Coumarin anticoagulants

Uses

The main use of the coumarins is in the treatment and prevention of thromboembolic disease, including deep vein thrombosis, pulmonary embolism, and cerebral embolism from cardiac and other sources.

Protein C, another vitamin K-dependent serine protease zymogen in plasma, is a regulatory protein that, when activated, limits the activity of two activated procoagulant co-factors, factors Va and VIIIa. Heterozygotes for hereditary isolated protein C deficiency tend to develop a thrombotic disease which has been successfully treated with long-term coumarins [ , ]. Apparently, the balance between the activities of protein C and the procoagulant factors (II, VII, IX, and X), which is disturbed in protein C-deficient patients, is restored during long-term treatment with coumarins.

Non-anticoagulant uses

Warfarin reduces calcium deposition in spontaneously degenerated bioprosthetic valves [ ].

Both direct and indirect antitumor actions of anticoagulants have been postulated on the basis of experimental findings in animals [ ]. Warfarin given alone or in combination with cytostatic drugs reduces the size of fibrosarcomas in animals [ , ] and of osteosarcomas in man [ ], and consequently prolongs survival times in both. The Cooperative Studies Program of the Veterans Administration Medical Research Service suggests that, among patients with various tumors, those with small-cell carcinoma of the lung had a significantly longer survival (about 4 years instead of 2) when warfarin was added to standard treatment [ ]. There were no differences in survival between warfarin-treated and control groups for advanced non-small-cell lung cancers, colorectal, head and neck, and prostate cancers [ ]. However, the contention that cancer morbidity and/or mortality is reduced in anticoagulated patients is not substantiated by the results of a retrospective study of 378 patients who had been taking anticoagulants for about 10 years on average [ ].

It has been suggested that warfarin may have a specific effect on tumor-cell growth via the inhibition of protein synthesis [ ]. Another explanation of the effect of the drug is a reduction in the co-adherence of tumor cells, which renders them more vulnerable to the actions of defence mechanisms. The reduction in co-adherence is thought to be caused by an inhibitory action of anticoagulants on the fibrin network that is vital for tumor cell growth. This hypothesis, which is supported by dose dependence of the inhibitory effect of both heparin and warfarin [ ], has been studied in a controlled randomized trial of the use of streptokinase after surgery for tumors of the large bowel [ ].

Supercoumarins

Anticoagulant pesticides are used widely in agricultural and urban rodent control. The emergence of warfarin-resistant strains of rats led to the introduction of a group of anticoagulant rodenticides variously referred to as “supercoumarins”, “superwarfarins”, “single dose” rodenticides, or “long-acting” rodenticides [ ]. This group includes the second generation 4-hydroxycoumarins brodifacoum, bromadiolone, difenacoum, and flocoumafen and the indanedione derivatives chlorophacinone and diphacinone; these drugs typically have longer half-lives than warfarin.

The greater potency and duration of action of long-acting anticoagulant rodenticides is attributed to: (i) their greater affinity for vitamin K epoxide reductase; (ii) their ability to disrupt the vitamin K epoxide cycle at more than one point; (iii) hepatic accumulation; and (iv) unusually long half-lives, due to high lipid solubility and enterohepatic recirculation.

There have been several published case reports of accidental or intentional poisoning due to these drugs [ , ], which can result in prolonged coagulopathy [ , ].

Most cases of anticoagulant rodenticide exposure involve young children, and so the amounts ingested are almost invariably small. In contrast, intentional ingestion of large quantities of long-acting anticoagulant rodenticides can cause anticoagulation for several weeks or months [ ]. Occupational exposure has also been reported.

Substantial ingestion produces epistaxis, gingival bleeding, widespread bruising, hematomas, hematuria with flank pain, menorrhagia, gastrointestinal bleeding, rectal bleeding, and hemorrhage into any internal organ; anemia can result. Spontaneous hemoperitoneum has been described. Severe blood loss can result in hypovolemic shock, coma, and death. The first clinical signs of bleeding can be delayed and patients can remain anticoagulated for several days (warfarin) or days, weeks, or months (long-acting anticoagulants) after ingesting large amounts.

In 10 762 children aged 6 years and under who unintentionally took single doses of brodifacoum, there were no deaths or major adverse effects, although 67 reported evidence of coagulopathy [ ]. There were minor and moderate adverse effects in 38 and 54 children, respectively. About half of the children received some form of gastrointestinal decontamination, which had no effect on the distribution of outcomes but caused adverse effects in 42 patients.

Ingestion of a small amount of a superwarfarin does not require specific therapy [ , ]. However, if the coagulopathy due to a superwarfarin is severe, large doses of vitamin K may be required [ ]. Recombinant activated factor VII has been successfully used to treat superwarfarin poisoning [ ].

There are now sufficient data in young children exposed to anticoagulant rodenticides to conclude that routine measurement of the international normalized ratio (INR) is unnecessary. In all other cases, the INR should be measured 36–48 hours after exposure. If the INR is normal at this time, even in the case of long-acting formulations, no further action is required. If active bleeding occurs, prothrombin complex concentrate (which contains factors II, VII, IX, and X) 50 units/kg, or recombinant activated factor VII 1.2–4.8 mg, or fresh frozen plasma 15 ml/kg (if no concentrate is available) and phytomenadione 10 mg intravenously (100 micrograms/kg for a child) should be given. If there is no active bleeding and the INR is under 4.0, no treatment is required; if the INR is 4.0 or higher phytomenadione 10 mg should be given intravenously. When cases of unexplained acquired coagulopathy and selective deficiency of vitamin K-dependent clotting factors occur in patients in the absence of liver disease or inhibitors, physicians should consider the possibility of superwarfarin poisoning as a cause.

General adverse effects and adverse reactions

Hemorrhage and, exceptionally, hemorrhagic skin necrosis are the major adverse reactions to the coumarins. Administration during pregnancy can cause an embryopathy. Allergic reactions are extremely rare. Tumor-inducing effects have not been reported.

In the common database of the German spontaneous reporting system, 1164 reports of adverse drug reactions were registered that had been attributed to therapy with vitamin K antagonists during the period from 1990 to 2002 (phenprocoumon: 91%; warfarin: 8.3%; acenocoumarol: 0.9%) [ ]. Among these reactions a reduction in prothrombin time was the most common (15%), followed by gastrointestinal hemorrhage (13%), cerebral hemorrhage (9.1%), melena (7.4%), and increased hepatic enzymes (7.3%). Unspecified hemorrhage, intracranial hemorrhage, and hematomas accounted for 6.0% each, hepatitis 5.7%, and hematuria 4.9%. There were 42 reports (3.0%) of skin necrosis and seven of hepatic necrosis. Altogether, there were 609 cases of drug-induced hemorrhage. On average 47 cases of hemorrhage were attributed to phenprocoumon or warfarin each year from 1990, with spikes in the numbers of cases in 1997 (n = 107) and 2002 (n = 110). During the entire period the amount of prescribing increased continuously. Total sales reached 132 million defined daily doses (DDDs) in 1997 and 190 million DDDs in 2001. It was therefore not surprising that the number of adverse reactions reports increased.

Cardiovascular

Vasodilatory effects on the coronary arteries, peripheral veins, and capillaries, with purple toes as one of the most obvious consequences [ , ], have been reported. Sensations of cold may be due to increased loss of body heat caused by peripheral vasodilatation [ ].

Cholesterol embolization, which promptly improves after the drug is withdrawn [ ], may explain the purple-toe phenomenon.

Matrix gamma-carboxyglutamic acid protein, a vitamin K-dependent protein, is a potent in vivo inhibitor of arterial calcification [ ] and Growth Arrest Specific Gene 6 (Gas-6), which is also vitamin K-dependent, protects the vasculature by effects on vascular smooth muscle cell apoptosis and movement [ ]. Unlike the coagulant factors, which are carboxylated in the liver, these two proteins are carboxylated in blood vessels, but both are inhibited by warfarin. Furthermore, polymorphisms of the VKORC1 gene (the CC and CT genotypes) confer nearly twice the risk of vascular events (stroke, coronary artery disease, aortic dissection) and are associated with lower levels of osteocalcin and Protein Induced in Vitamin K Absence or Antagonism II (PIVKA-II, a des-γ-carboxyprothrombin) than those with the TT genotype [ ]. In this context it has therefore been suggested that three factors may influence arterial calcification: (a) the level of expression of matrix γ-carboxyglutamic acid protein; (b) vitamin K status; and (c) mutations or polymorphisms that affect the activities of either γ-glutamyl carboxylase or vitamin K epoxide reductase complex [ ]. One might therefore expect long-term warfarin treatment to be associated with an increased risk of vascular disease.

Calciphylaxis (vascular calcification, thrombosis, and skin necrosis) has been attributed to warfarin in a patient with diabetes mellitus [ ]. However, in 116 patients with diabetes and hypertension, half of whom were taking warfarin, there was no effect on systolic blood pressure or pulse pressure [ ].

Arterial thrombosis has been attributed to the prothrombotic effect of warfarin in a patient with protein S deficiency [ ].

Nervous system

Apart from cerebral hemorrhage, coumarins have no direct adverse effects on the nervous system.

Sensory systems

The incidence of ocular bleeding has been studied in 210 patients taking warfarin and in 210 sex-matched and age-matched controls [ ]. The incidences of ocular bleeding were 11% in the patients and 3.8% in the controls; the risk was higher in older than in younger patients and was five times higher in patients with hypertension.

A large subconjunctival hemorrhage occurred in a 76-year-old woman with an INR of 10; the authors suggested that such abnormalities could herald more serious risks [ ].

Endocrine

An antithyroid effect of bishydroxycoumarin has been suspected [ ].

Metabolism

Dicoumarol has been reported to have a uricosuric effect [ ].

Hematologic

Gastrointestinal

Gastrointestinal complications due to coumarins are limited to hemorrhage, such as small bowel obstruction due to bleeding [ ].

• A 53-year-old woman developed abdominal pain and vomiting while taking warfarin after aortic and mitral valve surgery. There was jejunal narrowing consistent with a stricture, probably as a result of submucosal bleeding. Warfarin was withdrawn and she was given heparin, with complete resolution of symptoms.

Liver

Only a few cases of hepatic injury have been documented in patients taking coumarins. In some of them, rechallenge caused a relapse [ ]. The usual presentation was with a cholestatic illness, beginning about 10 days after the coumarin anticoagulant was started, sometimes associated with eosinophilia. In Switzerland, oral anticoagulants were involved in only 11 of 674 reports of drug-induced liver disease collected during 1981–95. Seven cases were due to phenprocoumon and three to acenocoumarol. The severity ranged from asymptomatic rises in liver enzymes to cholestatic hepatitis. The interval between administration and onset of symptoms varied from 2 days to several weeks or months [ ].

Subacute liver failure necessitating orthotopic liver transplantation has been reported with phenprocoumon [ ].

• A 39-year-old woman developed idiopathic thrombosis of the posterior tibial vein. Oral contraceptives and resistance to activated protein C were identified as risk factors. After initial treatment with intravenous heparin, she was given phenprocoumon and the oral contraceptive was withdrawn. After 4 months she developed subacute liver failure and phenprocoumon was withdrawn immediately. Autoimmune disease, viral hepatitis, toxic causes, and Budd–Chiari syndrome were excluded. Despite symptomatic treatment, she deteriorated further and orthotopic liver transplantation was performed. Histopathology of the explanted liver further excluded ischemic liver cell necrosis and Budd–Chiari syndrome.

It has been hypothesized that the hepatotoxicity that has been attributed to warfarin in experimental animals is due to the formation of a toxic metabolite, o-hydroxyphenylacetylacetaldehyde, which is formed when warfarin is metabolized to 3-hydroxycoumarin, rather than 7-hydroxycoumarin; the former is more likely to be formed in patients with reduced CYP2A6 activity [ ].

Urinary tract

In one case, multisystem abnormalities, including renal insufficiency, were caused by cholesterol embolization. Withdrawal of the anticoagulant resulted in dramatically improved renal function [ ].

Skin

Maculopapular rashes with cross-sensitivity between coumarin derivatives have been reported [ ]. Non-pruritic purpuric skin eruptions, histologically presenting as vasculitis and reappearing on rechallenge with warfarin or acenocoumarol, have been described [ ].

Musculoskeletal

Preliminary results that suggested coumarin-associated reduction in bone density have not been confirmed. In a study from the Osteoporotic Fractures Research Group, 6201 postmenopausal women who were either users (n = 149) or non-users of warfarin were assessed for fractures and bone mineral density. Over 2 years, the two groups had similar age-adjusted heel and hip bone mineral density measurements. During an average of 3.5 years, non-traumatic, non-vertebral fractures occurred in 10% of warfarin users and 9.3% of non-users [ ].

However, bones that fracture during oral anticoagulant therapy require more time to form adequate amounts of callus. This may be explained by an anticoagulant-induced increase in the size of fracture hematoma. However, the observation that warfarin inhibits calcification inside artificial hearts implanted in calves adds to the body of evidence indicating that coumarin depresses not only coagulation factors, but also other (Gla)-containing proteins, for example osteocalcin, a shortage of which may also delay skeletal calcification [ ]. Osteocalcin is a non-collagenous bone matrix protein containing gamma-carboxyglutamic acid, the synthesis of which is vitamin K-dependent. During oral anticoagulation with phenprocoumon, osteocalcin concentrations were lower than in control subjects, whereas the proportion of non-carboxylated osteocalcin was significantly higher than in healthy subjects [ ]. Since osteocalcin concentrations reflect bone formation, but not bone resorption activity, the reduced serum total osteocalcin concentrations during oral anticoagulation do not necessarily imply that bone loss occurs in these patients. However, reduced bone formation and impaired gamma-carboxylation of osteocalcin in patients treated with phenprocoumon can be clinically important in circumstances such as fracture healing or when there is pre-existing bone disease. The authors suggested that vitamin D regulates the synthesis of vitamin K-dependent bone protein, but no significant effect of the duration of phenprocoumon therapy on parathormone and vitamin D concentrations has been observed.

The effects of warfarin on bone mineral density have been assessed in 5533 men aged 65 years and older [ ]. Warfarin users and 5212 non-users had similar baseline bone mineral density at the hip and spine and similar annualized bone loss at the total hip as 2683 non-users during a mean follow-up of 5.1 years; the risk of non-spinal fractures was similar in the two groups (adjusted HR = 1.06; 95% CI = 0.68, 1.65).

Immunologic

Skin-test reactivity, that is induration and tissue factor generation by monocytes, is reduced by therapeutic doses of oral anticoagulants, but lymphocyte transformation activity is not. This constitutes the rationale for the use of oral anticoagulants in the treatment of immune diseases characterized by fibrin deposition, such as allograft rejection and lupus nephritis [ ].

Coumarins can rarely cause hypersensitivity reactions. Leukocytoclastic vasculitis has been reported [ , ].

• A 48-year-old man who had taken warfarin sodium for 2 months developed acute renal failure and reddish purplish macules on his hypogastric regions and legs. Kidney biopsy showed allergic interstitial nephritis and a punch skin biopsy showed a leukocytoclastic vasculitis. Both biopsies also contained large numbers of eosinophils, highly suggestive of a drug-induced reaction.

• A leukocytoclastic vasculitis has been attributed to acenocoumarol in a 62-year-old woman because of the close temporal relation between exposure to the drug and the onset of the symptoms, and spontaneous resolution of the lesions after acenocoumarol was withdrawn.

Pregnancy

Antithrombotic treatment during pregnancy carries a well-established and substantial risk for both mother and fetus [ ]. The mother has an increased chance of abortion and of perinatal bleeding complications.

Contraceptive counselling must be given to all women who need anticoagulants. Recommendations for the use of anticoagulants during pregnancy have been reassessed in various publications [ ]. In cases of previous venous thrombosis and/or pulmonary embolism, two reasonable approaches are possible. One can either use low-dose heparin throughout pregnancy followed by an oral anticoagulant postpartum for 4–6 weeks, or one can choose to initiate clinical surveillance combined with periodic venous non-invasive tests followed by an oral anticoagulant postpartum for 4–6 weeks. If venous thrombosis occurs, heparin should be used until term, withdrawn immediately before delivery, and then both heparin and an oral anticoagulant can be started postpartum. When pregnancy is planned in patients who are taking long-term oral anticoagulants, the physician should either replace the oral anticoagulant with heparin before conception is attempted, or perform frequent pregnancy tests and substitute oral anticoagulant for heparin when pregnancy is achieved.

In patients with artificial heart valves, management in pregnancy is problematic, because the efficacy of heparin has not been established. Two approaches have been recommended. One can use heparin in a therapeutic dosage throughout pregnancy or use heparin until the 13th week, followed by an oral anticoagulant until the middle of the third trimester, and then heparin until delivery.

Teratogenicity

The teratogenic effects of warfarin (“warfarin embryopathy”) include chondrodysplasia punctata, frontal bossing, a short neck, low birth weight, short limbs, polydactyly, and respiratory difficulty secondary to choanal atresia [ , ]. Neurological effects include optic atrophy and microcephaly [ , ], the Dandy-Walker malformation and agenesis of the corpus callosum [ ], and cerebral hemispheric atrophy and porencephaly [ ]. Neurodevelopmental delay has also been described [ ].

• A 27-year-old man, whose mother had taken warfarin throughout her pregnancy, had nasal and digital hypoplasia and thoracic kyphosis, delayed developmental milestones, bilateral deafness, and mild intellectual impairment. An MRI brain scan was normal but an MRI scan of the spinal cord showed cord thinning at mid C2 and compression at C6 secondary to a posterior disc bulge; there was spinal cord compression at T4–T6 with focal thoracic kyphosis and posterior disc bulging.

In a study based on observational data collected by a Teratology Information Service on 1186 pregnant women, of whom 173 had taken a coumarin anticoagulant (almost all phenprocoumon) and 1013 had not been exposed to potential teratogens, the crude rate for live births was higher in the controls (0.91 versus 0.53) and there were lower crude rates for induced abortion (0.02 versus 0.22) and spontaneous abortion (0.07 versus 0.25) [ ]. When correction was introduced for potential biases, the rates for spontaneous abortion increased to 0.16 and 0.42 respectively. When adjustment was made for delayed entry, the rates of induced abortion were 0.04 (controls) and 0.29 (warfarin) and for spontaneous abortion 0.05 and 0.36 respectively.

Vitamin K antagonists, including hydroxycoumarin derivatives and indan-1,3-dione-derived drugs, can be teratogenic and can also induce bleeding in the fetus [ , , , ]. Adverse fetal outcomes occur in about one-third of pregnancies after either oral anticoagulants or heparin [ ]. This surprisingly high figure, which is often quoted, should be tempered by the consideration that these data are largely (if not exclusively) derived from women with heart valves taking relatively large doses of warfarin in order to maintain the INR at around 4. Some have suggested that the risk of fetal embryopathy is lower in women taking lower doses for prophylaxis or treatment of deep venous thrombosis[ ].

The pattern of teratogenicity of congeners of the vitamin K antagonist group is generally known as warfarin embryopathy, also referred to as the fetal-warfarin syndrome and Conradi–Hünermann syndrome, although the latter term also covers an identical hereditary disorder. The pattern is now considered to represent a specific group of malformations occurring in some fetuses exposed to a vitamin K antagonist during the first trimester, with a critical period during the sixth to twelfth weeks of gestation. The minimal criteria for the diagnosis include either nasal hypoplasia or stippled epiphyses. In severe cases the forehead is bossed and the nose is sunken, with deep grooves between the alae nasi and the tip of the nose; other skeletal deformities can also be present. Half the affected children have upper airway obstruction, secondary to underdeveloped cartilage. Radiography shows stippling (caused by abnormal focal calcification of the epiphyseal regions), preferentially in the axial skeleton, for example the proximal femur, and in the calcanei. Children with milder defects can show catch-up growth, and stippling can disappear after the first year of life; however, in severe cases the nose remains small and sunken. It is questionable whether mental retardation is part of the syndrome, since it has only been observed in cases of exposure for at least two trimesters.

The belief that the incidence of the overt syndrome is of the order of 5% seems to have been confirmed by a review of 186 studies: among 1325 pregnant women who took anticoagulant drugs, 970 were allocated to warfarin and the incidence of warfarin embryopathy was 4.6% [ ]. However, in a prospective survey of 72 pregnant women with cardiac valve prostheses, the incidence of coumarin embryopathy was 25% in the 12 pregnancies in which heparin was not substituted for acenocoumarol until after the 7th week, and 30% in the 37 pregnancies in which acenocoumarol was given throughout pregnancy. There were no signs of coumarin embryopathy when acenocoumarol was withdrawn from the sixth to the twelfth weeks of gestation and the women were treated instead with heparin [ ].

Oral anticoagulation may cause warfarin embryopathy by inhibiting post-translational carboxylation of proteins needed in the normal ossification process. Intensity of treatment appears to be of importance, since there were no cases of warfarin embryopathy in 44 consecutive children of 42 mothers exposed in the first trimester, but who had prothrombin times prolonged by 40–60% [ , ]. Experiments in rats in which highly intensive long-term anticoagulant therapy produced excessive mineralization disorders favor this causal relation [ ].

Another possible adverse reaction is the occurrence of central nervous system anomalies in fetuses exposed to vitamin K antagonists at any time during pregnancy. The anomalies may result from fetal intracerebral hemorrhage with scarring. There are two patterns: one consists of dorsal midline dysplasia, expressed as agenesis of the corpus callosum, Dandy-Walker malformation, midline cerebral atrophy, or possible encephalocele; the other consists of ventral midline dysplasia, characterized by optic atrophy. The patients are never completely normal on follow-up, and the resulting personal and social burdens are considerable. According to a review [ ], there were central nervous system anomalies in 26 (2.7%) of 970 pregnancies in which warfarin was used. Neonates can also have hemorrhagic complications if a pregnant mother takes an oral anticoagulant near term. The neonatal liver is immature, and concentrations of vitamin K-dependent coagulation factors are low. Although maternal warfarin concentrations are in the therapeutic range, bleeding can occur in neonates. Warfarin in particular should be avoided beyond 36 weeks of gestation [ ].

Other consequences of oral anticoagulation during pregnancy are spontaneous abortion, stillbirth, and premature birth [ ].

Lactation

In contrast to heparin, coumarins are secreted into the breast milk, but it has long been known that prothrombin activity in the plasma of neonates whose mothers take coumarins is not significantly reduced [ ] and that warfarin does not have anticoagulant effects in breast-fed infants when given to nursing mothers [ , ]. These conclusions are subject to the reservation that in some of the studies the dose of anticoagulant was low [ ]. Acenocoumarol-treated breastfeeding mothers can as a rule safely breastfeed their infants [ , ]; nevertheless, it is prudent to check the infant’s prothrombin time in such cases.

There are differences between infants in their sensitivity to coumarins. Some experts therefore recommended weekly oral administration of 1 mg of vitamin K ₁ to the child if the mother is taking a coumarin and breastfeeding [ ].

Genetic

There is a strong association between CYP2C9 variant alleles and warfarin dosage requirements [ ], and the CYP2C9*2 and CYP2C9*3 variant alleles, which are associated with reduced enzyme activity, have been associated with significant reductions in mean warfarin dosage requirements [ ]. The possession of a variant allele may also be associated with an increased risk of adverse effects [ ]. The CYP2C9 genotype is also relevant to acenocoumarol [ ]. Other polymorphisms, such as those that affect CYP3A4 or CYP1A2, may also be important [ ]. The molecular basis of warfarin resistance is unclear but could be due to unusually high CYP2C9 activity (pharmacokinetic resistance) or to abnormal activity of vitamin K epoxide reductase (pharmacodynamic resistance) [ ].

Further anecdotal data have shown that reduced CYP2C9 activity is associated with the presence of the rare CYP2C9*11 allele [ ].

• An 18-year-old man, who was highly sensitive to acenocoumarol was genotyped for functionally defective alleles in the CYP2C9 and VKORC1 genes, as were members of his family. The proband had the CYP2C9*3 allele, a CYP2C9*11 allele, and the VKORC1 AA diplotype, which were all traced back through the parental lines. Thus, acenocoumarol sensitivity in this subject resulted from inheritance of multiple functionally defective alleles in both the CYP2C9 and VKORC1 genes.

In 201 patients who were genotyped for polymorphisms in 29 genes related to warfarin pharmacodynamics and pharmacokinetics, polymorphisms in or flanking the genes VKORC1, CYP2C9, CYP2C18, CYP2C19, PROC, APOE, EPHX1, CALU, GGCX, and ORM1-ORM2, and haplotypes of VKORC1, CYP2C9, CYP2C8, CYP2C19, PROC, F7, GGCX, PROZ, F9, NR1I2, and ORM1-ORM2 were significantly associated with dose [ ]. The associations with VKORC1, CYP2C9, CYP2C18, and CYP2C19 remained significant after correction for multiple testing, but the associations with CYP2C18 and CYP2C19 were explained by linkage disequilibrium with CYP2C9*2 and/or CYP2C9*3. PROC and APOE were both significantly associated with dose after correction within each gene. In a multiple regression model, VKORC1, CYP2C9, PROC, and the non-genetic predictors age, body weight, drug–drug interactions, and indication for treatment jointly accounted for 62% of the variance in warfarin dose.

The authors of this paper were perhaps unduly optimistic about the potential contribution of these findings to the clinical use of warfarin, as others have been [ ]. Some have been less impressed [ , ]. It has been commented that while there is evidence of clinical validity of both VKORC1 and CYP2C9 genes in predicting stable warfarin doses (an intermediate outcome), there is little or no evidence that VKORC1 and CYP2C9 testing will reduce the risk of severe bleeding events [ ]. The American College of Medical Genetics has concluded (February 2008) that “there is insufficient evidence, at this time, to recommend for or against routine CYP2C9 and VKORC1 testing in warfarin-naive patients” [ ].

A pharmacogenetic algorithm has been developed based on a study in 1015 patients, in whom the independent predictors of therapeutic dose were: VKORC1 polymorphism 1639/3673G>A (− 28% per allele); body surface area (+ 11% per 0.25 m ² ); CYP2C9*3 (− 33% per allele); CYP2C9*2 (− 19% per allele); age (− 7% per decade); target international normalized ratio (INR) (+ 11% per 0.5 unit increase); amiodarone use (− 22%); smoker status (+ 10%); race (− 9%); and current thrombosis (+ 7%) [ ]. The algorithm explained just over 50% of the variability in the warfarin dose in a sample of 292 individuals in the cohorts. A clinical equation explained only about 20% of the variability. In a prospective study of the use of the algorithm in patients initiating warfarin therapy, two had a major hemorrhage. The algorithm is available on the web [ ].

African–Americans have slightly higher warfarin dosage requirements. Polymorphisms in the gene encoding apolipoprotein E (APOE) may partly explain this variability by altering transport of vitamin K to the liver, as has been shown in a prospective study of 232 individuals (52% Caucasian and 48% African–American) [ ]. In multivariable analyses, the presence of the ε4 allele was associated with a statistically significantly higher dose of warfarin among African–Americans (45 versus 35 mg) but not in Caucasians (38 versus 35 mg). In addition, the warfarin maintenance dose increased in African–Americans according to the genotypes that have previously associated with differential hepatic chylomicron clearance (ε2/ε2 or ε2/ε3, 30 mg; ε3/ε3 35 mg; ε3/ε4 or ε4/ε4, 45 mg), although the ε4/ε4 genotype was rare and not clearly associated with higher doses. The association of APOE with warfarin dosing was independent of CYP2C9 and VKORC1 polymorphisms.

In a prospective study in 362 patients with INRs of 2–3 the maintenance dose of warfarin was significantly related to CYP2C9 genotype in Caucasians but not in African–Americans; among the former, variant carriers (CYP2C9*2 and CYP2C9*3) needed 31 mg/week while wild-type carriers required 38 mg/week, even after adjustment for possible confounding factors [ ]. Among the African–Americans there was no difference based on CYP2C9 genotype.

In another comparison of CYP2C9 and VKORC1 1173C/T genotypes and the risk of hemorrhage among African–Americans and European–Americans there were 44 major and 203 minor episodes of hemorrhage during 555 person-years in 446 patients (mean age 61 years, 50% men, 227 African–Americans) [ ]. The variant CYP2C9 genotype conferred an increased risk of major hemorrhage (HR = 3.0; 95% CI = 1.1, 8.0) but not minor hemorrhage (HR = 1.3; 95% CI = 0.8, 2.1). The risk of major hemorrhage was 5.3 times higher (95% CI = 0.4, 64) before stabilization of therapy, 2.2 times higher (95% CI = 0.7, 6.5) after stabilization, and 2.4 times higher (95% CI = 0.8, 7.4) during all periods when anticoagulation was not stable. The variant VKORC1 1173C/T genotype did not confer a significant increase in the risk of major hemorrhage (HR = 1.7; 95% CI = 0.7, 4.4) or minor hemorrhage (HR = 0.8; 95% CI = 0.5, 1.3).

The distribution of genotypes of CYP2C9*2, CYP2C9*3, and VKORC1 Asp36Tyr genotypes in Ethiopians has been reported: 13/150 were heterozygous for CYP2C9*2; 7/150 were heterozygous for CYP2C9*3; and 39/154 were heterozygous and 3/154 were homozygous for the Asp36Tyr polymorphism in VKORC1, which confers warfarin resistance [ ].

In contrast to African–Americans, Asian patients require a much lower maintenance dose than Caucasians. In a study of five single nucleotide polymorphisms of the vitamin K epoxide reductase complex subunit 1 gene (VKORC1) and the CYP2C9*3 variant 108 Korean patients with atrial fibrillation the genotypic frequencies of VKORC1 + 1173CT and CYP2C9*1/*3 were 18% and 10% respectively; VKORC1 + 1173CC and CYP2C9*3/*4 were found in one patient each [ ]. Patients with at least one copy of the VKORC1 + 1173C allele or the H7 (group B) haplotype required a significantly higher dose of warfarin (n = 20; 5.5 mg/day) than those who were homozygous for the + 1173 T allele or the H1 (group A) haplotype (3.8 mg/day). There were also statistically significant differences in warfarin dose between those with the CYP2C9*1/*1 variant (4.3 mg/day) and those with the genotypes CYP2C9*1/*3 and CYP2C9*3/*4 (2.7 mg/day).

Of 66 Korean patients in whom CYP2C19 polymorphisms were evaluated, 25 were homozygous for the wild type, four had heterozygous mutations at both loci, and others had mutations on either the CYP2C19*2 or *3 locus [ ]. There was a higher incidence of bleeding complications in those with a higher allele frequency of CYP2C19. However, the distribution of polymorphisms was similar to that in a Caucasian population.

In 191 patients taking warfarin, half of whom were Malays and half Chinese, two CYP2C9 genotypes were detected; 93% had CYP2C9 1/1 and 7% CYP2C9 1/3 [ ]. Warfarin doses were higher in patients with the former genotype but patients with the latter genotype had a higher rate of serious and life-threatening episodes of bleeding (15 versus 6.2 per 100 patients per 6 months).

There is also a rare Ala-10 mutation in the propeptide of factor IX, which leads to an increased risk of bleeding when starting anticoagulation with coumarin anticoagulants [ , ].

An effect on acenocoumarol dose requirements appears to be absent for the CYP2C9*2 allele and the consequences for phenprocoumon metabolism have not yet been established. In 1124 patients from the Rotterdam Study who took acenocoumarol or phenprocoumon there was a statistically significant difference in the first INR between patients with variant genotypes and those with the wild type [ ]. Almost all acenocoumarol-treated patients with a variant genotype had a significantly higher mean INR and a higher risk of an INR of 6.0 or over during the first 6 weeks of treatment and there was a clear genotype–dose relation. Individuals with one or more CYP2C9*2 or CYP2C9*3 alleles required a significantly lower dose of acenocoumarol than wild-type patients. In patients taking phenprocoumon there were no significant differences between variant genotypes and the wild type genotype. The authors concluded that phenprocoumon is a clinically useful alternative in patients carrying the CYP2C9*2 and CYP2C9*3 alleles.

Age

Age is generally considered as an important susceptibility factor for bleeding during oral anticoagulant treatment [ ]. The frequency of bleeding was higher in elderly patients in four of five cohort studies of warfarin-related bleeding, published over a 10-year period [ , ]. In one study [ ], age was the only significant independent risk factor for subdural hemorrhage, apart from the intensity of anticoagulation, whereas age was only of borderline significance for intracerebral hemorrhage. Various explanations could account for the effect of age: co-medication, co-morbidity, increased sensitivity to warfarin, and impaired vascular integrity. Elderly patients, especially those with moderate hypertension, cerebral thrombosis, or a latent gastrointestinal ulcer, should be supervised closely [ ].

The association of older age with the risk of hemorrhage in patients with atrial fibrillation, whether or not they are taking warfarin, has been assessed in 13 509 adults with non-valvular atrial fibrillation [ ]. There were 170 major hemorrhages during 15 300 person-years of warfarin therapy and 162 major hemorrhages during 15 530 person-years without warfarin therapy. Rates of hemorrhage rose with older age, with an average increase of 1.2 (95% CI = 1.0, 1.4) per older age category in patients taking warfarin and 1.5 (95% CI = 1.3, 1.8) in those not taking warfarin. Rates of intracranial hemorrhage were significantly higher in those aged 80 and older (adjusted rate ratio = 52; 95% CI = 51, 53 for those taking warfarin; adjusted rate ratio = 55; 95% CI = 52, 59 for those not taking warfarin). Thus, older age increases the risk of major hemorrhage, particularly intracranial hemorrhage, in patients with atrial fibrillation, whether or not they are taking warfarin.

Dental extraction

In 58 women and 92 men, mean age 66 years, the mean INR was 2.5 (range 0.9–4.2) and 49 (33%) had an INR over 2.5; 10 bled after dental extraction, five of whom had an INR over 2.5; they were all managed conservatively and none was admitted to hospital [ ].