00940
2013
AOPXXX10.1177/1060028013500940Mosholder et alThe Annals of Pharmacotherapy
Research Report
Bleeding Events Following Concurrent Use of Warfarin and Oseltamivir by Medicare Beneficiaries
Annals of Pharmacotherapy 47(11) 1420–1428 © The Author(s) 2013 Reprints and permissions: sagepub.com/journalsPermissions.nav DOI: 10.1177/1060028013500940 aop.sagepub.com
Andrew D. Mosholder, MD, MPH1, Judith A. Racoosin, MD, MPH1, Stephanie Young, MPH2, Michael Wernecke, BA3, Azadeh Shoaibi, MS, MHS1, Thomas E. MaCurdy, PhD3,4, Christopher Worrall, BS5, and Jeffrey A. Kelman, MD, MMSc5
Abstract Background: During the 2009 H1N1 influenza pandemic, the UK Medicines and Healthcare Products Regulatory Agency received case reports suggesting a potentiation of warfarin anticoagulation by the antiviral drug oseltamivir. We evaluated this putative interaction using Medicare data. Objective: To determine the frequency of bleeding following addition of oseltamivir or comparator drugs among Medicare beneficiaries taking warfarin. Methods: This was a retrospective cohort evaluation using Medicare nationwide data. Cohort members were Medicare Parts A, B, and D beneficiaries from June 30, 2006 to October 31, 2010 receiving warfarin for at least 1 month prior to a concomitant drug of interest (oseltamivir, ampicillin, trimethoprim–sulfamethoxazole (TMP-SMX), and angiotensin-converting enzyme (ACE) inhibitors). Bleeding within 14 days of new prescriptions for oseltamivir or comparators was identified using inpatient or emergency department ICD-9 (International Classification of Diseases, ninth revision) discharge diagnosis codes for gastrointestinal hemorrhage, epistaxis, hematuria, and intracranial bleeding. Patients with bleeding within 30 days preceding the prescription concomitant to warfarin were excluded. Results: With concomitant ACE inhibitors as reference, adjusted odds ratios (ORs) for any bleeding events within 14 days were 1.47 (95% confidence interval [CI] = 1.08-1.88), 1.24 (95% CI = 0.97-1.57), and 2.74 (95% CI = 2.53-3.03), for warfarin plus ampicillin, oseltamivir, and TMP-SMX, respectively. In a sensitivity analysis, adjusted ORs over a 7-day period were 1.89 (95% CI = 1.29-2.59), 1.47 (95% CI = 1.06-2.02), and 3.07 (95% CI = 2.76-3.49) for warfarin plus ampicillin, oseltamivir, and TMP-SMX, respectively. Conclusions: Bleeding with oseltamivir plus warfarin was not significantly increased over a 14-day observation period; a sensitivity analysis showed a statistically significant increase over a 7-day period; in contrast, the data consistently showed the known tendency of TMP-SMX to potentiate the effects of warfarin. The results should be interpreted with the limitations of this approach in mind, including the inability to control for unmeasured confounders. Keywords warfarin, oseltamivir, drug–drug interaction, influenza, bleeding Received February 22, 2013; revised May 9, 2013 and June 11, 2013; accepted June 13, 2013.
The Food and Drug Administration (FDA) Amendments Act of 2007 mandates that FDA develop a system for postmarket risk identification and analysis of drugs using electronic health care data (eg, insurance claims data, electronic health records data) from disparate data sources, both federal and private. In response to the Congressional mandate, FDA has initiated pilot programs to develop the scientific methodologies, data infrastructure, governance, and privacy policies to conduct active medical product safety surveillance in large observational databases. In the summer of 2009, with the advent of the H1N1 influenza pandemic, FDA’s Center for Drug Evaluation and Research strengthened its baseline monitoring of antiviral drug adverse effects. One enhancement was to leverage the
SafeRx project, a collaboration between FDA, the Office of the Assistant Secretary for Planning and Evaluation, and the 1
Center for Drug Evaluation and Research, U.S. Food and Drug Administration, Silver Spring, MD, USA 2 University of California, Los Angeles, CA, USA 3 Acumen LLC, Burlingame, CA, USA 4 Stanford University, Stanford, CA, USA 5 Centers for Medicare & Medicaid Services, Baltimore, MD, USA Corresponding Author: Andrew D. Mosholder, MD, MPH, FDA Center for Drug Evaluation and Research, Office of Pharmacovigilance and Epidemiology, Division of Epidemiology II, 10903 New Hampshire Avenue, Silver Spring, MD 20993-0002, USA. Email: [email protected]
Downloaded from aop.sagepub.com at NORTHERN ARIZONA UNIVERSITY on June 23, 2015
1421
Mosholder et al Centers for Medicare & Medicaid Services (CMS), to use Medicare data for near real-time medical product safety surveillance. In August 2009, the UK Medicines and Healthcare Products Regulatory Agency reported receiving postmarket case reports suggesting an interaction between warfarin and oseltamivir; however, it felt that the reports were not conclusive.1 This observation by the UK Medicines and Healthcare Products Regulatory Agency followed an earlier report by Health Canada in 2006 describing 19 spontaneous postmarket pharmacovigilance reports suggesting a possible interaction between oseltamivir and warfarin, resulting in higher international normalized ratios (INRs).2 More recently, in a case series of 15 patients treated concomitantly with oseltamivir and warfarin, 7 patients showed increased prothrombin time INRs from previously stable values; 3 of these 7 had clinical bleeding, and 1 of the 7 had another INR increase when rechallenged with the combination.3 Contrasting these case reports, a formal drug–drug interaction study, in which 20 chronic warfarin users were administered concomitant oseltamivir for 5 days, did not find evidence for an interaction.4 The vitamin K antagonist anticoagulant warfarin is stereoselectively metabolized by a variety of cytochrome P450 enzymes (2C9, 2C19, 2C8, 2C18, 1A2, and 3A4), with 2C9 providing most of the metabolic activity; the consequence is that warfarin is subject to many metabolic drug–drug interactions.5 In contrast to warfarin, neither oseltamivir nor its active metabolite oseltamivir carboxylate is a substrate for any cytochrome P450 isozymes,6 nor do they inhibit any cytochrome P450 isozymes.7 Neither oseltamivir nor its active metabolite has been shown to interfere with coagulation.4 Accordingly, there is no readily identifiable biologically plausible mechanism for a potentiation of warfarin anticoagulation by oseltamivir. Given the inconsistency between the case reports suggestive of a potential warfarin– oseltamivir interaction, and the clinical pharmacological profile of oseltamivir showing a minimal potential risk for drug interactions, FDA thought it useful to conduct an evaluation in the Medicare database seeking clinical evidence of a putative warfarin–oseltamivir interaction. We gathered surveillance data on Medicare beneficiaries’ exposure to warfarin and selected concomitant drugs. A CMS contractor, Acumen LLC, performed the data analyses. The purpose of the analysis was to determine the frequency of bleeding events following the combination of warfarin plus oseltamivir or selected comparator drugs, among Medicare Part D (prescription drug coverage) beneficiaries.
Methods Use of CMS data for postmarket surveillance activities has been approved by FDA’s Research Involving Human Subjects Committee. This retrospective cohort evaluation included Medicare Parts A, B, and D beneficiaries enrolled
at least 6 months from June 30, 2006 to October 31, 2010 who received warfarin for at least 1 month prior to a prescription for a concomitant drug of interest (oseltamivir or the comparator drugs ampicillin, trimethoprim–sulfamethoxazole [TMP-SMX], or angiotensin-converting enzyme inhibitors [ACEIs]). Patients prescribed more than one of the comparator drugs of interest were grouped by the first such drug received. Excluding patients prescribed a second drug of interest during the risk period would have eliminated events attributable to the second concomitant drug, but would have based eligibility on a postexposure event, which can introduce bias8; also, we expected few such patients. Two other antiviral drugs indicated for influenza, rimantadine and zanamivir, have not been used frequently in the Medicare population and were not included as comparators. Ampicillin and TMP-SMX were selected for their use in acute febrile illnesses, albeit for bacterial rather than viral infections. In addition, TMP-SMX interferes with warfarin metabolism,4 thereby providing a “positive control.” The class of ACEI was selected as a “negative control” because it is widely used in the Medicare population and is not labeled for a warfarin interaction4; thus it was chosen as the baseline comparator. We identified bleeding events occurring within 14 days of a new prescription for concomitant oseltamivir or a comparator using International Classification of Diseases, ninth revision (ICD-9) codes for epistaxis, gastrointestinal bleeding, hematuria, and intracranial bleeding. A combined event category enumerated patients with any of these events. Patients with any such code in the 30 days preceding the prescription concomitant to warfarin were excluded. A 14-day window for bleeding events was selected because there was a concern that it could take several days for a concomitant drug to interfere with warfarin metabolism to the extent that a bleeding outcome would occur and be observed and documented as an insurance claim; one recent observational study of warfarin and interacting antibiotics showed increased gastrointestinal bleeding after 5 or more days of co-prescribing.9 Additionally, because the course of oseltamivir varies from 5 to 10 days depending on whether it is used as treatment or prophylaxis, 14 days was considered to be an interval more likely to capture potentially related events. However, we also performed a sensitivity analysis with a shorter window of 7 days. To capture the most serious clinical events, we included Medicare Parts A and B discharge diagnosis codes from only inpatient stays and emergency department visits. Table 1 shows the list of selected bleeding events with the corresponding ICD-9 codes. We used multivariate logistic regression to estimate odds ratios (ORs) as approximations of relative risks of developing major bleeds within 2 weeks, adjusted for important potential confounders including demographic factors (agegroup, gender, geographical region, and income subsidy); recent hospitalization or emergency room visit;
Downloaded from aop.sagepub.com at NORTHERN ARIZONA UNIVERSITY on June 23, 2015
1422
Annals of Pharmacotherapy 47(11)
Table 1. Selected Hemorrhagic Events With Corresponding ICD-9 (International Classification of Diseases, Ninth Revision) Codes. Hemorrhagic Event Epistaxis Gastrointestinal bleeding
Hematuria Intracranial bleeding
ICD-9 Codes 784.7 530.21 531.2, 531.2x, 531.4, 531.4x, 531.6, 531.6x 532.0, 532.00, 532.01, 532.2, 532.2x, 532.4, 532.4x, 532.6, 532.6x 533.0, 533.00, 533.01, 533.2, 533.2x, 533.4, 533.4x, 533.6, 533.6x 534.0, 534.00, 534.01, 534.2, 534.2x, 534.4, 534.4x, 534.6, 534.6x 535.01, 535.11, 535.21, 535.31, 535.41, 535.51, 535.61, 537.83, 537.84 562.02, 562.03, 562.12, 562.13 569.85, 569.86 578, 578.x 599.7, 599.7x 430 431 432, 432.x
concomitant drugs potentially associated with bleeding (nonsteroidal anti-inflammatory drugs, selective serotonin reuptake inhibitors, proton pump inhibitors, platelet inhibitors); comorbid medical conditions including those that could affect bleeding (liver disease, cystitis, ulcer, diverticulosis, and other chronic medical conditions); warfarin duration; and drugs known to interact with warfarin. Except where otherwise indicated, covariates were assessed over a 6-month look-back period, consistent with the required minimum duration of Medicare enrollment; concomitant medications were assessed over a 30-day look-back period. To manage the large number of drugs known to interact with warfarin, those drugs were identified and grouped according to Holbrook et al’s10 systematic review of warfarin–drug interactions into the following categories: highly probable potentiator, probable potentiator, highly probable inhibitor, and probable inhibitor. The statistical significance level was determined at the .05 level, and all the analyses were performed using Stata version 11 (StataCorp, College Station, TX). Confidence intervals on the observed frequencies of patients with events in each cohort were calculated using the bootstrap method.11
Results Table 2 displays the cohort size and demographics, and Table 3 displays the numbers of outcome events. The main results are displayed in Table 4, with adjusted ORs for each drug combination cohort compared with the negative control (warfarin + ACEI) for the combined endpoint, and
stratified by bleeding site. Full results for covariates in the 14-day interval are presented in Tables A1 to A4 in the appendix. The size of the warfarin–oseltamivir and warfarin–ampicillin cohorts were about 1/10 the size of the warfarin–TMP-SMX and warfarin–ACEI cohorts. As evidenced in the table, over 14 days patients taking warfarin–oseltamivir had small, nonstatistically significant elevations in the likelihood of the combined endpoint, as well as gastrointestinal bleeding and epistaxis compared with patients taking warfarin + ACEI (OR = 1.24, 95% confidence interval [CI] = 0.97-1.57; OR = 1.31, 95% CI = 0.92-1.80; OR = 1.63, 95% CI = 0.84-2.70, respectively). In contrast, patients taking warfarin–TMP-SMX, the positive control cohort, were significantly more likely to have had the combined endpoint (OR = 2.74, 95% CI = 2.53-3.03), gastrointestinal bleeding (OR = 2.53, 95% CI = 2.27-2.93), intracranial bleeding (OR = 1.67, 95% CI = 1.25-2.40), epistaxis (OR = 2.07, 95% CI = 1.66-2.61), and hematuria (OR = 4.21, 95% CI = 3.49-5.12) compared with the negative control ACEI. The magnitudes of the associations of warfarin–TMP-SMX with the combined endpoint and gastrointestinal bleeding were higher than the magnitude of the associations of warfarin–ampicillin or warfarin–oseltamivir with the combined outcome and gastrointestinal bleeding, based on the nonoverlapping confidence intervals for the ORs in these comparisons. In addition, the patients taking ampicillin–warfarin had statistically significantly elevated likelihoods of having the combined bleeding endpoint as well as hematuria (OR = 1.47, 95% CI = 1.08-1.88; OR = 2.97, 95% CI = 1.97-4.37), respectively. Table 4 also summarizes unadjusted results and the 7-day sensitivity analysis. In comparison with the 14-day data shown, the results using 7 days as the risk window were notable for a significantly elevated adjusted OR for the association of warfarin–oseltamivir use with epistaxis (OR = 2.69, 95% CI = 1.14-4.60), and the combined endpoint (OR = 1.47, 95% CI = 1.06-2.02). As with the results based on the 14-day exposure window, the OR for warfarin–TMP-SMX with combined bleeding had a confidence interval that did not overlap with those for the combined bleeding 7-day ampicillin and oseltamivir ORs (7-day warfarin–TMP-SMX OR = 3.07, 95% CI = 2.76-3.49). Full results for covariates in the 14-day risk window are presented in Tables A1 to A4 in the appendix. With respect to other potentially interacting drugs, we found no statistically significant effects on bleeding events from warfarin inhibitors or potentiators, although bleeding events were increased with selective serotonin reuptake inhibitors, antiplatelet therapy, and nonsteroidal anti-inflammatory drugs (Table A3 in the appendix).
Discussion Although warfarin can produce bleeding from any site, gastrointestinal bleeding is the most frequently reported
Downloaded from aop.sagepub.com at NORTHERN ARIZONA UNIVERSITY on June 23, 2015
1423
Mosholder et al Table 2. Summary of Demographics. ACE Inhibitors Demographics
n
Total Gender Male Female Age category (years) 84 Low-income subsidy (LIS)a No LIS LIS with 15% copay LIS with high copay LIS with low copay LIS with zero copay Region Northeast Southeast South Midwest West Other or missing
Ampicillin
Oseltamivir n
TMP-SMX
%
n
%
%
n
%
158 724
100.0
9150
100.0
13 406
100.0
143 076
100.0
71 754 86 970
45.2 54.8
2900 6250
31.7 68.3
4958 8448
37.0 63.0
47 191 95 885
33.0 67.0
17 768 50 161 61 868 28 927
11.2 31.6 39.0 18.2
1041 2010 3416 2683
11.4 22.0 37.3 29.3
2041 3468 4538 3359
15.2 25.9 33.9 25.1
21 168 35 899 51 458 34 551
14.8 25.1 36.0 24.1
100 007 2968 17 498 26 773 11 478
63.0 1.9 11.0 16.9 7.2
4074 99 723 1413 2841
44.5 1.1 7.9 15.4 31.0
6231 155 1118 2135 3767
46.5 1.2 8.3 15.9 28.1
70 110 2212 15 469 25 457 29 828
49.0 1.5 10.8 17.8 20.8
32 394 46 275 14 782 44 289 20 758 226
20.4 29.2 9.3 27.9 13.1 0.1
2083 2937 837 2324 937 32
22.8 32.1 9.1 25.4 10.2 0.3
3264 3961 1208 3431 1519 23
24.3 29.5 9.0 25.6 11.3 0.2
25 137 47 098 14 892 38 270 17 540 139
17.6 32.9 10.4 26.7 12.3 0.1
Abbreviations: ACE, angiotensin-converting enzyme; TMP-SMX, trimethoprim–sulfamethoxazole. a Eligibility for Part D low-income subsidy is used to proxy for income. Poorer beneficiaries are required to pay smaller copays.
Table 3. Outcome Counts and Rates per Thousand Beneficiaries (14-Day Risk Window). Combined Cohort ACE inhibitor Ampicillin Oseltamivir TMP-SMX
Gastrointestinal
Intracranial
Epistaxis
Hematuria
Cohort Size
Count
Rate
Count
Rate
Count
Rate
Count
Rate
Count
Rate
158 724 9150 13 406 143 076
744 63 66 1747
4.69 6.89 4.92 12.21
417 31 39 911
2.63 3.39 2.91 6.37
66 2 4 112
0.42 0.22 0.30 0.78
126 5 13 199
0.79 0.55 0.97 1.39
152 26 11 560
0.96 2.84 0.82 3.91
Abbreviations: ACE, angiotensin-converting enzyme; TMP-SMX, trimethoprim–sulfamethoxazole.
hemorrhagic complication of warfarin anticoagulation in postmarket surveillance,12 which is consistent with our findings. Compared with the other concomitant drugs assessed, the positive control cohort, warfarin–TMP-SMX, had significantly elevated adjusted ORs for gastrointestinal bleeding events and the combined bleeding events. We did not find compelling evidence to support an interaction between oseltamivir and warfarin resulting in increased bleeding events among Medicare beneficiaries. In the 14-day results, OR point estimates for some outcomes were elevated for the warfarin–oseltamivir cohort, but they were not statistically significant. In the 7-day sensitivity analysis, warfarin–oseltamivir showed a modest but
statistically significant increase in combined events, with the highest point estimate among individual events observed for warfarin–oseltamivir and epistaxis. However, patients in our sample being treated for influenza with oseltamivir might have nasal discharge, which could increase the risk of epistaxis in that time frame; one case series found epistaxis was associated with influenza.13 We speculate that the association of ampicillin specifically with hematuria may reflect patients receiving ampicillin for urinary tract infections, which can produce hematuria. We focused on a 14-day risk window based on the projected time it might take for a drug interaction with warfarin to manifest clinically and be detected in an insurance claims
Downloaded from aop.sagepub.com at NORTHERN ARIZONA UNIVERSITY on June 23, 2015
1424
Annals of Pharmacotherapy 47(11)
Table 4. Main Results of All Models (7-Day and 14-Day, Adjusted and Unadjusted). Cohort
Combined
Gastrointestinal
Unadjusted odds ratio (95% confidence interval), 7 days ACE inhibitor 1.00 (1.00, 1.00) 1.00 Ampicillin 1.87 (1.28, 2.57) 1.47 Oseltamivir 1.24 (0.90, 1.71) 1.28 TMP-SMX 2.93 (2.64, 3.32) 2.44 Adjusted odds ratio (95% confidence interval), 7 days ACE inhibitor 1.00 (1.00, 1.00) 1.00 Ampicillin 1.89 (1.29, 2.59) 1.46 Oseltamivir 1.47 (1.06, 2.02) 1.53 TMP-SMX 3.07 (2.76, 3.49) 2.56 Unadjusted odds ratio (95% confidence interval), 14 days ACE inhibitor 1.00 (1.00, 1.00) 1.00 Ampicillin 1.47 (1.08, 1.88) 1.29 Oseltamivir 1.05 (0.83, 1.33) 1.11 TMP-SMX 2.62 (2.43, 2.89) 2.43 Adjusted odds ratio (95% confidence interval), 14 days ACE inhibitor 1.00 (1.00, 1.00) 1.00 Ampicillin 1.47 (1.08, 1.88) 1.26 Oseltamivir 1.24 (0.97, 1.57) 1.31 TMP-SMX 2.74 (2.53, 3.03) 2.53
Intracranial
Epistaxis
Hematuria
(1.00, 1.00) (0.88, 2.16) (0.80, 1.92) (2.18, 2.85)
1.00 0.50 1.01 1.97
(1.00, 1.00) (0.00, 1.85) (0.00, 2.72) (1.28, 3.16)
1.00 0.93 2.12 2.08
(1.00, 1.00) (0.00, 2.18) (0.91, 3.72) (1.52, 2.93)
1.00 3.86 0.44 5.00
(1.00, 1.00) (2.15, 6.13) (0.00, 1.02) (3.88, 6.62)
(1.00, 1.00) (0.85, 2.13) (0.96, 2.34) (2.23, 3.07)
1.00 0.47 1.21 1.97
(1.00, 1.00) (0.00, 1.84) (0.00, 3.11) (1.23, 3.16)
1.00 1.08 2.69 2.36
(1.00, 1.00) (0.00, 2.45) (1.14, 4.60) (1.72, 3.52)
1.00 3.84 0.50 5.14
(1.00, 1.00) (2.13, 6.15) (0.00, 1.15) (3.96, 6.92)
(1.00, 1.00) (0.87, 1.82) (0.78, 1.51) (2.20, 2.74)
1.00 0.53 0.72 1.88
(1.00, 1.00) (0.00, 1.41) (0.15, 1.73) (1.43, 2.70)
1.00 0.69 1.22 1.75
(1.00, 1.00) (0.25, 1.33) (0.61, 1.95) (1.41, 2.25)
1.00 2.97 0.86 4.10
(1.00, 1.00) (2.01, 4.35) (0.34, 1.36) (3.41, 4.85)
(1.00, 1.00) (0.84, 1.77) (0.92, 1.80) (2.27, 2.93)
1.00 0.45 0.71 1.67
(1.00, 1.00) (0.00, 1.22) (0.14, 1.76) (1.25, 2.40)
1.00 0.83 1.63 2.07
(1.00, 1.00) (0.29, 1.60) (0.84, 2.70) (1.66, 2.61)
1.00 2.97 0.99 4.21
(1.00, 1.00) (1.97, 4.37) (0.41, 1.54) (3.49, 5.12)
Abbreviations: ACE, angiotensin-converting enzyme; TMP-SMX, trimethoprim–sulfamethoxazole.
system; additionally, previous work suggested that increases in bleeding events via warfarin drug–drug interactions were most prominent in the 6- to 15-day period.7 The adjusted ORs with warfarin–TMP-SMX were statistically significant at both 7 and 14 days for every bleeding outcome analyzed. This consistent pattern was unique to TMP-SMX. Although the combined events ORs for oseltamivir at 7 days, and for ampicillin at 7 and 14 days, were statistically significant, the confidence intervals for those ORs had no overlap with the corresponding OR confidence intervals for warfarin–TMP-SMX. Because the elevated ORs observed with warfarin-TMPSMX showed the capability of the analysis to identify the positive control, we have some reassurance that this drug– drug interaction assessment had the potential to discover a major effect on bleeding by the combination of warfarin and oseltamivir if there was one. However, this assessment has important limitations. Our results, especially from a statistical standpoint, could be consistent with a modest association with bleeding events. However, it seems unlikely from our results that oseltamivir could have an effect comparable to TMP-SMX. Our results also are inconsistent with an effect as frequent as that described in the case series by Lee et al,3 a discrepancy for which we do not have a ready explanation, although the specific clinical bleeding events in the 3 patients from their case series (bloody sputum in 2, and bloody ascites in a third) would not have met our outcome definition. Another methodologic limitation was our
exclusive reliance on diagnosis billing codes, without examination of medical records to validate the occurrence of the specific outcomes; however, 2 studies in other settings that assessed sets of ICD-9 gastrointestinal disorder codes similar to our gastrointestinal bleeding category showed generally good positive predictive values.14,15 Similarly, we had to rely on prescription claims as a surrogate for drug exposure. Some patients may have filled osletamivir prescriptions for future use (“stockpiling”) and been inappropriately considered exposed to oseltamivir. However, during the 2009 pandemic H1N1 influenza outbreak, prescribing trends for oseltamivir closely paralleled the epidemic curve for influenza-like illnesses, suggesting that stockpiling of prescriptions was likely not common.16 Episodes of bleeding that did not receive medical attention, such as fatal intracranial hemorrhage that resulted in out of hospital death, could not be ascertained. Additionally, elevations of clotting times, as were reported in postmarketing surveillance and by Lee et al,3 could not be detected in the Medicare database. The analysis is also limited by sample size for some drug–drug combinations. For example, the warfarin–oseltamivir cohort was substantially smaller than the ACEI and TMP-SMX comparator cohorts. The numbers of events were also smaller for the nongastrointestinal bleeding sites. Another challenge is the need to control for the many potential confounders in the setting of warfarin use including the many drugs that interact with warfarin metabolism, the many drugs that may potentiate bleeding
Downloaded from aop.sagepub.com at NORTHERN ARIZONA UNIVERSITY on June 23, 2015
1425
Mosholder et al on their own, and the many health conditions that may predispose to bleeding. We attempted to address these confounders in the model, but the potential for unmeasured confounding certainly remains. In particular, the modest increase in bleeding events observed over the 7-day window with oseltamivir could be due to confounding factors related to acute influenza, such as changes in diet, elevated temperature, concomitant medications, gastrointestinal function, and the like, which could alter INR in the absence of a true drug–drug interaction. Such effects might explain some of the reported cases of increased INR or bleeding with the warfarin–oseltamivir combination. There is evidence that influenza infection enhances coagulation.17,18 Speculatively, such an effect could mask a potentiation of warfarin anticoagulation by oseltamivir, although only with its use in treatment of influenza rather than prophylaxis. Our analysis was not capable of distinguishing prophylactic use from treatment of influenza. The results of an assessment such as this are considered along with the body of evidence that FDA holds on a given safety concern. One can foresee 3 potential outcomes: •• The body of evidence does not suggest an elevated risk and standard monitoring is continued.
•• The body of evidence suggests an elevated risk, but questions remain about the strength or validity of the safety concern, and a more in-depth assessment may be needed. •• The body of evidence regarding the elevated risk is so strong that FDA proceeds to a regulatory action. In the case of the putative warfarin–oseltamivir interaction, the body of evidence supports continuation of standard monitoring. In conclusion, our analysis of claims for bleeding events among Medicare patients co-prescribed oseltamivir and warfarin failed to show clear evidence of a drug–drug interaction, consistent with the reported absence of a pharmacokinetic interaction between these drugs.4 In the same analysis, we successfully identified the predicted potentiation of anticoagulation by concomitant TMP-SMX, evidenced by the increased risk of bleeding events. Although future observational studies might be able to address some limitations of the present analysis, on balance, we felt that with its large size and substantial capture of warfarin-treated patients, the Medicare database functioned well in providing relevant information regarding the potential drug interaction in a timely fashion.
Appendix Table A1. Patient Demographics.a Adjusted Odds Ratio for Outcome (95% Confidence Interval) Patient Demographics Age bracket (years) 84 Gender Male Female Low-income subsidy (LIS) No LIS LIS with 15% copay LIS with high copay LIS with low copay LIS with zero copay Region Northeast Southeast South Midwest West Other or unknown
Combined
Gastrointestinal
Intracranial
Epistaxis
Hematuria
1.00 1.04 1.31 1.66
(1.00, 1.00) (0.91, 1.20) (1.13, 1.49) (1.38, 1.90)
1.00 1.26 1.59 2.11
(1.00, 1.00) (1.02, 1.60) (1.29, 1.93) (1.67, 2.59)
1.00 1.52 1.94 2.88
(1.00, 1.00) (0.89, 3.52) (1.08, 5.02) (1.53, 6.73)
1.00 0.70 0.89 0.95
(1.00, 1.00) (0.48, 0.99) (0.64, 1.27) (0.65, 1.33)
1.00 0.87 1.09 1.26
(1.00, 1.00) (0.66, 1.13) (0.85, 1.36) (0.98, 1.60)
1.00 0.82
(1.00, 1.00) (0.75, 0.90)
1.00 1.13
(1.00, 1.00) (1.00, 1.31)
1.00 0.96
(1.00, 1.00) (0.72, 1.33)
1.00 0.93
(1.00, 1.00) (0.71, 1.14)
1.00 0.42
(1.00, 1.00) (0.36, 0.48)
1.00 0.93 1.11 1.08 1.11
(1.00, 1.00) (0.66, 1.30) (0.95, 1.25) (0.97, 1.20) (0.97, 1.26)
1.00 0.80 1.22 1.11 1.17
(1.00, 1.00) (0.43, 1.32) (1.02, 1.45) (0.95, 1.32) (0.97, 1.38)
1.00 1.11 1.12 1.31 0.98
(1.00, 1.00) (0.00, 2.81) (0.61, 1.86) (0.86, 2.04) (0.56, 1.76)
1.00 1.20 1.05 0.89 0.81
(1.00, 1.00) (0.45, 2.19) (0.70, 1.45) (0.63, 1.21) (0.56, 1.20)
1.00 0.91 0.96 1.07 1.16
(1.00, 1.00) (0.45, 1.53) (0.71, 1.23) (0.88, 1.29) (0.93, 1.50)
1.01 1.00 1.03 0.99 0.90 0.34
(0.90, 1.13) (1.00, 1.00) (0.88, 1.20) (0.89, 1.10) (0.77, 1.01) (0.00, 1.15)
0.88 1.00 0.96 0.84 0.82 0.64
(0.75, 1.03) (1.00, 1.00) (0.79, 1.14) (0.72, 0.97) (0.68, 0.97) (0.00, 2.25)
0.78 1.00 0.92 1.12 1.01 0.00
(0.50, 1.19) (1.00, 1.00) (0.54, 1.56) (0.76, 1.58) (0.59, 1.57) (0.00, 0.00)
1.16 1.00 1.16 1.21 1.23 0.00
(0.87, 1.58) (1.00, 1.00) (0.77, 1.72) (0.95, 1.59) (0.82, 1.72) (0.00, 0.00)
1.32 1.00 1.18 1.21 0.90 0.00
(1.08, 1.67) (1.00, 1.00) (0.93, 1.53) (1.01, 1.46) (0.72, 1.16) (0.00, 0.00)
a
Full results for primary models predicting frequency of hemorrhagic events within 14 days of a drug cohort prescription among Medicare beneficiaries on warfarin for at least 30 days.
Downloaded from aop.sagepub.com at NORTHERN ARIZONA UNIVERSITY on June 23, 2015
1426
Annals of Pharmacotherapy 47(11)
Table A2. Patient Hospital Encounters.a Adjusted Odds Ratio for Outcome (95% Confidence Interval) Hospital Encounters
Combined
Gastrointestinal
Hospitalizations in past 30 days 0 1.00 (1.00, 1.00) 1 1.16 (1.02, 1.34) 2+ 1.34 (1.05, 1.70) Emergency room visits in past 30 days 0 1.00 (1.00, 1.00) 1 1.45 (1.29, 1.61) 2+ 1.93 (1.57, 2.32) Hospitalizations in 180 to 31 days prior 0 1.00 (1.00, 1.00) 1 1.09 (0.94, 1.22) 2+ 1.11 (0.95, 1.33) Emergency room visits in 180 to 31 days prior 0 1.00 (1.00, 1.00) 1 1.06 (0.94, 1.19) 2+ 1.25 (1.09, 1.48)
Intracranial
Epistaxis
Hematuria
1.00 1.10 1.58
(1.00, 1.00) (0.93, 1.28) (1.22, 2.09)
1.00 0.83 1.36
(1.00, 1.00) (0.48, 1.39) (0.53, 2.57)
1.00 1.50 1.14
(1.00, 1.00) (1.05, 2.10) (0.63, 2.11)
1.00 1.20 0.80
(1.00, 1.00) (0.94, 1.52) (0.45, 1.23)
1.00 1.44 1.71
(1.00, 1.00) (1.23, 1.68) (1.32, 2.23)
1.00 1.97 2.42
(1.00, 1.00) (1.24, 2.92) (1.18, 4.00)
1.00 1.20 1.94
(1.00, 1.00) (0.83, 1.59) (1.17, 2.66)
1.00 1.53 2.42
(1.00, 1.00) (1.21, 1.83) (1.67, 3.39)
1.00 1.08 1.14
(1.00, 1.00) (0.92, 1.26) (0.90, 1.43)
1.00 0.84 1.13
(1.00, 1.00) (0.55, 1.27) (0.62, 1.97)
1.00 1.01 1.00
(1.00, 1.00) (0.63, 1.42) (0.71, 1.52)
1.00 1.16 1.13
(1.00, 1.00) (0.89, 1.46) (0.82, 1.58)
1.00 1.06 1.15
(1.00, 1.00) (0.91, 1.24) (0.94, 1.41)
1.00 1.01 0.83
(1.00, 1.00) (0.64, 1.55) (0.47, 1.39)
1.00 1.02 1.61
(1.00, 1.00) (0.72, 1.40) (1.09, 2.20)
1.00 1.09 1.33
(1.00, 1.00) (0.87, 1.34) (1.03, 1.75)
a Full results for primary models predicting frequency of hemorrhagic events within 14 days of a drug cohort prescription among Medicare beneficiaries on warfarin for at least 30 days.
Table A3. Patient Medications.a Adjusted Odds Ratio for Outcome (95% Confidence Interval) Warfarin duration (days) 30-89 90-179 180+ Concomitant medicationsb Highly probable potentiators Probable potentiators Highly probable inhibitors Probable inhibitors Platelet inhibitors NSAIDs PPI SSRI
Combined
Gastrointestinal
Intracranial
Epistaxis
Hematuria
1.00 0.98 0.98
(1.00, 1.00) (0.87, 1.12) (0.87, 1.13)
1.00 0.99 0.99
(1.00, 1.00) (0.84, 1.19) (0.82, 1.17)
1.00 0.81 0.92
(1.00, 1.00) (0.56, 1.37) (0.58, 1.62)
1.00 1.22 1.27
(1.00, 1.00) (0.92, 1.87) (0.92, 1.99)
1.00 0.96 0.91
(1.00, 1.00) (0.76, 1.26) (0.69, 1.20)
1.01 0.96 0.90 1.03 1.48 1.24 0.91 1.11
(0.92, 1.11) (0.89, 1.04) (0.71, 1.07) (0.89, 1.18) (1.29, 1.72) (1.08, 1.43) (0.83, 1.00) (1.00, 1.22)
0.98 1.07 0.92 1.06 1.66 1.39 0.95 1.18
(0.87, 1.09) (0.96, 1.22) (0.74, 1.14) (0.89, 1.21) (1.38, 2.03) (1.14, 1.66) (0.85, 1.08) (1.02, 1.33)
0.94 0.87 1.23 1.18 0.79 1.39 0.85 1.23
(0.69, 1.28) (0.64, 1.14) (0.57, 2.04) (0.59, 1.83) (0.27, 1.56) (0.71, 2.08) (0.56, 1.19) (0.91, 1.74)
1.10 0.95 0.94 1.46 2.03 1.02 0.89 1.12
(0.85, 1.42) (0.78, 1.21) (0.49, 1.39) (1.03, 1.98) (1.29, 2.82) (0.61, 1.48) (0.67, 1.18) (0.86, 1.41)
1.03 0.78 0.78 0.79 1.08 0.98 0.88 0.95
(0.88, 1.24) (0.67, 0.89) (0.52, 1.08) (0.57, 1.05) (0.74, 1.47) (0.68, 1.26) (0.73, 1.02) (0.77, 1.11)
Abbreviations: NSAIDs, nonsteroidal anti-inflammatory drugs; PPI, proton pump inhibitor; SSRI, selective serotonin reuptake inhibitor. a Full results for primary models predicting frequency of hemorrhagic events within 14 days of a drug cohort prescription among Medicare beneficiaries on warfarin for at least 30 days. b Drugs are categorized using Holbrook et al’s10 systematic review of warfarin interactions.
Downloaded from aop.sagepub.com at NORTHERN ARIZONA UNIVERSITY on June 23, 2015
1427
Mosholder et al Table A4. Patient Health Characteristics.a Patient Health Characteristics Chronic health conditions Chronic liver disease Chronic renal failure Additional comorbidities Acute renal failure Acute myocardial infarction, 30-0 days prior Acute myocardial infarction, 180-31 days prior Calculus of kidney Calculus of urinary tract Cancer, kidney/prostrate/ bladder Cerebral arterial occlusion Cystitis Dementia Diverticulosis Heart failure, 30-0 days prior Heart failure, 180-31 days prior Hypertension Ischemic stroke Pyelonephritis Metastatic cancer Pneumonia Pneumothorax/pleural effusion Severe cancer Ulcer
Adjusted Odds Ratio for Outcome (95% Confidence Interval) Combined
Gastrointestinal
Intracranial
Epistaxis
Hematuria
1.35 1.24
(1.05, 1.73) (1.12, 1.37)
1.47 1.40
(1.10, 1.91) (1.22, 1.60)
1.48 1.14
(0.17, 3.07) (0.73, 1.64)
1.60 1.11
(0.71, 2.52) (0.87, 1.45)
1.12 1.06
(0.65, 1.72) (0.87, 1.30)
1.04 1.26
(0.90, 1.21) (0.96, 1.60)
1.05 1.22
(0.86, 1.29) (0.85, 1.64)
1.11 0.22
(0.60, 1.62) (0.00, 0.69)
1.15 1.49
(0.74, 1.69) (0.68, 2.30)
0.89 1.50
(0.59, 1.14) (0.75, 2.53)
1.04
(0.80, 1.30)
0.99
(0.72, 1.31)
3.20
(1.47, 5.48)
0.74
(0.32, 1.29)
0.90
(0.46, 1.47)
1.18 1.47 1.27
(0.92, 1.44) (0.68, 2.34) (1.08, 1.49)
0.99 1.25 1.06
(0.67, 1.34) (0.23, 2.47) (0.79, 1.40)
0.91 0.00 1.05
(0.21, 1.83) (0.00, 0.00) (0.49, 1.78)
0.60 1.40 0.72
(0.11, 1.24) (0.00, 4.67) (0.40, 1.19)
1.83 1.49 1.60
(1.32, 2.46) (0.61, 2.62) (1.24, 1.99)
1.05 1.19 1.07 1.10 1.12
(0.92, 1.18) (0.99, 1.41) (0.92, 1.24) (0.95, 1.24) (0.98, 1.23)
1.07 0.74 1.07 1.22 1.29
(0.92, 1.27) (0.51, 0.97) (0.91, 1.28) (1.00, 1.43) (1.09, 1.46)
1.34 1.10 1.41 0.93 0.65
(0.86, 2.00) (0.41, 1.87) (0.87, 2.20) (0.42, 1.68) (0.43, 1.02)
1.25 1.00 0.86 0.96 1.28
(0.90, 1.65) (0.51, 1.64) (0.60, 1.22) (0.52, 1.38) (0.97, 1.67)
0.83 2.28 1.06 0.97 0.87
(0.63, 1.04) (1.77, 3.09) (0.82, 1.33) (0.69, 1.28) (0.71, 1.09)
1.15
(1.04, 1.27)
1.13
(0.99, 1.29)
1.42
(0.93, 2.06)
1.30
(1.02, 1.76)
1.03
(0.85, 1.24)
1.05 0.95 1.12 1.35 1.00 1.10
(0.89, 1.24) (0.85, 1.07) (0.79, 1.50) (1.03, 1.67) (0.85, 1.15) (0.97, 1.26)
1.00 0.92 1.36 1.15 1.02 1.21
(0.80, 1.25) (0.77, 1.06) (0.89, 2.04) (0.75, 1.57) (0.82, 1.24) (1.00, 1.43)
0.83 1.41 0.82 1.73 1.18 1.09
(0.31, 1.50) (0.87, 2.12) (0.00, 2.23) (0.46, 3.84) (0.54, 2.09) (0.65, 1.89)
1.13 0.73 0.41 1.71 0.94 1.21
(0.74, 1.74) (0.50, 1.00) (0.00, 1.11) (0.74, 2.88) (0.57, 1.33) (0.82, 1.74)
1.18 1.07 1.05 1.61 0.95 0.82
(0.78, 1.64) (0.85, 1.33) (0.52, 1.57) (1.09, 2.33) (0.69, 1.26) (0.62, 1.11)
1.13 1.58
(0.88, 1.45) (1.27, 1.95)
1.22 1.96
(0.87, 1.79) (1.45, 2.48)
0.88 0.94
(0.12, 2.03) (0.00, 1.99)
0.93 1.68
(0.34, 1.78) (0.78, 2.71)
1.13 0.86
(0.71, 1.74) (0.39, 1.47)
a Full results for primary models predicting frequency of hemorrhagic events within 14 days of a drug cohort prescription among Medicare beneficiaries on warfarin for at least 30 days.
Authors’ Note The views expressed in this manuscript represent the opinions of the authors, and do not necessarily represent the views of the US Food and Drug Administration, the Centers for Medicare & Medicaid Services, or the US Department of Health and Human Services. This work was presented in abstract form at the 27th International Conference on Pharmacoepidemiology, August 14-17, 2011, Chicago, Illinois. An abstract was published in Pharmacoepidemiology and Drug Safety, August 2011, Volume 20, Issue Supplement S1.
Acknowledgments We thank the Centers for Medicare & Medicaid Services for their operational leadership and analytic support, and the Health and
Human Services Office of the Assistant Secretary for Planning and Evaluation for their financial support of this project.
Declaration of Conflicting Interests The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the SafeRx Project, a joint initiative of the Centers for Medicare & Medicaid Services (CMS), US Food and Drug Administration (FDA), and the US Department of
Downloaded from aop.sagepub.com at NORTHERN ARIZONA UNIVERSITY on June 23, 2015
1428
Annals of Pharmacotherapy 47(11)
Health and Human Services (HHS) Office of the Assistant Secretary for Planning and Evaluation (ASPE). Acumen LLC is a contractor to CMS.
References 1. MHRA suspected adverse drug reaction (ADR) analysis: influenza antivirals—Tamiflu (oseltamivir) and Relenza (zanamivir), August 21, 2009. http://www.mhra.gov.uk/ home/groups/pl-p/documents/websiteresources/con054667. pdf. Accessed October 19, 2011. 2. Health Products and Food Branch, Marketed Health Products Directorate. Can Adverse React Newslett. 2006;16:1, 1-2. 3. Lee SH, Kang HR, Jung JW, et al. Effect of oseltamivir on bleeding risk associated with warfarin therapy: a retrospective review. Clin Drug Investig. 2012;32:131-137. 4. Davies BE, Aceves-Baldo P, Lennon-Chrimes S, Brewster M. Effect of oseltamivir treatment on anticoagulation: a crossover study in warfarinized patients. Br J Clin Pharmacol. 2010;70:834-843. 5. Bristol-Myers Squibb Company. Coumadin [package insert]. http://packageinserts.bms.com/pi/pi_coumadin.pdf. Accessed October 19, 2011. 6. Dutkowski R, Thakrar B, Froehlich E, Suter P, Oo C, Ward P. Safety and pharmacology of oseltamivir in clinical use. Drug Saf. 2003;26:787-801. 7. He G, Massarella J, Ward P. Clinical pharmacokinetics of the prodrug oseltamivir and its active metabolite Ro 64-0802. Clin Pharmacokinet. 1999;37:471-484. 8. Rothman KJ, Greenland S. Cohort studies. In: Rothman KJ, Greenland S, Lash TL, eds. Modern Epidemiology. 3rd ed. Philadelphia, PA: Lippincott Williams & Wilkins; 2008:100-110. 9. Schelleman H, Bilker WB, Brensinger CM, Han X, Kimmel SE, Hennessy S. Warfarin with fluoroquinolones, sulfonamides, or azole antifungals: interactions and the risk of
hospitalization for gastrointestinal bleeding. Clin Pharmacol Ther. 2008;84:581-588. 10. Holbrook AM, Pereira JA, Labiris R, et al. Systematic overview of warfarin and its drug and food interactions. Arch Intern Med. 2005;165:1095-1106. 11. Efron B. Bootstrap methods: another look at the jackknife. Ann Stat. 1979;7:1-26. 12. Wysowski DK, Nourjah P, Swartz L. Bleeding complications with warfarin use: a prevalent adverse effect resulting in regulatory action. Arch Intern Med. 2007;167:1414-1419. 13. Duque V, Vaz J, Mota V, Morais C, Da Cunha S, MeliçoSilvestre A. Clinical manifestations of pandemic (H1N1) 2009 in the ambulatory setting. J Infect Dev Ctries. 2011;5:658663. 14. Raiford DS, Pérez Gutthann S, García Rodríguez LA. Positive predictive value of ICD-9 codes in the identification of cases of complicated peptic ulcer disease in the Saskatchewan hospital automated database. Epidemiology. 1996;7:101-104. 15. Cattaruzzi C, Troncon MG, Agostinis L, García Rodríguez LA. Positive predictive value of ICD-9th codes for upper gastrointestinal bleeding and perforation in the Sistema Informativo Sanitario Regionale database. J Clin Epidemiol. 1999;52:499-502. 16. Borders-Hemphill V, Mosholder A. U.S. utilization pat terns of influenza antiviral medications during the 2009 H1N1 influenza pandemic. Influenza Other Respi Viruses. 2012;6:e129-e133. 17. Madjid M, Aboshady I, Awan I, Litovsky S, Casscells SW. Influenza and cardiovascular disease: Is there a causal relationship? Tex Heart Inst J. 2004;31:4-13. 18. Keller TT, van der Sluijs KF, de Kruif MD, et al. Effects on coagulation and fibrinolysis induced by influenza in mice with a reduced capacity to generate activated protein C and a deficiency in plasminogen activator inhibitor type 1. Circ Res. 2006;99:1261-1269.
Downloaded from aop.sagepub.com at NORTHERN ARIZONA UNIVERSITY on June 23, 2015
2013
AOPXXX10.1177/1060028013500940Mosholder et alThe Annals of Pharmacotherapy
Research Report
Bleeding Events Following Concurrent Use of Warfarin and Oseltamivir by Medicare Beneficiaries
Annals of Pharmacotherapy 47(11) 1420–1428 © The Author(s) 2013 Reprints and permissions: sagepub.com/journalsPermissions.nav DOI: 10.1177/1060028013500940 aop.sagepub.com
Andrew D. Mosholder, MD, MPH1, Judith A. Racoosin, MD, MPH1, Stephanie Young, MPH2, Michael Wernecke, BA3, Azadeh Shoaibi, MS, MHS1, Thomas E. MaCurdy, PhD3,4, Christopher Worrall, BS5, and Jeffrey A. Kelman, MD, MMSc5
Abstract Background: During the 2009 H1N1 influenza pandemic, the UK Medicines and Healthcare Products Regulatory Agency received case reports suggesting a potentiation of warfarin anticoagulation by the antiviral drug oseltamivir. We evaluated this putative interaction using Medicare data. Objective: To determine the frequency of bleeding following addition of oseltamivir or comparator drugs among Medicare beneficiaries taking warfarin. Methods: This was a retrospective cohort evaluation using Medicare nationwide data. Cohort members were Medicare Parts A, B, and D beneficiaries from June 30, 2006 to October 31, 2010 receiving warfarin for at least 1 month prior to a concomitant drug of interest (oseltamivir, ampicillin, trimethoprim–sulfamethoxazole (TMP-SMX), and angiotensin-converting enzyme (ACE) inhibitors). Bleeding within 14 days of new prescriptions for oseltamivir or comparators was identified using inpatient or emergency department ICD-9 (International Classification of Diseases, ninth revision) discharge diagnosis codes for gastrointestinal hemorrhage, epistaxis, hematuria, and intracranial bleeding. Patients with bleeding within 30 days preceding the prescription concomitant to warfarin were excluded. Results: With concomitant ACE inhibitors as reference, adjusted odds ratios (ORs) for any bleeding events within 14 days were 1.47 (95% confidence interval [CI] = 1.08-1.88), 1.24 (95% CI = 0.97-1.57), and 2.74 (95% CI = 2.53-3.03), for warfarin plus ampicillin, oseltamivir, and TMP-SMX, respectively. In a sensitivity analysis, adjusted ORs over a 7-day period were 1.89 (95% CI = 1.29-2.59), 1.47 (95% CI = 1.06-2.02), and 3.07 (95% CI = 2.76-3.49) for warfarin plus ampicillin, oseltamivir, and TMP-SMX, respectively. Conclusions: Bleeding with oseltamivir plus warfarin was not significantly increased over a 14-day observation period; a sensitivity analysis showed a statistically significant increase over a 7-day period; in contrast, the data consistently showed the known tendency of TMP-SMX to potentiate the effects of warfarin. The results should be interpreted with the limitations of this approach in mind, including the inability to control for unmeasured confounders. Keywords warfarin, oseltamivir, drug–drug interaction, influenza, bleeding Received February 22, 2013; revised May 9, 2013 and June 11, 2013; accepted June 13, 2013.
The Food and Drug Administration (FDA) Amendments Act of 2007 mandates that FDA develop a system for postmarket risk identification and analysis of drugs using electronic health care data (eg, insurance claims data, electronic health records data) from disparate data sources, both federal and private. In response to the Congressional mandate, FDA has initiated pilot programs to develop the scientific methodologies, data infrastructure, governance, and privacy policies to conduct active medical product safety surveillance in large observational databases. In the summer of 2009, with the advent of the H1N1 influenza pandemic, FDA’s Center for Drug Evaluation and Research strengthened its baseline monitoring of antiviral drug adverse effects. One enhancement was to leverage the
SafeRx project, a collaboration between FDA, the Office of the Assistant Secretary for Planning and Evaluation, and the 1
Center for Drug Evaluation and Research, U.S. Food and Drug Administration, Silver Spring, MD, USA 2 University of California, Los Angeles, CA, USA 3 Acumen LLC, Burlingame, CA, USA 4 Stanford University, Stanford, CA, USA 5 Centers for Medicare & Medicaid Services, Baltimore, MD, USA Corresponding Author: Andrew D. Mosholder, MD, MPH, FDA Center for Drug Evaluation and Research, Office of Pharmacovigilance and Epidemiology, Division of Epidemiology II, 10903 New Hampshire Avenue, Silver Spring, MD 20993-0002, USA. Email: [email protected]
Downloaded from aop.sagepub.com at NORTHERN ARIZONA UNIVERSITY on June 23, 2015
1421
Mosholder et al Centers for Medicare & Medicaid Services (CMS), to use Medicare data for near real-time medical product safety surveillance. In August 2009, the UK Medicines and Healthcare Products Regulatory Agency reported receiving postmarket case reports suggesting an interaction between warfarin and oseltamivir; however, it felt that the reports were not conclusive.1 This observation by the UK Medicines and Healthcare Products Regulatory Agency followed an earlier report by Health Canada in 2006 describing 19 spontaneous postmarket pharmacovigilance reports suggesting a possible interaction between oseltamivir and warfarin, resulting in higher international normalized ratios (INRs).2 More recently, in a case series of 15 patients treated concomitantly with oseltamivir and warfarin, 7 patients showed increased prothrombin time INRs from previously stable values; 3 of these 7 had clinical bleeding, and 1 of the 7 had another INR increase when rechallenged with the combination.3 Contrasting these case reports, a formal drug–drug interaction study, in which 20 chronic warfarin users were administered concomitant oseltamivir for 5 days, did not find evidence for an interaction.4 The vitamin K antagonist anticoagulant warfarin is stereoselectively metabolized by a variety of cytochrome P450 enzymes (2C9, 2C19, 2C8, 2C18, 1A2, and 3A4), with 2C9 providing most of the metabolic activity; the consequence is that warfarin is subject to many metabolic drug–drug interactions.5 In contrast to warfarin, neither oseltamivir nor its active metabolite oseltamivir carboxylate is a substrate for any cytochrome P450 isozymes,6 nor do they inhibit any cytochrome P450 isozymes.7 Neither oseltamivir nor its active metabolite has been shown to interfere with coagulation.4 Accordingly, there is no readily identifiable biologically plausible mechanism for a potentiation of warfarin anticoagulation by oseltamivir. Given the inconsistency between the case reports suggestive of a potential warfarin– oseltamivir interaction, and the clinical pharmacological profile of oseltamivir showing a minimal potential risk for drug interactions, FDA thought it useful to conduct an evaluation in the Medicare database seeking clinical evidence of a putative warfarin–oseltamivir interaction. We gathered surveillance data on Medicare beneficiaries’ exposure to warfarin and selected concomitant drugs. A CMS contractor, Acumen LLC, performed the data analyses. The purpose of the analysis was to determine the frequency of bleeding events following the combination of warfarin plus oseltamivir or selected comparator drugs, among Medicare Part D (prescription drug coverage) beneficiaries.
Methods Use of CMS data for postmarket surveillance activities has been approved by FDA’s Research Involving Human Subjects Committee. This retrospective cohort evaluation included Medicare Parts A, B, and D beneficiaries enrolled
at least 6 months from June 30, 2006 to October 31, 2010 who received warfarin for at least 1 month prior to a prescription for a concomitant drug of interest (oseltamivir or the comparator drugs ampicillin, trimethoprim–sulfamethoxazole [TMP-SMX], or angiotensin-converting enzyme inhibitors [ACEIs]). Patients prescribed more than one of the comparator drugs of interest were grouped by the first such drug received. Excluding patients prescribed a second drug of interest during the risk period would have eliminated events attributable to the second concomitant drug, but would have based eligibility on a postexposure event, which can introduce bias8; also, we expected few such patients. Two other antiviral drugs indicated for influenza, rimantadine and zanamivir, have not been used frequently in the Medicare population and were not included as comparators. Ampicillin and TMP-SMX were selected for their use in acute febrile illnesses, albeit for bacterial rather than viral infections. In addition, TMP-SMX interferes with warfarin metabolism,4 thereby providing a “positive control.” The class of ACEI was selected as a “negative control” because it is widely used in the Medicare population and is not labeled for a warfarin interaction4; thus it was chosen as the baseline comparator. We identified bleeding events occurring within 14 days of a new prescription for concomitant oseltamivir or a comparator using International Classification of Diseases, ninth revision (ICD-9) codes for epistaxis, gastrointestinal bleeding, hematuria, and intracranial bleeding. A combined event category enumerated patients with any of these events. Patients with any such code in the 30 days preceding the prescription concomitant to warfarin were excluded. A 14-day window for bleeding events was selected because there was a concern that it could take several days for a concomitant drug to interfere with warfarin metabolism to the extent that a bleeding outcome would occur and be observed and documented as an insurance claim; one recent observational study of warfarin and interacting antibiotics showed increased gastrointestinal bleeding after 5 or more days of co-prescribing.9 Additionally, because the course of oseltamivir varies from 5 to 10 days depending on whether it is used as treatment or prophylaxis, 14 days was considered to be an interval more likely to capture potentially related events. However, we also performed a sensitivity analysis with a shorter window of 7 days. To capture the most serious clinical events, we included Medicare Parts A and B discharge diagnosis codes from only inpatient stays and emergency department visits. Table 1 shows the list of selected bleeding events with the corresponding ICD-9 codes. We used multivariate logistic regression to estimate odds ratios (ORs) as approximations of relative risks of developing major bleeds within 2 weeks, adjusted for important potential confounders including demographic factors (agegroup, gender, geographical region, and income subsidy); recent hospitalization or emergency room visit;
Downloaded from aop.sagepub.com at NORTHERN ARIZONA UNIVERSITY on June 23, 2015
1422
Annals of Pharmacotherapy 47(11)
Table 1. Selected Hemorrhagic Events With Corresponding ICD-9 (International Classification of Diseases, Ninth Revision) Codes. Hemorrhagic Event Epistaxis Gastrointestinal bleeding
Hematuria Intracranial bleeding
ICD-9 Codes 784.7 530.21 531.2, 531.2x, 531.4, 531.4x, 531.6, 531.6x 532.0, 532.00, 532.01, 532.2, 532.2x, 532.4, 532.4x, 532.6, 532.6x 533.0, 533.00, 533.01, 533.2, 533.2x, 533.4, 533.4x, 533.6, 533.6x 534.0, 534.00, 534.01, 534.2, 534.2x, 534.4, 534.4x, 534.6, 534.6x 535.01, 535.11, 535.21, 535.31, 535.41, 535.51, 535.61, 537.83, 537.84 562.02, 562.03, 562.12, 562.13 569.85, 569.86 578, 578.x 599.7, 599.7x 430 431 432, 432.x
concomitant drugs potentially associated with bleeding (nonsteroidal anti-inflammatory drugs, selective serotonin reuptake inhibitors, proton pump inhibitors, platelet inhibitors); comorbid medical conditions including those that could affect bleeding (liver disease, cystitis, ulcer, diverticulosis, and other chronic medical conditions); warfarin duration; and drugs known to interact with warfarin. Except where otherwise indicated, covariates were assessed over a 6-month look-back period, consistent with the required minimum duration of Medicare enrollment; concomitant medications were assessed over a 30-day look-back period. To manage the large number of drugs known to interact with warfarin, those drugs were identified and grouped according to Holbrook et al’s10 systematic review of warfarin–drug interactions into the following categories: highly probable potentiator, probable potentiator, highly probable inhibitor, and probable inhibitor. The statistical significance level was determined at the .05 level, and all the analyses were performed using Stata version 11 (StataCorp, College Station, TX). Confidence intervals on the observed frequencies of patients with events in each cohort were calculated using the bootstrap method.11
Results Table 2 displays the cohort size and demographics, and Table 3 displays the numbers of outcome events. The main results are displayed in Table 4, with adjusted ORs for each drug combination cohort compared with the negative control (warfarin + ACEI) for the combined endpoint, and
stratified by bleeding site. Full results for covariates in the 14-day interval are presented in Tables A1 to A4 in the appendix. The size of the warfarin–oseltamivir and warfarin–ampicillin cohorts were about 1/10 the size of the warfarin–TMP-SMX and warfarin–ACEI cohorts. As evidenced in the table, over 14 days patients taking warfarin–oseltamivir had small, nonstatistically significant elevations in the likelihood of the combined endpoint, as well as gastrointestinal bleeding and epistaxis compared with patients taking warfarin + ACEI (OR = 1.24, 95% confidence interval [CI] = 0.97-1.57; OR = 1.31, 95% CI = 0.92-1.80; OR = 1.63, 95% CI = 0.84-2.70, respectively). In contrast, patients taking warfarin–TMP-SMX, the positive control cohort, were significantly more likely to have had the combined endpoint (OR = 2.74, 95% CI = 2.53-3.03), gastrointestinal bleeding (OR = 2.53, 95% CI = 2.27-2.93), intracranial bleeding (OR = 1.67, 95% CI = 1.25-2.40), epistaxis (OR = 2.07, 95% CI = 1.66-2.61), and hematuria (OR = 4.21, 95% CI = 3.49-5.12) compared with the negative control ACEI. The magnitudes of the associations of warfarin–TMP-SMX with the combined endpoint and gastrointestinal bleeding were higher than the magnitude of the associations of warfarin–ampicillin or warfarin–oseltamivir with the combined outcome and gastrointestinal bleeding, based on the nonoverlapping confidence intervals for the ORs in these comparisons. In addition, the patients taking ampicillin–warfarin had statistically significantly elevated likelihoods of having the combined bleeding endpoint as well as hematuria (OR = 1.47, 95% CI = 1.08-1.88; OR = 2.97, 95% CI = 1.97-4.37), respectively. Table 4 also summarizes unadjusted results and the 7-day sensitivity analysis. In comparison with the 14-day data shown, the results using 7 days as the risk window were notable for a significantly elevated adjusted OR for the association of warfarin–oseltamivir use with epistaxis (OR = 2.69, 95% CI = 1.14-4.60), and the combined endpoint (OR = 1.47, 95% CI = 1.06-2.02). As with the results based on the 14-day exposure window, the OR for warfarin–TMP-SMX with combined bleeding had a confidence interval that did not overlap with those for the combined bleeding 7-day ampicillin and oseltamivir ORs (7-day warfarin–TMP-SMX OR = 3.07, 95% CI = 2.76-3.49). Full results for covariates in the 14-day risk window are presented in Tables A1 to A4 in the appendix. With respect to other potentially interacting drugs, we found no statistically significant effects on bleeding events from warfarin inhibitors or potentiators, although bleeding events were increased with selective serotonin reuptake inhibitors, antiplatelet therapy, and nonsteroidal anti-inflammatory drugs (Table A3 in the appendix).
Discussion Although warfarin can produce bleeding from any site, gastrointestinal bleeding is the most frequently reported
Downloaded from aop.sagepub.com at NORTHERN ARIZONA UNIVERSITY on June 23, 2015
1423
Mosholder et al Table 2. Summary of Demographics. ACE Inhibitors Demographics
n
Total Gender Male Female Age category (years) 84 Low-income subsidy (LIS)a No LIS LIS with 15% copay LIS with high copay LIS with low copay LIS with zero copay Region Northeast Southeast South Midwest West Other or missing
Ampicillin
Oseltamivir n
TMP-SMX
%
n
%
%
n
%
158 724
100.0
9150
100.0
13 406
100.0
143 076
100.0
71 754 86 970
45.2 54.8
2900 6250
31.7 68.3
4958 8448
37.0 63.0
47 191 95 885
33.0 67.0
17 768 50 161 61 868 28 927
11.2 31.6 39.0 18.2
1041 2010 3416 2683
11.4 22.0 37.3 29.3
2041 3468 4538 3359
15.2 25.9 33.9 25.1
21 168 35 899 51 458 34 551
14.8 25.1 36.0 24.1
100 007 2968 17 498 26 773 11 478
63.0 1.9 11.0 16.9 7.2
4074 99 723 1413 2841
44.5 1.1 7.9 15.4 31.0
6231 155 1118 2135 3767
46.5 1.2 8.3 15.9 28.1
70 110 2212 15 469 25 457 29 828
49.0 1.5 10.8 17.8 20.8
32 394 46 275 14 782 44 289 20 758 226
20.4 29.2 9.3 27.9 13.1 0.1
2083 2937 837 2324 937 32
22.8 32.1 9.1 25.4 10.2 0.3
3264 3961 1208 3431 1519 23
24.3 29.5 9.0 25.6 11.3 0.2
25 137 47 098 14 892 38 270 17 540 139
17.6 32.9 10.4 26.7 12.3 0.1
Abbreviations: ACE, angiotensin-converting enzyme; TMP-SMX, trimethoprim–sulfamethoxazole. a Eligibility for Part D low-income subsidy is used to proxy for income. Poorer beneficiaries are required to pay smaller copays.
Table 3. Outcome Counts and Rates per Thousand Beneficiaries (14-Day Risk Window). Combined Cohort ACE inhibitor Ampicillin Oseltamivir TMP-SMX
Gastrointestinal
Intracranial
Epistaxis
Hematuria
Cohort Size
Count
Rate
Count
Rate
Count
Rate
Count
Rate
Count
Rate
158 724 9150 13 406 143 076
744 63 66 1747
4.69 6.89 4.92 12.21
417 31 39 911
2.63 3.39 2.91 6.37
66 2 4 112
0.42 0.22 0.30 0.78
126 5 13 199
0.79 0.55 0.97 1.39
152 26 11 560
0.96 2.84 0.82 3.91
Abbreviations: ACE, angiotensin-converting enzyme; TMP-SMX, trimethoprim–sulfamethoxazole.
hemorrhagic complication of warfarin anticoagulation in postmarket surveillance,12 which is consistent with our findings. Compared with the other concomitant drugs assessed, the positive control cohort, warfarin–TMP-SMX, had significantly elevated adjusted ORs for gastrointestinal bleeding events and the combined bleeding events. We did not find compelling evidence to support an interaction between oseltamivir and warfarin resulting in increased bleeding events among Medicare beneficiaries. In the 14-day results, OR point estimates for some outcomes were elevated for the warfarin–oseltamivir cohort, but they were not statistically significant. In the 7-day sensitivity analysis, warfarin–oseltamivir showed a modest but
statistically significant increase in combined events, with the highest point estimate among individual events observed for warfarin–oseltamivir and epistaxis. However, patients in our sample being treated for influenza with oseltamivir might have nasal discharge, which could increase the risk of epistaxis in that time frame; one case series found epistaxis was associated with influenza.13 We speculate that the association of ampicillin specifically with hematuria may reflect patients receiving ampicillin for urinary tract infections, which can produce hematuria. We focused on a 14-day risk window based on the projected time it might take for a drug interaction with warfarin to manifest clinically and be detected in an insurance claims
Downloaded from aop.sagepub.com at NORTHERN ARIZONA UNIVERSITY on June 23, 2015
1424
Annals of Pharmacotherapy 47(11)
Table 4. Main Results of All Models (7-Day and 14-Day, Adjusted and Unadjusted). Cohort
Combined
Gastrointestinal
Unadjusted odds ratio (95% confidence interval), 7 days ACE inhibitor 1.00 (1.00, 1.00) 1.00 Ampicillin 1.87 (1.28, 2.57) 1.47 Oseltamivir 1.24 (0.90, 1.71) 1.28 TMP-SMX 2.93 (2.64, 3.32) 2.44 Adjusted odds ratio (95% confidence interval), 7 days ACE inhibitor 1.00 (1.00, 1.00) 1.00 Ampicillin 1.89 (1.29, 2.59) 1.46 Oseltamivir 1.47 (1.06, 2.02) 1.53 TMP-SMX 3.07 (2.76, 3.49) 2.56 Unadjusted odds ratio (95% confidence interval), 14 days ACE inhibitor 1.00 (1.00, 1.00) 1.00 Ampicillin 1.47 (1.08, 1.88) 1.29 Oseltamivir 1.05 (0.83, 1.33) 1.11 TMP-SMX 2.62 (2.43, 2.89) 2.43 Adjusted odds ratio (95% confidence interval), 14 days ACE inhibitor 1.00 (1.00, 1.00) 1.00 Ampicillin 1.47 (1.08, 1.88) 1.26 Oseltamivir 1.24 (0.97, 1.57) 1.31 TMP-SMX 2.74 (2.53, 3.03) 2.53
Intracranial
Epistaxis
Hematuria
(1.00, 1.00) (0.88, 2.16) (0.80, 1.92) (2.18, 2.85)
1.00 0.50 1.01 1.97
(1.00, 1.00) (0.00, 1.85) (0.00, 2.72) (1.28, 3.16)
1.00 0.93 2.12 2.08
(1.00, 1.00) (0.00, 2.18) (0.91, 3.72) (1.52, 2.93)
1.00 3.86 0.44 5.00
(1.00, 1.00) (2.15, 6.13) (0.00, 1.02) (3.88, 6.62)
(1.00, 1.00) (0.85, 2.13) (0.96, 2.34) (2.23, 3.07)
1.00 0.47 1.21 1.97
(1.00, 1.00) (0.00, 1.84) (0.00, 3.11) (1.23, 3.16)
1.00 1.08 2.69 2.36
(1.00, 1.00) (0.00, 2.45) (1.14, 4.60) (1.72, 3.52)
1.00 3.84 0.50 5.14
(1.00, 1.00) (2.13, 6.15) (0.00, 1.15) (3.96, 6.92)
(1.00, 1.00) (0.87, 1.82) (0.78, 1.51) (2.20, 2.74)
1.00 0.53 0.72 1.88
(1.00, 1.00) (0.00, 1.41) (0.15, 1.73) (1.43, 2.70)
1.00 0.69 1.22 1.75
(1.00, 1.00) (0.25, 1.33) (0.61, 1.95) (1.41, 2.25)
1.00 2.97 0.86 4.10
(1.00, 1.00) (2.01, 4.35) (0.34, 1.36) (3.41, 4.85)
(1.00, 1.00) (0.84, 1.77) (0.92, 1.80) (2.27, 2.93)
1.00 0.45 0.71 1.67
(1.00, 1.00) (0.00, 1.22) (0.14, 1.76) (1.25, 2.40)
1.00 0.83 1.63 2.07
(1.00, 1.00) (0.29, 1.60) (0.84, 2.70) (1.66, 2.61)
1.00 2.97 0.99 4.21
(1.00, 1.00) (1.97, 4.37) (0.41, 1.54) (3.49, 5.12)
Abbreviations: ACE, angiotensin-converting enzyme; TMP-SMX, trimethoprim–sulfamethoxazole.
system; additionally, previous work suggested that increases in bleeding events via warfarin drug–drug interactions were most prominent in the 6- to 15-day period.7 The adjusted ORs with warfarin–TMP-SMX were statistically significant at both 7 and 14 days for every bleeding outcome analyzed. This consistent pattern was unique to TMP-SMX. Although the combined events ORs for oseltamivir at 7 days, and for ampicillin at 7 and 14 days, were statistically significant, the confidence intervals for those ORs had no overlap with the corresponding OR confidence intervals for warfarin–TMP-SMX. Because the elevated ORs observed with warfarin-TMPSMX showed the capability of the analysis to identify the positive control, we have some reassurance that this drug– drug interaction assessment had the potential to discover a major effect on bleeding by the combination of warfarin and oseltamivir if there was one. However, this assessment has important limitations. Our results, especially from a statistical standpoint, could be consistent with a modest association with bleeding events. However, it seems unlikely from our results that oseltamivir could have an effect comparable to TMP-SMX. Our results also are inconsistent with an effect as frequent as that described in the case series by Lee et al,3 a discrepancy for which we do not have a ready explanation, although the specific clinical bleeding events in the 3 patients from their case series (bloody sputum in 2, and bloody ascites in a third) would not have met our outcome definition. Another methodologic limitation was our
exclusive reliance on diagnosis billing codes, without examination of medical records to validate the occurrence of the specific outcomes; however, 2 studies in other settings that assessed sets of ICD-9 gastrointestinal disorder codes similar to our gastrointestinal bleeding category showed generally good positive predictive values.14,15 Similarly, we had to rely on prescription claims as a surrogate for drug exposure. Some patients may have filled osletamivir prescriptions for future use (“stockpiling”) and been inappropriately considered exposed to oseltamivir. However, during the 2009 pandemic H1N1 influenza outbreak, prescribing trends for oseltamivir closely paralleled the epidemic curve for influenza-like illnesses, suggesting that stockpiling of prescriptions was likely not common.16 Episodes of bleeding that did not receive medical attention, such as fatal intracranial hemorrhage that resulted in out of hospital death, could not be ascertained. Additionally, elevations of clotting times, as were reported in postmarketing surveillance and by Lee et al,3 could not be detected in the Medicare database. The analysis is also limited by sample size for some drug–drug combinations. For example, the warfarin–oseltamivir cohort was substantially smaller than the ACEI and TMP-SMX comparator cohorts. The numbers of events were also smaller for the nongastrointestinal bleeding sites. Another challenge is the need to control for the many potential confounders in the setting of warfarin use including the many drugs that interact with warfarin metabolism, the many drugs that may potentiate bleeding
Downloaded from aop.sagepub.com at NORTHERN ARIZONA UNIVERSITY on June 23, 2015
1425
Mosholder et al on their own, and the many health conditions that may predispose to bleeding. We attempted to address these confounders in the model, but the potential for unmeasured confounding certainly remains. In particular, the modest increase in bleeding events observed over the 7-day window with oseltamivir could be due to confounding factors related to acute influenza, such as changes in diet, elevated temperature, concomitant medications, gastrointestinal function, and the like, which could alter INR in the absence of a true drug–drug interaction. Such effects might explain some of the reported cases of increased INR or bleeding with the warfarin–oseltamivir combination. There is evidence that influenza infection enhances coagulation.17,18 Speculatively, such an effect could mask a potentiation of warfarin anticoagulation by oseltamivir, although only with its use in treatment of influenza rather than prophylaxis. Our analysis was not capable of distinguishing prophylactic use from treatment of influenza. The results of an assessment such as this are considered along with the body of evidence that FDA holds on a given safety concern. One can foresee 3 potential outcomes: •• The body of evidence does not suggest an elevated risk and standard monitoring is continued.
•• The body of evidence suggests an elevated risk, but questions remain about the strength or validity of the safety concern, and a more in-depth assessment may be needed. •• The body of evidence regarding the elevated risk is so strong that FDA proceeds to a regulatory action. In the case of the putative warfarin–oseltamivir interaction, the body of evidence supports continuation of standard monitoring. In conclusion, our analysis of claims for bleeding events among Medicare patients co-prescribed oseltamivir and warfarin failed to show clear evidence of a drug–drug interaction, consistent with the reported absence of a pharmacokinetic interaction between these drugs.4 In the same analysis, we successfully identified the predicted potentiation of anticoagulation by concomitant TMP-SMX, evidenced by the increased risk of bleeding events. Although future observational studies might be able to address some limitations of the present analysis, on balance, we felt that with its large size and substantial capture of warfarin-treated patients, the Medicare database functioned well in providing relevant information regarding the potential drug interaction in a timely fashion.
Appendix Table A1. Patient Demographics.a Adjusted Odds Ratio for Outcome (95% Confidence Interval) Patient Demographics Age bracket (years) 84 Gender Male Female Low-income subsidy (LIS) No LIS LIS with 15% copay LIS with high copay LIS with low copay LIS with zero copay Region Northeast Southeast South Midwest West Other or unknown
Combined
Gastrointestinal
Intracranial
Epistaxis
Hematuria
1.00 1.04 1.31 1.66
(1.00, 1.00) (0.91, 1.20) (1.13, 1.49) (1.38, 1.90)
1.00 1.26 1.59 2.11
(1.00, 1.00) (1.02, 1.60) (1.29, 1.93) (1.67, 2.59)
1.00 1.52 1.94 2.88
(1.00, 1.00) (0.89, 3.52) (1.08, 5.02) (1.53, 6.73)
1.00 0.70 0.89 0.95
(1.00, 1.00) (0.48, 0.99) (0.64, 1.27) (0.65, 1.33)
1.00 0.87 1.09 1.26
(1.00, 1.00) (0.66, 1.13) (0.85, 1.36) (0.98, 1.60)
1.00 0.82
(1.00, 1.00) (0.75, 0.90)
1.00 1.13
(1.00, 1.00) (1.00, 1.31)
1.00 0.96
(1.00, 1.00) (0.72, 1.33)
1.00 0.93
(1.00, 1.00) (0.71, 1.14)
1.00 0.42
(1.00, 1.00) (0.36, 0.48)
1.00 0.93 1.11 1.08 1.11
(1.00, 1.00) (0.66, 1.30) (0.95, 1.25) (0.97, 1.20) (0.97, 1.26)
1.00 0.80 1.22 1.11 1.17
(1.00, 1.00) (0.43, 1.32) (1.02, 1.45) (0.95, 1.32) (0.97, 1.38)
1.00 1.11 1.12 1.31 0.98
(1.00, 1.00) (0.00, 2.81) (0.61, 1.86) (0.86, 2.04) (0.56, 1.76)
1.00 1.20 1.05 0.89 0.81
(1.00, 1.00) (0.45, 2.19) (0.70, 1.45) (0.63, 1.21) (0.56, 1.20)
1.00 0.91 0.96 1.07 1.16
(1.00, 1.00) (0.45, 1.53) (0.71, 1.23) (0.88, 1.29) (0.93, 1.50)
1.01 1.00 1.03 0.99 0.90 0.34
(0.90, 1.13) (1.00, 1.00) (0.88, 1.20) (0.89, 1.10) (0.77, 1.01) (0.00, 1.15)
0.88 1.00 0.96 0.84 0.82 0.64
(0.75, 1.03) (1.00, 1.00) (0.79, 1.14) (0.72, 0.97) (0.68, 0.97) (0.00, 2.25)
0.78 1.00 0.92 1.12 1.01 0.00
(0.50, 1.19) (1.00, 1.00) (0.54, 1.56) (0.76, 1.58) (0.59, 1.57) (0.00, 0.00)
1.16 1.00 1.16 1.21 1.23 0.00
(0.87, 1.58) (1.00, 1.00) (0.77, 1.72) (0.95, 1.59) (0.82, 1.72) (0.00, 0.00)
1.32 1.00 1.18 1.21 0.90 0.00
(1.08, 1.67) (1.00, 1.00) (0.93, 1.53) (1.01, 1.46) (0.72, 1.16) (0.00, 0.00)
a
Full results for primary models predicting frequency of hemorrhagic events within 14 days of a drug cohort prescription among Medicare beneficiaries on warfarin for at least 30 days.
Downloaded from aop.sagepub.com at NORTHERN ARIZONA UNIVERSITY on June 23, 2015
1426
Annals of Pharmacotherapy 47(11)
Table A2. Patient Hospital Encounters.a Adjusted Odds Ratio for Outcome (95% Confidence Interval) Hospital Encounters
Combined
Gastrointestinal
Hospitalizations in past 30 days 0 1.00 (1.00, 1.00) 1 1.16 (1.02, 1.34) 2+ 1.34 (1.05, 1.70) Emergency room visits in past 30 days 0 1.00 (1.00, 1.00) 1 1.45 (1.29, 1.61) 2+ 1.93 (1.57, 2.32) Hospitalizations in 180 to 31 days prior 0 1.00 (1.00, 1.00) 1 1.09 (0.94, 1.22) 2+ 1.11 (0.95, 1.33) Emergency room visits in 180 to 31 days prior 0 1.00 (1.00, 1.00) 1 1.06 (0.94, 1.19) 2+ 1.25 (1.09, 1.48)
Intracranial
Epistaxis
Hematuria
1.00 1.10 1.58
(1.00, 1.00) (0.93, 1.28) (1.22, 2.09)
1.00 0.83 1.36
(1.00, 1.00) (0.48, 1.39) (0.53, 2.57)
1.00 1.50 1.14
(1.00, 1.00) (1.05, 2.10) (0.63, 2.11)
1.00 1.20 0.80
(1.00, 1.00) (0.94, 1.52) (0.45, 1.23)
1.00 1.44 1.71
(1.00, 1.00) (1.23, 1.68) (1.32, 2.23)
1.00 1.97 2.42
(1.00, 1.00) (1.24, 2.92) (1.18, 4.00)
1.00 1.20 1.94
(1.00, 1.00) (0.83, 1.59) (1.17, 2.66)
1.00 1.53 2.42
(1.00, 1.00) (1.21, 1.83) (1.67, 3.39)
1.00 1.08 1.14
(1.00, 1.00) (0.92, 1.26) (0.90, 1.43)
1.00 0.84 1.13
(1.00, 1.00) (0.55, 1.27) (0.62, 1.97)
1.00 1.01 1.00
(1.00, 1.00) (0.63, 1.42) (0.71, 1.52)
1.00 1.16 1.13
(1.00, 1.00) (0.89, 1.46) (0.82, 1.58)
1.00 1.06 1.15
(1.00, 1.00) (0.91, 1.24) (0.94, 1.41)
1.00 1.01 0.83
(1.00, 1.00) (0.64, 1.55) (0.47, 1.39)
1.00 1.02 1.61
(1.00, 1.00) (0.72, 1.40) (1.09, 2.20)
1.00 1.09 1.33
(1.00, 1.00) (0.87, 1.34) (1.03, 1.75)
a Full results for primary models predicting frequency of hemorrhagic events within 14 days of a drug cohort prescription among Medicare beneficiaries on warfarin for at least 30 days.
Table A3. Patient Medications.a Adjusted Odds Ratio for Outcome (95% Confidence Interval) Warfarin duration (days) 30-89 90-179 180+ Concomitant medicationsb Highly probable potentiators Probable potentiators Highly probable inhibitors Probable inhibitors Platelet inhibitors NSAIDs PPI SSRI
Combined
Gastrointestinal
Intracranial
Epistaxis
Hematuria
1.00 0.98 0.98
(1.00, 1.00) (0.87, 1.12) (0.87, 1.13)
1.00 0.99 0.99
(1.00, 1.00) (0.84, 1.19) (0.82, 1.17)
1.00 0.81 0.92
(1.00, 1.00) (0.56, 1.37) (0.58, 1.62)
1.00 1.22 1.27
(1.00, 1.00) (0.92, 1.87) (0.92, 1.99)
1.00 0.96 0.91
(1.00, 1.00) (0.76, 1.26) (0.69, 1.20)
1.01 0.96 0.90 1.03 1.48 1.24 0.91 1.11
(0.92, 1.11) (0.89, 1.04) (0.71, 1.07) (0.89, 1.18) (1.29, 1.72) (1.08, 1.43) (0.83, 1.00) (1.00, 1.22)
0.98 1.07 0.92 1.06 1.66 1.39 0.95 1.18
(0.87, 1.09) (0.96, 1.22) (0.74, 1.14) (0.89, 1.21) (1.38, 2.03) (1.14, 1.66) (0.85, 1.08) (1.02, 1.33)
0.94 0.87 1.23 1.18 0.79 1.39 0.85 1.23
(0.69, 1.28) (0.64, 1.14) (0.57, 2.04) (0.59, 1.83) (0.27, 1.56) (0.71, 2.08) (0.56, 1.19) (0.91, 1.74)
1.10 0.95 0.94 1.46 2.03 1.02 0.89 1.12
(0.85, 1.42) (0.78, 1.21) (0.49, 1.39) (1.03, 1.98) (1.29, 2.82) (0.61, 1.48) (0.67, 1.18) (0.86, 1.41)
1.03 0.78 0.78 0.79 1.08 0.98 0.88 0.95
(0.88, 1.24) (0.67, 0.89) (0.52, 1.08) (0.57, 1.05) (0.74, 1.47) (0.68, 1.26) (0.73, 1.02) (0.77, 1.11)
Abbreviations: NSAIDs, nonsteroidal anti-inflammatory drugs; PPI, proton pump inhibitor; SSRI, selective serotonin reuptake inhibitor. a Full results for primary models predicting frequency of hemorrhagic events within 14 days of a drug cohort prescription among Medicare beneficiaries on warfarin for at least 30 days. b Drugs are categorized using Holbrook et al’s10 systematic review of warfarin interactions.
Downloaded from aop.sagepub.com at NORTHERN ARIZONA UNIVERSITY on June 23, 2015
1427
Mosholder et al Table A4. Patient Health Characteristics.a Patient Health Characteristics Chronic health conditions Chronic liver disease Chronic renal failure Additional comorbidities Acute renal failure Acute myocardial infarction, 30-0 days prior Acute myocardial infarction, 180-31 days prior Calculus of kidney Calculus of urinary tract Cancer, kidney/prostrate/ bladder Cerebral arterial occlusion Cystitis Dementia Diverticulosis Heart failure, 30-0 days prior Heart failure, 180-31 days prior Hypertension Ischemic stroke Pyelonephritis Metastatic cancer Pneumonia Pneumothorax/pleural effusion Severe cancer Ulcer
Adjusted Odds Ratio for Outcome (95% Confidence Interval) Combined
Gastrointestinal
Intracranial
Epistaxis
Hematuria
1.35 1.24
(1.05, 1.73) (1.12, 1.37)
1.47 1.40
(1.10, 1.91) (1.22, 1.60)
1.48 1.14
(0.17, 3.07) (0.73, 1.64)
1.60 1.11
(0.71, 2.52) (0.87, 1.45)
1.12 1.06
(0.65, 1.72) (0.87, 1.30)
1.04 1.26
(0.90, 1.21) (0.96, 1.60)
1.05 1.22
(0.86, 1.29) (0.85, 1.64)
1.11 0.22
(0.60, 1.62) (0.00, 0.69)
1.15 1.49
(0.74, 1.69) (0.68, 2.30)
0.89 1.50
(0.59, 1.14) (0.75, 2.53)
1.04
(0.80, 1.30)
0.99
(0.72, 1.31)
3.20
(1.47, 5.48)
0.74
(0.32, 1.29)
0.90
(0.46, 1.47)
1.18 1.47 1.27
(0.92, 1.44) (0.68, 2.34) (1.08, 1.49)
0.99 1.25 1.06
(0.67, 1.34) (0.23, 2.47) (0.79, 1.40)
0.91 0.00 1.05
(0.21, 1.83) (0.00, 0.00) (0.49, 1.78)
0.60 1.40 0.72
(0.11, 1.24) (0.00, 4.67) (0.40, 1.19)
1.83 1.49 1.60
(1.32, 2.46) (0.61, 2.62) (1.24, 1.99)
1.05 1.19 1.07 1.10 1.12
(0.92, 1.18) (0.99, 1.41) (0.92, 1.24) (0.95, 1.24) (0.98, 1.23)
1.07 0.74 1.07 1.22 1.29
(0.92, 1.27) (0.51, 0.97) (0.91, 1.28) (1.00, 1.43) (1.09, 1.46)
1.34 1.10 1.41 0.93 0.65
(0.86, 2.00) (0.41, 1.87) (0.87, 2.20) (0.42, 1.68) (0.43, 1.02)
1.25 1.00 0.86 0.96 1.28
(0.90, 1.65) (0.51, 1.64) (0.60, 1.22) (0.52, 1.38) (0.97, 1.67)
0.83 2.28 1.06 0.97 0.87
(0.63, 1.04) (1.77, 3.09) (0.82, 1.33) (0.69, 1.28) (0.71, 1.09)
1.15
(1.04, 1.27)
1.13
(0.99, 1.29)
1.42
(0.93, 2.06)
1.30
(1.02, 1.76)
1.03
(0.85, 1.24)
1.05 0.95 1.12 1.35 1.00 1.10
(0.89, 1.24) (0.85, 1.07) (0.79, 1.50) (1.03, 1.67) (0.85, 1.15) (0.97, 1.26)
1.00 0.92 1.36 1.15 1.02 1.21
(0.80, 1.25) (0.77, 1.06) (0.89, 2.04) (0.75, 1.57) (0.82, 1.24) (1.00, 1.43)
0.83 1.41 0.82 1.73 1.18 1.09
(0.31, 1.50) (0.87, 2.12) (0.00, 2.23) (0.46, 3.84) (0.54, 2.09) (0.65, 1.89)
1.13 0.73 0.41 1.71 0.94 1.21
(0.74, 1.74) (0.50, 1.00) (0.00, 1.11) (0.74, 2.88) (0.57, 1.33) (0.82, 1.74)
1.18 1.07 1.05 1.61 0.95 0.82
(0.78, 1.64) (0.85, 1.33) (0.52, 1.57) (1.09, 2.33) (0.69, 1.26) (0.62, 1.11)
1.13 1.58
(0.88, 1.45) (1.27, 1.95)
1.22 1.96
(0.87, 1.79) (1.45, 2.48)
0.88 0.94
(0.12, 2.03) (0.00, 1.99)
0.93 1.68
(0.34, 1.78) (0.78, 2.71)
1.13 0.86
(0.71, 1.74) (0.39, 1.47)
a Full results for primary models predicting frequency of hemorrhagic events within 14 days of a drug cohort prescription among Medicare beneficiaries on warfarin for at least 30 days.
Authors’ Note The views expressed in this manuscript represent the opinions of the authors, and do not necessarily represent the views of the US Food and Drug Administration, the Centers for Medicare & Medicaid Services, or the US Department of Health and Human Services. This work was presented in abstract form at the 27th International Conference on Pharmacoepidemiology, August 14-17, 2011, Chicago, Illinois. An abstract was published in Pharmacoepidemiology and Drug Safety, August 2011, Volume 20, Issue Supplement S1.
Acknowledgments We thank the Centers for Medicare & Medicaid Services for their operational leadership and analytic support, and the Health and
Human Services Office of the Assistant Secretary for Planning and Evaluation for their financial support of this project.
Declaration of Conflicting Interests The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the SafeRx Project, a joint initiative of the Centers for Medicare & Medicaid Services (CMS), US Food and Drug Administration (FDA), and the US Department of
Downloaded from aop.sagepub.com at NORTHERN ARIZONA UNIVERSITY on June 23, 2015
1428
Annals of Pharmacotherapy 47(11)
Health and Human Services (HHS) Office of the Assistant Secretary for Planning and Evaluation (ASPE). Acumen LLC is a contractor to CMS.
References 1. MHRA suspected adverse drug reaction (ADR) analysis: influenza antivirals—Tamiflu (oseltamivir) and Relenza (zanamivir), August 21, 2009. http://www.mhra.gov.uk/ home/groups/pl-p/documents/websiteresources/con054667. pdf. Accessed October 19, 2011. 2. Health Products and Food Branch, Marketed Health Products Directorate. Can Adverse React Newslett. 2006;16:1, 1-2. 3. Lee SH, Kang HR, Jung JW, et al. Effect of oseltamivir on bleeding risk associated with warfarin therapy: a retrospective review. Clin Drug Investig. 2012;32:131-137. 4. Davies BE, Aceves-Baldo P, Lennon-Chrimes S, Brewster M. Effect of oseltamivir treatment on anticoagulation: a crossover study in warfarinized patients. Br J Clin Pharmacol. 2010;70:834-843. 5. Bristol-Myers Squibb Company. Coumadin [package insert]. http://packageinserts.bms.com/pi/pi_coumadin.pdf. Accessed October 19, 2011. 6. Dutkowski R, Thakrar B, Froehlich E, Suter P, Oo C, Ward P. Safety and pharmacology of oseltamivir in clinical use. Drug Saf. 2003;26:787-801. 7. He G, Massarella J, Ward P. Clinical pharmacokinetics of the prodrug oseltamivir and its active metabolite Ro 64-0802. Clin Pharmacokinet. 1999;37:471-484. 8. Rothman KJ, Greenland S. Cohort studies. In: Rothman KJ, Greenland S, Lash TL, eds. Modern Epidemiology. 3rd ed. Philadelphia, PA: Lippincott Williams & Wilkins; 2008:100-110. 9. Schelleman H, Bilker WB, Brensinger CM, Han X, Kimmel SE, Hennessy S. Warfarin with fluoroquinolones, sulfonamides, or azole antifungals: interactions and the risk of
hospitalization for gastrointestinal bleeding. Clin Pharmacol Ther. 2008;84:581-588. 10. Holbrook AM, Pereira JA, Labiris R, et al. Systematic overview of warfarin and its drug and food interactions. Arch Intern Med. 2005;165:1095-1106. 11. Efron B. Bootstrap methods: another look at the jackknife. Ann Stat. 1979;7:1-26. 12. Wysowski DK, Nourjah P, Swartz L. Bleeding complications with warfarin use: a prevalent adverse effect resulting in regulatory action. Arch Intern Med. 2007;167:1414-1419. 13. Duque V, Vaz J, Mota V, Morais C, Da Cunha S, MeliçoSilvestre A. Clinical manifestations of pandemic (H1N1) 2009 in the ambulatory setting. J Infect Dev Ctries. 2011;5:658663. 14. Raiford DS, Pérez Gutthann S, García Rodríguez LA. Positive predictive value of ICD-9 codes in the identification of cases of complicated peptic ulcer disease in the Saskatchewan hospital automated database. Epidemiology. 1996;7:101-104. 15. Cattaruzzi C, Troncon MG, Agostinis L, García Rodríguez LA. Positive predictive value of ICD-9th codes for upper gastrointestinal bleeding and perforation in the Sistema Informativo Sanitario Regionale database. J Clin Epidemiol. 1999;52:499-502. 16. Borders-Hemphill V, Mosholder A. U.S. utilization pat terns of influenza antiviral medications during the 2009 H1N1 influenza pandemic. Influenza Other Respi Viruses. 2012;6:e129-e133. 17. Madjid M, Aboshady I, Awan I, Litovsky S, Casscells SW. Influenza and cardiovascular disease: Is there a causal relationship? Tex Heart Inst J. 2004;31:4-13. 18. Keller TT, van der Sluijs KF, de Kruif MD, et al. Effects on coagulation and fibrinolysis induced by influenza in mice with a reduced capacity to generate activated protein C and a deficiency in plasminogen activator inhibitor type 1. Circ Res. 2006;99:1261-1269.
Downloaded from aop.sagepub.com at NORTHERN ARIZONA UNIVERSITY on June 23, 2015
