Patients start out with uncomplicated atrial fibrillation, then cycle
between health states until death occurs or the 20-year period ends. The length
of each cycle is 30 days. Temporary health states (eg, elevated liver function
tests and minor hemorrhages) are not depicted. The health states are equivalent
for aspirin, warfarin, and ximelagatran treatment, but the probabilities,
costs, and quality of life vary with treatment. Any health state could lead
to death, including death from liver failure (not shown).
Bars indicate range of cost per additional quality-adjusted life-year
(QALY) of ximelagatran compared with warfarin as determined in sensitivity
analyses over plausible ranges for variables. Upper and lower limits of variables
examined in sensitivity analyses are indicated. Arrows indicate that the cost
per QALY exceeds $200 000. INR indicates international normalized ratio.
Above the horizontal line, ximelagatran costs less than $50 000
per quality-adjusted life-year (QALY); below the line, warfarin costs less
than $50 000 per QALY. To the left of the nearly vertical line, both
of these anticoagulants cost more than $50 000 per QALY but aspirin is
O’Brien CL, Gage BF. Costs and Effectiveness of Ximelagatran for Stroke Prophylaxis in Chronic Atrial Fibrillation. JAMA. 2005;293(6):699–706. doi:10.1001/jama.293.6.699
Author Affiliations: Washington University
School of Medicine, St Louis, Mo.
Context Recent trials have found that ximelagatran and warfarin are equally
effective in stroke prevention for patients with atrial fibrillation. Because
ximelagatran can be taken in a fixed, oral dose without international normalized
ratio monitoring and may have a lower risk of hemorrhage, it might improve
quality-adjusted survival compared with dose-adjusted warfarin.
Objective To compare quality-adjusted survival and cost among 3 alternative therapies
for patients with chronic atrial fibrillation: ximelagatran, warfarin, and
Design Semi-Markov decision model.
Patients Hypothetical cohort of 70-year-old patients with chronic atrial fibrillation,
varying risk of stroke, and no contraindications to anticoagulation therapy.
Main Outcome Measures Quality-adjusted life-years (QALYs) and costs in US dollars.
Results For patients with atrial fibrillation but no additional risk factors
for stroke, both ximelagatran and warfarin cost more than $50 000 per
QALY compared with aspirin. For patients with additional stroke risk factors
and low hemorrhage risk, ximelagatran modestly increased quality-adjusted
survival (0.12 QALY) at a substantial cost ($116 000 per QALY) compared
with warfarin. For ximelagatran to cost less than $50 000 per QALY it
would have to cost less than $1100 per year or be prescribed to patients who
have an elevated risk of intracranial hemorrhage (>1.0% per year of warfarin)
or a low quality of life with warfarin therapy.
Conclusion Assuming equal effectiveness in stroke prevention and decreased hemorrhage
risk, ximelagatran is not likely to be cost-effective in patients with atrial
fibrillation unless they have a high risk of intracranial hemorrhage or a
low quality of life with warfarin.
The 2.3 million persons in the United States with atrial fibrillation1,2 have a 5-fold increased risk of ischemic
stroke.3,4 Randomized controlled
trials have found that warfarin reduces the risk of ischemic stroke in atrial
fibrillation by 65%.5 However, warfarin is
prescribed for only half of patients with atrial fibrillation who are appropriate
anticoagulation candidates.6- 8 Warfarin
therapy is hampered by drug and food interactions, slow onset of action, requirement
of regular monitoring, and individual variability in metabolism.9,10
Ximelagatran, an oral, direct thrombin inhibitor, has been developed
to address these shortcomings. Unlike warfarin, ximelagatran has no known
food or drug interactions and has a consistent pharmacokinetic profile.11- 14 Ximelagatran
does not therefore require dose adjustment. Recently, 2 large randomized trials
showed that ximelagatran is as effective as warfarin in stroke prevention
and may cause less bleeding among patients with chronic atrial fibrillation.15,16 However, both trials noted that 6%
to 7% of patients taking ximelagatran developed liver function abnormalities
and 3 of 6948 participants died with possible liver failure.17
With a more favorable pharmacokinetic profile, equal efficacy in stroke
prevention, and probable lower risk of bleeding, ximelagatran may increase
quality-adjusted survival compared with warfarin. However, it is unclear whether
this improvement justifies the additional cost and offsets rare liver toxicity.
In this analysis, we compare the projected quality-adjusted survival and costs
of ximelagatran, warfarin, and aspirin in patients with chronic atrial fibrillation.
Using a semi-Markov model,18 we performed
a decision analysis comparing 3 treatments: aspirin, adjusted-dose warfarin
with an international normalized ratio (INR) of 2 to 3, and fixed-dose ximelagatran
(36 mg twice per day) in patients with atrial fibrillation. We expressed our
results in terms of risk of adverse events, quality-adjusted life expectancy,
2003 US dollars, and incremental cost-effectiveness ratios. Our base case
consisted of a hypothetical cohort of 70-year-old patients with atrial fibrillation,
a moderate risk of stroke, and no contraindications to anticoagulant therapy.
The permanent health states in the model included healthy with atrial
fibrillation, ischemic stroke (fatal, major, mild, or reversible), transient
ischemic attack (TIA), hemorrhage (fatal, intracranial [ICH], or major or
minor noncerebral), recurrent or combined events, and death (Figure 1). As described below, utilities and costs were applied
to each of the outcomes over their expected duration. For all treatments,
we quantified quality-adjusted life expectancy, risk of adverse events, and
net cost over a maximum of 20 years.
Mortality rates in the model were adjusted for aging (beginning at age
70 years), presence of atrial fibrillation, and antithrombotic therapy. Median
survival was 11.7 years with aspirin, 13.3 years with warfarin, and 13.4 years
with ximelagatran.7,19- 26
Risk of Liver Function Abnormalities. We estimated
the rate of elevated liver function test results on ximelagatran to be 1.0%
per month for the first 6 months and 0.08% per month subsequently.15- 17 We assumed that 1
in 2300 patients taking ximelagatran would develop fatal hepatic damage17 and that there would be a monthly 0.035% risk of
elevated liver function test results with warfarin15 or
Hemorrhage Risk. We quantified hemorrhage risk
based on rates in the SPORTIF (Stroke Prevention Using an Oral Thrombin Inhibitor
in Atrial Fibrillation) trials. In our analysis, ICH included hemorrhagic
strokes and subdural hematomas. The annual rate of ICH in SPORTIF III and
V was 0.4% in participants randomized to receive warfarin (22 ICHs in 5652
patient-years of warfarin).15,16 The
rate of major hemorrhage was 2.5% per year of warfarin therapy. Based on the
pooled SPORTIF III and V data, the relative risk of major hemorrhage (including
ICH) with ximelagatran vs warfarin was 0.74 (95% confidence interval [CI],
0.57-0.97), and the combined rate of major and minor hemorrhages with ximelagatran
was 32.0% vs 39.1% with warfarin.15,16
In other randomized trials, the relative risk of major hemorrhage with
aspirin compared with warfarin was 0.59.27 In
our baseline model, patients who had a major hemorrhage while taking ximelagatran
or warfarin stopped the anticoagulant and began aspirin.28
Ischemic Stroke Risk. We quantified stroke
risk based on a validated prediction rule, CHADS2, in which stroke
rate depends on the presence of the following risk factors: congestive heart
failure, hypertension, age older than 75 years, diabetes mellitus, and history
of stroke or TIA (Table 1).29,55 In the base case, stroke rates were
52% lower for patients treated with warfarin rather than aspirin,27 and ximelagatran and warfarin had identical stroke
rates.15,16 In the base case,
we assumed a moderate risk of stroke (4.5% per year with aspirin therapy)
and that 28% of neurologic ischemic events were TIAs.30- 33 The
rate of stroke and TIA increased by a factor of 1.4 per decade of life, compounded
Stroke and Hemorrhage Severity. We classified
initial ischemic stroke in 4 categories of severity: fatal, moderate to severe
neurologic residua, mild neurologic residua, and no residual deficit (Table 1).30- 39 Similarly,
we classified major hemorrhage as fatal, nonfatal ICH, nonfatal extracranial
major hemorrhage, and nonfatal extracranial minor hemorrhage (Table 1).30- 37,39,56- 63 Nonfatal
extracranial major hemorrhages affected quality of life for 1 month, while
ICH had neurologic residua (or was fatal). In the model, minor hemorrhages
decreased quality of life for only 2 days.
To calculate quality-adjusted survival, we multiplied the probabilities
of adverse events by quality-of-life estimates, known as utilities (Table 1).45 By
definition, death from any cause had a utility of 0. We obtained the utility
for warfarin from our previous survey of 83 patients with atrial fibrillation.
Seventy of these patients were able to rate their quality of life while taking
warfarin, including prothrombin time monitoring and changes in diet or lifestyle,
with a mean value of 0.987.45 The average utility
of aspirin therapy was 0.998.
To estimate the utility of ximelagatran therapy, we conducted a 1-time
e-mail survey of the Anticoagulation-Thromboembolism Research Consortium,
a group of approximately 30 physicians involved in antithrombotic clinical
management and research, of whom 12 responded. We also surveyed 7 decision-analysts
who have published in the area of antithrombotic therapy, of whom 3 responded;
2 of these respondents had usable results. Respondents gave 2 estimates for
the utility of ximelagatran, 1 for the first 6 months of therapy (with monthly
liver function monitoring) and 1 for subsequent therapy (with liver function
monitoring every 3 months). Based on the values of 0.987 for warfarin utility
and 0.998 for aspirin utility, respondents recommend a mean utility for ximelagatran
of 0.989 (95% CI, 0.986-0.991) during the first 6 months and 0.994 (95% CI,
0.993-0.996) thereafter. For patients who developed liver function abnormalities,
0.002 quality-adjusted life-year (QALY) was subtracted to account for additional
blood tests and physician visits.
For each treatment, we projected net cost over 20 years. Future costs
and life-years were discounted at 3% per year. Because we were interested
in the incremental cost-effectiveness of one option vs another, rather than
absolute costs, we excluded medical costs unrelated to antithrombotic therapy,
hemorrhage, or neurological ischemia. Costs reflected the perspective of a
health maintenance organization or insurance company that covered inpatient
and outpatient medical care and prescription costs but did not pay for indirect
costs (eg, lost wages). Costs were expressed in 2003 US dollars.64
Adverse Events. Cost of a minor hemorrhage
was based on remuneration for an expanded problem-focused physician visit
(Current Procedural Terminology [CPT ] code 99213).49 We estimated the cost of a major extracranial hemorrhage
based on Medicare remuneration for the diagnosis-related group associated
with gastrointestinal hemorrhage.51 We calculated
costs for stroke, TIA, and ICH by using the median value of published studies
and Medicare remuneration, and we estimated the cost of fatal hepatic failure
(Table 1).51- 54
Drug Costs. We calculated the $545 annual cost
of warfarin therapy by combining its annual prescription cost48 with
Medicare reimbursement for 14 INR tests and minimal established patient office
visits (CPT code 99211) per year (Table 1).49 In sensitivity analyses,
we examined patients initiating warfarin, to whom we added a cost of up to
8 INRs and up to 8 physician visits during the first month based on our experience
Based on the cost of clopidogrel (average US price, $4.39 per day) and
the cost of ximelagatran in Germany (Exanta, AstraZeneca, London, England;
€4.50 per day), we estimated the drug cost of ximelagatran to be $5 per
day ($1825 per year).48 The total annual cost
of ximelagatran therapy included a liver function test50 and
minimal established patient office visit on initiation of treatment, then
monthly visits and liver function measurements for 6 months. After the initial
6 months of therapy, the cost of ximelagatran included only 1 liver function
test and minimal established patient office visits every 2 to 3 months. The
cost associated with elevated liver function test results included 2 problem-oriented
physician office visits (CPT code 99212)49 and
2 additional liver function tests (Table 1).
We performed 1-way sensitivity analyses of the variables in the decision
model over their plausible ranges (Table 1).
In a 2-way sensitivity analyses analysis, we calculated cost-effectiveness
ratios of ximelagatran over combinations of stroke and ICH risk. In first-order
Monte Carlo simulations, we randomly sampled (with replacement) 10 000
times a set of utilities from 70 patients we have previously reported on who
had atrial fibrillation and usable utility values45 and
simulated outcomes using uniform distributions of all variables.
All analyses were performed with SMLTREE.66
Under base-case conditions, the quality-adjusted life expectancy in
70-year-old atrial fibrillation patients at moderate risk of stroke (4.5%
per year of aspirin) and lower risk of ICH (0.4% per year of warfarin) was
9.51 QALYs with ximelagatran therapy, 9.39 QALYs with warfarin therapy, and
8.58 QALYs with aspirin therapy. The use of ximelagatran would yield only
0.12 QALY more than warfarin, at a cost of approximately $116 000 per
QALY (Table 2).
We examined how each variable affected quality-adjusted survival and
cost for all plausible values (Figure 2).
The most influential variables were ICH risk, effectiveness of stroke prophylaxis,
risk of hepatic damage from ximelagatran, patient utilities for warfarin,
and cost of ximelagatran.
ICH Risk. As risk of ICH increased, the marginal
cost per QALY of ximelagatran compared with warfarin decreased from $116 000
to $44 000 (Table 2). Using $50 000
per additional QALY as a threshold for cost-effectiveness, we found that ximelagatran
was cost-effective compared with warfarin for patients with an ICH risk greater
than 1% per year of warfarin (Figure 3).
If, however, ximelagatran did not reduce the risk of hemorrhage relative to
warfarin, then the cost of ximelagatran exceeded $250 000 per QALY compared
with warfarin, regardless of hemorrhage risk. In contrast, for patients at
high risk of ICH who would otherwise be prescribed aspirin, ximelagatran could
improve survival by approximately 0.8 QALY (Table
2), at a cost of $17 400 per QALY.
Ischemic Stroke Risk. The cost-effectiveness
of ximelagatran was extremely sensitive to its effectiveness: if ximelagatran
reduced the relative risk of stroke by greater than 6% vs warfarin, then it
would cost less than $50 000 per QALY. Assuming equal effectiveness,
for patients with atrial fibrillation with the lowest risks of stroke (0.8%
per year of aspirin) and ICH (0.4% per year of warfarin), ximelagatran modestly
increased survival (by 0.12 QALY) compared with aspirin, at a cost of $167 000
per QALY. For these low-risk patients, warfarin yielded the fewest QALYs (10.43).
For patients with a higher risk of stroke (5.5% per year of aspirin), warfarin
cost $1000 or less per QALY compared with aspirin, and ximelagatran cost $9000
per QALY compared with aspirin but more than $100 000 per QALY compared
Hepatic Failure. If only 1 in 10 000 patients
initiating ximelagatran developed permanent hepatic failure (rather than our
base-case estimate of 1 in 2300), then the cost per QALY of ximelagatran compared
with warfarin would be $82 290. Even if none of the patients taking ximelagatran
develop permanent hepatic failure, the cost per QALY of ximelagatran compared
with warfarin would slightly exceed $75 000.
Patient Utility. In 58 of 70 patients we previously
surveyed,45 ximelagatran cost greater than
$50 000 per QALY gained compared with either aspirin or warfarin therapy.
In these patients, the average utility for warfarin was 0.996 (range, 0.97-1.00).
The 12 patients whose cost per additional QALY for ximelagatran was less than
$50 000 compared with warfarin had a median utility for warfarin of 0.954
Monitoring. If the cost of performing an INR
test was only $5 or if only 6 INRs per year were required to monitor warfarin
therapy, then ximelagatran would cost approximately $140 000 per QALY
gained vs warfarin. If the INR test cost $50 or if 24 INRs per year were required,
then ximelagatran would cost approximately $70 000 per QALY gained. Neither
cost nor frequency of liver function monitoring had a significant impact on
Age. We considered patients aged 65 to 90 years.
The quality-adjusted survival of a 65-year-old patient was 11.1 years with
ximelagatran, and the cost per QALY with ximelagatran vs warfarin was $100 020.
The quality-adjusted survival of a 90-year-old was only 3.49 years with ximelagatran,
3.47 years with warfarin, and 3.34 years with aspirin. In nonagenarians, ximelagatran
cost more than $150 000 per QALY compared with warfarin, and warfarin
cost $8400 per QALY compared with aspirin.
Costs. The costs and utilities of adverse events,
including major and minor stroke, major and minor hemorrhage, TIA, liver failure,
and elevation of liver function parameters did not significantly affect cost-effectiveness.
In contrast, ximelagatran would cost less than $50 000 per QALY compared
with warfarin if its annual price was less than $1100 and less than $75 000
per QALY if its annual price was $1300 or less.
2- and 3-Way Sensitivity Analyses. We examined
2- and 3-way sensitivity analyses of key variables. These analyses confirmed
that low stroke rates favored aspirin therapy and that high ICH risks favored
ximelagatran (Figure 3). The most important
3-way sensitivity analysis corroborated the relevance of variability in utilities
for patients at moderate risk of stroke (4.5% per year of aspirin) and high
ICH risk (1.2% per year of warfarin). On average, ximelagatran was cost-effective
($44 000 per QALY) in these patients (Figure
3), but when we examined the 70 patients whom we previously surveyed,45 ximelagatran cost $75 000 or less per QALY in
44 of them; the other 26 had high utilities for warfarin.
To assess the combined effect of the precision of the model variables,
we compared ximelagatran with warfarin in Monte Carlo simulations. In the
base case, ximelagatran had a greater quality-adjusted survival 49% of the
time and warfarin was better 51% of the time. For patients with a higher ICH
risk, ximelagatran had a greater quality-adjusted survival 52% of the time.
We found that switching therapy for a patient with atrial fibrillation
and low bleeding risk from warfarin to ximelagatran would increase survival
modestly (0.12 QALY) at a substantial cost. In the base case, the cost per
QALY gained would be $116 000, which exceeds the usual limits for a cost-effective
therapy.67 However, in sensitivity analyses,
we identified 2 important subgroups of patients with atrial fibrillation for
whom ximelagatran would be cost-effective: patients whose utility for warfarin
was low (<0.97) and patients whose risk of ICH was greater than 1.0% per
year. For these patients, ximelagatran is likely to cost less than $50 000
Previous work has shown that the major risk factors for ICH are hypertension,
tobacco use, high fall risk, advanced age, prior stroke, prior bleeding, white-matter
hyperintensities (leukoaraiosis) on brain imaging, and neuropsychiatric impairment
(eg, dementia).67- 70 Patients
with 2 or more of these risk factors are likely to have an ICH risk exceeding
1% per year of warfarin therapy.71 In these
patients, ximelagatran would be cost-effective compared with warfarin (assuming
that ximelagatran does reduce the risk of ICH compared with warfarin). Presently,
many patients with these factors take aspirin (or avoid antithrombotic therapy),8 and they have the greatest potential gain from ximelagatran—almost
1.0 QALY compared with aspirin therapy.
Estimating the risk of ICH is one way clinicians could identify patients
who may derive benefit from ximelagatran. Quantifying the utility of warfarin
is another way. In patients with a low risk of ICH, ximelagatran was cost-effective
in 12 (17%) of 70 patients, all of whom rated the utility of warfarin to be
less than 0.97. Clinicians could estimate their patients’ preferences
qualitatively or quantify them formally using the standard gamble, the time-tradeoff
method,45 or a decision aid.72,73
Although risk of ICH and patient preference for warfarin were key determinants
of the tradeoff between ximelagatran and warfarin, patient stroke risk was
relatively unimportant, except that at the lowest stroke rate, neither anticoagulant
was cost-effective compared with aspirin. At all other stroke rates, warfarin
and ximelagatran were cost-effective compared with aspirin.
The cost-effectiveness of ximelagatran was sensitive to the effectiveness
of ximelagatran vs warfarin in stroke prevention (Figure 2). Taken together, SPORTIF III and V found equal stroke
rates in patients with atrial fibrillation randomized to receive either ximelagatran
or warfarin.15,16 Thus, we assumed
equal effectiveness in the decision model. Individually, however, the 2 trials
had conflicting results. In SPORTIF III, ximelagatran had a 29% lower rate
of ischemic stroke or systemic embolism15;
in SPORTIF V, which was double-blind, warfarin had a 28% lower rate of ischemic
stroke or systemic embolism.16 In sensitivity
analysis, if ximelagatran reduced the rate of ischemic stroke by more than
6% vs warfarin, as is possible with poor warfarin management, then ximelagatran
would cost less than $50 000 per year.
The probability of permanent hepatic failure during ximelagatran use
also significantly affected cost and benefits. Three possible hepatic deaths
of 6948 ximelagatran-treated patients have been reported from clinical trials.17 However, trial participants were ideal candidates
for ximelagatran therapy—patients with known liver disease were excluded.
In clinical practice, where patients are less carefully selected and monitored,
the risk of liver failure is probably greater.
Ximelagatran has been approved for use in several European countries
(primarily for short-term use in the prevention of venous thromboembolism
in orthopedic surgery) but was denied US Food and Drug Administration (FDA)
approval in October 2004. After reviewing the SPORTIF trials and other data,
the FDA Advisory Committee concluded that the risks of ximelagatran outweigh
the benefits.74 We found that ximelagatran
could improve the quality-adjusted survival of patients who had a low utility
for warfarin or a high bleeding risk at an acceptable cost, but this assumes
that it is prescribed only in patients with low risk of hepatotoxicity and
is carefully monitored. Not only would patients with alcoholism or liver disease
need to avoid ximelagatran, but so would patients with impaired renal function
because the active metabolites of ximelagatran are renally cleared.
Our analysis has a number of important potential limitations. The efficacies
used in the base case are based on randomized, controlled clinical trials
in which compliance, monitoring, and follow-up are better than in general
clinical practice. Therefore, we may have overestimated the benefits of both
anticoagulants, because lapses in compliance are common, especially among
the very elderly.75 Additionally, because ximelagatran
has a shorter half-life than warfarin or aspirin, noncompliant patients taking
ximelagatran may be especially susceptible to lapses in adherence. A final
limitation is our extrapolation of results from clinical trials lasting only
1 to 3 years. Rates of adverse events may vary over the long term.
In conclusion, although we found that ximelagatran can increase quality-adjusted
survival compared with warfarin, the increment is modest in patients at low
risk of ICH. At a price of $1700 per year or greater, ximelagatran would cost
more than $100 000 per QALY compared with warfarin in patients with atrial
fibrillation who have a moderate risk of stroke and low ICH risk. Our study
yielded these findings despite assuming equal effectiveness in stroke prevention
and a decreased ICH risk with ximelagatran therapy over 20 years. Unless ximelagatran
costs less than $1100 per year, it will be cost-effective only in patients
with very low utility for warfarin (<0.97) or in patients at high risk
of ICH. If additional randomized trials with ximelagatran are conducted in
the atrial fibrillation population, they should preferentially recruit these
Corresponding Author: Brian F. Gage, MD,
MSc, General Medical Sciences, Washington University School of Medicine, Campus
Box 8005, 660 S Euclid Ave, St Louis, MO 63110 (email@example.com).
Author Contributions: Drs O’Brien and
Gage had full access to their Markov model and the data they collected and
had access to summary data from the SPORTIF trials. Both authors take responsibility
for the integrity of the data and the accuracy of the data analysis.
Study concept and design: O’Brien, Gage.
Acquisition of data: O’Brien, Gage.
Analysis and interpretation of data: O’Brien,
Drafting of the manuscript: O’Brien,
Critical revision of the manuscript for important
intellectual content: O’Brien, Gage.
Statistical analysis: O’Brien, Gage.
Obtained funding: Gage.
Study supervision: Gage.
Financial Disclosures: None reported.
Funding/Support: This study was supported by
Agency for Healthcare Research and Quality grant R01 HS10133 and by the Aetna
Quality Care Research Fund.
Role of the Sponsors: The study’s sponsors
had no role in the design and conduct of the study, in the collection, analysis,
and interpretation of the data, and in the preparation, review, or approval
of the manuscript.
Acknowledgment: We thank Deepak Voora, MD,
and Paul Milligan, RPh, who critiqued an early version of the manuscript.