to the editor: Drs. Wolf and Singer¹ review clinical trials demonstrating the benefit of warfarin (Coumadin) anticoagulation therapy in patients with nonvalvular atrial fibrillation. I would like to comment on the generalizability of these trials to typical patients seen in primary care practice.²

First, the six clinical trials referenced in the article did not randomize a substantial number of very old patients. The mean age of the patients varied from 67 to 73 years, with an overall mean of 69 years. Probably less than 100 patients over 80 years of age were randomized to receive warfarin treatment. Therefore, based on these clinical trials, the efficacy of warfarin therapy in the very old population is speculative.

The best trial available that addresses the very old population is the Stroke Prevention in Atrial Fibrillation (SPAF-II) trial,³ which was not reviewed by Drs. Wolf and Singer. Nearly 200 patients with a mean age of 80 years were randomized to receive warfarin therapy; a similar number of patients were randomized to receive aspirin therapy. The results were largely negative, as there were no significant differences between aspirin and warfarin in any measured outcome (stroke, transient ischemic attack, mortality). This trial admitted patients who were considerably older than those in any of the previous studies, and the results suggest that warfarin simply may not be as effective in preventing stroke in patients over 80 years of age.

The safety of warfarin anticoagulation therapy in highly selected patients was demonstrated in the first six trials. However, in the SPAF-II trial, the rate of serious bleeding was higher in very old patients taking warfarin; the rate of intracerebral hemorrhage was 1.8 percent per year, with a mortality rate of 71 percent. The absolute reduction of 1.2 percent in the annual rate of thromboembolic stroke was overshadowed by serious bleeding side effects.

My second concern also has to do with generalizability of findings, although this concern is unrelated to patient age. In the six clinical trials reported by Drs. Wolf and Singer, over 90 percent of patients with nonvalvular atrial fibrillation were excluded from participating. “Patients enrolled in these trials were carefully screened for the slightest characteristics that might place them at any risk of bleeding and were more likely to be more compliant than the average.”⁴ It is difficult to know how much one can generalize from results documented in less than 10 percent of the population with nonvalvular atrial fibrillation. Patients may have clear contraindications to warfarin therapy. Many other patients with co-morbidities, on multiple medications, prone to falls, etc., present unique and difficult decisions.

Multiple randomized trials have demonstrated that highly selected patients under 75 or 80 years of age who have nonvalvular atrial fibrillation may benefit from warfarin anticoagulation therapy, at least in the short term. Long-term data are not available for any age group. A recent meta-analysis found the margin between benefit and harm to be very thin.⁵ No clinical trial evidence indicates that frail patients or those over 80 years of age will benefit more from warfarin therapy than from treatment with aspirin. Although I agree with Drs. Wolf and Singer that it is important to identify patients who may benefit from warfarin treatment, I believe the decision to initiate anticoagulation therapy should be made after the physician has taken into consideration the age of the patient and has weighed heavily an individual's overall vigor and co-morbid conditions.

in reply: In patients with atrial fibrillation, anticoagulation therapy reduces the risk of ischemic stroke by two thirds. This large effect has been demonstrated by a remarkable, consistent set of randomized trials.^1,2 In nearly all of the trials, the risk of major hemorrhage attributable to anticoagulation therapy was small. In particular, the risk of intracranial hemorrhage was less than 0.5 percent per year. Economic models using these results demonstrate very favorable estimates of cost-effectiveness for long-term anticoagulation therapy.³

Nonetheless, anticoagulation therapy can be complicated and is risky. While there is an abundance of favorable evidence from randomized trials, few studies have evaluated the use of anticoagulation therapy for atrial fibrillation in usual clinical practice. This is not unique for atrial fibrillation but is typical of the evidence bearing on many therapeutic interventions. It is likely that the safety profile observed in the trials will not be reproduced if anticoagulation therapy is applied indiscriminately to patients with atrial fibrillation. However, it is simply wrong to assert that patients entered into the clinical trials could not have had the “slightest . . . risk of bleeding,” even if one could define such elusive criteria. Patients in the trials were representative of a large proportion of patients with atrial fibrillation. Probably more important to preserving safety, the trials managed anticoagulation therapy in an organized manner. The success of anticoagulation depends on maintaining the International Normalized Ratio (INR) in the appropriate range, which is 2.0 to 3.0 for atrial fibrillation.⁴ To reproduce the results of the randomized trials, physicians have to provide high-quality management of anticoagulation therapy.

The decision to initiate anticoagulation therapy is particularly difficult when the patient is elderly. The risk of intracranial hemorrhage appears to increase with age, but the risk of ischemic stroke in atrial fibrillation without anticoagulation therapy increases even faster (in terms of absolute risk).^1,5 Dr. Ackermann cites the SPAF II study as evidence against the net benefit of anticoagulation therapy in elderly patients with atrial fibrillation. Problems in the design and conduct of SPAF II make it difficult to interpret the comparison of warfarin versus aspirin.⁶ The finding in SPAF II of a nearly 2 percent per year risk of intracranial hemorrhage among anticoagulated patients over 75 years of age is worrisome. However, this risk was based on only seven events, some of which were the result of the high INR target used in SPAF II (ratio of 2.0 to 4.5).^7,8 The pooled analysis of the first five trials accumulated as many person-years among patients over 75 years of age as were accumulated in SPAF II and observed only a 0.3 percent per year risk of intracranial hemorrhage. The benefit, in terms of reduced risk of ischemic stroke, was more than 10 times larger. More recently, the SPAF III randomized trial demonstrated that anticoagulation targeted at an INR of 2.0 to 3.0 was far superior to the use of aspirin (as well as very-low-dose warfarin) among high-risk, often elderly patients with atrial fibrillation.⁹ There was as well no increase in rates of major bleeding in those randomized to a target INR of 2.0 to 3.0.

Future research should help identify more specifically those patients with atrial fibrillation who definitely need anticoagulation therapy and those who can safely forgo such treatment. Until then, patients with atrial fibrillation who are over 65 years of age, as well as younger patients who have risk factors, are strong candidates for anticoagulation therapy. Clearly, anticoagulation therapy is less attractive for very frail or cognitively-impaired patients. However, a high level of vigor is not a prerequisite for anticoagulation therapy. Most elderly persons will benefit from avoiding a stroke. The main issue is whether compliance with warfarin dosing and frequent INR tests will be adequate to ensure safe management of anticoagulation.

editor's note: The following letter was received after Drs. Wolf and Singer had written their reply.

to the editor: I wish to comment on the issue of warfarin anticoagulation in elderly persons with atrial fibrillation.¹

As was ably pointed out in the article, there is little doubt that elderly persons with atrial fibrillation should receive anticoagulation therapy with only unusual exceptions and that maintenance INR values should be modified for age. No other treatment puts quality of life first so clearly. I applaud the authors' effort to draw attention to this important treatment.

However, I would like to point out why some physicians may resist initiating anticoagulation therapy in older patients: they may have done so before and seen frightening complications. Such complications may arise when conventional methods of initiating anticoagulation are used. Many resources, such as Goodman and Gilman's Principles of Pharmacologic Therapeutics⁵ and Drug Facts and Comparisons,⁶ advise initiating warfarin treatment at 5 to 10 mg for two to four days and adjustment thereafter. In a frail elderly patient with a low serum albumin level and plenty of other drugs competing for binding sites there, the final maintenance dose can be as low as 0.5 mg, in my experience. Average maintenance doses for elderly patients are around 3.6 mg.^2,3 Initiating warfarin therapy with 5 to 10 mg for two to four days can lead to practically complete anticoagulation and high risk of hemorrhagic catastrophe. Having once observed or caused such a catastrophe, a physician is unlikely to be convinced to start anticoagulation therapy in the next frail 90-year-old patient he or she sees, no matter what the physician just read in American Family Physician or elsewhere.

At least in this instance, the Physician's Desk Reference⁴ may have better advice: “It is recommended that Coumadin therapy be initiated with a dose of 2 to 5 mg per day with dosage adjustment based on the results of the INR determination.” General admonitions like “start low and go slow” don't stand up in all situations, and some drugs should not be under-dosed in geriatric patients, such as antibiotics and diuretics. However, all too often this practice is ignored in situations in which it is most imperative, such as in warfarin treatment.

In summary, compared with young persons, elderly persons with atrial fibrillation don't need warfarin less often, but they need less warfarin, especially at first.