Systematic Review of Cost-Effectiveness Research of Stroke Evaluation and Treatment
To the Editor:
Holloway and colleagues’ review of cost-effectiveness studies in stroke evaluation and treatment1 may have inadvertently introduced major biases by the selection criteria used for inclusion of studies. They decided to include only studies that used quality-adjusted life-years (QALYs) as the indicator of health effect. The justification for this criterion is not given. By doing this, cost-effectiveness studies that used indicators such as lives saved or strokes avoided are excluded, and the authors do not provide information on study exclusions to allow the reader to assess the potential bias created.
The review is biased in two ways. First, the use of QALYs is inappropriate in many areas of stroke evaluation and management where measures of diagnostic accuracy, patient satisfaction, or reduction in symptoms are of relevance. It is noteworthy that the review excluded consideration of the most effective intervention for stroke management—organized stroke care and rehabilitation—for which reviews of cost-effectiveness studies have been performed.2 3 Thus, the review is biased in describing the range of cost-effectiveness studies in stroke.
Second, the review provides biased estimates of cost-effectiveness. To illustrate this bias, consider the use of anticoagulation for patients with nonvalvular atrial fibrillation. The cost-effectiveness studies they present show that warfarin dominates among high- and medium-risk patients but in low-risk patients has a very high cost per QALY. The authors concluded that anticoagulation was the preferred option for all but the low-risk patients.
A cost-effectiveness study comparing anticoagulation only, anticoagulation or aspirin, or aspirin only that reported cost per stroke prevented was excluded but comes to remarkably different conclusions.4 In this study, the cost per stroke prevented (which may arguably be a more relevant outcome than a QALY to most patients and doctors) was substantially lower for the aspirin-only regimen at US$1300 (Table⇓).
The most effective treatment is anticoagulation for those who can tolerate it and aspirin for the remainder, as this prevents 1300 strokes a year if complications are low, and even in complications are high still prevents more strokes than simply giving everyone aspirin—but this approach ignores the higher costs involved in anticoagulation. If complications of anticoagulation are high, which tends to be the case in older patients, the aspirin only policy is the most cost-effective option.
I hope the authors of this review will consider updating it by using more appropriate inclusion criteria and thereby arriving at more relevant decisions to aid clinicians and policy makers.
- Copyright © 1999 by American Heart Association
Holloway RG, Benesch CG, Rahilly CR, Courtright CE. A systematic review of cost-effectiveness research of stroke evaluation and treatment. Stroke.. 1999;30:1340–1349.
Keith RA. Rehabilitation after stroke: cost-effectiveness analyses. J R Soc Med.. 1996;89:631–633.
Gustafsson C, Asplund K, Britton M, Norrving B, Olsson B, Marke LA. Cost effectiveness of primary stroke prevention in atrial fibrillation: Swedish national perspective. BMJ.. 1992;305:1457–1460.
We thank Dr Ebrahim for taking an interest in our article. He states that our analysis was biased because we excluded studies that did not report quality-adjusted life-years as the health effect. We excluded such studies because we wanted a measure of cost-effectiveness that we could compare across health care interventions: the cost-per-QALY metric allows for such comparisons. Including studies using different outcomes (eg, cost per stroke avoided) would limit our ability to make such comparisons. We agree with Dr Ebrahim that measures of diagnostic accuracy, patient satisfaction, and reduction of symptoms are relevant in evaluating the effectiveness of diagnostic and management strategies of patients with stroke. We also believe that the QALY effectively incorporates those attributes into a single health measure of both quality and quantity. Cost-effectiveness studies that use measures other than QALYs, however, are important and should be the subject of future systematic reviews. We consider our project an important step in improving the evidence base upon which cost-effectiveness data can be used by clinicians and policy makers.
Dr Ebrahim also asserts that we provided biased estimates of cost-effectiveness and refers to one of our cited studies.R1 This study addressed the cost-effectiveness of warfarin compared with aspirin or no therapy in patients with nonvalvular atrial fibrillation. He states that we “conclude that anticoagulation was the preferred option for all but low-risk patients.” In our article, we presented only the data on the cost-effectiveness of warfarin compared with aspirin, since this was a primary objective of the referenced article. By not presenting the aspirin data, we never meant to imply that aspirin was not cost-effective in some patients. In fact, Gage and colleaguesR1 found that aspirin was the preferred therapy in low-risk patients if their estimated stroke rate was 1.1% per year—a finding more in keeping with the data from the article provided by Dr Ebrahim.R2 We also note that the table provided by Dr Ebrahim uses the terms “low risk” and “high risk” to describe the risk of bleeding on anticoagulation; our article used these terms to describe the annual risk of stroke.
A major conclusion in our study was that analyses employing methodologically sound methods and studying the same condition (ie, screening and treating for asymptomatic carotid stenosis) yielded very different estimates of cost-effectiveness. Further research is needed to determine why such similarly framed questions could lead to such disparate results. These results suggest, however, that readers of such analyses should exhibit a healthy skepticism toward any study making cost-effectiveness claims. Therefore, as we indicate, it might be premature to use the results from such analyses to develop stroke policy and guidelines. Such a conclusion is relevant to aid clinicians and policy makers in making decisions.