Polymorphisms in VKORC1 have more impact than CYP2C9 polymorphisms on early warfarin International Normalized Ratio control and bleeding rates

Authors


Correspondence: Dr R. C. Tait, Department of Haematology, 3rd Floor Macewen Building, Glasgow Royal Infirmary, Glasgow G4 OSF, UK. E-mail: Campbell.Tait@ggc.scot.nhs.uk

Summary

Poor warfarin control with resultant high International Normalized Ratios (INRs) and bleeding events is most common during the first months of treatment. The effects of genetic polymorphisms at the vitamin K epoxide reductase [VKORC1] and cytochrome P450 2C9 [CYP2C9] loci have been increasingly acknowledged as contributory factors of enhanced warfarin sensitivity. In our prospective, blinded study, 557 patients (49·1% male, mean age 65·4 years, range 18–91 years) commencing warfarin (target INR 2·5) were genotyped and monitored through the first 3 months of anticoagulation. Homozygosity for the −1639 G>A single nucleotide functional promoter polymorphism of the VKORC1 gene (genotype AA; 14·5% of cases) was associated with a significantly shortened time to therapeutic INR ≥ 2 (P < 0·01), reduced stable warfarin dose (P < 0·01), and an increased number of INRs > 5 (P < 0·001) and occurrence of bleeding events (P < 0·01) during the first month, as compared to the GG genotype. CYP2C9 genetic variations *2 and *3 were not associated with significant effect on these factors. Neither VKORC1 nor CYP2C9 polymorphisms influenced these parameters beyond the first month of treatment. These findings imply possible benefits of assessing VKORC1 polymorphisms prior to anticoagulation, particularly as a low dose induction regime in VKORC1 AA individuals appears to reduce the incidence of high INRs.

Warfarin remains the most widely used anticoagulant for prevention and treatment of thromboembolic disease. However its narrow therapeutic window and significant inter-individual variation demand regular monitoring of the International Normalized Ratio (INR). The most important side effect of warfarin is major bleeding, which is most common during the first 90 d of treatment whilst patients are being established on warfarin and INR control is least stable (Palareti et al, 1996). The use of validated induction algorithms and frequent INR monitoring may improve early INR control and are recommended in UK guidelines (Baglin et al, 2007; Keeling et al, 2011).

Age, ethnicity, co-morbid disease and co-medications are known to influence warfarin dose (Carlquist & Anderson, 2011). However, it is also acknowledged that genetic factors may explain up to 50% of the variability in warfarin dose requirements between individuals. (Flockhart et al, 2008). This is largely due to several single nucleotide polymorphisms (SNPs) in the cytochrome P450, family 2, subfamily C, polypeptide 9 gene (CYP2C9) and the vitamin K epoxide reductase subunit 1 gene (VKORC1). Warfarin pharmacokinetics are strongly influenced by the metabolism of the more active S enantiomer by the cytochrome CYP2C9 enzyme. In particular, two genetic variants (CYP2C9*2 which results in an amino acid substitution Arg144Cys, and CYP2C9*3 which results in an Ile359Leu substitution) are associated with slower warfarin metabolism and lower warfarin maintenance dose requirements than the wild type CYP2C9*1-encoded enzyme (Margaglione et al, 2000; Caldwell et al, 2007). The prevalence of these SNPs differs between racial groups (Thacker et al, 2008) and their effect on warfarin dosing variability has been estimated to be approximately 17% per *2 allele and 38% per *3 allele (Millican et al, 2007). The pharmacodynamic effects of warfarin on the INR are influenced by SNPs in VKORC1, the primary target of warfarin. In particular, the non-coding VKORC1 SNP -1639G>A is associated with a haplotype which has significantly reduced enzyme activity in the presence of warfarin (Rieder et al, 2005; Yuan et al, 2005). Homozygosity for this SNP is associated with up to 50% reduction in VKORC1 enzyme activity compared to wild type, and a significant increased sensitivity to warfarin (Oldenburg et al, 2007a,b). This SNP has been reported to explain 27% (range 15–34%) of warfarin dosing variability (Wadelius et al, 2005).

The recognition of a role for pharmacogenetics in warfarin prescribing has gained momentum and an enlarging body of complex warfarin induction algorithms, including CYP2C9 and VKORC1 genotype, have been developed in an attempt to improve efficiency and safety of warfarin dosing (Sconce et al, 2005; Voora et al, 2005; Millican et al, 2007; Lenzini et al, 2008; Klein et al, 2009). These models have been shown to predict 48–79% of variability in therapeutic warfarin maintenance dose. They compare favourably to pre-treatment predictions based on demographic and clinical data alone, but are similar to predictions based solely on warfarin doses and corresponding INR results during the first week of warfarin therapy which have been estimated to explain 68–77% of variance in maintenance dose (Ferder et al, 2010; Le Gal et al, 2010).

Limited studies have directly examined the effect of these warfarin-sensitizing SNPs on clinical outcomes such as increased incidence of bleeding events (Margaglione et al, 2000; Higashi et al, 2002; Lenzini et al, 2008). This has lead to a lack of a definitive opinion regarding whether the analysis of a patient's genetics would have a significant impact on safety of warfarin usage (Flockhart et al, 2008; Mannucci et al, 2008; Thacker et al, 2008; Rosove & Grody, 2009). There is also a lack of evidence concerning the cost effectiveness of genotype-led dosing and there is a fear that utilizing this testing strategy will potentially lull clinicians into a false sense of security and compromise “conscientious monitoring” (Rosove & Grody, 2009).

Four small randomized controlled trials have compared genotype guided versus standard warfarin dosing in patients commencing oral anticoagulation (Hillman et al, 2005; Anderson et al, 2007; Burmester et al, 2011; Caraco et al, 2008). Results of a systematic review examining genotype guided warfarin dosing trials did not find a significant difference in its primary end point of improvement of time in therapeutic range or reduction in adverse bleeding events. However, pharmacogenetic-guided dosing more effectively approximated stable doses, thereby demonstrating some beneficial effect (Kangelaris et al, 2009). Given the lack of high quality evidence for clinical benefit and the significant logistical and cost issues of providing rapid genetic results, the consensus remains that ‘further data is required’ before pharmacogenetic prescribing can be routinely recommended (Carlquist & Anderson, 2011; Keeling et al, 2011).

In order to assess the potential significance of VKORC1-1639A, CYP2C9 *2 and *3 SNPs on clinically relevant outcomes we undertook a prospective observational study of INR control and bleeding events during the first 3 months of warfarin therapy.

Patients, materials and methods

Patients

Consecutive patients commencing warfarin with target INR range 2·0–3·0 and intended duration of warfarin ≥3 months were recruited at two hospital anticoagulant clinics in Glasgow. Warfarin induction dosing was undertaken according to local policies, which recommended a 10 mg or age-adjusted induction algorithm (Fennerty et al, 1984; Roberts et al, 1999) for inpatients and those with acute thrombotic events or low dose 5 or 2 mg induction regimes (Oates et al, 1998; Tait & Sefcick, 1998) for elective warfarin initiation as an outpatient. Clinicians responsible for warfarin initiation and on-going dose adjustment were blinded to genotype results. Patient selection was not restricted by indication for warfarin or comorbidities. However, patients under 18 years or receiving amiodarone as co-medication were excluded. The study was approved by the Greater Glasgow and Clyde local ethics committee.

Genotyping

DNA samples, prepared from 150 μl anticoagulated blood using the Applied Biosystems Prism semi-automated 6100 method (Applied Biosystems, Warrington, UK), were genotyped for CYP2C9 and VKORC1 SNPs using TaqMan®, Drug Metabolism Genotyping Assays on the StepOnePlus Real-Time polymerase chain reaction machine (Applied Biosystems). Details of the Drug Metabolism genotyping assay kits are given in Table 1.

Table 1. Identification of drug metabolism genotyping assay kits
Common nameAssay IDBase changeGene symbolNCBI SNP referenceCelera ID
  1. ID, identification code; NCBI, National Center for Biotechnology Information; SNP single nucleotide polymorphism. All assay kits were provided by Applied Biosystems, Life Technologies, Paisley, UK.

CYP2C9*2C_25625805_10430C>TCYP2C9rs1799853hCV25625805
CYP2C9*3C_27104892_101075A>CCYP2C9rs1057910hCV27104892
VKORC1 G>AC_30403261_20-1639G>AVKORC1rs9923231hCV30403261

Data collection

All patient demographic and clinical data and all warfarin doses and INR results were obtained from the local anticoagulant database (DAWNAC). This included age, sex, ethnicity, indication for anticoagulation, co-medications and relevant medical history as well as high INR results and adverse clinical events.

Main outcome measures

The primary outcome measure was the prevalence of significantly above-range INRs (pre-defined as INR > 5) as a surrogate marker of increased bleeding risk. These events were analysed separately according to their occurrence during days 1–30 or days 31–90. Four secondary outcome measures included rate of bleeding events, time to first therapeutic INR (INR ≥ 2), stable warfarin maintenance dose and time to stable INR – defined as the first of 2 days, 7 days apart, since warfarin commenced on which INR had been within range without a change in dosing. Bleeding events were categorized as major (serious/life threatening) or clinically significant non-major bleeding according to previously published recommendations (Schulman & Kearon, 2005). Major bleeds were defined as fatal bleeding or symptomatic bleeding in a critical organ, such as intracranial, intraspinal, intraocular, retroperitoneal, intra-articular or pericardial or intramuscular with compartment syndrome and/or bleeding causing a fall in haemoglobin level of 20 g/l or more, or leading to transfusion of two or more units of red cells. Clinically significant non-major bleeding was defined as bleeding requiring medical attention or attendance at hospital, but not fulfilling criteria for major bleeding.

Statistical analysis

Power calculations, assuming a 20% prevalence for warfarin sensitizing SNPs and 12% incidence of primary events during the first month of warfarin therapy, indicated that 636 patients would be required to provide 80% power to detect a doubling of risk of primary outcome events, at P value < 0·05, associated with a warfarin sensitizing SNP compared to wild type genotype. Data were tested using the chi-square analysis of categorical variables, log rank test for time to event analyses and Mood's Median/Mann–Whitney U test for non-parametric data. Confidence intervals were set at 95%. All analyses were performed using Mini-TAB statistical package.

Results

Baseline characteristics

During the study recruitment period 557 patients (median age 69 years, range 18–93 years) were enrolled. Five hundred and eleven patients completed 3 months of warfarin treatment, 30 completed >1 but <3 months and 16 patients completed <1 month. The prevalence of warfarin-sensitizing SNPs in the study population was 38·4% (i.e. 196 patients with at least one of the following rare alleles CYP2C9*2, or *3 or homozygous for VKORC1 AA), exceeding our initial prediction, thus ensuring that the study was still adequately powered despite not quite meeting the initial recruitment target. Baseline patient characteristics and allelic frequencies are illustrated in Table 2. For each of the three variants studied, homozygotes and heterozygotes were in Hardy–Weinberg equilibrium suggesting no hidden bias in choice of patients. Each genotype group was equally distributed on the basis of sex, age and induction regime.

Table 2. Study patient demographics and genotypes
Total patients (557)n (%)
Demographics
Male sex283 (50·8)
Caucasian546 (98·0)
Indication
Atrial fibrillation251 (45·1)
Pulmonary embolism107 (19·2)
Deep vein thrombosis107 (19·2)
Other thrombosis35 (6·3)
Other indication33 (5·9)
Induction regime
<5 mg50 (8·9)
5 mg158 (28·4)
>5 mg329 (59·1)
Not known20 (3·6)
Genetic variations
VKORC1
AA81 (14·5)
GA266 (47·8)
GG210 (37·7)
CYP2C9*2
RR (wild type)423 (75·9)
RC128 (22·9)
CC6 (1·1)
CYP2C9*3
II (wild type)490 (87·9)
IL65 (11·7)
LL2 (0·4)
Patients with no warfarin-sensitizing alleles
(GG, RR, II)137 (24·6)

Out of range INRs

INRs > 5 were recorded in 12·7% (n = 71) patients overall during the first month of treatment. In the case of VKORC1 there was a significant association (P < 0·001) between out of range INR and genotype (Table 3). There was no association between increased INR and CYP2C9 *2 and *3 genotypes compared to *1 wild type (P = 0·081 and 0·144 respectively).

Table 3. Incidence of out of range INRs according to CYP2C9 and VKORC1 single nucleotide polymorphisms
GenotypeDays 1–30 (557 patients)Days 31–90 (511 patients)
INR > 5INR > 5
n (total patients)%n (total patients)%
  1. a

    P values by chi-square analysis for INRs > 5 in patients with at least one CYP2C9*2 or *3 allele compared to wild type patients (*1*1) with no *2 or *3 variant alleles.

  2. b

    P values by chi-square analysis for INR > 5 comparing against VKORC1 GA genotype.

  3. c

    P values by chi-square analysis for INR > 5 comparing against VKORC1 GG genotype.

CYP2C9*1 homozygous (RR and II)39 (370)10·514 (339)4·1
CYP2C9*2 (RC or CC)

23 (134)

P = 0·081a

17·2

4 (121)

P = 0·688a

3·3
CYP2C9*3 (IL or LL)

12 (68)

P = 0·144a

17·6

3 (66)

P = 0·878a

4·5
VKORC1 AA

26 (81)

= <0·001b

P = <0·001c

32

2 (70)

P = 0·876b

P = 0·425c

2·9
VKORC1 GA

33 (266)

P = 0·013c

12·4

8 (248)

P = 0·303c

3·2
VKORC1 GG12 (210)5·710 (193)5·2

INRs > 5 were recorded in 3·9% (n = 20) patients during months 2–3 of therapy. This lower rate of supra-therapeutic INRs is to be expected as patients become established on therapy. Of these high INRs, nine of the 20 cases (45%) occurred in patients who had neither VKORC1 AA genotype nor had either of the CYP2C9 variants.

Bleeding events

The total incidence of bleeding events recorded during the first month of anticoagulation was lower than predicted (1·97% overall, n = 11; three major bleeds and eight clinically significant non-major bleeds). The numbers of bleeding events recorded in the VKORC1 AA genotype group was significantly higher at 4·9% (n = 4) compared to 0·47% (n = 1) in the GG genotype group [P = 0·009]. The bleeding rate in the GA genotype group (2·3%) was not significantly different from that in either AA or GG groups. The incidence of bleeding events did not correlate with CYP2C9 genotype.

Days to INR ≥ 2, days to stable warfarin dose, final stable warfarin dose

As anticipated, warfarin-sensitizing SNPs reduced the time taken to reach therapeutic INR. Patients with VKORC1 AA haplotype reached an INR of ≥2 more quickly when compared independently against GA and GG genotypes by log rank test (median number of days: AA 4 d, GA 5 d and GG 5 d; AA versus GA P < 0·01, AA versus GG P < 0·01, GA versus GG P = 0·04). There was no significant difference in median time to INR ≥ 2 between CYP2C9 *2, *3 and wild type groups (*2 variants 5 d, *3 variants 4·5 d and wild type 5 d).

During the study period 440 patients (78·9%) reached stable anticoagulation as defined by the criteria detailed in Patients, materials and methods. Contrary to previous reports in the literature (Higashi et al, 2002; Voora et al, 2005), genotype variations did not appear to alter time to reach steady state INR.

VKORC1 and CYP2C9*2 genotype both appeared to be associated with final stable warfarin dose. Patients with variant alleles had a lower final dose requirement; VKORC1 median dose AA 2 mg, GA 4 mg and GG 4·5 mg [P < 0·01] and CYPC9*2 variant 3·6 mg, wild type 4·0 mg [P < 0·01]. There was no significant difference in final dose between CYP2C9*3 variant and wild type groups; CYP2C9*3 variant 4·2 mg and wild type 4·0 mg [P = 0·675]. There was evidence of a cumulative impact of genetic variations inversely correlating with the final stable warfarin dose (Fig 1).

Figure 1.

Stable warfarin maintenance dose requirements according to the number of warfarin-sensitizing single nucleotide polymorphisms. Box depicts the inter-quartile range; horizontal line and vertical whiskers represent the median, 2·5 and 97·5 centile ranges, respectively.

Choice of induction regime

Finally we performed a focused analysis of the VKORC1 AA genotype patients to assess whether the induction regime utilized (2, 5 or 10 mg) influenced the number of out of range INRs during the first month of therapy. Out of 78 VKORC1 AA patients, one of eight (12·5%) using a 2 mg induction regime experienced an INR > 5 during the first month of therapy. In contrast, 25 of 70 (35·7%) using a ≥5 mg induction regime experienced an INR > 5. It is notable that patients given such low dose induction were usually >80 years old, so the low incidence of AA patients with INR > 5 is even more relevant.

Discussion

Individual responses to anticoagulation using Vitamin K antagonists are a reflection of a number of clinical and genetic factors. The contributory effect of pharmacogenetics has become increasingly recognized, however, to date there is no consensus to support the adoption of genetic testing into clinical practice.

This large prospective study has examined the impact of CYP2C9 and VKORC1 polymorphisms on both laboratory and clinical outcome variables in a real world setting. Examination of these two gene loci generally confirm previous epidemiological analyses reporting the incidence of warfarin-sensitizing genetic variants within a predominantly Caucasian population (Oldenburg et al, 2007a). Our findings also support published evidence demonstrating the effect that warfarin-sensitizing genes have on reducing stable maintenance warfarin dose.

Our data suggests that the VKORC1 genotypes GA and especially AA confer a more striking effect on early warfarin control, significantly influencing time to first INR in therapeutic range and incidence of supra-therapeutic INRs. However, it should be noted that of 71 INRs > 5 occurring in the first month only 26 of these individuals had the “high risk” AA genotype – supporting the fact that genotype is only one of a number of factors resulting in an increased predisposition to poor warfarin control. The study also shows a trend towards an increased number of early bleeding events in the AA genotype cohort, even allowing for the fact that the total number of bleeds was lower than expected and the study was not powered to detect a statistically significant difference at this level.

Our findings suggest that any impact of VKORC1 and CYP2C9 SNPs on variability of warfarin response is predominantly limited to the first month of treatment. After this time the predictive impact is largely superseded by other influences that predict INR response, such as dose history and previous INRs, as has been reported by others (Ferder et al, 2010). The effect of VKORC1 was significantly stronger than CYP2C9, although it must also be acknowledged that the prevalence of individuals homozygous or compound heterozygous for CYP2C9 variants is relatively low. It is recognized that the impact of CYP2C9 is often not evident until after early induction i.e. not reflected in early INR response (Ferder et al, 2010) and it is possible that by this stage other contributory factors, such as age, weight, diet and other drugs, may mask the effect of the genetic variants.

Certain limitations to the study design may have resulted in an underestimate of supra-therapeutic INRs and bleeding events. The study population comprised patients attending the outpatient anticoagulant clinic who had either been referred from the inpatient setting or from primary care. Any patients who had been commenced on warfarin as an inpatient and subsequently suffered an early bleed or who experienced multiple very early high INRs, and may have had warfarin prematurely stopped for either of these reasons, would have never been brought to our attention as a result of not being referred to the outpatient clinic. The use of standard loading protocols was beneficial in preventing clinician variability bias to the study, however, patients commencing warfarin electively began treatment on low dose loading regimes (2 or 5 mg), which again would be likely to reduce the incidence of bleeding events relative to the high dose loading regimes that are utilized as a more standard approach in the inpatient setting.

Our findings do support the potential value of assessing the VKORC1 genotype. We recognize that there are logistical problems with implementing warfarin genetic testing. Firstly, most patients who require warfarin therapy need it immediately but currently provision of genetic results in a timely fashion is difficult. Secondly, the number of patients with SNPs conferring extreme risk of over anticoagulation are relatively small. However, given the high rate of supratherapeutic INRs in the AA genotype (which is associated with an average daily maintenance dose of around 2 mg), the potential to decrease this risk by using a 2 mg induction regime may justify screening for this particular genotype, particularly in the patients who are commenced on the drug semi-electively (e.g. asymptomatic atrial fibrillation). However, large-scale interventional studies would be required to confirm the benefit and assess the cost-effectiveness of such a management strategy.

In conclusion, we have shown that in our study population VKORC1 polymorphisms are more influential than CYP2C9 with regard to time to therapeutic INR and development of high INRs, and that beyond the first month of therapy both of these genotype groups have less effect on the risk of high INRs. Consideration could be given to screening for VKORC1 AA homozgygous individuals so they might benefit from a very low dose induction regimen.

Acknowledgements

KL and CT designed the study and wrote the paper. AME, PT and CT assisted with patient recruitment while KL, assisted by RS, undertook the data collection and analysis. DG performed the genotyping. All authors contributed to the final manuscript and would like to thank Dr Simon Rinaldi for statistical support.

Ancillary