS score that drops hypertension and heart failure would likely improve its discriminatory power, but requires further study. 
4. Warfarin anticoagulation did not reduce stroke risk in this study. In fact, there was a strong tendency to more frequent cerebrovascular events, particularly in the elderly. Whether this surprising finding was caused by a direct effect of warfarin on the rate of hemorrhagic stroke, or that patients who were prescribed warfarin were already an inherently high-risk group, the fact remains that one needs to be very cautious when prescribing anticoagulation to these patients. It would appear that in some patients identifyed as being at higher stroke risk (particularly older patients), the risks of anticoagulation still outweigh the benefits. 

The upshot of this study is that CHADS2 appears to function adequately in identifying patients at low-risk for stroke, and provides support to a clinical decision to withhold warfarin in that context. However, its ability to identify high-risk patients suitable for anticoagulation is highly questionable, as 3 of its 5 component factors (HTN, heart failure and possibly age) perform poorly in the dialysis population. How to approach the issue of anticoagulation in such patients remains a gray area.