Risk of Hemorrhage and Treatment Costs Associated With Warfarin Drug Interactions in Patients With Atrial Fibrillation

Clinical Therapeutics/Volume 34, Number 7, 2012

Risk of Hemorrhage and Treatment Costs Associated With Warfarin Drug Interactions in Patients With Atrial Fibrillation Dong-Churl Suh, MBA, PhD1; Winnie W. Nelson, PharmD, MS2; Jiyoon C. Choi, PharmD2; and Insun Choi, PhD3 1

Chung-Ang University College of Pharmacy, Seoul, South Korea; 2Janssen Scientific Affairs, Raritan, New Jersey; and 3Rutgers University School of Pharmacy, Piscataway, New Jersey

ABSTRACT Background: Drug interactions with warfarin are common and may be responsible for increased patient morbidity and treatment costs. Objectives: To assess the usage patterns of drugs that potentiate warfarin’s anticoagulant activity and discuss their associated relationship with both risk of hemorrhage and treatment costs among warfarin users with atrial fibrillation (AF). Methods: A nested case– control study of long-term warfarin-treated AF patients was conducted using a health insurance claims database. Patients with a hemorrhagic event (cases) were matched to control patients using the incidence density sampling method. Drugpotentiating warfarin effects were identified within 30 days before the hemorrhagic event. Conditional logistic regression was used to calculate the association between use of potentiating drugs and hemorrhage risk. Mean treatment costs and CIs were calculated using the bootstrap method and tested using the t-test. Factors associated with treatment costs were determined using generalized linear models with the log-link function and ␥ distribution. Results: Approximately 80% of AF patients were prescribed at least 1 warfarin-potentiating medication while taking warfarin. Patients who used these medications had a 26% higher risk of hemorrhage compared with those who did not use these drugs. Likelihood of hemorrhagic events was significantly increased with the use of potentiating drugs from the following therapeutic classes: anticoagulants (odds ratio [OR] ⫽ 1.91), anti-infectives (OR ⫽ 1.76), antiplatelets (OR ⫽ 1.56), and analgesics (OR ⫽ 1.33). The risk also increased when patients took ⱖ3 therapeutic classes of interacting medications (OR ⫽ 1.62–1.85). Among patients with a hemorrhagic event, patients who were prescribed potentiating drugs had higher hemorrhagerelated treatment costs ($1359) compared with those

July 2012

patients without prescriptions for warfarin-potentiating drugs ($691; P ⬍ 0.001). Conclusions: Warfarin-potentiating drugs were commonly used among AF patients on warfarin. The use of potentiating drugs increased the risk of a hemorrhage, leading to higher treatment costs. More frequent monitoring or alternative anticoagulant therapies are needed to avoid frequent warfarin drug interactions. (Clin Ther. 2012;34:1569–1582) © 2012 Elsevier HS Journals, Inc. All rights reserved. Keywords: atrial fibrillation, hemorrhage, treatment costs, warfarin

INTRODUCTION The cost of treating atrial fibrillation (AF) in the United States is estimated at $6.7 billion per year.1 Warfarin is most commonly used in patients with AF for stroke prevention, and its efficacy has been well established.2 However, the usefulness of warfarin therapy is limited because of its narrow therapeutic index and widely variable patient response, which can make patient management challenging. Warfarin has been shown to interact with a large number of medications, and the majority of warfarin patients are prescribed these medications concomitantly with warfarin.3,4 Thus, warfarin is frequently cited as one of the leading drugs involved in serious drug adverse events.5 Adverse events are mostly hemorrhage-related and can range from minor bleeds, such as epistaxis, to more serious major bleeds, such as gastrointestinal (GI) hemorrhages. An increased risk of bleeding is expected from warfarin use alone,6 not taking into account specific patient factors. For most patients prescribed warAccepted for publication May 23, 2012. http://dx.doi.org/10.1016/j.clinthera.2012.05.008 0149-2918/$ - see front matter © 2012 Elsevier HS Journals, Inc. All rights reserved.

1569

Clinical Therapeutics farin, comorbidities and interacting medications are unavoidable. Patient factors that may affect therapy and alter the risk of bleeding must be given special consideration.7–9 Although minor bleeding events may merely inconvenience most patients and require little to no treatment, major bleeding events are significantly more dangerous, difficult to treat, and frequently require patient hospitalization.10,11 The economic burden of hemorrhages because of the use of warfarin-potentiating drugs is considerable in the United States. In particular, warfarin-related major hemorrhagic events are responsible for a significant portion of the total AF-related treatment costs, with studies reporting mean costs as high as $10,000 to $15,000 per event, depending on the nature and severity of the hemorrhage.12–14 A specialized management approach has been shown to be cost-effective, because of a reduction in both the incidence of stroke and the frequency of adverse events.15 Anticoagulation therapy and the safety issues surrounding its use are being re-examined in light of the recent Joint Commission’s call to action to “reduce the likelihood of patient harm associated with the use of anticoagulation therapy,” as stated in the National Patient Safety Goals.16 The clinical effects associated with warfarin drug interactions are already well recognized; however, there is little information concerning the effect of warfarin drug interactions on treatment costs. Assessing these costs will help to fully characterize the consequences of warfarin therapy and may help to justify additional expenditures and efforts to optimize anticoagulant therapy. By identifying prescribing patterns and patient risk in the real-world setting, prescribers may be able to focus on specific groups of medications that may warrant increased vigilance, rather than rely primarily on medication alert systems. The objectives of this study were to determine the proportion of patients prescribed long-term warfarin therapy together with ⱖ1 warfarin-potentiating medications, to examine the association between the use of potentiating drugs and the risk of hemorrhagic events, and to estimate hemorrhage-related treatment costs. We also sought to identify specific patient factors that influenced treatment costs.

METHODS Data Source A nested case– control study was conducted using data from the Medtstat MarketScan from January 1,

1570

2004 to September 30, 2009. This database, which holds longitudinal information for patients from several employers geographically distributed throughout the United States, includes patients receiving insurance coverage from large self-insured employers, smaller health insurance plans, and Medicare supplemental insurance paid for by employers, including services provided under Medicare-covered payment, employerpaid portion, and any out-of-pocket expenses. The database includes paid claims for prescription medications, medical services and procedures with associated diagnosis codes, and demographic characteristics on all patients. Information pertaining to prescription drugs includes date dispensed, drug quantity, and intended days’ supply. Diagnoses corresponding to medical encounters were identified using the International Classification of Diseases-9-Clinical Modification (ICD-9-CM) system for diagnoses and hospital-based procedures, whereas the Current Procedural Terminology diagnostic coding system was used for procedures performed in the outpatient setting.17

Study Design and Population Inclusion and Exclusion Criteria

Eligible warfarin users ⱖ18 years old, with at least 2 medical encounters resulting in an AF diagnosis between January 1, 2005, and June 30, 2008, were identified for inclusion in the study. During this period, identified patients were required to have initiated warfarin therapy and to have had at least 1 additional prescription for warfarin within 6 months after the initial warfarin prescription. This algorithm was used to include only AF patients who were treated with long-term warfarin. Patients were also required to have continuous medical and prescription insurance coverage for 1 year before warfarin initiation (ie, history period) and during the entire follow-up period. Patients were excluded if they were treated with warfarin during the history period, or if they had a cardioversion performed during the history period because they were presumed to have received anticoagulation therapy previously. Patients were further excluded if they had any of the following conditions during the study period: malignancy (ICD-9 codes 140 –239 or neoplastic medications); pregnancy (ICD-9 codes 630 – 677, V22, V23, V24, V27, or V72.4); immunodeficiency (ICD-9 codes 042, 043, 044, 079.53, 279.19, 795.71, 795.8, v08, 996.8, V42, V43, V49.83, V58.44, or E878 or the medications cyclosporine, tacrolimus, sirolimus,

Volume 34 Number 7

D.-C. Suh et al.

Table I. International Classification of Disease-9th Revision-Clinical Modification (ICD-9-CM) codes used to detect hemorrhage. Events Intracranial hemorrhage Gastrointestinal hemorrhage*

Other hemorrhage*

ICD-9-CM Code(s) 430–432.9 455.2, 455.5, 455.8, 456.0, 456.20, 459.0, 530.7, 530.8, 531.00–.01, 531.20–.21, 531.40–.41, 531.60–61, 532.00–.01, 532.20–.21, 532.40– .41, 532.60–.61, 533.00–.01, 533.20–.21, 533.40–.41, 533.60–.61, 534.00–.01, 534.20–.21, 534.40–.41, 534.60–.61, 535.01, 535.11, 535.21, 535.31, 535.41, 535.51, 535.61, 537.8, 562.02–.03, 562.12– .13, 568.81, 569.3, 569.85, 578.0, 578.1, 578.9 719.1, 786.3, 423.0, 593.81, 784.7, 784.8, 599.7, 623.8, 626.2, 626.6

*Primary diagnosis only.

mycophenolate, daclizumab, basiliximab, or azathioprine); alcohol or drug abuse (ICD-9 codes 303–305); abnormal coagulation profile (ICD-9 codes 790.92 or 286.9); or trauma (ICD-9 codes 800 – 839 or 849 –929).

event.16 –18 Using this methodology, a patient who was initially a control patient could later become a study patient if that patient subsequently experienced a hemorrhage and, therefore, could be matched with ⬎1 study patient.

Study Population

Medications Potentiating Warfarin Effects

Study patients were those who had a hemorrhagic event at any point between the warfarin initiation date and the end of the follow-up period. The date of the first hemorrhagic event was defined as the index date. Control patients were those patients who did not experience a hemorrhagic event until the date of the initial hemorrhagic event for the corresponding study patient. Control patients were matched to study patients by hemorrhagic events during the 1-year history period by age ⫾ 5 years, gender, daily dose of warfarin, days of warfarin exposure, previous hospitalizations or emergency room visits, number of outpatient visits, CHADS2 score (a simple mnemonic score estimating the risk of stroke, based on the presence of Congestive heart failure, Hypertension, Age ⱖ75, Diabetes mellitus, and previous Stroke), and specific comorbidities, including stroke, cardiovascular disease, hypertension, anemia, diabetes, and GI disorders. Each study patient was matched to a maximum of 4 control patients using the incidence density sampling method. When ⬎4 possible control matches were available, 4 were randomly selected from the cohort of at-risk patients on the study patient’s index date. The incidence density sampling method treated patients as control patients until the occurrence of a hemorrhagic

July 2012

Based on published drug interaction references,19,20 we identified 223 medications that potentiated warfarin’s effects. The interactions were assessed by identifying agents that either increased the anticoagulant effect of warfarin or increased the risk of bleeding through alternate mechanisms. Identified medications were grouped into 21 therapeutic classes based on the classification system used in the American Hospital Formulary Service (AHFS) Drug Information database.21 Concomitant exposure to warfarin and the interacting agent was defined as a prescription for a potentiating medication in which at least 1 of the supply days overlapped with warfarin use during the 30-day period before the hemorrhagic event. Hemorrhagic types were classified into 3 categories: intracranial hemorrhage, GI hemorrhage, and other hemorrhages (as listed in Table I). To ensure comparable results with other studies, ICD-9-codes for hemorrhagic events were selected using codes validated in a previous study.22 In this previous study, ICD-9 codes for hemorrhagic events were validated by reviewing the patient’s charts, specifically looking for laboratory results, imaging studies, or physician documentation backing the administrative hemorrhage claim. These ICD-9 codes were also sim-

1571

Clinical Therapeutics ilar to those employed by other studies that investigated the risk of hemorrhage.19,20

Statistical Analysis Differences in patient demographic characteristics between matched study patients and controls were tested using the paired t-test for continuous variables, and McNemar’s ␹2 test and Cochran’s Q test for categorical variables.23 Conditional logistic regression was used to calculate the odds of a hemorrhagic event after adjusting for independent study variables. This odds ratio (OR) can be interpreted as an unbiased estimate of the rate ratio because of the incidence density sampling method used in a nested case– control study design.23–25 Treatment costs were calculated during the 3-month time frame after the occurrence of the index hemorrhagic event. Among matched patients (N ⫽ 3228), 1752 were available for cost calculation— after the exclusion of control patients who had a hemorrhagic event within 3 months after their match date with a study patient (n ⫽ 754) and those who could not be observed for at least 3 months after the index hemorrhagic event (n ⫽ 722). The incidence density sampling method classified patients as controls until a hemorrhage occurrence, and then it subsequently classified that patient as a study patient if that patient later experienced a hemorrhage.23,24 To eliminate the inclusion of hemorrhage-related treatment costs incurred by control patients who experienced a hemorrhage within 3 months of their being matched with a study patient, all patients identified using these exclusionary criteria were removed from the cost calculations, along with their corresponding matched study and control patient(s). Total costs were composed of hemorrhage-related costs, cardiovascular disease (CVD)-related costs, and other treatment costs. Hemorrhage-related costs were those costs associated with medical services involved in hemorrhage treatment (listed in Table I) and the prescription drug costs for hemostatic agents. CVD costs were the costs incurred with CVD diagnoses based on ICD-9 codes 390 – 459 and prescribed cardiovascular drug costs (AHFS code 24:00). Other treatment costs were those costs not included in hemorrhage-related costs and CVD costs. All costs were adjusted to 2009 US dollars using the consumer price index. Costs associated with hemorrhage were calculated for observable study patients (n ⫽ 520). The factors

1572

influencing hemorrhage-related costs were identified using a generalized linear model with log-link function and ␥ distribution, adjusting for age, gender, use of potentiating medications, therapeutic class of potentiating medications, number of different potentiating therapeutic classes used, number of outpatient visits, warfarin daily dose, and presence of anemia. The measures of goodness-of-fit of several models with distinct combinations of the study variables were compared with the saturated model, and the best fit of the multivariable model was selected based on the Akaike information criterion and ⫺2 log likelihood ratio.26 –28 A CI for the mean cost estimate was calculated using the bootstrap method.29 Sensitivity analyses were performed to assess the robustness of the study results. First, we changed the period of time used to identify drugs that potentiated warfarin’s effects from the base case of 30 days before the index hemorrhagic event to both 7 and 14 days. In the second sensitivity analysis, we included patients who underwent cardioversion in our eligible study cohort, reassessing the change in risk from interacting medications and any changes in hemorrhage-related treatment costs. All statistical analyses were conducted using SAS version 9.2 software (SAS Institute Inc., Cary, North Carolina).

RESULTS The study cohort consisted of 7971 eligible patients. Of these patients, 744 patients who experienced a hemorrhage were matched with 2484 patients without hemorrhage on the date of each patient’s initial hemorrhagic event (Figure 1). Subsequent to matching, there were no significant differences in patient demographic characteristics between the study and control patient groups. The average age and daily warfarin dose were similar between the 2 groups: the mean age was ⬃70 years and the mean daily dose was 2.8 mg/d (Table II).

Usage Patterns of Warfarin-potentiating Drugs Table III presents the usage patterns of warfarinpotentiating drugs and the OR for hemorrhagic events. The proportion of patients who used warfarin-potentiating drugs was high among both patient groups. A greater proportion of patients with a hemorrhagic event used a potentiating drug (83.3%) compared with the patients without a hemorrhage (79.1%). Among these potentiating agents, the most commonly used agents were lipid-lowering agents, followed by antihy-

Volume 34 Number 7

D.-C. Suh et al.

Patients who had two or more AF diagnoses and had at least one additional prescription for warfarin within 6 months after the first warfarin prescription between 1/1/2005 and 06/30/2008. N = 198,697 Exlcude patients who did not have continuous medical and prescription insurance coverage from 1 year before the index date (the fist hemorrhage) to at least 1 year after the index date. N = 111,943 Exlcude patients who had intervion such as cardioversion during the history period. N = 4263 Exlcude patients who underwent warfarin treatment during the history period. N = 3125 Exlcude patients who had one of the following conditions during the study period: 1. Malignancy (ICD-9 code 140-239) or used anti-neoplastic agents (n = 47,989) 2. Pregnancy (n = 119) 3. Immunodeficiency disease, organ transplant, or used a systemic immunosuppressive therapy (n = 6711) 4. Alochol or drug abuse (n = 1033) 5. Abnormal coagulation profile or other unspecifeid coagulation defects (n - 2254) 6. Trauma (ICD-9 code 800-839 or 849-929; n = 8133) N = 66,239 Exlcude patients who did not have a diagnosis of AF before the initiation of warfarin. N = 5156 Eligible patients N = 7971 Matched patients Patients with hemorrhage (n = 744) and patients withour hemorrhage (n = 2484) N = 3228 Excluded control patients with hemorrhagic events within 3 months after matching (n = 754) and patients who could not be observed for at least 3 months after the index date (n = 722). N = 1476 Patients used for cost calculations N = 1752

Figure 1. Patient selection flowchart. AF ⫽ atrial fibrillation; ICD-9 ⫽ International Classification of Diseases-9th Revision.

pertensive agents, GI drugs, analgesics, and anti-infectives. Additionally, 56% of study patients with hemorrhages were found to be using multiple potentiating medications from different therapeutic classes. Patients who used any potentiating medications had a 26% higher risk of hemorrhage than those patients who did not these medications. The following therapeutic classes of medications had the most significant effect on odds of hemorrhage occurrence: anticoagulants (OR ⫽ 1.91; 95% CI, 1.20 –2.41), anti-infectives (OR ⫽ 1.76; 95% CI, 1.39 –2.13), antiplatelets (OR ⫽ 1.56; 95% CI, 1.18 –1.68), and analgesics (OR ⫽ 1.33; 95% CI, 1.07–2.24). When patients were taking mul-

July 2012

tiple potentiating medications from different therapeutic classes, the risk of hemorrhage increased. If patients were using ⬎3 warfarin-potentiating classes of drugs, the likelihood of hemorrhagic events increased by 62% to 85% (OR ⫽ 1.62 for 3 therapeutic classes to OR ⫽ 1.85 for ⱖ4 therapeutic classes).

Total Mean Treatment Costs The total mean treatment costs (with a 95% CI) for patients with and without hemorrhages are shown in Figure 2. Patients who experienced a hemorrhage had mean treatment costs of $7493, whereas control patients (without hemorrhage) incurred total mean treat-

1573

Clinical Therapeutics

Table II. Characteristics of warfarin users with atrial fibrillation after matching.* Characteristics Total patients Age, mean (SD), y 18–⬍65 65–⬍75 75–⬍85 ⱖ85 Gender Male Female Type of hemorrhage during history period Intracranial hemorrhage GI bleeding Other bleeding Warfarin daily dose, mean (SD)‡ ⱕ5 mg/d ⬎5–7 mg/d ⬎7 mg/d

Patients With Hemorrhage No. (%)

Patients Without Hemorrhage No. (%)

744 (100.0) 70.2 (11.2) 249 (33.5) 191 (25.7) 245 (32.9) 59 (7.9)

2484 (100.0) 69.7 (11.1) 851 (34.3) 616 (24.8) 856 (34.5) 161 (6.5)

P† 0.938 0.487

1.000 413 (55.5) 331 (44.5) 68 (9.1) 4 (0.5) 25 (3.4) 48 (6.5) 2.8 (1.4) 710 (95.4) 32 (4.3) 2 (0.3) 577 (77.6)

1382 (55.6) 1102 (44.4) 184 (7.4) 9 (0.4) 77 (3.1) 104 (4.2) 2.8 (1.4) 2402 (96.7) 79 (3.2) 3 (0.1) 1913 (77.0)

237 (31.9) 129 (17.3) 378 (50.8)

865 (34.8) 430 (17.3) 1189 (47.9)

1.000

CHADS2 score 0 1 2–3 ⱖ4

196 (26.3) 274 (36.8) 230 (30.9) 44 (5.9)

738 (29.7) 952 (38.3) 693 (27.9) 101 (4.1)

1.000

Comorbidities Stroke Cardiovascular disease§ Hypertension Anemia Diabetes GI disorders

86 (11.6) 273 (36.7) 352 (47.3) 65 (8.7) 115 (15.5) 105 (14.1)

220 (8.9) 856 (34.5) 1111 (44.7) 140 (5.6) 307 (12.4) 264 (10.6)

1.000 1.000 1.000 1.000 1.000 1.000

Hospitalization or emergency room visit No. of outpatient visits 0–6 7–12 ⱖ13

1.000

0.498 1.000

1.000

CHADS2 ⫽ a simple mnemonic score estimating the risk of stroke, based on the presence of Congestive heart failure, Hypertension, Age ⱖ75, Diabetes mellitus, and previous Stroke; GI ⫽ gastrointestinal. *Patients with and without hemorrhage were matched using history of hemorrhage, age, gender, history of hemorrhage, use of medical services, CHADS2 scores and comorbidity. † P value was calculated using McNemar’s ␹2, Cochran’s Q test, or paired t-test based on the type of variables. ‡ Average daily dose of warfarin was calculated by dividing the sum of warfarin doses (ie, strengths per unit multiplied by number of days supplied for that unit) by total days’ supply. § Cardiovascular diseases included myocardial infarction, ischemic heart disease, and congestive heart failure.

1574

Volume 34 Number 7

D.-C. Suh et al.

Table III. Usage patterns of drugs potentiating warfarin effects and odds ratios (ORs) for hemorrhagic events.* Characteristics

Patients With Hemorrhage (n ⫽ 744)

Patients Without Hemorrhage (n ⫽ 2484)

Total Patients

No. (%)

No. (%)

OR (95% CI)

P

Use of potentiating medications Nonusers Any potentiating medication users

124 (16.7) 620 (83.3)

518 (20.9) 1966 (79.1)

1 1.26 (1.00–1.57)

0.047

Therapeutic classes Lipid-lowering agents Antihypertensive agents Gastrointestinal drugs Analgesics Anti-infectives Thyroids and antithyroids Antiplatelets Antidepressants Antiarrhythmics Anticoagulants Others‡

273 (36.7) 226 (30.4) 168 (22.6) 164 (22.0) 141 (19.0) 117 (15.7) 95 (12.8) 93 (12.5) 42 (5.6) 32 (4.3) 74 (10.0)

777 (31.3) 766 (30.8) 520 (20.9) 421 (16.9) 292 (11.8) 414 (16.7) 195 (7.9) 238 (9.6) 137 (5.5) 67 (2.7) 220 (8.7)

1.18 (0.98–1.06) 0.87 (0.71–1.19) 0.96 (0.78–1.65) 1.33 (1.07–2.24) 1.76 (1.39–2.13) 0.89 (0.70–2.07) 1.56 (1.18–1.68) 1.28 (0.97–1.55) 1.07 (0.74–3.04) 1.91 (1.20–2.41) 1.04 (0.77–1.10)

0.081 0.162 0.714 0.01 ⬍.001 0.335 0.002 0.08 0.725 0.007 0.783

No. of therapeutic classes None 1 2 3 ⱖ4

124 (16.7) 205 (27.6) 172 (23.1) 142 (19.1) 101 (13.6)

518 (20.9) 787 (31.7) 612 (24.6) 349 (14.0) 218 (8.8)

1 1.06 (0.82–1.48) 0.652 1.12 (0.86–2.15) 0.398 1.62 (1.21–2.55) 0.001 1.85 (1.34–2.36) ⬍0.001

OR of Hemorrhage†

*Medications were identified during the 30 days before the index date (hemorrhage). † Odds ratios (ORs) were calculated using conditional logistic regression and were calculated for patients who were matched based on the study variables listed in the report. ‡ Others included herbals/supplements, anticholinergics, anti-Parkinson agents, anticonvulsants, antidiabetics, steroid hormones, stimulants, stimulants, antigout agents, and leukotriene inhibitors.

ment costs of $4151, representing 80.5% higher costs for patients who experienced a hemorrhage. The mean cost of treating the hemorrhage itself was $1247, with the remaining costs composed of CVD and other costs. Specific amounts for CVD and other costs were also higher among patients with hemorrhage compared with controls ($3385 vs $2623 and $2861 vs $1525, respectively). Table IV presents the treatment costs associated with hemorrhages. Patients who used warfarin-potentiating medications had higher mean hemorrhage-related costs ($1359) compared with patients without a drug interaction ($691). These costs were entirely composed of medical costs, with no prescription costs attributable to hemorrhage treatment. The cost ratios for factors associated with hemorrhage-related treatment costs are shown in Table V.

July 2012

Patients ⬎85 years of age had 67% higher adjusted hemorrhage-related costs compared with patients ⬍65 years of age. Patients who used any potentiating drugs had 67% higher adjusted hemorrhage-related treatment costs than patients without any potentiating drugs. Female patients had 40% higher hemorrhage-related costs than male patients (cost ratio ⫽ 1.40; 95% CI, 1.05–1.89). Patients taking a dose of warfarin ⬎7 mg/d experienced a decrease in costs by 88% compared with patients taking ⱕ5 mg/d (cost ratio ⫽ 0.12; 95% CI, 0.02– 0.93). Patients who had 7 to 12 outpatient visits per year had 49% higher hemorrhage-related costs than those patients who had 0 to 6 outpatient visits per year. Patients who had anemia had hemorrhage-related treatment costs that were almost 3 times higher than those for pa-

1575

Clinical Therapeutics

$8000

$7493 ($6626–$8555)

$7000

$1247 ($1028-$1540)

$6000

NA

$5000

$3385 ($2769–$4206)

$4000

$4151 ($3564–$5224)

P = 0.129

$2623 ($2117–$3681)

$3000 $2000

$2861 ($2403–$3614)

$1000 $0

P < 0.001

$1525 ($1336–$2077)

Patients with Hemorrhage (N = 520)

Patients without Hemorrhage (N = 1232)

Bleeding Costs

CVD Costs

Other Costs

Figure 2. Total treatment costs for patients with versus without hemorrhage (mean and 95% confidence interval).

tients without anemia (cost ratio ⫽ 2.96; 95% CI, 1.86 – 4.72).

Sensitivity Analyses Two sensitivity analyses were performed to assess the robustness of the study design. The first analysis

tested different observation periods used to detect the use of potentiating medications before the index hemorrhagic event, assessing differences in the risk of hemorrhagic events by the number of therapeutic classes. The range of ORs between the number of therapeutic classes was similar for the base case (OR ⫽ 1.06 –1.85),

Table IV. Treatment costs associated with hemorrhage.* All Patients With Hemorrhage (n ⫽ 520)

Patients Using WarfarinPotentiating Drugs (n ⫽ 433)

Patients Not Using Warfarin-Potentiating Drugs (n ⫽ 87)

Costs

Mean

Bootstrap (95% CI)

Mean

Bootstrap (95% CI)

Mean

Bootstrap (95% CI)

P†

Total costs Medical costs Inpatient Outpatient Emergency visit

$1247

($1028–$1540)

$1359

($1097–$1718)

$691

($512–$961)

⬍0.001

$690 $483 $74

($491–$977) ($413–$598) ($55–$107)

$809 $472 $77

($566–$1157) ($398–$609) ($57–$116)

$96 $537 $57

($40–$193) ($381–$765) ($13–$217)

⬍0.001 0.588 0.625

Prescription costs

$0

($0–$0)

$0

($0–$1)

$0

($0–$0)

0.549

*Hemorrhage-associated treatment costs included costs associated with services rendered in each medical facility with a diagnosis for hemorrhage and costs for hemostatic agents dispensed during the 3-month period after the index hemorrhage. Costs were adjusted for 2009 values using the 2009 price index. † P value represented differences in average hemorrhage-associated costs between patients with and without drug interactions.

1576

Volume 34 Number 7

D.-C. Suh et al.

Table V. Factors associated with hemorrhage-related treatment costs. Adjusted Cost Ratios†

Univariate Cost Ratios Variables*

Cost Ratio

(95% CI)

P

Cost Ratio

(95% CI)

P

Age, y 18–⬍65 65–⬍75 75–⬍85 ⱖ85

1 1.23 1.47 1.98

(0.86–1.75) (1.05–2.06) (1.19–3.28)

0.257 0.026 0.008

1 0.94 0.94 1.67

(0.66–1.35) (0.65–1.35) (1.02–2.73)

0.740 0.741 0.040

Gender Male Female

1 1.59

(1.21–2.09)

0.001

1 1.40

(1.05–1.89)

0.024

Use of potentiating medications Nonusers Any potentiating users

1 1.97

(1.37–2.83)

⬍0.001

1 1.67

(1.17–2.38)

0.004

Therapeutic class of potentiating medications Lipid-lowering agents Antihypertensive agents Analgesics Gastrointestinal drugs Anti-infectives Thyroids and antithyroids Antiplatelets Antidepressants Antiarrhythmics Anticoagulants Others

1.16 1.02 0.94 1.11 0.80 0.99 1.83 0.73 0.74 1.76 0.75

(0.87–1.56) (0.74–1.41) (0.66–1.34) (0.78–1.58) (0.56–1.13) (0.67–1.45) (1.21–2.75) (0.48–1.13) (0.39–1.38) (0.90–3.45) (0.46–1.22)

0.312 0.891 0.746 0.573 0.204 0.950 0.004 0.163 0.338 0.098 0.244

No. of therapeutic classes of potentiating medications None 1 2 3 ⱖ4

1 2.45 1.90 1.38 1.92

(1.61–3.74) (1.23–2.92) (0.87–2.17) (1.18–3.13)

⬍0.001 0.004 0.170 0.009

No. of outpatient visits 0–6 7–12 ⱖ13

1 1.13 0.84

(0.79–1.63) (0.63–1.13)

0.508 0.260

1 1.49 0.93

(1.04–2.14) (0.69–1.25)

0.029 0.614

Warfarin daily dose‡ ⱕ5 mg/d ⬎5–7 mg/d ⬎7 mg/d

1 0.52 0.08

(0.27–1.00) (0.01–0.70)

0.051 0.023

1 0.72 0.12

(0.39–1.33) (0.02–0.93)

0.294 0.043

Comorbidities Anemia

3.05

(1.91–4.88)

⬍0.001

2.96

(1.86–4.72)

⬍0.001

*All variables were identified from the date of warfarin initiation to the hemorrhage event date, except for drug– drug interactions and hemorrhage variables. † Cost ratios were estimated using a generalized linear model with log-link function and ␥ distribution. ‡ Average daily dose of warfarin was calculated by dividing sum of warfarin doses (ie, strengths per unit ⫻ days supplied for that unit) by total days of supply.

July 2012

1577

Clinical Therapeutics the 14-day observation period (OR ⫽ 1.03–1.80), and the 7-day observation period (OR ⫽ 1.04 –2.00). The base-case analysis showed the smallest range between ORs, indicating that the observation periods used in the study produced the most stable results. The second analysis was performed to assess different exclusion criteria by including patients who underwent cardioversion within the study period, resulting in 8618 patients in the study cohort. The ORs of the base-case analysis were not significantly different from those of the cohort, including cardioversion patients (ORs ⫽ 1.01 for 1 therapeutic class, 1.29 for 2 therapeutic classes, 1.63 for 3 therapeutic classes, and 1.90 for 4 therapeutic classes) (data not reported).

DISCUSSION This study found that patients who used warfarin-potentiating drugs had significantly more hemorrhagic events (OR ⫽ 1.26) than patients who did not use these drugs. The majority of study patients (⬃80%) were found to be taking at least 1 medication that potentiated the effects of warfarin. The highest likelihood of hemorrhagic events was found among patients who took anticoagulants (OR ⫽ 1.91), followed by antiinfectives (OR ⫽ 1.76), antiplatelets (OR ⫽ 1.56), and analgesics (OR ⫽ 1.33). This study also found that treatment costs related to hemorrhagic events among patients who used any warfarin-potentiating medication were 43% higher than among those patients who did not use any potentiating drugs, after adjusting for patient demographic characteristics and comorbidities. Previous estimates of the concomitant use of warfarin with interacting medications were consistent with the study’s findings, despite the use of different study designs and inconsistent definitions of interacting drugs.3,4,30,31 Although our study evaluated only potentiating drugs, previous studies that assessed medications that both increased and decreased hemorrhage risk provided similar estimates. The majority of warfarin interactions were reported to increase the risk of bleeding,32 and this might partially account for the consistency of study results. Additionally, because the majority of previous studies focused primarily on hospitalized patients, they did not accurately estimate the rate of concomitant use of warfarin-potentiating medications in an outpatient setting, where patients might not be as carefully monitored as hospitalized patients and might

1578

experience a higher number of interactions. Finally, each study included different warfarin-potentiating drugs, because the list of warfarin medication interactions varies significantly among published sources.33 This study found that the majority of patients using warfarin were also prescribed a drug that interacts with warfarin. The risk of hemorrhage was elevated when multiple potentiating drugs from different therapeutic classes were used together. An incremental increase in risk was observed in patients who were prescribed ⬎1 therapeutic class of potentiating medication. Despite the fact that warfarin is well known for its tendency to interact with many different medications, interactions continue to occur in real-life practice. A prescriber might be unaware of all the potential concomitant interactions of medications prescribed to a specific patient because the patient might be under the care of multiple physicians. Clinicians’ knowledge of potential interactions might also play a role, because it was shown that clinicians’ recognition of interacting medications could be improved by the use of automated drug-interaction alerts.34 Medication alert systems for prescribers and pharmacists are available; these typically send alerts whenever a prescribed interaction is detected. However, given the multitude of possible warfarin interactions, a perceived problem with drug alerts is the incidence of too many false-positive warnings, which can induce “alert fatigue.”35 Furthermore, because anticoagulation therapy is a vital part of AF treatment that significantly improves patient prognosis, the benefit of therapy will often outweigh the risks of bleeding when interactions are noted. Given the large number of potential warfarin interactions and frequent alerts, it is possible that many alerts will ultimately be ignored. Other studies investigated the risks associated with the coadministration of warfarin with various classes of therapeutic agents, and their findings were broadly in line with the results of this study,22,36 – 40 One study, conducted in a Finnish hospital, found that the likelihood of hemorrhage was 2.57 times higher when warfarin was coadministered with a nonsteroidal anti-inflammatory drug (NSAID), and 3.10 times higher when it was administered with a cyclooxygenase-2 inhibitor.36 Overall, this study noted a 62% increase in the risk of hemorrhage when warfarin was administered with an antiplatelet, a percentage that was very close to the 56% increase in risk observed with the

Volume 34 Number 7

D.-C. Suh et al. same class of agents in this study. However, unlike this study, the Finnish study did not find this increase in risk to be significant, possibly because of switching of medications and inaccurate recording.36 The present study found that the likelihood of hemorrhage was not significantly increased (OR ⫽ 1.18; 95% CI, 0.98 –1.06) when warfarin was administered concomitantly with a statin. This result was supported by data from a Canadian nested case– control study, which also found no impact of concomitant statin and warfarin use, when linked administrative databases from Ontario were analyzed (OR ⫽ 0.91; 95% CI, 0.77–1.07).37 Previous studies also agreed with the results of this study, showing that anti-infectives20,38 and analgesics36,39,40 increased the risk of bleeding when taken with warfarin. Both anti-infectives and analgesics are usually prescribed as a relatively short course of treatment. As a result, clinicians might not consider their interaction to be clinically important enough to consider the use of an alternative agent. However, within a short time frame, major changes in a patient’s international normalized ratio (INR) can occur, placing the patient at greater risk of hemorrhage. This phenomenon was observed in a previous study assessing anticoagulation levels among warfarin patients; significantly elevated INR levels were reported among hospitalized patients who began treatment with new drugs, particularly anti-infectives.41 In general, medications taken on a long-term basis with warfarin did not show much change in the risk of hemorrhage,3,37,42 possibly because regular use of a medication was usually associated with a one-time consistent change in INR. Although warfarin has a narrow therapeutic index, changes in INR levels can be managed through a series of dose adjustments. This might also be related to the “healthy user effect,” whereby users of long-term preventive medications, such as antihypertensives or lipid-lowering agents, might be more attentive to personal health risks and, as a result, might show better anticoagulation control. Patients who used warfarin-potentiating drugs had at least 67% higher hemorrhage-related treatment costs compared with patients who did not use such drugs, after adjusting for patient demographic characteristics and comorbidities. This suggested that drug interactions were related to both the incidence and severity of hemorrhagic events. The increase in severity could be related to the difficulty of treating the hemor-

July 2012

rhage, as well as the site of the hemorrhage. For example, a common interaction between NSAIDs and warfarin was shown to specifically increase the risk of GI hemorrhage to a greater extent than other types of hemorrhage.40,43 GI bleeding events are considered major hemorrhagic events and require additional resources to treat.44 Independent of the site of bleeding, the general effect of many of the interactions with warfarin is the elevation of INR levels into supratherapeutic ranges (INR ⱖ 4.0), which is intuitively related to both the incidence and severity of bleeding.45 These results had strong implications for the need to improve both warfarin prescribing practices and therapy management. Currently, the quality of anticoagulation care is suboptimal, because patients were shown to spend a relatively low percentage of their time (55%) within a therapeutic INR range, which places patients at greater risk of either clot formation or bleeding.46 This might influence the rate of major bleeding events among warfarin patients, reported at 2.3% per year, which was somewhat higher than the 1.1% per year rate reported for the alternative option of daily aspirin therapy.7 Careful management of warfarin therapy could help to maintain INR levels and decrease the rate of significant bleeding events, and it could also significantly reduce costs by decreasing the number of hospitalizations and emergency room visits.12,47 However, the requirement for frequent monitoring and management requires additional resources and time. Alternative anticoagulants are available in the form of direct thrombin inhibitors and direct Factor Xa inhibitors.48 Our results provided an estimate of avoidable costs when preventive measures were taken to improve anticoagulation management of AF, and they will help managed care decision makers evaluate the impact of drug interactions when considering new management techniques. Sensitivity analyses were performed to assess the robustness of the study results. In the first sensitivity analysis, the time period used to identify drugs potentiating warfarin before the hemorrhagic event was varied. The base-case analysis assessed the risk of hemorrhage based on potentiating medications used within 30 days before hemorrhage. We changed this observation period to both 7 and 14 days to assess whether our results significantly changed, and we found similar hemorrhage risks to those of the base case. In general, a trend of increasing risk was seen with the use of a higher number of warfarin-potentiating therapeutic classes,

1579

Clinical Therapeutics consistent with clinical expectations. The base-case analysis thus appeared appropriate because the alteration of the observation period did not drastically change the results, and the trends of the ORs were consistent with clinical expectations. In the second sensitivity analysis, we did not exclude the 4263 patients who underwent cardioversion during the history period. The base-case analysis excluded these patients because patients undergoing cardioversion are expected to receive several weeks’ worth of anticoagulation therapy49 and, therefore, patients could not be considered new warfarin users. The results of the base-case analysis of excluding patients who underwent cardioversion procedures did not differ from those obtained when including these patients, supporting the robustness of the original study design. Although this study was conducted using a large number of patients observed from a usual care practice setting, its design was not without some limitations regarding both the estimate of risk and calculation of treatment costs. Because of the nature of claims data, the study was unable to capture the use of any over-the-counter (OTC) agents, of which many might have potentially interacted with warfarin (especially many of the riskier agents, such as the analgesics, aspirin, NSAIDs, and acetaminophen). Thus, we might have underestimated the frequency of the interaction of warfarin and the drugs potentiating its efficacy, but we might have also overestimated the risk of bleeding for other agents because concurrent OTC analgesic use might have increased the risk of bleeding that was erroneously attributed to the prescribed agent identified via claims data. We were also unable to assess the potential effects of both patient diet and warfarin adherence on the effectiveness of warfarin. Furthermore, we did not include variables for anticoagulation control as measured by INR because the information was not available in health insurance claims databases. Although the study was adjusted for patient clinical characteristics and the use of medical services and medications, the adjustment was made with observable variables, which could be obtained via health insurance claims data. Selection of treatment by unobserved severity of the disease was not controlled in the analysis. In addition, the study had an essential selection bias, because physicians might avoid prescribing potential interactive medications

1580

to patients who were at a high risk of bleeding from receiving combination therapy. In this case, the true increase in risk of hemorrhage from combination therapy would be higher than what was observed in our study. Moreover, hemorrhagic events were identified using ICD-9-CM codes, which might have underestimated the true incidence of hemorrhagic events, but the use of ICD-9 codes ensured that each event was of clinical significance. This study only established an association between the use of potentiating agents and hemorrhagic events, so estimates of hemorrhage risk should be cautiously interpreted, given the multifactorial nature of bleeding events. Often, there may be more than one interaction that precipitates an event and, in some cases, that event may be unrelated to the identified interaction, such as that resulting from physical injury. Furthermore, the calculation of treatment costs associated with hemorrhage was based on a short follow-up period of 3 months after the hemorrhagic event. Finally, the treatment costs assessed in this study represented direct medical costs and, therefore, underestimated total societal costs because they were based solely on the amounts actually paid by health insurance companies to health care providers and did not include indirect costs (eg, lost productivity). The database also did not include payment by Medicaid recipients, the uninsured, or out-of-pocket payments.

CONCLUSIONS Patients who used warfarin-potentiating medications had a 26% higher risk of hemorrhage compared with those patients who did not use these drugs. Warfarin drug interactions occurred in approximately 80% of patients included in our study. The risk of hemorrhage was significantly higher among patients who used anticoagulants, antiplatelets, anti-infectives, and analgesics, as well as among those who used multiple warfarin-potentiating medications. In addition, hemorrhage-related treatment costs were significantly higher among patients who used warfarin-potentiating medications ($1359) than among those who did not use these medications ($691). In an effort to decrease patient treatment costs and improve outcomes, prescribers should be vigilant when assessing all medications that patients are prescribed, and they should also be aware of the agents with the strongest warfarin-potentiating effects to minimize risk of patient hemorrhage.

Volume 34 Number 7

D.-C. Suh et al.

ACKNOWLEDGMENTS This study was performed with funding from Janssen Scientific Affairs, LLC. Drs. Suh, Nelson, and J.C. Choi contributed to study design and interpretation; Dr. Suh contributed to literature search and writing, Dr. I. Choi contributed to data analysis.

12.

13.

CONFLICTS OF INTEREST Drs. Suh and I. Choi have no conflicts of interest to declare. Drs. Nelson and J.C. Choi have ownership interests (stock) in Johnson and Johnson, and are employees of Janssen Scientific Affairs, LLC.

REFERENCES 1. Coyne K, Paramore C, Grandy S. Assessing the direct costs of treating nonvalvular atrial fibrillation in the United States. Value Health. 2006;9:348 –356. 2. Hirsh J, Dalen J, Anderson DR, et al. Oral anticoagulants: mechanism of action, clinical effectiveness, and optimal therapeutic range. Chest. 2001;119(Suppl 1):8S–21S. 3. Wittkowsky AK, Boccuzzi S, Wogen J. Frequency of concurrent use of warfarin with potentially interacting drugs. Pharmacotherapy. 2004;24:1668 –1674. 4. Kotirum S, Chaiyakunapruk N, Jampachaisri K. Utilization review of concomitant use of potentially interacting drugs in Thai patients using warfarin therapy. Pharmacoepidemiol Drug Saf. 2006;16:216 –222. 5. Gurwitz JH, Field TS, Harrold LR, et al. Incidence and preventability of adverse drug events among older persons in the ambulatory setting. JAMA. 2003;289:1107–1116. 6. Johnsen S, Sorenson H, Mellemkjoer L, Blot W, et al. Hospitalisation for upper gastrointestinal bleeding associated with use of oral anticoagulants. Thromb Haemost. 2001;86:563–568. 7. Levine MN, Raskob G, Beyth RJ, et al. Hemorrhagic complications of anticoagulant treatment (The Seventh ACCP Conference on Antithrombotic and Thrombolytic Therapy). Chest. 2004;126:287S–310S. 8. Gasse C, Hollowell J, Meier CR, Haefeli WE. Drug interactions and risk of acute bleeding leading to hospitalisation or death in patients with chronic atrial fibrillation treated with warfarin. Thromb Haemost. 2005;94:537–543. 9. Lindh J, Holm L, Dahl M. Incidence and predictors of severe bleeding during warfarin treatment. J Thromb Thrombolysis. 2008;25:151–159. 10. Metlay J, Hennessy S, Localio A. Patient reported receipt of medication instructions for warfarin is associated with reduced risk of serious bleeding events. J Gen Intern Med. 2008;10:1589 –1594. 11. Fihn SD, Callahan CM, Martin DC, et al. The risk for and severity of bleeding complications in elderly patients treated with warfa-

July 2012

14.

15.

16.

17. 18.

19.

20.

21.

22.

23. 24.

25. 26.

rin. The National Consortium of Anticoagulation Clinics. Ann InternMed.1996;124:970–979. Kim M, Metlay J, Cohen A. Hospitalization costs associated with warfarin-related bleeding events among older community-dwelling adults. Pharmacoepidemiol Drug Saf. 2010;19:731–736. Bullano MF, Willey V, Hauch O, et al. Longitudinal evaluation of health plan cost per venous thromboembolism or bleed event in patients with a prior venous thromboembolism event during hospitalization. J Manag Care Pharm. 2005;11:663– 673. Abdelhafiz A, Wheeldon N. Use of resources and cost implications of stroke prophylaxis with warfarin for patients with nonvalvular atrial fibrillation. Am J Geriatr Pharmacother. 2003;1:53– 60. Sullivan PW, Arant T, Ellis S, et al. The cost effectiveness of anticoagulation management services for patients with atrial fibrillation and at high risk of stroke in the US. Pharmacoeconomics. 2006;24:1021–1033. Breslow NE, Day NE. Statistical Methods in Cancer Research, Volume 1—The Analysis of Case-Control Studies. Lyon, France: International Agency for Research on Cancer; 1980. Rothman K, Greenland S. Modern Epidemiology. 2nd ed. Philadelphia, PA: Lippincot Williams & Wilkins; 1998. Etminan M. Pharmacoepidemiology II: The nested casecontrol study-a novel approach in pharmacoepidemiologic research. Pharmacotherapy. 2004;24:1105–1109. Boulanger L, Hauch O, Friedman M, et al. Warfarin exposure and the risk of thromboembolic and major bleeding events among Medicaid patients with atrial fibrillation. Ann Pharmacother. 2006;40:1024 –1029. Zhang K, Young C, Berger J. Administrative claims analysis of the relationshipbetweenwarfarinuseandriskofhemorrhageincluding drug-drug and drug-disease interactions. J Manag Care Pharm. 2006:12:640–648. American Society of Health-System Pharmacists. American Hospital Formulary Service Drug Information. Bethesda, MD: American Society of Health-System Pharmacists; 2009. Arnason T, Wells PS, Walraven Cv, Forster AJ. Accuracy of coding for possible warfarin complications in hospital discharge abstracts. Thrombosis Res. 2006;118:253–262. Agresti A. Categorical Data Analysis. 2nd ed. New York, NY: John Wiley & Sons; 2002. Lubin JH, Gail MH. Biased selection of controls for case-control analyses of cohort studies. Biometrics. 1984;40: 63–75. Langholz B, Goldstein L. Risk set sampling in epidemiologic cohort studies. Statist Sci. 1996;11:35–53. Akaike H. Information theory and an extension of the maximum likelihood principle. In: Petrov BN, Caski F, eds. Proceedings of the Second International Symposium on Information Theory. Budapest: Akademiai Kiado; 1973:267–281.

1581

Clinical Therapeutics 27. Collett D. Modelling Survival Data in Medical Research. New York: Chapman & Hall; 1994. 28. Burnham KP, Anderson DR. Model Selection and Multimodel Inference: A Practical Information-Theoretic Approach. New York: Springer-Verlag; 2002. 29. Efron B, Tibshirani RJ. An Introduction to the Bootstrap. Boca Raton, FL: Chapman & Hall/CRC; 1993. 30. Howard PA, Ellerbeck EF, Engelman KK, Patterson KL. The nature and frequency of potential warfarin drug interactions that increase the risk of bleeding in patients with atrial fibrillation. Pharmacoepidemiol Drug Saf. 2002;11:569 –576. 31. Verhovsek M, Motlagh B, Crowther M. Quality of anticoagulation and use of warfarin-interacting medications in long-term care: a chart review. BMC Geriatr. 2008;8:1– 6. 32. Holbrook A, Pereira J, Labiris R. Systemic overview of warfarin and its drug and food interactions. Arch Intern Med. 2005;165:1095–1106. 33. Anthony M, Romero K, Malone D. Warfarin interactions with substances listed in Drug Information Compendia and in the FDA-approved label for warfarin sodium. Clin Pharmacol Therapeut. 2009;86: 425– 429. 34. Glassman P, Simon B, Belperio P. Improving recognition of drug interactions: benefits and barriers to using automated alerts. Med Care. 2002;40:1161–1171. 35. Peterson J, Bates DW. Preventable medication errors: identifying and eliminating serious drug interactions. J Am Pharm Assoc. 2001;41: 156 –160. 36. Hauta-aho M, Tirkkonen T, Valhlberg T. The effect of drug interactions on bleeding risk associated with warfarin therapy in hospitalized patients. Ann Med. 2009; 41:619 – 628. 37. Douketis JD, Melo M, Bell CM, Mamdani MM. Does statin therapy decrease the risk for bleeding in

1582

38.

39.

40.

41.

42.

43.

patients who are receiving warfarin? Am J Med. 2007;120:369.e369 – 369.e314. Angaran DM, Dias VC, Arom KV, et al. The comparative influence of prophylactic antibiotics on the prothrombin response to warfarin in the postoperative prosthetic cardiac valve patient. Cefamandole, cefazolin, vancomycin. Ann Surg. 1987; 206:155–161. Battistella M, Mamdami MM, Juurlink DN, et al. Risk of upper gastrointestinal hemorrhage in warfarin users treated with nonselective NSAIDs or COX-2 inhibitors. Arch Intern Med. 2005;165:189 –192. Delaney JA, Opatrny L, Brophy JM, Suissa S. Drug drug interactions between antithrombotic medications and the risk of gastrointestinal bleeding. CMAJ. 2007;177:347–351. Siguret V, Esquirol C, Debray M, et al. Excess antivitamin K in elderly hospitalised patients aged over 70: a one-year prospective survery. Presse Medicale. 2003;32:972–977. Anderson R, Divers C, Von Hennigs I. Drug interaction exposure in chronic atrial fibrillation patients maintained on warfarin. J Thromb Thrombolysis. 2005;19:123–124. Garcia Rodriguez L, Tolosa L. Risk of upper gastrointestinal complica-

44.

45.

46.

47.

48.

49.

tions among users of traditional NSAIDs and COXIBs in the general population. Gastroenterology. 2007; 132:498 –506. Long A, Bendz L, Horvath M, et al. Characteristics of ambulatory anticoagulant adverse drug events: a descriptive study. Thromb J. 2010;8:1– 8. Ansell J, Hirsh J, Hylek E, et al. Pharmacology and management of vitamin k antagonists: American College of Chest Physicians evidence-based clinical practiceguidelines(8thedition).Chest.2008; 133:160s–198s. Baker W, Cios D, Sander D, Coleman C. Meta-analysis to assess the quality of warfarin control in atrial fibrillation patients in the United States. J Manag Care Pharm. 2009;15: 244 –252. Chiquette E, Amato MB, H. Comparison of an anticoagulation clinic with usual medical care. Arch Intern Med. 1998;158:1641–1647. Lam Y, Ma T, Yan B. Alternatives to chronic warfarin therapy for the prevention of stroke in patients with atrial fibrillation. Int J Cardiol. 2010; 150:4 –11. National Institute for Health and Clinical Excellence. The Management Of Atrial Fibrillation (NICE Clinical Guideline 36). London: National Health Service, UK; 2006.

Address for correspondence: Dong-Churl Suh, MBA, PhD, College of Pharmacy, Chung-Ang University, Heukseok-ro, Dongjak-gu, Seoul, South Korea. E-mail: [email protected]

Volume 34 Number 7