UBC Faculty Research and Publications

Genotype-guided versus traditional clinical dosing of warfarin in patients of Asian ancestry: a randomized… Syn, Nicholas L; Wong, Andrea L; Lee, Soo-Chin; Teoh, Hock-Luen; Yip, James W L; Seet, Raymond C; Yeo, Wee T; Kristanto, William; Bee, Ping-Chong; Poon, LM; Marban, Patrick; Wu, Tuck S; Winther, Michael D; Brunham, Liam R; Soong, Richie; Tai, Bee-Choo; Goh, Boon-Cher Jul 10, 2018

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


52383-12916_2018_Article_1093.pdf [ 1.26MB ]
JSON: 52383-1.0368868.json
JSON-LD: 52383-1.0368868-ld.json
RDF/XML (Pretty): 52383-1.0368868-rdf.xml
RDF/JSON: 52383-1.0368868-rdf.json
Turtle: 52383-1.0368868-turtle.txt
N-Triples: 52383-1.0368868-rdf-ntriples.txt
Original Record: 52383-1.0368868-source.json
Full Text

Full Text

RESEARCH ARTICLE Open AccessGenotype-guided versus traditional clinicaldosing of warfarin in patients of Asianancestry: a randomized controlled trialNicholas L. Syn1,2†, Andrea Li-Ann Wong1,2†, Soo-Chin Lee1,2, Hock-Luen Teoh3, James Wei Luen Yip4,Raymond CS Seet3,13, Wee Tiong Yeo4, William Kristanto4, Ping-Chong Bee5, LM Poon1, Patrick Marban1,Tuck Seng Wu6, Michael D. Winther7, Liam R. Brunham8,9, Richie Soong2,10, Bee-Choo Tai11and Boon-Cher Goh1,2,12*AbstractBackground: Genotype-guided warfarin dosing has been shown in some randomized trials to improve anticoagulationoutcomes in individuals of European ancestry, yet its utility in Asian patients remains unresolved.Methods: An open-label, non-inferiority, 1:1 randomized trial was conducted at three academic hospitals in South EastAsia, involving 322 ethnically diverse patients newly indicated for warfarin (NCT00700895). Clinical follow-up was 90 days.The primary efficacy measure was the number of dose titrations within the first 2 weeks of therapy, with a meannon-inferiority margin of 0.5 over the first 14 days of therapy.Results: Among 322 randomized patients, 269 were evaluable for the primary endpoint. Compared withtraditional dosing, the genotype-guided group required fewer dose titrations during the first 2 weeks (1.77 vs. 2.93, difference −1.16, 90% CI −1.48 to −0.84, P < 0.001 for both non-inferiority and superiority). The percentage oftime within the therapeutic range over 3 months and median time to stable international normalized ratio (INR)did not differ between the genotype-guided and traditional dosing groups. The frequency of dose titrations(incidence rate ratio 0.76, 95% CI 0.67 to 0.86, P = 0.001), but not frequency of INR measurements, was lower at 1,2, and 3 months in the genotype-guided group. The proportions of patients who experienced minor or majorbleeding, recurrent venous thromboembolism, or out-of-range INR did not differ between both arms. For predictingmaintenance doses, the pharmacogenetic algorithm achieved an R2 = 42.4% (P < 0.001) and mean percentage errorof −7.4%.Conclusions: Among Asian adults commencing warfarin therapy, a pharmacogenetic algorithm meets criteria forboth non-inferiority and superiority in reducing dose titrations compared with a traditional dosing approach, andperforms well in prediction of actual maintenance doses. These findings imply that clinicians may consider applying apharmacogenetic algorithm to personalize initial warfarin dosages in Asian patients.Trial registration: ClinicalTrials.gov NCT00700895. Registered on June 19, 2008.Keywords: Pharmacogenetics, Pharmacogenomics, Precision medicine, CYP2C9, Cytochrome P450, VKORC1, Warfarin,Anticoagulants, Anticoagulation, Polymorphism* Correspondence: phcgbc@nus.edu.sg†Nicholas L. Syn and Andrea Li-Ann Wong contributed equally to this work.1Department of Haematology-Oncology, National University Cancer Institute,Singapore, Singapore2Cancer Science Institute of Singapore, National University of Singapore,Singapore, SingaporeFull list of author information is available at the end of the article© The Author(s). 2018 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, andreproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link tothe Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver(http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.Syn et al. BMC Medicine  (2018) 16:104 https://doi.org/10.1186/s12916-018-1093-8BackgroundWhile effective in preventing thromboembolic events,clinical application of warfarin is characterized by anarrow therapeutic index and often requires multipledose titrations especially during the first few weeks oftherapy. Well-managed warfarin therapy is associatedwith a reduction in the risk of complications [1], yetthe majority of patients do not achieve long-termstable international normalized ratio (INR) within thetherapeutic range [2], indicating the difficulty in iden-tifying an optimal maintenance dose for individualpatients. A growing body of evidence has emergedindicating that the cytochrome P450 2C9 (CYP2C9)and Vitamin K epoxide reductase complex subunit 1(VKORC1) genotypes are associated with maintenancedose requirements, accounting for up to 40–45% ofthe inter-individual variability, depending on the pop-ulations and specific polymorphisms studied [3–5].Accordingly, since 2007, the United States Food andDrug Administration product label for warfarin hasbeen updated to reflect the potential value of incorp-orating genetic information into dose selection. Mostmajor contemporary clinical trials and meta-analysescomparing genotype-guided dosing to routine clinicalpractice or clinically guided algorithms have employedsurrogate outcomes and were not powered to demon-strate a difference in clinical endpoints [6–13].To date, the utility of genotype-guided dosing re-mains unresolved, particularly in Asian populations,since most randomized studies have thus far been per-formed in predominantly Caucasian cohorts. Variationin the epidemiology of VKORC1 and CYP2C9 geneticpolymorphisms across different ancestral populationscould impact the performance of pharmacogeneticallytailored dosing strategies [6, 7, 14]. The VKORC1 H1/H1 haplotype, which confers high sensitivity to war-farin, is present in 74%, 42%, and 7% of self-identifiedChinese, Malay, and Indian patients, respectively, whilethe CYP2C9*3 allele, which is associated with the poormetabolizer phenotype, is present in 7%, 9%, and 18%of patients, respectively [15]. On average, Asian pa-tients homozygous for less-sensitive VKORC1 haplo-types (H7, H8, or H9) and wild-type for CYP2C9 willrequire more than 3.5 times the maintenance dosageneeded by patients with the VKORC1 H1/H1 haplotypeand a copy of the CYP2C9*3 allele [15], highlighting apotential pitfall of empirical dose initiation and titration.Consequently, the application of pharmacogenetics toprovide tailored doses to patients of Asian ancestry isparticularly compelling. Accordingly, this randomized trialwas conducted to test whether a pharmacogeneticallybased dosing algorithm, which was developed from aracially diverse Asian cohort [16], is non-inferior to trad-itional clinical dosing.MethodsThe ethics review committees at participating centersapproved the study protocol (Additional file 1). The studywas conducted in accordance with Good Clinical Practiceguidelines, and patients provided written informed con-sent prior to enrollment. All serious adverse events werereported to the Domain Specific Review Board and theMedical Clinical Research Committee, Ministry of Health,in accordance with published guidelines. The study is reg-istered at ClinicalTrials.gov (Identifier: NCT00700895).Study designThis open-label, non-inferiority, randomized trial wasconducted in three large tertiary hospitals in South EastAsia. Randomization was computer generated with a 1:1allocation ratio, and patients were allocated to the treat-ment arms by means of sequentially numbered, opaque,and sealed envelopes.Eligibility criteria were age 18 years or older, a new indi-cation for long-term anticoagulation with warfarin, andtransaminases less than three times the upper limit ofnormal and bilirubin within normal range. Exclusion cri-teria were uncontrolled hypertension, peptic ulcer disease,previous history of liver disease, malabsorption syndromeor chronic diarrheal conditions, or any other medicalconditions deemed unfit for warfarin administration basedon the clinical judgment of primary treating physicians.Patients were not allowed to start warfarin before enrol-ment in the study. Demographic, clinical, and laboratorymeasurements were collected at baseline. Patient geno-types were determined through pyrosequencing as previ-ously described [15, 16], and data on race or ethnicity wasself-reported.InterventionThe study intervention period comprised of a doseinitiation period (first 3 days) and a dose adjustmentperiod (remainder of study). All patients were initiatedon low-molecular weight heparins at the point ofrandomization. The expected turnaround time forgenotyping was 2 days and warfarin was initiated on thethird day in both groups. Patients randomized to thegenotype-guided dosing strategy received their tailoreddose for 3 consecutive days. This was calculated usingan algorithm which takes into account the presence ofthe CYP2C9*3 allele, VKORC1 381 genotype, age, andweight [16]. The VKORC1 381 T > C single nucleotidepolymorphism is in complete linkage disequilibriumwith –1639G > A, and has been shown to discriminatethe H1 and H7 haplotypes in Asian individuals [15, 16].If genotype results were unavailable by the first sched-uled dose of warfarin (day 1), the patient would betreated with the traditional dosing approach. Patientsrandomized to the traditional dosing approach wereSyn et al. BMC Medicine  (2018) 16:104 Page 2 of 10initiated using a standardized loading dose regimen usedby the National University Hospital AnticoagulationClinic consisting of per os warfarin 5 mg on days 1 and2, followed by 3 mg on day 3. If the patient was morethan 75 years of age, the dose on day 2 was lowered to4 mg (Additional file 1). To account for instances whena different dose than that pre-specified was administeredduring the warfarin initiation period, deviations from theprotocol-specified dose were considered as dose adjust-ments. During the first 14 days, there were threemandatory INR checks on day 6, between days 7 and 9,and between days 12 and 14. Based on these INRmeasurements, warfarin dose titrations in both groupsas well as decisions to stop low-molecular weightheparin treatment were made according to usual clinicalpractice and centralized at the anticoagulation clinics(Additional file 1: Appendices 1 and 2). Included pa-tients were followed up until day 90 after warfarin initi-ation. The number and frequency of follow-up visitswere according to dosing tables that simulate real-worldclinical practice (Additional file 1). If urgent anticoagula-tion was needed, patients on both study arms receivedlow molecular weight heparin till INR reached the thera-peutic range of 1.9 to 3.1 to avoid warfarin-inducedthrombosis due to inhibition of Protein S and C. To avoidvariability from different warfarin sources, Marevan® tab-lets supplied by GlaxoSmithKline (Douglas ManufacturingLtd., AK, NZ) were used throughout this study.OutcomesThe primary outcome was the number of dose titrationsperformed up to end of week 2 (day 14). Patients cen-sored prior to day 14 were excluded from the modifiedintention-to-treat set due to insufficient data on dosetitrations, INR, and other anticoagulation parameters forthe evaluation of primary and secondary endpoints.Secondary outcomes were time to stable INR, defined asthe number of days from warfarin initiation to attainingtherapeutic INR (≥ 1.9 and ≤ 3.1) for the latter of twoconsecutive measurements that are at least 7 days apart;percentage of time spent within the therapeutic range(PTTR), which was estimated using the linearinterpolation method of Rosendaal et al. [17]; incidence ofdose adjustments and INR monitoring during follow-up;and the proportions of patients who had a bleeding epi-sode (classified as minor or major [18]), recurrent venousthromboembolism, and any measured INR value < 1.9 or >3.1. The PTTR was included in June 2016 as a secondaryoutcome by way of protocol amendment following a meet-ing with the Scientific Review Committee for the Surveil-lance and Pharmacogenomics Initiative for Adverse DrugReactions (SAPhIRE) program, who recommended thatreporting of this endpoint would facilitate between-trialcomparisons and enable meta-analyses of similar trials.Statistical analysisThe trial was powered to establish whethergenotype-guided warfarin dose administration wasnon-inferior to traditional clinical dosing for the primaryendpoint of number of dose titrations within the first2 weeks of therapy. Based on previous data [19], thesample size was estimated assuming a conservativebetween-group mean difference of 1.0 and a commonstandard deviation of 1.4 dose titrations. Therefore, with80% power and a one-sided type I error of 5%, a samplesize of 270 would be able to demonstrate non-inferiorityof the genotype-guided group for a predefinednon-inferiority margin of 0.5 dose titrations. Assumingup to 15% loss to follow-up before day 14, a minimumof 320 patients was deemed necessary. If the upperbound of the 90% confidence interval (CI) of the differ-ence in treatment (genotype-guided vs. traditional dos-ing) was lesser than 0.5, the null hypothesis would berejected, which would signify that the genotype-guidedstrategy was non-inferior to the traditional dosingapproach. When non-inferiority was proven, a two-tailedt test with an alpha value of 0.05 was used for superioritytesting.All other secondary endpoints were tests of superior-ity of genotype-guided dosing versus traditional dosing,and significance was defined as a two-tailed nominalP < 0.05. Time to stable INR was evaluated using theKaplan–Meier method, and the log-rank test was usedto compare differences. Percentage of time within thetherapeutic range was compared using two-sample ttests. Mixed effects Poisson regression models wereused to estimate incidence rate ratios (IRRs) for com-paring the number of dose adjustments and INR mea-surements between interventions, while accounting forpossible intra-subject correlation of count data whichwere measured at 1, 2, and 3 months. To account forthe reduced follow-up time among patients who with-drew or discontinued the trial before day 90, we usedan exposure variable in the Poisson regression for thenumber of days on trial. Predicted incidences of doseadjustments and INR measurements were estimated viaStata’s post-estimation command, immediately afterfitting Poisson regression models. Differences in theproportions of patients who experienced minor ormajor bleeding, recurrent venous thromboembolism,and INR < 1.9 or > 3.1 were quantified using relativerisks, with P values provided by Fisher’s exact test, and95% CIs obtained from exact binomial distributions.Finally, the performance of the genotype-guided war-farin dosing model was evaluated using the Pearson’sproduct-moment correlation, with 95% CIs computedbased on Fisher’s transformation, mean percentageerror, root mean squared error, and Bland–Altmananalysis.Syn et al. BMC Medicine  (2018) 16:104 Page 3 of 10All analyses were performed on a modifiedintention-to-treat basis and without imputation. Statisticalanalyses were performed in Stata version 13.0 (STATACorp., College Station, TX, USA).ResultsFrom May 11, 2007, through July 14, 2016, a total of 334patients were screened, of whom 322 were randomized(159 to the genotype-guided group and 163 to the trad-itional dosing group) (Fig. 1). Baseline characteristicsand genotypic distributions were well-balanced betweenboth groups. Patients had a median age of 60 years(range, 19–89), and the majority of patients were male(58.4%) and of Chinese race (61.2%) (Table 1). Genotyperesults were available within the first 4 days for 147 of159 (92.5%) patients randomized to the pharmacogenet-ics arm, and therefore these patients successfullyreceived the first genotype-tailored dose on days 3 or 4as scheduled in the protocol. Specifically, 88 (55.3%), 34(21.4%), 14 (8.8%), and 11 (6.9%) patients had genotyperesults returned on days 1 through 4, respectively. Theremaining 12 patients (7.5%) randomized to the pharma-cogenetics arm were switched to traditional dosing asgenotype results were not available by day 5.In the primary analysis only patients who receivedwarfarin treatment for at least 14 days were included.Thus, 133 (83.6%) and 136 (83.4%) patients from thegenotype-guided and traditional dosing groups, respect-ively, were included in the modified intention-to-treatset (reasons for censoring are shown in Fig. 1). Clinicaldemographics and genotypic frequencies among patientswho discontinued warfarin before 14 days of therapy aredetailed in Additional file 2: Table S1, and baseline char-acteristics were relatively similar as compared to theoverall population. The causes of death of four patientsin the traditional dosing group were cardiac arrest,retroperitoneal bleed, hospital-associated pneumonia,and advanced cancer. Median duration of warfarin ther-apy was comparable between the two groups, and was90.0 days (interquartile range (IQR) 83.8–90.0 days) inthe traditional dosing group and 90.0 days (IQR 77.0–90.0 days) in the genotype-guided group.Fig. 1 Flow of participants through the study of genotype-guided versus traditional-dosing of warfarin. aFurther tests were negative for thrombus.bPotential drug interaction with concomitant corticosteroid medications. cPatients were started on conventional dose of warfarin while awaitinggenotype resultsSyn et al. BMC Medicine  (2018) 16:104 Page 4 of 10Primary outcomeThe average number of dose titrations performed up tothe 14th day was 1.77 (95% CI 1.55 to 2.00) in thegenotype-guided group versus 2.93 (95% CI 2.63 to 3.24)in the traditional dosing group (mean difference −1.16,90% CI −1.48 to −0.84). Thus, both non-inferiority (P <0.001), according to the pre-specified definition, andsuperiority (P < 0.001) of the genotype-guided dosingalgorithm over the traditional dosing algorithm wasestablished (Fig. 2). This difference in mean number ofdose titrations corresponds to an IRR of 0.60 (95% CI0.51 to 0.70, two-sided P < 0.001) in favor of thegenotype-guided dosing algorithm.Secondary outcomesThe effect of warfarin therapy on INR trajectories isdepicted in Fig. 3a. The median time to stable INR,defined as the number of days from randomization tothe latter of two consecutive measurements that are atleast 7 days apart, was 36 days (IQR 20–74 days) in theTable 1 Baseline clinical characteristics and demographicsTraditional dosing (n = 163) Genotype-guided dosing (n = 159)Age, mean (SD), years 59.4 (14.5) 58.4 (14.3)Male, No. (%) 88 (54.0) 100 (62.9)Weight, mean (SD), kg 66.9 (16.8) 67.3 (14.1)Race, No. (%)Chinese 98 (60.1) 99 (62.3)Malay 39 (23.9) 32 (20.1)Indian 17 (10.4) 14 (8.8)Others 9 (5.5) 14 (8.8)CYP2C9 genotype, No./total (%)Presence of *3 allele 11/160 (6.9) 7/158 (4.4)VKORC1–381 genotype, No./total (%)C/C 91/162 (56.2) 97/159 (61.0)C/T 47/162 (29.0) 43/159 (27.0)T/T 24/162 (14.8) 19/159 (12.0)Indication, No./total (%)Atrial fibrillation 55/160 (34.4) 61/156 (39.1)Stroke 11/160 (6.9) 11/156 (7.1)Deep vein thrombosis 44/160 (27.5) 42/156 (26.9)Pulmonary embolism 19/160 (11.9) 17/156 (10.9)Left ventricular thrombus 17/160 (10.6) 18/156 (11.5)Others 26/160 (16.3) 14/156 (9.0)Amiodarone, No./total (%) 3/159 (1.9) 7/156 (4.5)Low-molecular weight heparins, No./total (%) 78/159 (49.1) 88/157 (56.1)Medical history, No./total (%)Stroke 16/160 (10.0) 10/157 (6.4)Deep vein thrombosis 7/160 (4.4) 4/157 (2.6)Pulmonary embolism 2/160 (1.3) 3/157 (1.9)Myocardial infarction 8/160 (5.0) 17/157 (10.8)Congestive heart failure 21/160 (13.1) 18/157 (11.5)Hypertension 86/160 (53.8) 92/157 (58.6)Type 2 diabetes mellitus 58/160 (36.3) 56/157 (35.7)Centre, No. (%)National University Hospital, Singapore 144 (88.3) 144 (90.6)University of Malaya Medical Centre, Malaysia 15 (9.2) 15 (9.4)Tan Tock Seng Hospital, Singapore 4 (2.5) 0 (0.0)Syn et al. BMC Medicine  (2018) 16:104 Page 5 of 10genotype-guided group versus 37 days (IQR 22–76 days)in the traditional dosing group. A total of 103 (77.4%) pa-tients in the genotype-guided group achieved stable INR ascompared with 108 (79.4%) in the traditional dosing groups,and the rate of attaining stable INR was not statistically dif-ferent between groups (genotype-guided vs. traditional dos-ing HR 1.00, 95% CI 0.76 to 1.31, P = 0.99) (Fig. 3b). Therewas no evidence of difference in the percentage of time inthe therapeutic range (based on the pre-specified INR targetof 1.9–3.1) over the follow-up period (Fig. 3a). The percent-age of time in the pre-specified therapeutic range were60.0% (95% CI 56.1% to 64.0%) in the genotype-guidedgroup compared with 57.1% (95% CI 53.2% to 61.0%) inthe traditional dosing group (mean difference 2.9%, 95%CI –2.6% to 8.4%, P = 0.29). Based on a post-hoc targetINR range of 2.0–3.0, the percentage of time in the thera-peutic range was 52.5% (95% CI 48.5% to 56.5%) in thegenotype-guided group compared with 47.1% (95% CI43.0% to 51.1%) in the traditional dosing group (mean dif-ference 5.4%, 95% CI –0.2% to 11.1%, P = 0.059).The number of dose adjustments and INR measurementsgenerally decreased over the first through third months oftreatment (Fig. 3c, d). The frequency of dose adjustmentswas significantly lower in the genotype-guided group overthe entire duration of treatment (4.51 ± 2.20 vs. 6.06 ± 2.93,IRR 0.76, 95% CI 0.67 to 0.86, P = 0.001) compared to thetraditional dosing group, after accounting for variation inbetween-individual exposure time and within-individualcorrelations in repeated measurements using a log-linearmixed effects Poisson model. The frequency of INRmeasurements did not differ significantly between thegenotype-guided group versus the traditional dosing groupover the follow-up period (8.63 ± 4.26 vs. 9.48 ± 4.05, IRR0.91, 95% CI 0.82 to 1.01, P = 0.076).Minor bleeding complications occurred in 8/132 (6.1%,95% CI 2.7% to 11.6%) patients in the genotype-guidedgroup versus 8/135 (5.9%, 95% CI 2.6% to 11.3%) patientsin the traditional dosing group (RR 1.02, 95% CI 0.40 to2.64, P = 0.96); major bleeding complications occurred in5/133 (3.8%, 95% CI 1.2% to 8.6%) and 5/136 (3.7%, 95%CI 1.2% to 8.4%) patients, respectively (RR 1.02, 95% CI0.30 to 3.45; P = 0.97); and recurrent venous thrombo-embolism was documented in 2/132 (1.5%, 95% CI 0.2%to 5.4%) and 1/135 (0.7%, 95% CI 0.02% to 4.1%) patients,respectively (RR 2.05, 95% CI 0.19 to 22.3, P = 0.55).Furthermore, an INR value of less than 1.9 was recordedat least once in 129/132 (97.7%, 95% CI 93.5% to 99.5%) inthe genotype-guided group versus 128/135 (94.8%, 95% CI89.6% to 97.9%) in the traditional dosing group (RR 1.03,95% CI 0.98 to 1.08, P = 0.21), whereas a measured INR ofgreater than 3.1 occurred in 59/132 (44.7%, 95% CI 36.0%to 53.6%) and 60/135 (44.4%, 95% CI 35.9% to 53.2%),respectively (RR 1.01, 95% CI 0.77 to 1.31, P = 0.97). Thus,no significant differences in these safety outcomes weredetected between the genotype-guided and traditionaldosing regimens.The predictive performance of the pharmacogeneticmaintenance dose model was also evaluated (Fig. 4a).Based on available data, the predicted daily maintenancedosages correlated positively with actual documenteda bFig. 2 a Number of dose titrations within first 2 weeks of therapy. Dark horizontal lines indicate median values. The circle represents the mean.The top line of the box indicates the 75th percentile, and the bottom line of the box indicates the 25th percentile. The top and bottom whiskersindicate the 97.5th and 2.5th percentiles, respectively. b Non-inferiority and superiority comparison for the primary endpoint of mean differencein number of dose titrations within first 2 weeks of therapy. Error bars indicate two-sided 90% or 95% CI, respectively. Since the upper bound ofthe 90% CI of the difference in treatment (genotype-guided vs. traditional dosing) was less than 0.5, the genotype-guided strategy was non-inferior to the traditional dosing approach. The upper bound of the 95% CI did not exceed 0, indicating that superiority was also demonstratedSyn et al. BMC Medicine  (2018) 16:104 Page 6 of 10stable dosages (R2 = 42.4%, 95% CI 31.9% to 52.4%, P <0.001) with a root mean-squared error of 1.10 mg and amean percentage error of −7.4% (Fig. 4b), indicating alow level of positive forecast bias and a respectable levelof predictive accuracy.DiscussionWarfarin and its analogues have been used as oral antico-agulants for more than 60 years, and many institutionsworldwide still employ an empirical dose initiation proto-col despite the known inter-individual variability in doserequirements and anticoagulation outcomes. In this studyinvolving patients with a new indication for warfarintherapy, the number of dose titrations in the first 2 weeks,and also throughout the follow-up period, was lower inthe genotype-guided group than in the traditional dosinggroup. Furthermore, the genotype dosing algorithmaccurately predicted the maintenance dose requirementsin patients who achieved stable INR. Our findings areconsistent with results from the COUMAGEN-I trial [10],which showed a similar advantage for accurate predictionof stable doses and frequency of dosing adjustments in thepharmacogenetically guided arm, but similar outcomes interms of anticoagulation control parameters such as thefraction of out-of-range INRs.The finding that percentage of time spent within thetherapeutic range (PTTR) was not statistically differentbetween the two groups was similar to that observed inthe recent COAG trial [6], but different from that in theEU-PACT [7], COUMAGEN-II [8], and GIFT trials [13].Although widely interpreted as failure of genotype-guideddosing, a major confounder when interpreting theseendpoints are the incidence of dose adjustments and INRmonitoring performed in the genotype-guided group andin the control group. Given that dosing titrations wereperformed more frequently in our control arm than in thegenotype-guided arm, this could have inflated the PTTRin the traditional dosing group and diminished anya bc dFig. 3 Secondary endpoints in the study. a Median international normalized ratio (INR) trajectory and 20–80th percentile bands over a 90-dayperiod. b Kaplan–Meier failure functions for the proportion of patients who achieved stable INR, which was not significantly different betweentreatment groups. Spikes on the Kaplan–Meier curves represent censoring. c and d Number of dose titrations and INR monitoring at 1, 2, and3 months, predicted using STATA’s post-estimation commandSyn et al. BMC Medicine  (2018) 16:104 Page 7 of 10apparent benefit of genotype-guided dosing. Notwith-standing, our trial was not designed to answer whethergenotype-guided dosing improves anticoagulation controlwhen controlled for the number of dose adjustments.Future trials may therefore consider incorporating thispotential confounder as an adjustment or stratificationvariable into their statistical analysis plans.Some investigators have advocated that warfarin dos-ing algorithms should be population specific and evalu-ated in populations similar to those from which theywere developed [14, 20]. Therefore, although severalgenotype-based dosing algorithms have been proposed[21–25], a strength of this study is the selection of analgorithm [16] developed and validated in a cohort thatis racially comparable to the current study population.The use of clinical algorithms for dose initiation anddose adjustment were applied in the EU-PACT andCOAG studies, which enrolled predominantly Caucasianand Black populations. These dosing algorithms have notbeen validated in Asian populations, and therefore werenot used in this study. Moreover, the fact that PTTR ofthe control group was comparable between our study andthe clinical algorithm dosing groups in the Western stud-ies would suggest no difference in outcomes with applica-tion of clinical algorithm-based dosing in this study.From the viewpoint of clinical applicability, the findingsof this study are representative of and generalizable to anethnogeographically diverse Asian population; the Chineseand Indian patients in these studies are mostly migrantsfrom China and South India, and the Malay patients areindigenous to the islands of the Indonesian archipelago,including Malaya, Sumatra, and Java [15, 16]. Reductionof frequency of dose titrations (the primary endpoint)using genotype-based algorithms is highly desirable in thecontext of Asia, where long distances from rural or subur-ban areas to healthcare facilities poses a barrier to optimalanticoagulation therapy.The use of a non-inferiority design in this study de-serves mention. Firstly, no comparative data about thecapability of a genotype-guided dosing strategy in redu-cing the number of dose titrations was available at the timeof conceptualization, although early observational andretrospective studies suggested that genotyping may havevalue in informing dose selection [26]. Considering that, atthe time this trial was conceived, there was no prospectivedata comparing pharmacogenetically guided dosing versustraditional dosing, it was arguably a reasonable concern thatpharmacogenetically guided dosing could be worse thantraditional dosing in terms of the number of dose adjust-ments required in the first 2 weeks. As such, anon-inferiority null hypothesis that genotype-guided dosingcould be worse than traditional dosing was arguably justifi-able and valid at that time. Therefore, the trial was designedto demonstrate that pharmacogenetically guided dosingwas not less efficacious than conventional dosing, and assecondary endpoints, to test whether a genotype-guided al-gorithm accurately predicts maintenance dose requirementsand improves other markers of anticoagulation control.There are several limitations of this study. Firstly, we didnot evaluate a combination of a loading-dose algorithma bFig. 4 a Scatterplot of predicted versus actual maintenance dosage. The solid line indicates the line of equivalence, while the dashed linerepresents the linear fit between algorithm-predicted and actual maintenance warfarin dosages. The plot includes only patients who haveachieved stable international normalized ratio (INR), which is defined as attaining therapeutic INR (≥ 1.9 and≤ 3.1) for two consecutivemeasurements that are at least 7 days apart. b Bland–Altman assessment of pharmacogenetic dosing model’s predictive performance.Shaded area indicates 95% confidence limitsSyn et al. BMC Medicine  (2018) 16:104 Page 8 of 10and dose-revision algorithm in the genotype-guided arm,which may be responsible for the PTTR advantage ob-served in the EU-PACT study [7]. Simulations integratinga pharmacokinetic-pharmacodynamic model [27, 28] andgenotype frequencies present in a Han-Chinese cohort infact suggests that the deployment of genetically informedloading doses and dose revisions is superior to a clinicallyguided dosing regimen [29]. Nevertheless, this study wasnot designed to test a difference in the PTTR, which asmentioned earlier, may be confounded by imbalance inthe number of dose titrations performed in the twogroups, nor was it powered to detect differences in theoutcome of bleeding and re-thrombosis. Other limitationsinclude the lack of adjustment for multiplicity among thesecondary endpoints; its open-label design, whichpotentially introduces ascertainment bias; and a lack ofpre-specified adjusted or subgroup analyses, for example,assessment of endpoints according to ethnic grouping,which may have afforded further information on the utilityof genotype-guided warfarin dosing. Furthermore, ap-proximately 16% of patients were excluded because theydid not continue warfarin for 14 days, although it shouldbe noted that this attrition rate is in line with the expecteddropout rate of approximately 15% that was accounted forin our sample size calculations. Moreover, the study wasdesigned as a multicenter clinical trial yet the majority(86.7%) of patients were enrolled at a single tertiary carecenter due to slow accrual in the other centers. This there-fore limits the generalizability of our results.ConclusionsIn this randomized, non-inferiority clinical trial thatincluded 322 adults of South East Asian ancestry,genotype-guided dosing reduced the number of dose titra-tions during the first 2 weeks compared to traditionaldosing (1.77 vs. 2.93) while maintaining similar INR timewithin therapeutic ranges. The reduction in frequency ofdose revisions persisted over the 90-day follow-up period(incidence rate ratio 0.76). The genotype-guided algorithmalso accurately predicted maintenance dose requirements.These findings imply that clinicians treating Asian patientsmay consider applying a pharmacogenetic algorithm topersonalize initial warfarin dosages.Additional filesAdditional file 1: Statistical Analysis Plan (SAP) and Study Protocol.(PDF 307 kb)Additional file 2: Table S1. Baseline characteristics of patients excludedfrom primary analysis. (DOCX 25 kb)AbbreviationsCYP2C9: cytochrome P450 2C9; INR: International normalized ratio;PTTR: percentage of time spent in therapeutic range; VKORC1: Vitamin Kepoxide reductase complex 1AcknowledgementsWe are grateful to Hiarul Fariz Hairol, Jesinda Pauline, Swee Siang Ng, KingXin Koh, Mohamed Hafiz Ridhwan Bin Yahy, Zuan Yu Mok, Michelle Rozario,Sili Tan, and Fei Fei Gan for technical and administrative support.FundingThis investigator-initiated trial was funded by the Singapore Ministry ofHealth’s National Medical Research Council (NMRC/CSA/021/2010 andNMRC/CSA/0048/2013). The Surveillance and Pharmacogenomics Initiativefor Adverse Drug Reactions (SAPhIRE) program is supported by a StrategicPosition Funds Grant (SPF2014/001) from the Biomedical Research Councilof the Agency for Science, Technology and Research (A*STAR). This research is alsosupported by the National Research Foundation Singapore and the SingaporeMinistry of Education under their Research Centres of Excellence Initiative.Availability of data and materialsAvailable through a data use agreement from the corresponding author.Authors’ contributionsConcept and design: BCG, BCT, NLS, SCL. Acquisition, analysis, orinterpretation of data: NLS, ALW, BCG, BCT, SCL, PM, HLT, GY, WTY, MLP, TSW,LB, MW, PCB, RS, RCSS, WK. Drafting of the manuscript: NLS, BCG, SCL, BCT,LB, ALW. Critical revision of the manuscript for important intellectual content:BCG, SCL, BCT, NLS, LB, ALW. Statistical analysis: NLS, BCT. Administrative,technical, or material support: PM, TSW, MW. Study supervision: BCG, SCL.All authors read and approved the final manuscript.Ethics approval and consent to participateThe ethics review committees at participating centers approved the studyprotocol (Additional file 1). The study was conducted in accordance withGood Clinical Practice guidelines, and patients provided written informedconsent prior to enrollment. All serious adverse events were reported to theDomain Specific Review Board and the Medical Clinical Research Committee,Ministry of Health in accordance with published guidelines. The study isregistered at ClinicalTrials.gov (Identifier: NCT00700895).Consent for publicationObtained as part of informed consent taking.Competing interestsThe authors declare that they have no competing interests.Publisher’s NoteSpringer Nature remains neutral with regard to jurisdictional claims inpublished maps and institutional affiliations.Author details1Department of Haematology-Oncology, National University Cancer Institute,Singapore, Singapore. 2Cancer Science Institute of Singapore, NationalUniversity of Singapore, Singapore, Singapore. 3Division of Neurology,Department of Medicine, National University Health System, Singapore,Singapore. 4Department of Cardiology, National University Heart Centre,Singapore, Singapore. 5Department of Medicine, University of Malaya MedicalCentre, Kuala Lumpur, Malaysia. 6Department of Pharmacy, NationalUniversity Hospital, Singapore, Singapore. 7Genome Institute of Singapore,Agency for Science, Technology and Research, Singapore, Singapore.8Translational Laboratory in Genetic Medicine, Agency for Science,Technology and Research, Singapore, Singapore. 9Department of Medicine,Centre for Heart Lung Innovation, University of British Columbia, Vancouver,Canada. 10Department of Pathology, Yong Loo Lin School of Medicine,National University Health System, Singapore, Singapore. 11Saw Swee HockSchool of Public Health, National University of Singapore, Singapore,Singapore. 12Department of Pharmacology, Yong Loo Lin School ofMedicine, National University Health System, Singapore 119228, Singapore.13Department of Medicine, Yong Loo Lin School of Medicine, NationalUniversity of Singapore, Singapore, Singapore.Syn et al. BMC Medicine  (2018) 16:104 Page 9 of 10Received: 3 January 2018 Accepted: 5 June 2018References1. Passman R. Time in therapeutic range in warfarin-treated patients: is verygood good enough? JAMA. 2016;316:872.2. Pokorney SD, Simon DN, Thomas L, Gersh BJ, Hylek EM, Piccini JP, et al.Stability of international normalized ratios in patients taking long-termwarfarin therapy. JAMA. 2016;316:661.3. Jonas DE, McLeod HL. Genetic and clinical factors relating to warfarindosing. Trends Pharmacol Sci. 2009;30:375–86.4. Johnson JA, Cavallari LH. Warfarin pharmacogenetics. Trends CardiovascMed. 2015;25:33–41.5. Syn NL-X, Yong W-P, Lee S-C, Goh B-C. Genetic factors affecting drug dispositionin Asian cancer patients. Expert Opin Drug Metab Toxicol. 2015;11:1879–92.6. Kimmel SE, French B, Kasner SE, Johnson JA, Anderson JL, Gage BF, et al. Apharmacogenetic versus a clinical algorithm for warfarin dosing. N Engl JMed. 2013;369:2283–93.7. Pirmohamed M, Burnside G, Eriksson N, Jorgensen AL, Toh CH, Nicholson T,et al. A randomized trial of genotype-guided dosing of warfarin. N Engl JMed. 2013;369:2294–303.8. Anderson JL, Horne BD, Stevens SM, Woller SC, Samuelson KM, MansfieldJW, et al. A randomized and clinical effectiveness trial comparing twopharmacogenetic algorithms and standard care for individualizing warfarindosing (CoumaGen-II). Circulation. 2012;125:1997–2005.9. Brensinger CM, Kimmel SE. Genetic warfarin dosing tables versus algorithms.JACC. 2011;57:612–8. http://dx.doi.org/10.1016/j.jacc.2010.08.64310. Anderson JL, Horne BD, Stevens SM, Grove AS, Barton S, Nicholas ZP, et al.Randomized trial of genotype-guided versus standard warfarin dosing inpatients initiating oral anticoagulation. Circulation. 2007;116:2563–70.11. Stergiopoulos K, Brown DL. Genotype-guided vs clinical dosing of warfarinand its analogues. JAMA Intern Med. 2014;174:1330.12. Dahal K, Sharma SP, Fung E, Lee J, Moore JH, Unterborn JN, et al. Meta-analysis of randomized controlled trials of genotype-guided vs standarddosing of warfarin. Chest. 2015;148:701–10.13. Gage BF, Bass AR, Lin H, Woller SC, Stevens SM, Al-Hammadi N, et al.Effect of genotype-guided warfarin dosing on clinical events andanticoagulation control among patients undergoing hip or kneearthroplasty. JAMA. 2017;318:1115.14. Limdi NA, Brown TM, Yan Q, Thigpen JL, Shendre A, Liu N, et al. Raceinfluences warfarin dose changes associated with genetic factors.Blood. 2015;126:539–45.15. Lee S-C, Ng S-S, Oldenburg J, Chong P-Y, Rost S, Guo J-Y, et al. Interethnicvariability of warfarin maintenance requirement is explained by VKORC1genotype in an Asian population. Clin Pharmacol Ther. 2006;79:197–205.16. Tham L-S, Goh B-C, Nafziger A, Guo J-Y, Wang L-Z, Soong R, et al. Awarfarin-dosing model in Asians that uses single-nucleotide polymorphismsin vitamin K epoxide reductase complex and cytochrome P450 2C9. ClinPharmacol Ther. 2006;80:346–55.17. Rosendaal FR, Cannegieter SC, Van Der Meer FJM, Briet E. A method todetermine the optimal intensity of oral anticoagulant therapy. ThrombHaemost. 1993;69:236–9.18. Levine M, Gent M, Hirsh J, Leclerc J, Anderson D, Weitz J, et al. Acomparison of low-molecular-weight heparin administered primarily athome with unfractionated heparin administered in the hospital for proximaldeep-vein thrombosis. N Engl J Med. 1996;334:677–81.19. Kovacs MJ, Rodger M, Anderson DR, Morrow B, Kells G, Kovacs J, et al.Comparison of 10-mg and 5-mg warfarin initiation nomograms togetherwith low-molecular-weight heparin for outpatient treatment of acutevenous thromboembolism. A randomized, double-blind, controlled trial.Ann Intern Med. 2003;138:714–9.20. Kubo K, Ohara M, Tachikawa M, Cavallari LH, Lee MTM, Wen MS, et al.Population differences in S-warfarin pharmacokinetics among AfricanAmericans, Asians and whites: their influence on pharmacogenetic dosingalgorithms. Pharmacogenomics J. 2017;17(6):494–500.21. Gage BF, Eby C, Milligan PE, Banet GA, Duncan JR, McLeod HL. Use ofpharmacogenetics and clinical factors to predict the maintenance dose ofwarfarin. Thromb Haemost. 2003;91:87–94.22. Sconce EA, Khan TI, Wynne HA, Avery P, Monkhouse L, King BP, et al. Theimpact of CYP2C9 and VKORC1 genetic polymorphism and patientcharacteristics upon warfarin dose requirements: proposal for a new dosingregimen. Blood. 2005;106:2329–33.23. International Warfarin Pharmacogenetics Consortium, Klein TE, Altman RB,Eriksson N, Gage BF, Kimmel SE, Lee MT, Limdi NA, Page D, Roden DM,Wagner MJ, JJ CMD. Estimation of the warfarin dose with clinical andpharmacogenetic data. N Engl J Med. 2009;360:753–64.24. Lenzini P, Wadelius M, Kimmel SE, Anderson JL. Integration ofgenetic, clinical, and INR data to refine warfarin dosing. ClinPharmacol Ther. 2010;87:572–8.25. Avery PJ, Jorgensen A, Hamberg AK, Wadelius M, Pirmohamed M, Kamali F.A proposal for an individualized pharmacogenetics-based warfarin initiationdose regimen for patients commencing anticoagulation therapy. ClinPharmacol Ther. 2011;90:701–6.26. Voora D, Eby C, Linder MW, Milligan PE, Bukaveckas BL, McLeod HL, et al.Prospective dosing of warfarin based on cytochrome P-450 2C9 genotype.Thromb Haemost. 2005;93:700–5.27. Hamberg a-K, Dahl M-L, Barban M, Scordo MG, Wadelius M, Pengo V,et al. A PK-PD model for predicting the impact of age, CYP2C9, andVKORC1 genotype on individualization of warfarin therapy. ClinPharmacol Ther. 2007;81:529–38.28. Hamberg a-K, Wadelius M, Lindh JD, Dahl ML, Padrini R, Deloukas P, et al. Apharmacometric model describing the relationship between warfarin doseand INR response with respect to variations in CYP2C9, VKORC1, and age.Clin Pharmacol Ther. 2010;87:727–34.29. Syn NL-X, Lee S-C, Brunham LR, Goh B-C. Pharmacogenetic versusclinical dosing of warfarin in individuals of Chinese and African-American ancestry: assessment using data simulation. PharmacogenetGenomics. 2015;25:491–500.Syn et al. BMC Medicine  (2018) 16:104 Page 10 of 10


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items