Care is required with cost effectiveness approach

Editor—The argument in the letter from Taylor and Ebrahim starts from the premise that the cost of providing an anticoagulation service for patients with atrial fibrillation is 15 times higher than that of treating with aspirin.¹ The inference that anticoagulation must be shown to be 15 times as effective as aspirin in the prevention of stroke is not correct because it ignores the costs of treating patients who develop strokes.

The point estimate and confidence intervals of the treatment effect shown by Hellemons et al² could be combined with the estimated costs of both preventing and treating strokes. In this way a model of cost effectiveness related to different baseline risks could be constructed. This, however, is rather different from the way that Taylor and Ebrahim suggest using the cost ratio. It is indeed unlikely from the trial data that anticoagulation is 15 times as effective as aspirin, but this does not mean that anticoagulation must prevent 15 times as many strokes to be cost effective.

Since the overall cost of treating patients with stroke is likely to be higher than the cost of drugs for preventing stroke, a relatively small benefit from anticoagulation might lead to an overall cost benefit in comparison with aspirin. This will be increasingly likely as the baseline risk of stroke increases.

As authors of other letters in the cluster have noted, the confidence intervals of the relative risk found by Hellemons et al are not tight enough to rule out a moderate beneficial effect of anticoagulation in comparison with aspirin.³ Although cost should be considered in choosing between aspirin and anticoagulation for patients with atrial fibrillation (alongside individual patients' risks and preferences for treatment), it should not be used to demand that the statistical difference between treatments matches the differential costs of one part of the service provision.