Pharmacogenomics, the study of interactions of genetics

Transcription

1 A Randomized and Clinical Effectiveness Trial Comparing Two Pharmacogenetic Algorithms and Standard Care for Individualizing Warfarin Dosing (CoumaGen-II) Jeffrey L. Anderson, MD; Benjamin D. Horne, PhD, MPH; Scott M. Stevens, MD; Scott C. Woller, MD; Kent M. Samuelson, MD; Justin W. Mansfield, MD; Michelle Robinson, BS; Stephanie Barton, PharmD; Kim Brunisholz, MST; Chrissa P. Mower, BS; John A. Huntinghouse, BS; Jeffrey S. Rollo, BS; Dustin Siler, BA; Tami L. Bair, BS; Stacey Knight, PhD, MStat; Joseph B. Muhlestein, MD; John F. Carlquist, PhD Downloaded from by guest on July 19, 2018 Background Warfarin is characterized by marked variations in individual dose requirements and a narrow therapeutic window. Pharmacogenetics (PG) could improve dosing efficiency and safety, but clinical trials evidence is meager. Methods and Results A Randomized and Clinical Effectiveness Trial Comparing Two Pharmacogenetic Algorithms and Standard Care for Individualizing Warfarin Dosing (CoumaGen-II) comprised 2 comparisons: (1) a blinded, randomized comparison of a modified 1-step (PG-1) with a 3-step algorithm (PG-2) (N 504), and (2) a clinical effectiveness comparison of PG guidance with use of either algorithm with standard dosing in a parallel control group (N 1866). A rapid method provided same-day CYP2C9 and VKORC1 genotyping. Primary outcomes were percentage of out-of-range international normalized ratios at 1 and 3 months and percentage of time in therapeutic range. Primary analysis was modified intention to treat. In the randomized comparison, PG-2 was noninferior but not superior to PG-1 for percentage of out-of-range international normalized ratios at 1 month and 3 months and for percentage of time in therapeutic range at 3 months. However, the combined PG cohort was superior to the parallel controls (percentage of out-of-range international normalized ratios 31% versus 42% at 1 month; 30% versus 42% at 3 months; percentage of time in therapeutic range 69% versus 58%, 71% versus 59%, respectively, all P 0.001). Differences persisted after adjustment for age, sex, and clinical indication. There were fewer percentage international normalized ratios 4 and 1.5 and serious adverse events at 3 months (4.5% versus 9.4% of patients, P 0.001) with PG guidance. Conclusions These findings suggest that PG dosing should be considered for broader clinical application, a proposal that is being tested further in 3 major randomized trials. The simpler 1-step PG algorithm provided equivalent results and may be preferable for clinical application. Clinical Trial Registration URL: Unique identifier: NCT (Circulation. 2012;125: ) Key Words: anticoagulants clinical trial genetics pharmacogenetics warfarin Pharmacogenomics, the study of interactions of genetics with pharmacotherapy, is a promising area for applying genetics to personalized or precision medicine. 1 3 Warfarin is prescribed to over 2 million patients in the United States for prevention of thromboembolic events associated with atrial fibrillation, venous and arterial thrombosis, orthopedic surgery, and prosthetic heart valves. Unfortunately, clinical management is difficult because of a narrow therapeutic index and marked interpatient variability in drug pharmacokinetics and pharmacodynamics, which lead to unpredictable and variable (up to 10-fold or greater) dosing requirements. 4 Anticoagulation trials for nonrheumatic atrial fibrillation have determined the optimal prothrombin time international normalized ratio (INR) range to be 2 to 3 with ratios 2 increasing thrombotic events and those 4 increasing hemorrhagic events. 5,6 Received September 29, 2011; accepted February 17, From the Intermountain Healthcare (Intermountain Medical Center and LDS and McKay-Dee Hospitals), Murray, UT (J.L.A., B.D.H., S.M.S., S.C.W., K.M.S., J.W.M., M.R., S.B., K.B., C.P.M., J.A.H., J.S.R., D.S., T.L.B., S.K., J.B.M., J.F.C.); and University of Utah School of Medicine, Salt Lake City, UT (J.L.A., B.D.H., S.M.S., S.C.W., K.M.S., S.B., K.B., S.K., J.B.M., J.F.C.). The online-only Data Supplement is available with this article at /-/DC1. Correspondence to Jeffrey L. Anderson, MD, Intermountain Medical Center, Cardiovascular Department, 5121 S Cottonwood St, Murray, UT jeffrey.anderson@imail.org 2012 American Heart Association, Inc. Circulation is available at DOI: /CIRCULATIONAHA

2 1998 Circulation April 24, 2012 Downloaded from by guest on July 19, 2018 Editorial see p 1964 Clinical Perspective on p 2005 Genotypes of the cytochrome p450 isoform, CYP2C9, and the vitamin K epoxide reductase complex subunit 1, VKORC1, have been found to conjointly determine warfarin dose requirements The *2 (R144C) and *3 (I359L) allelic variants of CYP2C9 cause reductions in enzymatic activity of 30% and 80%, respectively, and increase bleeding risk. 9 Ten VKORC1 single nucleotide polymorphisms, many of which are tightly linked, and 5 inferred haplotypes determine low-, intermediate-, and high-dose requirements. 11,15 Together, these genotypes plus clinical characteristics predict approximately one-half of interindividual dose variability These observations suggest clinical applicability for CYP2C9 and VKORC1 genotyping. Indeed, the US Food and Drug Administration has revised drug-prescribing information for warfarin to include information regarding the effect of genetic makeup on drug dosing and to highlight the opportunity for healthcare providers to use genetic testing to improve the initial estimate of warfarin dose for individual patients. In the 2010 label revision, a table is provided with dose ranges by genotype with the intent to lower the risk of bleeding complications. 19 However, pharmacogenetic (PG)-guided warfarin dosing algorithms have not been adequately tested for their impact on clinical outcomes in prospective, controlled trials. In a previous randomized study (Applying Pharmacogenetic Algorithms to Individualize Dosing of Warfarin [CoumaGen], N 200 patients), 20 we found that PG-guided dosing predicted stable maintenance doses better than a standard fixeddose regimen. However, the end points of percentage of out-of-range INRs and time in therapeutic range (TTR) were not met. Nevertheless, patients in 2 prespecified genetic subgroups did experience end point reductions. In CoumaGen-II we further pursued PG guidance of warfarin dose initiation within a single, integrated healthcare system. Methods Study Design CoumaGen-II comprised 2 prospective clinical trial comparisons: (1) a blinded, randomized comparison of 2 refined PG-guided algorithms and (2) a clinical effectiveness comparison of PG-guided therapy with use of either algorithm with parallel, standard (STD; non-pg based) dosing. The study was approved by the institutional review boards of the 3 participating Intermountain Healthcare (Intermountain) hospitals (Intermountain Medical Center, LDS Hospital, and McKay-Dee Hospital) and registered on clinicaltrials.gov (NCT ). The primary end points of interest were percentage of out-of-range (% OOR) INRs and percentage of time in therapeutic range (% TTR) during the first month and, secondarily, after up to 3 months of warfarin therapy, by use of the method of linear interpolation. 21 Inclusion and Exclusion Criteria in the Randomized Study Inclusion in the randomized study required age 18 years, an indication for initiation of warfarin anticoagulation, and written informed consent. Women who were pregnant, lactating, or of child-bearing potential; patients participating in other investigational trials within 30 days; those taking rifampin within 3 weeks; those with severe comorbidities (eg, creatinine 2.5, hepatic insufficiency, active malignancy, advanced physiological age, expected survival 6 months); noncompliance risk; and those deemed inappropriate for PG-guided dosing for any other reason were excluded. Study Procedures in the Randomized Study Arms Qualifying, consenting patients underwent buccal swabbing for DNA testing, performed in the Cardiovascular Molecular and Genetic Research Laboratory at LDS Hospital. A rapid turnaround (median, 1 hour) melting curve analytic method was used as described previously 20,22 and in online-only Data Supplement Materials. Randomization used a mixed block scheme: blocks of 2, 4, 6, and 8 (each with 1:1 allocation to PG-1 and PG-2) were randomly ordered in superblocks of 20. The overall code was held by study genotyping laboratory personnel and pharmacy and was blinded to clinical personnel and patients. Subsequent to publication of our internally generated PG algorithms, 16,20 a major multicenter collaborative effort (International Warfarin Pharmacogenetics Consortium [IWPC]) derived and published a common algorithm to predict stable maintenance dose based on 5000 patients across broad geographic and ethnic/racial groups. 23 The IWPC algorithm, potentially more generalizable than these internal algorithms, also was found to be an accurate predictor of maintenance dose in our CoumaGen-I study cohort. Hence, we based our PG-1 (1 step) algorithm on the IWPC algorithm with minor modifications to accommodate different INR targets and smoking status, based on supplemental data from Gage et al 24 (see Figure I of the online-only Data Supplement Materials). Subsequent dose adjustments were made by use of standard, non-pg, INR-based algorithms (Figure I of the online-only Data Supplement Materials). 20,25 The alternative, 3-step algorithm (PG-2) included 2 further modifications of IWPC (see Figure II of the online-only Data Supplement Materials). (1) It ignored the CYP2C9 variant status for the first 2 days. (This was based on information suggesting that CYP2C9 affects elimination rates of s-warfarin but not initial drug sensitivity; thus, CYP2C9 might not impact initial dosing requirement, but it might become important only during the elimination phase of s-warfarin, ie, beginning at about day 3, given a drug half-life of 15 to 42 hours. 26 ) (2) It used a special dose-revision algorithm based on a day 4 (or day 5) INR after 3 (or 4) warfarin doses, generated by the International Dose-Revision Collaborators and shown to be more accurate than empirical dose revision by incorporating genetics, clinical information, and early patient dosing history and response. 27 Dose adjustment after 7 days used the validated INR-based Intermountain Healthcare Chronic Anticoagulation Clinic Protocol Algorithm for both PG-guided arms (see Figure III of the online-only Data Supplement Materials). 20 In both PG arms, genotype and clinical information (age, sex, height, weight, smoking status, race/ethnicity, relevant drugs, target INR, loading dose option) then were entered into a preprogrammed Excel spreadsheet, which calculated an initial and maintenance daily dose. The research pharmacist converted the calculated dose to the nearest 0.5-mg dose increment and placed the dose order under physician direction. Target turnaround time from enrollment to assigned dose order was 6 hours (maximum, 24 hours). These PG algorithms also accommodated the option of 1 loading (2 ) dose (eg, for orthopedic patients) or 2 loading doses (eg, to increase efficiency in attaining the therapeutic range and decrease hospital stay, 25 up to 15 mg per loading dose on days 1 to 2, with the exception of CYP2C9 variant carriers, in which a maximum of 10 mg was allowed). 20 The algorithm allowed for receipt of 1 preenrollment, empirical dose of warfarin (eg, mg) by adjusting doses on days 2 and 3 to achieve the calculated 3-day dose total (13.6% of enrollees). Protocol INRs were designated at baseline, on day 4 1; on days 8, 14, 21, and 30; then, monthly. Additional INRs were allowed at clinician discretion to achieve and maintain stable therapeutic INRs.

3 Anderson et al Pharmacogenetic Guidance of Warfarin Dosing 1999 Downloaded from by guest on July 19, 2018 Parallel Standard Dosing Cohort The parallel, standard-dosing patient control cohort was identified by a query of the electronic medical records databases of the 3 participating hospitals for the time interval spanning enrollment of the randomized PG-guided cohorts (July 2008 through December 2010). Patients 18 years of age initiating warfarin therapy with a baseline and at least 1 follow-up INR level between days 3 and 14 were selected. Initial dose selection and therapy modification was at individual Intermountain-credentialed physician/healthcare provider discretion. Standard management is non PG-based. A standard (fixed) initial maintenance dose of 5 mg/d is generally assumed, with loading doses and clinical-factor modifications not specified. However, the same standard INR-based dose-modification algorithm developed and promoted by Intermountain is generally recommended (Figure III of the online-only Data Supplement Materials). 20 Study Duration Study duration was 3 months or to the end of warfarin therapy if 3 months (eg, orthopedic patients generally were treated for 1 month). Primary study hypotheses were as follows: (1) The 3-step PGdosing algorithm (PG-2) will be noninferior to the 1-step dosing algorithm (PG-1) for % OOR INRs (1 ) and TTR (2 ) and will be superior in the combined wild-type and multiple variant genetic subgroups, and in the CYP2C9 variant subgroups, at 1 (1 ) and 3 months (2 ). (2) PG-dosing (PG-1 PG-2) will decrease % OOR (1 ) and increase TTR (2 ) in PG-guided (PG-1 PG-2) patients versus parallel controls. Secondary hypotheses were as follows: (1) PG-guided dosing will decrease the %INRs 4 and 1.5. (2) PG-guided dosing will decrease the composite of %INRs 4 and 1.5 and study relevant (ie, potentially treatment related) serious adverse events (death, myocardial infarction, stroke/transient ischemic attack, thromboembolism, and clinically significant bleeding events). (3) The PG-2 algorithm will be noninferior to the PG-1 algorithm for %INRs 4 and 1.5 or serious adverse events. (4) PG-guided dosing will decrease the number of INRs measured up to 3 months. (5) PG-guided dosing will predict stable maintenance dose better than the initial empirical dose selection. (6) Patients with one or more CYP2C9 variant will require a longer time to achieve a stable maintenance INR. Determination of Stable Maintenance Dose A stable maintenance dose was determined for study patients as the last stable dosing interval achieved on day 8 or later that was associated with 2 or more therapeutic INRs measured 1 week or more apart without a dosing change of 0.5 mg/d for doses 5 mg/d or 1 mg/d for doses 5 mg/d. Cases with abbreviated or unstable dosing patterns (see below) were excluded from analysis. For cases that closely approached stability, stable dose was estimated by interpolation/extrapolation by an experienced investigator (J.L.A.) blinded to treatment arm and genotype. Recruitment Targets and Study Power Based on power calculations and feasibility, the minimum recruitment target for the randomized, PG-guided comparison was set at 500 patients. All qualifying parallel control patients were included, anticipated to number For hypothesis 1, the power to exclude inferiority of PG-2 versus PG-1 at a margin ( ) of 5% points with 250 patients per group at a 2-sided 0.05 is 87%, assuming a common standard deviation of For hypothesis 2, a total of 500 PG-guided patients and 1000 controls yields 99% power at 2-sided 0.05 for a decrease in proportion of OOR INRs from 0.40 to 0.34 (proportions based on CoumaGen-I). 20 Statistical Analysis Comparisons between groups for primary and secondary end points were made by using unpaired t test, 2, or log-rank tests as appropriate. All consented, randomized patients who were successfully genotyped and received at least 1 dose of warfarin with at least 1 postdose INR were included in efficacy analyses (modified intention to treat). All patients receiving at least 1 dose of warfarin were included in safety (eg, bleeding) analyses until 1 week after the last warfarin dose or, in the case of comparisons between randomized and parallel control groups, until 3 months after warfarin initiation. Serious adverse events of primary interest were death, myocardial infarction, stroke/transient ischemic attack, thromboembolic event, and clinically significant bleeding. Univariate and multivariate odds ratios (with 95% confidence intervals) were calculated for discrete variables by use of logistic regression. Heterogeneity between subgroups was assessed using multivariate analysis of variance. Linear regression was used to adjust the primary end point for baseline demographics and to generate multivariable models for the observed stable maintenance warfarin dose entering genetic variants and relevant clinical variables. Noninferiority was tested by use of the equivalence test in SAS (Cary, NC). 28 To account for laboratory INR-measurement error, a 10% margin outside of the target range was allowed in determination of OOR values, ie, INRs 1.8, 3.3 for INR 2.5 target; 2.25, 3.85 for INR 3.0 target. Below therapeutic INRs were counted beginning on day 3 to allow for the usual lag in achieving a therapeutic INR. The noninferiority for the primary end point was set at an absolute %OOR INR for PG-2 no more than 5% greater than that of PG-1. For each patient, the proportion of OOR INRs of all INRs was determined, and this proportion then was averaged over all patients. Significance for this primary end point was set at P In case of a nonsignificant primary end point, results for secondary end points and subset analyses were to be considered exploratory and nominally significant at P Mean and mean absolute maintenance dose errors for PG-guided patients were compared with errors assuming a virtual standard empirical dose of 5 mg/d. Safety and Clinical Events Committees An independent Data and Safety Monitoring Committee tracked unblinded safety data. A separate independent Clinical Events Committee adjudicated clinical adverse events blinded to study arm. Results Patient Enrollment and Demographics A total of 504 patients were randomly assigned between the 2 PG arms (PG-1 257, PG-2 247). Baseline characteristics of the 2 groups were comparable, including genotype profile (Table 1). Of PG-guided patients in the current study, 477 fulfilled criteria for the efficacy analysis (modified intention to treat), and 488 who received at least 1 dose of warfarin were included in safety (eg, bleeding) analyses. The parallel, nonrandomized control group, initiated concurrently on warfarin at the same institutions, was slightly younger, less frequently carried a postorthopedic surgery diagnosis, and more frequently carried a diagnosis of coronary artery disease (Table 1). Randomized Comparison of the 2 PG-Dosing Algorithms The primary hypothesis in the randomized study of noninferiority for PG-2 compared with the PG-1 algorithm was verified for % OOR INR at 1 and 3 months and for % TTR at 3 months (Table 2). However, PG-2 was not superior to PG-1, either overall, or for the subgroup with no variants or multiple variants (OOR INRs 32.7% versus 31.5%, P 0.67), or for the subgroup with CYP2C9 variants alone (32.3% versus 32.3%, P 1.00). Results of secondary outcomes also were similar for PG-2 and PG-1 patients, including INR safety signals and clinical adverse events (probability value range, , Table 3). Accordingly, the 2 PG-guided

4 2000 Circulation April 24, 2012 Table 1. Baseline Patient Characteristics and Allelic Variant Frequencies Downloaded from by guest on July 19, 2018 Characteristic PG-1 Arm PG-2 Arm Combined PG Parallel Controls N Age, y, mean (range) 61.3 (20 90) 59.9 (18 85) 60.6 (18 90) 57.3 (18 99)* Sex, % female Weight, kg, mean (SD) 93.1 (23.9) 92.3 (24.5) 92.7 (24.2) 91.3 (28.0) Postoperative orthopedic, % * Deep vein thrombosis and/or * pulmonary embolism, % Atrial fibrillation, % * Other/uncertain indication, % * Ethnicity/race white, % H/o diabetes, % H/o hypertension, % Current tobacco use, % NA H/o CAD, % * CYP 2C9 *2 WT (CC), % NA CT, % NA TT, % CYP 2C9 *3 WT (AA), % NA AC, % NA CC, % NA VKORC WT (GG), % NA GA, % NA AA, % NA Any variant, % NA Treatment/follow-up duration, mean (SD), d 46.8 (30.5) 47.7 (30.3) 47.2 (30.4) 48.9 (33.5) H/o indicates history of; WT, wild type; CAD, coronary artery disease; PG, pharmacogenetics; and NA, not available. *P 0.01 between combined PG and control groups. groups were combined for comparative effectiveness analyses versus standard dosing in parallel controls. Genetic Determination of Stable Dose and PG-Guided Dose Prediction Stable maintenance dose in the combined PG group could be determined in 444 patients and varied inversely (by fold) and highly significantly (P 0.001) with the number of reduced-function variant alleles (Figure 1), confirming previous observations by many groups, including our own. 16,19 PG guidance much more accurately (P 0.001) predicted maintenance dose in wild-type (no variant) and multiple ( 1) variant groups than a virtual standard regimen (ie, 5 mg/d) (Figure 2): no-variant patients would require an upward Table 2. Primary End Point Results for Hypotheses 1 and 2 End Point PG-1 Group (n 245) PG-2 Group (n 232) P Value Combined PG Group (n 477) Parallel Controls (n 1866) P Value Out-of-range INRs (%)* at 30 d, mean (CI) 30.6 ( ) 31.8 ( ) Pni Ps ( ) 41.5 ( ) Out-of-range INRs (%)* up to 3 mo, mean (CI) 30.3 ( ) 30.3 ( ) Pni Ps ( ) 42.3 ( ) Time in therapeutic range (%) at 30 d, ( ) 67.5 ( ) Pni 0.12 Ps ( ) 58.4 ( ) mean (CI) Time in therapeutic range (%) up to 3 mo, 21 mean (CI) 71.7 ( ) 70.8 ( ) Pni Ps ( ) 58.6 ( ) INR indicates international normalized ratio; PG, pharmacogenetics; ni, noninferiority; s, superiority; and mitt, modified intention to treat. CI indicates 95% confidence interval. *Primary end point, mitt population. End point measurements begin on day 3 to allow for onset of drug effect.

5 Anderson et al Pharmacogenetic Guidance of Warfarin Dosing 2001 Table 3. Secondary End Point Results Downloaded from by guest on July 19, 2018 End Point PG-1 Group (n 250) PG-2 Group (n 238) P Value Combined PG Group Parallel Controls (n 1866) P Value Av. %INRs 4 or % 15.4% % 27.4% Av. %INRs % 4.28% % 4.52% 0.97 Av. %INRs % 11.1% % 22.9% CEC-adjudicated SAEs* No. of events NA NA No. (%) of patients 10 (4.0) 6 (2.5) (3.3) NA NA 90-d incident SAEs /n (%) patients /487 (4.5) 162/1726 (9.4) % patients with INR 4, 1.5 or CEC-SAE 53.6% 59.2% % NA NA No. of INRs measured, mean Days to first therapeutic INR, mean Days to stable maintenance dose, mean NA NA Mean absolute error (calculated vs achieved stable dose mg/week ) NA NA INR indicates international normalized ratio prothrombin time; SAE, serious adverse events; CEC, Clinical Events Committee; NA, not available; Av., average; No., number. *Study SAEs: adjudicated death, myocardial infarction, stroke, thromboembolic event, moderate or severe hemorrhage. After day 3, INRs available to assess in 479 patients; SAEs assessed in all 488 patients receiving warfarin. Incident SAEs for comparison with parallel controls determined by International Classification of Diseases, Ninth Edition, code search of SAEs of interest for new (incident) SAE diagnosis during follow-up. adjustment in weekly dose by an average of 10.5 mg, and multiple-variant patients would require downtitration by an average of 10.9 mg/wk, whereas essentially no adjustment on average was required after PG-guided initial dose selection. In contrast, doses in the 1 variant subgroup (most commonly in VKORC1 or CYP2C9*2), representing the average (median) population subset, were predicted well by both PG- and fixed-dosing methods (Figure 2). Similarly, the overall mean absolute error in individual patients for PG-calculated dosing was significantly less than virtual empirical (standard) dosing (P 0.001). By genetic subgroup, it was substantially reduced for no-variant and multiple-variant patients and was more modestly reduced but also significant for those with single variants (Figure 3). PG-predicted dosing was within 1 mg/d in 63.2% of the PG-guided patients with an established stable maintenance dose (n 444) in comparison with 37.6% of these patients based on an empirical dose of 5 mg/d (P 0.001). Linear regression analysis for the dependent variable stable warfarin dose determined that the 3 variant alleles accounted for 37.2% of variance in dose, and the clinical variables, for 13.4%. Together in a single model, all variables accounted for 49.4% of stable dose variance. Figure 1. Average stable maintenance warfarin doses (milligrams per week) by number of variant alleles. Numbers (%) of patients in each group are as follows: wild type (no variants), 104, 23.4%; 1 variant, 183, 41.2%; 2 variants, 125, 28.2%; 3 variants, 32, 7.2%. SEMs are 1.26 for wild type, and 0.88, 0.86, and 0.17 for 1, 2, and 3 variant groups, respectively. Dose differences across groups are highly significant (P 0.001). Figure 2. Mean dosing error by genotype: PG-guided versus standard (5 mg/d) initial dose assumption. Mean required adjustment in weekly maintenance dose (in milligrams) between initial (algorithm-predicted) and final (stable, titrated) dose in wild-type allelic patients (n 104), patients with one variant allele (n 183), and patients with 1 (range, 2 5) variant alleles (n 157) by PG-guided and assumed STD (5 mg/d) dosing algorithms. Significance levels of comparisons by algorithm were as follows: no variants, P 0.001; 1 variant, P 0.30; 1 variant, P SEMs for the PG bars were 1.16, 0.81, and PG indicates pharmacogenetics; STD, standard.

6 2002 Circulation April 24, 2012 Downloaded from by guest on July 19, 2018 Figure 3. Mean absolute dose error (in milligrams per week) by genotype: PG-guided versus standard (5 mg/d) initial dose assumption. Mean absolute error in weekly maintenance dose (in milligrams per week) for patients comparing individual initial (algorithm-predicted/assigned) and final (stable, titrated) dose in wild-type allelic patients (n 104), patients with 1 variant allele (n 183), and patients with 1 (range, 2 5) variant alleles (n 157) by PG-guided and by assumed STD (5 mg/d) dosing algorithms. Significance levels of comparisons by algorithm were as follows: no variants, P 0.001; 1 variant, P 0.03; 1 variant, P SEMs for the PG bars were 0.78, 0.54, and 0.50, respectively. PG indicates pharmacogenetics; STD, standard. Comparison of PG-Guided and Standard Warfarin Dosing In comparison with standard initial dosing, PG dosing was associated with large (11% absolute, 26% relative), highly significant reductions in the primary comparative effectiveness end point of % OOR INRs at 1 month (P adjusted for age, sex, orthopedic surgery, atrial fibrillation, deep vein thrombosis, pulmonary embolism, heart failure, valve replacement, and arterial thrombosis) and also up to 3 months (adjusted P 0.001). Results for the more traditional TTR metric followed in parallel, with absolute improvements in %TTR of 11% at 1 month (adjusted P 0.001) and 13% at 3 months (adjusted P 0.001) (Table 2, Figure 4). Percentage TTR was highly negatively correlated with % OOR INRs both at 30 days (r 0.816) and at 3 months (r 0.759). Time to first therapeutic INR tended to be shorter with PG guidance (Table 3). Given the large difference in the proportion of post orthopedic surgery patients in the PG-guided and parallel Figure 4. Percentage of OOR INRs and % TTR for PG-guided dosing algorithms and parallel (standard dosing) controls. All intertreatment group comparisons were highly significant (all P 0.001; see Table 2 for confidence intervals and individual probability values). OOR indicates out of range; INR, international normalized ratio; TTR, time in therapeutic range; and PG, pharmacogenetics. control groups, a stratified analysis was performed. An advantage to PG-guided therapy was observed in both strata: for the orthopedic group, % OOR INRs for PG (n 266) and STD therapy (n 275) groups were 31.3% versus 49.7% at 30 days and 31.5% versus 52.3% at up to 3 months. For the nonorthopedic group, % OOR INRs for PG (n 212) and STD therapy (n 1636) groups were 31.4% versus 40.2% at 30 days and 29.4% versus 40.6% at up to 3 months. All comparisons were highly significant in unadjusted and ageand sex-adjusted analyses (P 0.001, ANOVA). Adverse Event and Safety Observations In keeping with its advantage to reduce out-of-range INRs and increase TTR, PG-guided dosing was associated with fewer markedly high ( 4) or low ( 1.5) INRs (15.3% versus 27.4%, PG versus STD, P 0.001) (Table 3). Of note, this was driven by a reduction in very low INRs. Clinical Events Committee adjudicated adverse events occurring during therapy were low with PG guidance (3.3% of patients) and did not differ significantly between PG groups (Table 3). However, computed medical record screening for incident serious adverse events through 90 days discovered increases in hemorrhagic events, thromboembolic events, death, and any serious adverse events in parallel controls in comparison with PG-guided patients (4.5% versus 9.4%; adjusted relative risk, 0.44 [CI ]; P 0.001;Table 3 and online-only Data Supplement Table I). Fewer serious adverse events were found both in those with and without an orthopedic indication for anticoagulation (online-only Data Supplement Table II). Discussion Study Summary We report a prospective study comprising 2 clinical trial comparisons: (1) a blinded, randomized trial comparing a modified, 1-step with a more complex 3-step PG-guided dosing algorithm in 500 patients initiating warfarin therapy, and (2) a clinical effectiveness comparison of these 500 patients receiving PG-guided dosing with a large, parallel, nonrandomized real-world control population concurrently initiating warfarin at our institutions and managed routinely with non-pg, standard-dose therapy. In primary and secondary end point comparisons, the 2 PG algorithms were comparable, with the more complex PG-2 algorithm failing to show superiority. This suggests to us that the simpler PG-1 algorithm, based on the large, multicenter IWPC algorithm, 23 with modifications according to Gage et al, 24 may be preferable for general clinical practice. However, and importantly, PG guidance by use of either algorithm was highly superior to standard dosing in the comparative effectiveness study in achieving and maintaining therapeutic INRs, a well-validated measure of clinical effectiveness and safety, and in reducing serious bleeding and thromboembolic events and death. These findings have important implications for ongoing randomized trials and clinical practice. Relevant Previous and Ongoing Studies Several prospective studies have demonstrated the feasibility of initiating warfarin therapy based on PG-guided dose prediction algorithms. 20,29 32 However, only a few of these

7 Anderson et al Pharmacogenetic Guidance of Warfarin Dosing 2003 Downloaded from by guest on July 19, 2018 studies were randomized, controlled clinical trials, and these few have been limited by diverse issues, including study design and sample size. 20,32 Although CoumaGen-I, the largest (N 200) of these, predicted stable maintenance doses better than standard empirical dosing (5 mg/d), the end points of %OOR INRs and %TTR were not met. Importantly, however, patients in 2 prespecified genetic subgroups (ie, those with multiple variants, anticipated to require smaller doses, and those with no variants, anticipated to require larger doses) benefited from by PG guidance (P 0.03). Perceived limitations of CoumaGen I included smaller than postulated benefits of PG guidance, limited power (because of sample size), aggressive dose management by the hospital s anticoagulation management service in the STD arm, and dose variability (both genetic and nongenetic) unaccounted for by the PG algorithm. Hence, in planning CoumaGen-II, we replaced the previous PG algorithm with 2 refined, updated algorithms based on major international collaborative efforts. 23,24 We increased randomized patient numbers (from 200 to 500) to increase power for the interalgorithm comparison, and we identified almost 2000 parallel, non PG-guided controls to allow for a much more powerful comparison of PG guidance with standard (fixed-dose) therapy. Medco-Mayo recently reported a clinical effectiveness study of PG-guided warfarin initiation. 33 This study reported that hospitalizations for bleeding and thromboembolism were reduced by genotyping (hazard ratio 0.72). However, the design of this nonrandomized study has been criticized as being susceptible to physician treatment bias (Hawthorne effect). Further, the results appear to lack temporal plausibility because genotyping was not available for a minimum of 11 days (median, 32 days), whereas the contribution of genotype to INR-alone guidance has been shown to diminish rapidly over time. 20,27,34 Currently, there are 3 major ongoing randomized trials of warfarin initiation: Clarification of Optimal Anticoagulation Through Genetics (COAG), 35 Genetics Informatics Trial of Warfarin Therapy (GIFT), 36 and European Pharmacogenetic Approach to Coumarin Anticoagulant Therapy (EU-PACT) (clinicaltrials.gov/ct2/show/nct ), each with differing enrollment populations, study designs, and recruitment targets (N ). Results of these trials are of critical importance to recommendations for clinical application of PG guidance of warfarin therapy, but these results are not expected to become available until at least Mechanistic Considerations Reasons for the failure of the PG-2 algorithm to demonstrate superiority are uncertain. Possibilities include a greater impact of CYP2C9 genotype on response than expected, beginning with the first dose, and also the greater complexity of the PG-2 regimen for dose initiation, which may have led to greater noncompliance with dose calculations. The utility of the third element in PG-2, the incorporation of genetic (and INR history) into the day 4 dose refinement algorithm, has been established and validated in a multicenter experience, 27 but its incremental impact appears to be relatively modest. This third element also may have been susceptible to greater noncompliance, which we could not accurately track. Nevertheless, the substantial benefit in INR control was impressive in both PG groups. Despite these limitations, the results give substantial impetus to PG guidance in clinical practice. Clinical Implications As previously summarized, randomized clinical trial evidence to substantiate the clinical benefit of PG guidance of warfarin dosing is inconclusive. Although PG dosing has been shown to more accurately predict individual warfarin dose requirements, it previously has not demonstrated the ability to improve TTR and decrease thromboembolic and bleeding events. Accordingly, PG guidance has not been endorsed by relevant medical societies or incorporated into clinical guideline recommendations. The results of our comparative effectiveness study add important incremental information and impetus to the existing evidence base. Strictly interpreted, they support a management strategy incorporating genetics, rather than precisely defining the incremental contribution of genetics to benefit. We further interpret these results as both supporting enrollment in ongoing clinical trials wherever possible (which can further validate our controlled but nonrandomized observations) and otherwise also encouraging the consideration of our PG-guided management strategy for warfarin initiation as a clinical option (based on physician preference/expertise, patient characteristics, and genotyping availability). Improvement in quality of care to the degree we observed promises to be cost effective. 37 We do not view our results as sufficiently definitive to establish PG dosing as a general therapeutic mandate. Nevertheless, use of PG guidance in settings similar to CoumaGen-II is consistent with guidelines recently published and extensively referenced by the Clinical Pharmacogenetics Implementation Consortium. 38 If a physician decides to use PG-guided dosing, warfarin dose recommendations can be generated that incorporate genotype along with clinical information by use of publically available mechanisms. The 2 best validated PG algorithms are available on-line at 24 and PA &objCls Dataset#tabview tab2. 23 These 2 algorithms formed the basis of our PG algorithms and give generally similar results. Alternatively (although less accurately 39 ), dose selection can be estimated by genotype (but without regard to clinical information) from a table in the current warfarin druginformation brochure. 19 Study Strengths and Limitations Study strengths include its prospective, randomized design for the PG-algorithm comparison, including oversight by safety and events committees, and industry-independent funding. In addition, our rapid genotyping assay enabled PG guidance to be applied in real time (ie, same day; sample processing time 1 hour), 22 applicable to the real world. The study was underpowered to rigorously test for superiority of the PG-2 algorithm; however, we can exclude with 80% power an absolute decrease of 5 percentage points in % OOR INR. The parallel control population for the clinical effectiveness comparison had the strength of large numbers of patients who were treated in standard real-world fashion, in

8 2004 Circulation April 24, 2012 Downloaded from by guest on July 19, 2018 the same hospitals, during the same time interval, using the same general treatment recommendations, and (for hospitalized patients) the same anticoagulation management services. However, it should be emphasized that the control group was not randomized, and certain baseline characteristics differed (although robust advantages to PG guidance persisted after adjustment for these). Furthermore, the exact contribution of genotyping versus clinical factors and management differences in explaining the primary outcome results cannot be accurately determined. Also, some demographic and treatment-related factors (eg, dosing information) were not available for the parallel control group in our database. Finally, the study relied heavily on our usual healthcare system s many clinical personnel for managing warfarindosing protocols rather than exclusively on research personnel, and we are unable to determine the exact rate of compliance with study-designated dosing recommendations. However, noncompliance would likely tend to reduce rather than inflate advantages of PG guidance. Conclusions In a parallel-group, single-healthcare-system clinical effectiveness study, PG dosing of warfarin was associated with substantially reduced % OOR INRs and increased % TTR (both by 10 absolute percentage points, P 0.001) during the critical first 1 to 3 months of warfarin therapy. Furthermore, PG guidance was associated with fewer serious adverse events. More complex PG dosing (delaying CYP2C9 entry until day 3, using genetics to modify day 4 5 adjustment) did not add significantly to dosing efficiency, suggesting that the simpler PG-1 algorithm may be preferable for general clinical application. The findings in this clinical effectiveness study suggest that PG dosing may be worth consideration for broader application in clinical practice, a proposal that is being further tested in 3 ongoing randomized trials. Acknowledgments The authors gratefully acknowledge the clinical and laboratory assistance of the following individuals: Drs Marc Williams and Brent James for study design suggestions and Dr Heidi May and Brianna Ronnow for technical assistance. The Data and Safety Monitoring Committee consisted of Dr Gregory Elliott (chair) and Drs Frank Yanowitz and Robert Jenson. The Clinical Events Committee consisted of Dr Donald L. Lappé (chair) and Drs Robert Fowles and Brian Whisenant. Sources of Funding This study was funded by a grant from the Deseret Foundation and the Heart and Lung Institute, Intermountain Medical Center, Intermountain Healthcare, Murray, UT. None. Disclosures References 1. Evans WE, Relling MV. Pharmacogenomics: translating functional genomics into rational therapeutics. Science. 2001;286: Johnson JA. Drug target pharmacogenomics: an overview. Am J Pharmacogenomics. 2001;1: Anderson JL, Carlquist JF, Horne BD, Muhlestein JB. Cardiovascular pharmacogenomics: current status, future prospects. J Cardiovasc Pharmacol Ther. 2003;8: Voora D, McLeod HL, Eby C, Gage BF. The pharmacogenetics of coumarin therapy. Pharmacogenomics. 2005;6: Hylek EM, Skates SJ, Sheehan MA, Singer DE. An analysis of the lowest effective intensity of prophylactic anticoagulation for patients with nonrheumatic atrial fibrillation. N Engl J Med. 1996;124: Oden A, Fahlen M, Hart RG. Optimal INR for prevention of stroke and death in atrial fibrillation: a critical appraisal. Thromb Res. 2006;117: Crespi CL, Miller VP. The R144C change in CYP2C9*2 allele alters interaction of the cytochrome P450 with NADPH:cytochrome P450 oxidoreductase. Pharmacogenetics. 1997;7: Takanashi K, Tainaka H, Kobayashi K, Yasumori T, Hosakawa M, Chiba K. Cyp2c9 ile359 and leu359 variants: enzyme kinetic study with seven substrates. Pharmacogenetics. 2000;10: Aithal G, Day CP, Kesteven PJ, Daly AK. Association of polymorphisms in the cytochrome P450 CYP2C9 with warfarin dose requirement and risk of bleeding complications. Lancet. 1999;353: Higashi MK, Veenstra DL, Kondo LM, Wittkowsky AK, Srinouanprachanh SL, Farin FM, Rettie AE. Association between CYP2C9 genetic variants and anticoagulation-related outcomes during warfarin therapy. JAMA. 2002;287: D Andrea G, D Ambrosio RL, Di Perna P, Chetta M, Santacroce R, Brancaccio V, Grandone E, Margaglione M. A polymorphism in the VKORC1 gene is associated with interindividual variability in the doseanticoagulant effect of warfarin. Blood. 2005;105: Hillman MA, Wilke RA, Caldwell MD, Berg RL, Glurich I, Burmester JK. Relative impact of covariates in prescribing warfarin according to CYP2C9 genotype. Pharmacogenetics. 2004;14: Wadelius M, Sorlin K, Wallerman O, Karlsson J, Yue OY, Magnusson PK, Wadelius C, Melhus H. Warfarin sensitivity related to CYP2C9, CYP3A5, ABCB2(MDR1) and other factors. Pharmacogenomics J. 2004;4: Wadelius M, Chen LY, Downes K, Ghori J, Hunt S, Eriksson N, Wallerman O, Melhus H, Wadelius C, Bentley D, Deloukas P. Common VKORC1 and GGCX polymorphisms associated with warfarin dose. Pharmacogenomics J. 2005;5: Rieder MJ, Reiner AP, Gage BF, Nickerson DA, Eby CS, McLeod HL, Blough DK, Thummel KE, Veenstra DL, Rettie AE. Effect of VKORC1 haplotypes on transcriptional regulation and warfarin dose. N Engl J Med. 2005;352: Carlquist JF, Horne BD, Muhlestein JB, Lappe DL, Whiting BM, Kolek MJ, Clarke JL, James BC, Anderson JL. Genotypes of the cytochrome p450 isoform, CYP2C9, and the vitamin K epoxide reductase complex subunit 1 conjointly determine stable warfarin dose: a prospective study J Thromb Thrombolysis. 2006;22: Sconce EA, Khan TI, Wynne HA, Avery P, Monkhouse L, King BP, Wood P, Kesteven PJ, Daly AK, Kamali F. The impact of CYP2C9 and VKORC1 genetic polymorphisms and patient characteristics upon warfarin dose requirements: proposal for a new dosing regimen. Blood. 2005;106: Aquilante CL, Langaee TY, Lopez LM, Yarandi HN, Tromberg JS, Mohuczy D, Gaston KL, Waddell CD, Chirico MJ, Johnson JA. Influence of coagulation factor, vitamin K epoxide reductase complex subunit 1, and cytochrome P450 2C9 gene polymorphisms on warfarin dose requirements. Clin Pharmacol Ther. 2006;79: Highlights of prescribing information. Coumadin packageinserts.bms.com/pi/pi_coumadin.pdf. Accessed September 28, Anderson JL, Horne BD, Stevens SM, Grove AS, Barton S, Nicholas ZP, Kahn SFS, May HT, Samuelson KM, Muhlestein JB, Carlquist JF. Randomized trial of genotype-guided versus standard warfarin dosing in patients initiating oral anticoagulation. Circulation. 2007;116: Rosendaal FR, Cannegieter SC, van der Meer FJ, Briet E. A method to determine the optimal intensity of oral anticoagulant therapy. Thromb Haemost. 1993;69: Carlquist JF, McKinney J, Nicholas Z, Clark J, Kahn S, Horne BD, Muhlestein JB, Anderson JL. Rapid melting curve analysis for genetic variants that underlie inter-individual variability in stable warfarin dosing. J Thromb Thrombolysis. 2008;26: The International Warfarin Pharmacogenetics Consortium. Estimation of the warfarin dose with clinical and pharmacogenetic data. N Engl J Med. 2009;360: Gage BF, Eby C, Johnson JA, Deych E, Rieder MJ, Ridker PM, Milligan PE, Grice G, Lenzini P, Rettie AE, Aquilante CL, Grosso L, Marsh S,

9 Anderson et al Pharmacogenetic Guidance of Warfarin Dosing 2005 Downloaded from by guest on July 19, 2018 Langaee T, Farnett LE, Voora D, Veenstra DL, Glynn RJ, Barrett A, McLeod HL. Use of pharmacogenetic and clinical factors to predict the therapeutic dose of warfarin. Clini Pharmacol Ther. 2008;84: Kovacs MJ, Rodger M, Anderson DR, Morrow B, Kells G, Kovacs J, Boyle E, Wells PS. Comparison of 10 mg and 5 mg warfarin initiation nomograms together with low-molecular-weight heparin for outpatient treatment of acute venous thromboembolism. A randomized, doubleblind, controlled trial. Ann Intern Med. 2003;138: Schwarz UI, Ritchie MD, Bradford Y, Li C, Dudek SM, Frye-Anderson A, Kim RB, Roden DM, Stein CM. Genetic determinants of response to warfarin during initial anticoagulation. N Engl J Med. 2008;358: Lenzini PA, Wadelius M, Kimmel SE, Anderson JL, Jorgensen AL, Pirmohamed M, Caldwell MD, Limdi N, Burmester JK, Dowd MB, Angchaisuksiri P, Bass AR, Chen JJ, Eriksson N, Rane A, Lindh JD, Carlquist JF, Horne BD, Grice G, Milligan PE, Eby C, Shin J, Kim H, Kurnik D, Stein CM, McMillin G, Pendleton RC, Berg RL, Deloukas P, Gage BF. Integration of genetic, clinical, and INR data to refine warfarin dosing Clin Pharmacol Ther. 2010;87: Schuirman D. The two sided test procedure versus the power approach. J Pharmacokinet Biopharm. 1987;15: Voora D, Eby C, Linder MW, Milligan PE, Bukaveckas BL, McLeod HL, Maloney W, Clohisy J, Burnett RS, Grosso L, Gatchel SK, Gage BF. Prospective dosing of warfarin based on cytochrome P-450 2C9 genotype. Thromb Haemost. 2005;93: Millican E, Jacobsen-Lenzini PA, Milligan PE, Grosso L, Eby C, Deych E, Grice G, Clohisy JC, Barrack RL, Burnett SJ, Voora D, Gatchel S, Tiemeier A, Gage BF. Genetic-based dosing in orthopaedic patients beginning warfarin therapy. Blood. 2007;110: Lenzini PA, Grice GR, Milligan PE, Dowd MB, Subherwal S, Deych E, Eby CS, King CR, Porche-Sorbet RM, Murphy CV, Marchand R, Millican EA, Barrack RL, Clohisy JC, Kronquist K, Gatchel SK, Gage BF. Laboratory and clinical outcomes of pharmacogenetic vs. clinical protocols for warfarin initiation in orthopedic patients. J Thromb Haemost. 2008;6: Caraco Y, Blotnick S, Muzkat M. CYP2C9 genotype-guided warfarin prescribing enhances the efficacy and safety of anticoagulation: a prospective, randomized, controlled study. Clin Pharmacol Ther. 2008;83: Epstein RS, Moyer TP, Aubert RE, O Kane DJ, Xia F, Verbrugge RR, Gage BF, Teagarden JR. Warfarin genotyping reduces hospitalization rates: Results from the MM-WES (Medco-Mayo Warfarin Effectiveness study). J Am Coll Cardiol. 2010;22: Ferder NS, Eby CS, Deych E, Harris JK, Ridker PM, Milligan PE, Goldhaber SZ, King CR, Giri T, McLeod HL, Glynn RJ, Gage BF. Ability of VKORC1 and CYP2C9 to predict therapeutic warfarin dose during the initial weeks of therapy. J Thromb Haemost. 2010;8: French B, Joo J, Geller NL, Kimmel SE, Rosenberg Y, Anderson JL, Gage BF, Johnson JA, Ellenberg JH. Statistical design of personalized medicine interventions: the Clarification of Optimal Anticoagulation through Genetics (COAG) trial. Trials. 2010;11: Do EJ, Lenzini P, Eby CS, Bass AR, McMillin GA, Stevens SM, Woller SC, Pendleton RC, Anderson JL, Proctor P, Nunley RM, Davila-Roman V, Gage BF. Genetics informatics trial (GIFT) of warfarin to prevent deep vein thrombosis (DVT): rationale and study design. Pharmacogenomics J. May doi: /tpj Available at: com. Accessed September 28, Rose AJ, Berlowitz DR, Ash AS, Ozonoff A, Hylek EM, GOldhaber- Fiebert JD. The business case for quality improvement: oral anticoagulation for atrial fibrillation. Circ Cardiovasc Qual Outcomes. 2011;4: Johnson JA, Gong L, Whirl-Carrillo M, Gage BF, Scott SA, Stein CM, Anderson JL, Kimmel SE, Lee MT, Pirmohamed M, Wadelius M, Klein TE, Altman RB. Clinical pharmacogenetics implementation consortium guidelines for CYP2C9 and VKORC1 genotypes and warfarin dosing. Clin Pharmacol Ther. 2011;90: Finkelman BS, Gage BF, Johnson JA, Brensinger CM, Kimmel SE. Genetic warfarin dosing: tables versus algorithms. J Am Coll Cardiol. 2011;57: CLINICAL PERSPECTIVE Warfarin is characterized by marked variations in individual dose requirements and a narrow therapeutic window. Much of this variability is explained by common reduced-function variants in 2 genes, CYP2C9 and VKORC1. Pharmacogenetic (PG) guidance of warfarin initiation could improve dosing efficiency and safety, but evidence from clinical trials is still meager. To increase understanding of the potential role of PG guidance in warfarin dose initiation and to build on an earlier study (CoumaGen-I), we designed a second 3-month study (CoumaGen-II) with 2 parts: (1) a blinded, randomized comparison of a modified 1-step (PG-1) with a 3-step PG algorithm (PG-2) (N 504) and (2) a clinical effectiveness comparison of PG-guided with standard dosing in a parallel (nonrandomized) control group (N 1866) within the Intermountain Healthcare system. A rapid method provided same-day CYP2C9 and VKORC1 genotyping. In the randomized comparison, PG-2 was noninferior but not superior to PG-1 for the primary end point of percentage of out-of-range international normalized ratios at 1 month and at 3 months and for percentage of time in therapeutic range at 3 months, suggesting that the simpler 1-step PG algorithm may be preferable for clinical application. However, PG guidance (combined cohort) was substantially superior to standard dosing in the parallel control group for these end points (all P 0.001). PG guidance also reduced percentage of international normalized ratios 4 and 1.5 and serious adverse events. These clinical effectiveness findings suggest that PG dosing should be considered for broader clinical application, a proposal that is being tested further in 3 major randomized trials. The simpler 1-stage PG algorithm provided equivalent results and may be preferable for clinical application.

10 A Randomized and Clinical Effectiveness Trial Comparing Two Pharmacogenetic Algorithms and Standard Care for Individualizing Warfarin Dosing (CoumaGen-II) Jeffrey L. Anderson, Benjamin D. Horne, Scott M. Stevens, Scott C. Woller, Kent M. Samuelson, Justin W. Mansfield, Michelle Robinson, Stephanie Barton, Kim Brunisholz, Chrissa P. Mower, John A. Huntinghouse, Jeffrey S. Rollo, Dustin Siler, Tami L. Bair, Stacey Knight, Joseph B. Muhlestein and John F. Carlquist Downloaded from by guest on July 19, 2018 Circulation. 2012;125: ; originally published online March 19, 2012; doi: /CIRCULATIONAHA Circulation is published by the American Heart Association, 7272 Greenville Avenue, Dallas, TX Copyright 2012 American Heart Association, Inc. All rights reserved. Print ISSN: Online ISSN: The online version of this article, along with updated information and services, is located on the World Wide Web at: Data Supplement (unedited) at: Permissions: Requests for permissions to reproduce figures, tables, or portions of articles originally published in Circulation can be obtained via RightsLink, a service of the Copyright Clearance Center, not the Editorial Office. Once the online version of the published article for which permission is being requested is located, click Request Permissions in the middle column of the Web page under Services. Further information about this process is available in the Permissions and Rights Question and Answer document. Reprints: Information about reprints can be found online at: Subscriptions: Information about subscribing to Circulation is online at: