Background
Atrial fibrillation (AF) is a common heart rhythm disorder affecting about 2.4 million people in the US [
1,
2], and this number is projected to exceed 5.6 million by 2050 [
3]. AF is associated with a 5-fold greater risk of embolic stroke [
4,
5] and accounts for 75,000 to 100,000 strokes per year in the US [
4]. The risk of stroke due to AF can be reduced by about 50 % with oral anticoagulants such as warfarin [
6,
7], and current American Heart Association and American Stroke Association (AHA/ASA) guidelines recommend prophylactic therapy with warfarin for high and moderate risk AF patients [
8]. Despite strong evidence supporting its efficacy, adherence to warfarin therapy is low—among patients who started warfarin therapy for AF, more than 1 in 4 patients discontinue warfarin therapy within one year [
9,
10]. Low adherence to warfarin therapy is driven by patient concerns about potential bleeding events and the need for continued periodic blood tests (prothrombin time/International Normalized Ratio (INR)) to monitor the patient’s response to warfarin [
11‐
14].
Any strategy that would increase adherence to warfarin therapy among eligible AF patients has the potential to prevent fatal and non-fatal stroke events. Strategies for increasing adherence to warfarin that have been previously investigated include counseling [
15,
16], the use of decision aids [
16], as well as self-testing and self-management programs [
17]. Providing patients with information about their genetic-based risks also has the potential to improve adherence and ultimately clinical outcomes. A recent study reported that patients with a genetic diagnosis of familial hypercholesterolemia were ~50 % more adherent to treatment than were patients without a genetic diagnosis [
18]. Similarly, patients’ knowledge of genetic test results increased adherence (63 vs. 45 %) to statin therapy in the AKROBATS study [
19].
Since gene variants in the 4q25 region of the human genome are associated with increased risk of AF and stroke [
20], providing 4q25 genetic test results to patients might increase adherence to warfarin therapy. If this strategy could be cost effective, it might justify conducting a clinical trial to test the hypothesis that genetic test results would increase adherence to warfarin therapy, which may lead to a lower incidence of preventable strokes. In order to provide cost effectiveness estimates that could be used to justify or design such a clinical trial, we investigated whether the use of genetic information to increase adherence could be cost effective over a range of adherence rates.
Discussion
We compared the usual care strategy with a strategy that used genetic testing to motivate AF patients who declined physician prescribed warfarin therapy to reconsider and initiate therapy. We found that the test strategy could be cost effective even if only 2.1 % of test positive individuals adhered to warfarin therapy over a 5 year horizon. If less than 2 % of test positive individuals would adhere to warfarin therapy, this strategy becomes prohibitively expensive, costing more than $100,000 per QALY if adherence were to fall below 1 %. However, if adherence were to be greater than 5.3 %, this genetic test strategy would be cost-saving under the base-case assumptions.
A 5.3 % increase in adherence in response to genetic test information is smaller than what has been reported by in real-world studies. For example, among patients who were identified as having hypercholesterolemia based on their blood cholesterol levels, adherence to statin medication went up from 39 to 93 % following a genetic confirmation of their diagnosis [
18]. Similarly, among all-comers with a newly-prescribed statin therapy, adherence increased from 45 to 63 % following genetic testing [
19]. Thus, it seems that genetic testing results have the potential to increases in adherence of greater magnitude than that in the model we present. The specific effect of genetic testing on adherence to anticoagulation among patients with AF after genetic testing would need to be determined in a clinical study.
The model baseline parameters were based on published information. The effect of these parameters on the ICERs predicted by the model was modest. The low end of annual rate of stroke events among AF patients who use aspirin (but decline warfarin) resulted in the biggest ICER using baseline parameters. That is, if annual stroke rate among AF patients on aspirin was 3 %, the ICER would be ~ $47,000, however if the annual rate of stroke was at the high end (6 %) the genetic testing strategy would be dominating (cost-saving with better outcomes). Interestingly, if the annual rate of stroke among aspirin users were to be as low as 2 %, the number of strokes prevented by warfarin treatment would be offset by an equal number of major bleeds caused by warfarin treatment (1.5) regardless of treatment strategy. Conversely, a higher rate of strokes in subgroups of patients, which can be estimated by risk scores such as CHADS2, would make the test strategy more cost saving and gain more QALYs compared with the usual care strategy.
This study has several potential limitations. First, this study is based on a theoretical patient cohort and therefore used baseline parameters that were established in other real-life cohorts. Deviation from these baseline parameter estimates could affect the results. However, our sensitivity analyses established that variations in these parameters have only a small effect on the conclusions of this study. We found that the cost of the genetic test had the greatest influence on the range of total costs. We believe that the range of costs investigated ($50 to $200) was reasonable, because the genetic test considered in this study involved genotyping two single nucleotide polymorphisms (SNPs). And Medicare reimbursement for two SNPs was ~ $100 in 2013—this is based on the combined reimbursement for the Factor V Leiden SNP and for the Factor II (20210G > A) SNP [
31]. We have based all other medical costs on published 2005 estimates [
25]. However, inflating these costs to 2014 would result in more favorable outcome for the test strategy since the test strategy results in fewer events than the usual care strategy for any adherence rate. Second, we focused our analysis on adherence to warfarin as an oral anticoagulant. We believe an investigation of adherence to warfarin is important because warfarin remains first line therapy (recommendation 1A) in the 2014 AHA/ACC guidelines for the management of patients with atrial fibrillation [
30]. New oral anticoagulants (NOACs: dabigatran, rivaroxaban, and apixaban) which are now available as anticoagulant option for some AF patients received a 1B recommendation. Patient adherence to NOACs is not likely better than to warfarin [
31], presumably because bleeding event rates for warfarin are largely not different than that for most NOACs [
32]. Moreover, the lack of approved antidote to NOACs can deter some patients from adhering to therapy. Third, we assumed the frequency of individuals who are carriers of a 4q25 risk allele (test positive) to be 40 %. This assumption is based on the reported allele frequencies of rs2200733 and rs10033464 in populations of European ancestry. The fraction of test positive individuals could vary in populations with a different ethnic ancestry. For example, among Yoruban in Ibadan, Nigeria, the fraction of test positive patients would be 86 % [
33]. We elected to use the test positive information for European ancestry patients because most of the genetic association studies supporting the association of these SNPs with risk of AF were conducted in these populations. Our model could be easily adopted for other expected test positive fractions. Given these limitations, the results of this study should be cautiously considered when trying to extrapolate to real-life studies.
Competing interests
DS, LB, JL, AA, and JD are employees of Quest Diagnostics. MP declares no competing interests.
Authors’ contributions
DS, MP and JD participated in the design of the study. AA, JL and LB performed the statistical analysis. DS, MP and JD conceived of the study, and participated in its design and coordination. AA, DS, JL, JD, LB, and MP helped to draft the manuscript. All authors read and approved the final manuscript.