J.A. Leey et al. The American Journal of Geriatric Pharmacotherapy Cost-Effectiveness of Genotype-Guided Warfarin
Therapy for Anticoagulation in Elderly Patients
With Atrial Fibrillation
Julio A. Leey, MD, MSc1; Steve McCabe, MD, MSc2; Jennifer A. Koch, MD3; and 1Division of Endocrinology, Metabolism, and Lipid Research, Washington University School of Medicine, St. Louis, Missouri; 2School of Public Health and Information Sciences, University of Louisvil e, Louisvil e, Kentucky; 3Department of Medicine, University of Louisvil e, Louisvil e, Kentucky; and 4Department of Family and Geriatric Medicine, University of Louisvil e, Copyright Excerpta Medica, Inc, 2009
Not for Commercial
Background: In patients with atrial fibrillation (AF), anticoagulation with warfarin decreases the risk of embolic
stroke by >50%. Identification of genetic polymorphisms in enzymes involved in the metabolism of warfarin can partially predict the maintenance dose and thus potentially decrease the incidence of bleeding episodes secondary to Objectives: The objectives of this study were to evaluate the potential clinical and economic outcomes of
genotype-guided warfarin therapy in elderly patients newly diagnosed with AF and to identify a threshold in bleeding risk at which such therapy may be cost-effective.
Methods: A decision tree was designed to represent the medical decision (pharmacogenetic testing or not) and
the main clinical outcomes (embolic stroke, bleeding). Event rates of embolic stroke and bleeding complications were based on data from previously published clinical trials and an observational study, respectively; costs were from a third-party payer perspective; and utilities were from the patient perspective. It was assumed that use of pharma- cogenetic testing would not lead the clinician to make any potentially harmful modifications to the regimen.
Results: This analysis found that any reduction in major bleeding as a result of pharmacogenetic testing would
lead to improved utility. The higher costs of pharmacogenetic testing compared with no testing would be immedi- ately offset by any reduction in major bleeding.
Conclusions: In this decision analysis, genotype-guided warfarin therapy for anticoagulation in elderly patients
with AF was potentially cost-effective, and its benefits were closely related to efficacy in preventing bleeding events. Clinical trials testing the efficacy of genotype-guided warfarin therapy are warranted. (Am J Geriatr Pharmacother. 2009;7:197–203) 2009 Excerpta Medica Inc.
Key words: decision analysis, cost-effectiveness, warfarin, pharmacogenetics, bleeding, stroke.
Data in this paper were presented as a poster at the 2008 Annual Scientific Meeting of the American Geriatric Society, April 30–May 4, 2008, Washington, DC.
Accepted for publication July 6, 2009. 2009 Excerpta Medica Inc. All rights reserved. The American Journal of Geriatric Pharmacotherapy INTRODUCTION
method used to help make the best decisions in the face Atrial fibrillation (AF) affects 5% to 10% of the elderly of uncertainty. It is useful when a randomized trial is population, substantially increasing the risk of embolic not feasible due to ethical considerations, economic events such as stroke.1 If untreated, AF can have a cata- constraints, time pressure, changing technology, or lo- strophic outcome for the patient and result in an enor- gistic impossibility. The decision-analysis method in- mous cost of care. Anticoagulation with warfarin de- volves creating a model of the alternative decisions and creases the risk of embolic stroke by >50%; however, their consequences based on the probabilities of the warfarin poses a risk for major hemorrhagic events events or final states of health (ie, risk of embolic such as intracranial hemorrhage and gastrointestinal stroke, risk of major bleeding, risk of death or disabili- bleeding.2,3 Major and minor embolic strokes associat- ty) and patient preferences for each of the final states.10 ed with AF are estimated to cost $1,758,548 and Each final state is given a preference value, or utility. $19,352, respectively, per event, whereas severe lower The utility value, which is assigned by patients, ranges gastrointestinal hemorrhages are estimated to cost from 0 to 1, representing the worst and best possible outcomes, respectively. These values are then weighted Adjustment of the warfarin dose is cumbersome and by the probability of the events actually occurring. The depends on several factors, including demographic decision alternative with the highest expected value is characteristics (age, sex, and race), nutrition (diet and the recommended choice. Given the inputs into the vitamin K intake), medical history (heart disease, liver decision model and its structure, this method identifies disease, and concomitant medications), anthropomet- the best long-term choice. Adding the cost of each ric variables (weight, height, and body surface area), outcome makes it possible to weigh the economic im- and genetic factors (cytochrome P450 [CYP] 2C9 and pact of decisions based on patients’ preferences (utili- vitamin K epoxide reductase [VKORC1] genotypes).5 ties) and the probability of events.
CYP2C9 is the enzyme that metabolizes s-warfarin, the The objectives of this study were to evaluate the po- main enantiomer of warfarin. Carriers of some genetic tential clinical and economic outcomes of genotype- polymorphisms of CYP2C9 metabolize s-warfarin more guided warfarin therapy in elderly patients newly diag- slowly than do others, leading to elevated international nosed with AF and to identify a threshold in bleeding normalized ratios (INRs) at common initial doses of risk at which genotype-guided warfarin therapy may be warfarin.6 VKORC1, which is inhibited by warfarin, is cost-effective in these patients. One of the questions an enzyme that activates vitamin K, allowing the syn- investigated was whether all patients who are candidates thesis of several clotting factors. In carriers of some for warfarin therapy are likely to benefit from pharma- genetic polymorphisms of VKORC1, the enzyme has a cogenetic testing, even those with a low bleeding risk.
greater affinity for warfarin, requiring a lower dose to appropriately inhibit coagulation. It is possible to have polymorphisms of both of these enzymes, inducing re- A decision tree was created to simulate a common sce- nario in geriatric practice: a patient aged >65 years is Identification of these genetic polymorphisms newly diagnosed with nonvalvular AF, and the initia- through pharmacogenetic testing has been reported tion of warfarin therapy is clinically indicated (Figure
to have 50% to 60% accuracy in predicting the warfa- 1). There are 2 alternatives with respect to initiating
rin maintenance dose,5 potentially decreasing the inci- this therapy: pharmacogenetic testing and no testing. dence of bleeding secondary to elevated INRs. Routine Each option has 2 types of clinical outcome, the first assessment of genetic polymorphisms related to warfa- related to the incidence of embolic stroke and the sec- rin metabolism has the potential to improve clinical ond related to complications of warfarin use. In clinical management and decrease the likelihood of bleeding.7 practice, patients with mild or moderate embolic stroke However, genetic testing would represent an additional would most likely continue to take warfarin after recov- cost in an already expensive process. The efficacy and ery from the stroke; therefore, in the model, the warfa- economic outcomes of such genotype-guided therapy rin complications outcome was added after the occur- rence of embolic stroke. The warfarin complications Given the limited number of published clinical trials included were major bleeding episodes sufficiently sig- that have assessed the effectiveness of genotype-guided nificant to prompt hospital admission and blood trans- warfarin therapy,8,9 this scenario represents an appro- fusion, that involve a critical site (intracranial, intraspi- priate model for decision analysis. Decision analysis is a nal, intraocular), or that are fatal.
J.A. Leey et al. The American Journal of Geriatric Pharmacotherapy The American Journal of Geriatric Pharmacotherapy Given that most adjustment of the warfarin dose and the majority of complications typically occur in the first Table. Probabilities, costs, and utility values used in year of therapy, the decision tree considered clinical outcomes over a period of 12 months. The results were expressed in terms of the cost of adverse events (eg, death, embolic stroke, warfarin complications) from the perspective of a third-party payer (in 2003 US dol- lars) and in terms of utility from the patient’s perspec- tive, a measure of the desirability of the outcome.
Probability, Cost, and Utility Values
The event rate of embolic stroke was based on the outcomes of published clinical trials,11 and event rates of bleeding complications were based on outcomes from an observational study (Table).12 Costs and utili-
ty values were taken from previous cost-effectiveness studies of anticoagulation.4,11,13 The costs of intracra- nial hemorrhage, other severe bleeding, and stroke were estimated from 2003 diagnosis-related groups, as calculated and used in previous analyses.4,11 The yearly cost of INR monitoring was approximated on the as- sumptions that a patient initiating warfarin therapy would require frequent INR monitoring during the first months and that the frequency of monitoring would be slightly lower in the setting of pharmaco- genetic testing.4 The cost of testing was approximat- ed from the Medicare Reimbursement Estimate and rounded to $250, as in a previous analysis.13 It was assumed that pharmacogenetic testing would not lead the clinician to make potentially harmful dose adjustments, an assumption supported by data indicat- ing no difference in time spent within the therapeutic INR range among patients with genotype-guided ver- Sensitivity Analysis
INR = international normalized ratio.
A sensitivity analysis was performed using a range of *Other major bleeding = gastrointestinal bleeding and epistaxis. values for the risk of embolic stroke and the incidence of warfarin complications to examine the robustness of the model. Because the 2 main branches of the tree (phar- with increased susceptibility to warfarin intoxication and macogenetic testing and no testing) were assumed to those who are resistant to warfarin compared with the have similar distribution but different values, the sensi- general population. Because the degree of success in tivity analysis was based on 2 factors: the incidence of predicting the maintenance dose of warfarin would have embolic stroke and the incidence of warfarin complica- an effect on complication rates, rates of warfarin compli- tions. It was reasoned that the higher the rate of embolic cations also were included in the sensitivity analysis.
stroke, the more likely that pharmacogenetic testing All analyses were performed using TreeAge Pro 2006 would prevent this event; therefore, the rate of embolic (TreeAge Software, Inc., Williamstown, Massachusetts).
stroke needed to be included in the sensitivity analysis. Pharmacogenetic-guided dosing of warfarin has been reported to predict the maintenance dose in 50% to 60% Because embolic stroke is a relatively long-term out- of patients,5 particularly in 2 patient subgroups: those come that is not likely to be affected by early improve- J.A. Leey et al. The American Journal of Geriatric Pharmacotherapy ment in adjustment of the warfarin dose as a result of DISCUSSION
pharmacogenetic testing, the main differences between This analysis found that if pharmacogenetic testing re- the 2 options were the potentially reduced risk of sulted in less major bleeding in association with warfa- bleeding associated with testing and the increased cost rin anticoagulation in elderly patients newly diagnosed of testing. Analysis of the decision tree indicated that with AF, this option would have a higher expected utili- any reduction in major bleeding would be expected to ty and, given a slight improvement in the risk of bleed- result in an improvement in utility. The cost of pharma- ing, less overall cost than no testing.
cogenetic testing would offset the savings associated The decision tree considered clinical outcomes over a with a reduction in the need for INR monitoring as a period of 12 months, since beyond the first year of war- result of rapid dose adjustment, but any reduction in farin therapy, most patients would be expected to have major bleeding would favor the use of pharmacoge- reached a stable maintenance dose, and the risk of bleeding associated with overdose would be lower. Be- The greater cost associated with pharmacogenetic cause the use of pharmacogenetic testing would be ex- testing would be immediately offset by any reduction in pected to help achieve maintenance dosing rapidly, it major bleeding. The present analysis found that if ma- made sense for the model to focus on bleeding events jor bleeding were reduced by any percentage, the phar- within the first 12 months after the initiation of therapy. macogenetic testing option would be less expensive Whether the use of genotype-guided warfarin dosing and, as noted earlier, result in higher expected utility would have any effect on the incidence of stroke during (Figure 2).
the first weeks or months is not known. Considering stroke as a constant in the model ensured that the utility of pharmacogenetic testing was not overestimated.
The outcome of drug therapy is often unpredictable, ranging from a beneficial effect to a lack of efficacy to a serious adverse effect. Polymorphisms in single genes that code for enzymes involved in drug metabolism are a well-known cause of such unpredictability. The ability to identify genetic variations by pharmacogenetic testing has raised expectations concerning the ability to predict the appropriate maintenance dose and, therefore, the response. However, acceptance of pharmacogenetic- guided warfarin therapy will depend on clinical evidence, In a randomized clinical trial involving 206 patients, Anderson et al8 found that an algorithm based on phar- macogenetic and clinical factors improved the accuracy and efficiency of warfarin dosing, although the primary end point of reduction in out-of-range INRs was not achieved. In contrast, in a clinical trial in 283 patients, Caraco et al9 found that patients who underwent CYP2C9 genotyping spent more time within the therapeutic INR range. In a study by Klein et al,14 a pharmacogenetic- based algorithm accurately identified greater proportions Figure 2. Sensitivity analysis of the probability of bleeding. of patients requiring warfarin doses ≤21 or ≥49 mg per The probability of bleeding associated with the week to achieve the target INR compared with a clinical use of pharmacogenetic (PG) testing would have algorithm. Even if a clinical trial were to report a decrease to be less than the probability during the first year in the proportion of out-of-range INRs, this end point without PG testing to have greater utility, or ex- would not necessarily correspond to a reduction in embolic pected value. In this analysis, the probability of stroke, particularly if the incidence of stroke was relatively bleeding was assumed to be 0.072 during the first low. The present analysis found that lowering the inci- year of warfarin therapy,12 and the expected value dence of warfarin complications by as little as 0.1% (from 7.2% to 7.1%) would be associated with a cost benefit.
The American Journal of Geriatric Pharmacotherapy Two decision-analysis models of pharmacogenetic- CONCLUSIONS
oriented warfarin dosing have been published that were In this decision analysis, genotype-guided warfarin also sensitive to reductions in the rate of bleeding.15,16 therapy for anticoagulation in elderly patients with AF Both analyses found that this practice may be cost- was potentially favorable in those at high risk for bleed- effective in patients who are at high risk for hemor- ing. The benefits of this strategy would be closely re- rhage. As in the study by You et al,15 the present study lated to its efficacy in preventing bleeding events. focused on the 12 months after initiation of warfarin Further clinical studies are needed to fully assess the therapy, as most bleeding episodes occur during the impact of using pharmacogenetic testing to help pre- first months of therapy. The model indicated that pa- vent bleeding complications of warfarin therapy in el- tients at high risk of bleeding would benefit from phar- macogenetic testing. HEMORR2HAGES,17 a clinical score used to identify patients at risk for bleeding, may ACkNOwLEDgMENTS
be useful for stratifying bleeding risk.
This analysis was funded by the Department of Family The prevalence of CYP2C9 and VKORC1 variants and Geriatric Medicine at the University of Louisville, associated with slow warfarin metabolism and, there- Louisville, Kentucky. The authors thank the School of fore, a high risk of bleeding would be likely to affect the Public Health and Information Sciences at the Univer- present model. Schwarz et al18 found that the frequency sity of Louisville for providing the computer laboratory of A/A polymorphism for the VKORC1 gene was 11% in which the analysis was performed.
and 0% in the white and black populations, respectively. The authors have indicated that they have no con- In a population with a very low prevalence of these flicts of interest regarding the content of this article.
polymorphisms, use of pharmacogenetic testing would have no benefit, whereas the ability to titrate the warfa- REFERENCES
rin dose appropriately would be improved in those with 1. Fuster V, Rydén LE, Cannom DS, et al, for the Writing a higher prevalence of these polymorphisms.
Committee to Revise the 2001 Guidelines for the Manage- A study by the International Warfarin Pharmacoge- ment of Patients With Atrial Fibrillation of the American netics Consortium is currently under way to determine Col ege of Cardiology/American Heart Association Task the dosing effectiveness, efficiency, and tolerability of Force on Practice Guidelines and the European Society of genotype-based dosing of warfarin. The results are ex- Cardiology Committee for Practice Guidelines. ACC/ AHA/ESC 2006 guidelines for the management of pa- The decision tree used in this analysis had several tients with atrial fibrillation—executive summary: A report limitations. First, for clarity of the model, the utility of the American College of Cardiology/American Heart values for mild, moderate, and severe stroke were Association Task Force on Practice Guidelines and the merged into a single category. The costs of bleeding European Society of Cardiology Committee for Practice episodes were similarly merged into a single category Guidelines (Writing Committee to Revise the 2001 Guide- (mild, moderate, severe, and death), as were the costs lines for the Management of Patients With Atrial Fibrilla- of stroke. Second, the incidence of stroke was consid- tion) [published correction appears in J Am Coll Cardiol. ered a constant in both groups in the model, although 2007;50:562]. J Am Coll Cardiol. 2006;48:854–906.
it is possible that the incidence of stroke might be 2. Landefeld CS, Goldman L. Major bleeding in outpatients slightly lower in the pharmacogenetic-testing group if treated with warfarin: Incidence and prediction by factors testing led to a decrease in out-of-range INR values known at the start of outpatient therapy. Am J Med. due to underdosing of warfarin in patients who were carriers of a polymorphism for warfarin resistance. Fi- 3. Fanikos J, Grasso-Correnti N, Shah R, et al. Major bleed- nally, costs were for a 12-month period only, whereas ing complications in a specialized anticoagulation service. the sequelae of stroke and intracranial hemorrhage may Am J Cardiol. 2005;96:595–598.
require more extended care and, therefore, be more 4. Leigh JP, White RH. An economic model for adverse expensive. Because the efficacy of pharmacogenetic events and costs for oral anticoagulants used for atrial fi- testing in preventing bleeding has not yet been deter- brillation. Curr Med Res Opin. 2007;23:2071–2081.
mined, there is a need for clinical trials or detailed co- 5. Zhu Y, Shennan M, Reynolds KK, et al. Estimation of hort studies in which bleeding rates are compared in warfarin maintenance dose based on VKORC1 (-1639 patients whose anticoagulation is and is not guided by G>A) and CYP2C9 genotypes. Clin Chem. 2007;53: J.A. Leey et al. The American Journal of Geriatric Pharmacotherapy 6. Gage BF, Lesko LJ. Pharmacogenetics of warfarin: Regu- 12. Hylek E, Evans-Molina C, Shea C, et al. Major hemor- latory, scientific, and clinical issues. J Thromb Thromboly- rhage and tolerability of warfarin in the first year of thera- py among elderly patients with atrial fibrillation. Circula- 7. Sanderson S, Emery J, Higgins J. CYP2C9 gene variants, drug dose, and bleeding risk in warfarin-treated patients: 13. Campos-Outcalt D. Personalized medicine: The promise, A HuGEnet systematic review and meta-analysis. Genet the reality. J Fam Pract. 2007;56:621–626.
14. Klein TE, Altman RB, Eriksson N, et al, for the Interna- 8. Anderson JL, Horne BD, Stevens SM, et al, for the tional Warfarin Pharmacogenetic Consortium. Estima- Couma-Gen Investigators. Randomized trial of genotype- tion of the warfarin dose with clinical and pharmacoge- guided versus standard warfarin dosing in patients initi- netic data. N Engl J Med. 2009;360:753–764.
ating oral anticoagulation. Circulation. 2007;116:2563– 15. You JH, Chan FW, Wong RS, Cheng G. The potential clinical and economic outcomes of pharmacogenetics- 9. Caraco Y, Blotnick S, Muszkat M. CYP2C9 genotype- oriented management of warfarin therapy—a decision guided warfarin prescribing enhances the efficacy and analysis. Thromb Haemost. 2004;92:590–597.
safety of anticoagulation: A prospective randomized 16. Eckman MH, Rosand J, Greenberg SM, Gage BF. Cost- controlled study. Clin Pharmacol Ther. 2008;83:460– effectiveness of using pharmacogenetic information in warfarin dosing for patients with nonvalvular atrial fibril- 10. Richardson WS, Detsky AS, for the Evidence-Based lation. Ann Intern Med. 2009;150:73–83.
Medicine Working Group. Users’ guides to the medical 17. Gage BF, Yan Y, Milligan PE, et al. Clinical classification literature. VII. How to use a clinical decision analysis. B. schemes for predicting hemorrhage: Results from the What are the results and will they help me in caring for National Registry of Atrial Fibrillation (NRAF). Am my patients? JAMA. 1995;273:1610–1613.
11. O’Brien C, Gage BF. Costs and effectiveness of ximela- 18. Schwarz UI, Ritchie MD, Bradford Y, et al. Genetic de- gatran for stroke prophylaxis in chronic atrial fibrillation. terminants of response to warfarin during initial antico- agulation. N Engl J Med. 2008;358:999–1008.
Address correspondence to: Julio A. Leey, MD, MSc, Division of Endocrinology, Metabolism, and Lipid Research,
Washington University School of Medicine, 660 S. Euclid Avenue, Box 8127, St. Louis, MO 63110. E-mail: jleey@



Common/Trade Report Format Limit of Detection (ng/mL) Analyte Names (if different from Analyte name) 1 2-hydroxyethylflurazepam 2-hydroxyethylflurazepam (Flurazepam Metabolite)3 7-aminoclonazepam 7-aminoclonazepam (Clonazepam Metabolite)4 7-Aminoflunitrazepam 7-Aminoflunitrazepam (Flunitrazepam Metabolite) 5 Acetaminophen 6 Alprazolam 7 Amitriptyline 8 Amobar

ferent amounts of Shannon information, they have one thingin common: Entropy within a chunk is relatively low, en-We argue that the ability to find meaningful chunks in se-tropy at chunk boundaries is relatively high. Two kinds ofquential input is a core cognitive ability for artificial generalintelligence, and that the Voting Experts algorithm, whichevidence argue that this signature of chunk

Copyright © 2010 Find Medical Article