Warfarin

(CC) Image: David E. Volk
Structure of Warfarin.

Warfarin (IUPAC name 4-hydroxy-3-(3-oxo-1-phenylbutyl)-2H-chromen-2-one), also widely called coumadin, is an anticoagulant medication used prophylactically to suppress the formation of embolism and thrombosis from conditions such as atrial fibrillation, deep venous thrombosis, and pulmonary embolism.

Originally designed to be a rat poison, warfarin works as an anticoagulant by suppressing the enzyme epoxide reductase in the liver, thereby suppressing the formation of the reduced form of vitamin K epoxide, which is needed for the synthesis of many coagulation factors. As a drug, it is often sold as the sodium salt of warfarin.

Discovery

The dangers of feeding livestock spoiled sweet clover hay were known in the 1920’s and scientists at the University of Wisconsin-Madison were spearheading the problem on multiple fronts. R.A. Brink and W.K. Smith were attempting to breed a new variety of clover that was free of the toxic effect and Karl Paul Link's laboratory was attempting to isolate the killer compound. Farmers were already hurting due to the Great Depression and despite knowing they should not feed their livestock such hay, they could not afford to buy uncontaminated supplies. This disease that was affecting many livestock throughout the US became known as "sweet clover disease".

The urgency of this work was punctuated in the winter of 1933 when Ed Carlson, a farmer from Deer Park, Wisconsin, came to Madison for help carrying a milk can of uncoagulated blood, a dead cow in his truck and a sample of the hay contaminated with the spoiled sweet clover. This spurred Link on and in 1940 he finally published the source of the hemorrhagic factor in sweet clover, 3,3'-methylene-bis[4-hyfroxycoumarin], known as coumarin. The following year Link published the discovery of dicumarol, the oxidised form of coumarin that was present in the spoiled sweet clover; this became widely used as an oral anticoagulants for medical treatment.

Given the death of cattle from hemorrhaging, Link began to realise that anticoagulants had potential to be rodenticides.

I had an intuitive feeling that this might be a good thing. A pretty bad thing for rats, but a good thing for humans. But the idea didn't come overnight. It came into my head and the heads of everyone in the lab over a period of years.

Dicumarol turned out to be a poor rodenticide as it acted too slowly but by 1948 Link had patented another coumarin derivative as a rodenticide. Research for both dicumarol and the new rodenticide was funded by the Wisconsin Alumni Research Foundation (WARF) and the brand name for these anticoagulating rat posions was coined 'Warfarin' after WARF. Later, the rodenticide was promoted for clinical applications under the brand name 'Coumadin'.

Mechanism of action

Warfarin therapy reduces the vitamin K dependent cofactors II, VII, IX, and X and the vitamin K dependent Protein C. The level of factor II is thought to most influence coagulation. The levels of factor VII and Protein C fall the fastest after warfarin is started. With the exception of Factor IX, these factors are from either the extrinsic pathway or the final common pathway.

Pharmacokinetics

Metabolism

Pharmacogenomics

About 15% of the effect of warfarin can be explained by the amount of warfarin in the blood. About 50% of the effect of warfarin can be explained by genetic factors. The American Food and Drug Administration "highlights the opportunity for healthcare providers to use genetic tests to improve their initial estimate of what is a reasonable warfarin dose for individual patients". A systematic review concluded that as of early 2009, there is insufficient benefit in genetic testing.

VKORC1

Genetic polymorphisms in the vitamin K epoxide reductase complex 1 (VKORC1) gene explain 30% of the dose variation between patients: particular mutations make VKORC1 less susceptible to suppression by warfarin There are a main haplotypes that explain 25% of variation: low-dose haplotype group (A) and a high-dose haplotype group (B). For the three combinations of the haplotypes, the mean daily maintenance dose of warfarin was:

• A/A: 2.7+/-0.2 mg
• A/B: 4.9+/-0.2 mg
• B/B: 6.2+/-0.3 mg

VKORC1 polymorphisms also explain why African Americans are relatively resistant to warfarin (higher proportion of group B haplotypes), while Asian Americans are more sensitive (higher proportion of group A haplotypes).

CYP2C9

CYP2C9 is an isoenzyme of cytochrome P-450. Polymorphisms of CYP2C9 explain another 10% of variation in warfarin dosing, mainly among Caucasian patients as these variants are rare in African American and most Asian populations. A meta-analysis of mainly Caucasian patients found:

• CYP2C9*2 allele:
  • present in 12.2% of patients
  • mean reduction was in warfarin dose was 0.85 mg (17% reduction)
  • relative bleeding risk was 1.91
• CYP2C9*3 allele:
  • present in 7.9% of patients
  • mean reduction was in warfarin dose was 1.92 mg (37% reduction)
  • relative bleeding risk was 1.77

Availability

Generic preparations have similar pharmacology.

Dosage

Clinical practice guidelines are available for using warfarin.

"After 4 to 5 days of concomitant warfarin and heparin therapy, heparin is discontinued when the INR has been in the therapeutic range on two measurements approximately 24 h apart."

Loading regimens

A decision analysis "suggests that patients with intermediate and high/likely probabilities of PE benefit from preemptive anticoagulation".

Because of warfarin's difficult pharmacokinetics, researchers have proposed algorithms for warfarin loading.

Randomized controlled trials of warfarin loading algorithms

	Dosing	Study	Time till a therapeutic INR (days)	Rate of anticoagulation (INR=2-3)	Rate of over-coagulation
Kovacs 10 mg	10 mg/day for two days then specified adjusted dosing	Kovacs†	4.2	83% by day 5	9% within 4 weeks(INR>5) (most occurred after day 10)
		Quiroz‡ Mean age: 51 United States	5 (two consecutive INRs≥2)	56% at 5 days	0% within 5 days (INR>5)
		Anderson‡		69% at 5 days
		Farahmand‡ Mean age: 48 Iran		27% at 5 days	2% at 5 days (INR>5)
Harrison 10 mg	10 mg/day for one day then flexible adjusted dosing	Harrison¶		63% at 3.5 days	17% within 3.5 days (INR>4.8)
		Crowther¶		69% at 5 days	0% within 5 days (INR>5) 24% within 5 days (INR>3)
Kovacs 5 mg	5 mg/day for two days then specified adjusted dosing	Kovacs†	5.6	46% by day 5	11% within 4 weeks (INR>5) (most occurred after day 10)
		Farahmand‡ Mean age: 48 Iran		47% at 5 days	2% at 5 days (INR>5)
Harrison 5 mg	5 mg/day for one day then flexible adjusted dosing	Harrison¶		80% at 3.5 days	4% within 4.5 days (INR>4.8)
		Crowther¶		88% at 5 days	3% within 5 days (INR>5) 7% within 5 days (INR>3)
		Quiroz‡ Mean age: 51 United States	5 (two consecutive INRs≥2)	52% at 5 days	0% within 5 days (INR>5)
Anderson pharmacogenetic-guided algorithm	Based on CYP2C9 and VKORC1	Anderson†		70% at 5 days
Caraco pharmacogenetic-guided algorithm	Based on CYP2C9	Caraco	4.8
† Blinded study. ‡ Independent study not from original investigators. ¶ Same research group (Hamilton, Ontario). Notes: 1. INR. International Normalized Ratio 2. The Kovacs 5 mg algorithm is the same as the Harrison 5 mg algorithm except that where Harrison gave a range warfarin of dosages based on the INR, Kovacs specified a dose (usually at the high end of the range offered by Harrison).

Empiric dosing

On first look, the evidence table suggests that the Harrison 5 mg algorithm from Hamilton is the chest combination of efficacy and safety; however, two independent studies (Kovacs and Quiroz in 2006 from the Massachusetts General Hospital) have not been able to replicate the results of Hamilton group of Harrison and Crowther. One explanation may be that the personnel in the Harrison study were more expert and the flexibility in the algorithm allowed expression of their expertise. If so, then perhaps equally expert health care providers should use the Harrison 5 mg algorithm while other personnel should use the Kovacs 10 mg algorithm. Considering that the expertise of the Massachusetts General Hospital could not replicate the Hamilton results, perhaps most providers should use the Kovacs 10 mg algorithm, at least for inpatients.

A systematic review of the randomized controlled trials done through 2003 of 5 mg versus the 10 mg concluded that the Kovacs 10 mg regimen is best. This conclusion was largely based on the inability of the results of the Harrison 5 mg flexible algorithm to be replicated by Kovacs.

Clinical practice guidelines in 2004 by the American College of Chest Physicians concluded that either 5 or 10 loads are acceptable. The guidelines also state "if treatment is not urgent (eg, chronic stable atrial fibrillation), warfarin administration, without concurrent heparin administration, can be commenced out-of-hospital with an anticipated maintenance dose of 4 to 5 mg per day."

Since publication of the systematic review and clinical practice guidelines, a nonblinded, randomized controlled trial found no difference between algorithms, but also achieved less frequent anticoagulation. The reason is not clear as the Kovacs 10 mg algorithm is very specific on each dose of warfarin. The last trial is also unique in that all patients were receiving a fondaparinux bridge.

Additional algorithms:

• The Tait 5 mg regimen is for outpatient anticoagulation. Patients are given 5 mg of warfarin per day for 5 days and then the INR is checked on day 5 to determine further dosing. (summary)
• The Fennerty 10 mg regimen is an older regimen that has not been studied in an randomized controlled trial.

Pharmacogenetic guided dosing

Randomized controlled trials:

• In the COAG trial, pharmacogenetic-guided algorithm using 3CYP2C9 and VKORC1 was compared to an algorythm that projected doses of warfarin based on clinical parameters entered into a clinical prediction rule. The study lasted 4 weeks. The proportion of time that the International Normalized Ratio was in therapeutic range was not statistically significantly different in the two groups (45.2% vs 45.4%, respectively).
• In the EU-PACT trial, a pharmacogenetic-guided algorithm using 3CYP2C9 and VKORC1 was compared to usual care. The study lasted 3 months. The proportion of time that the International Normalized Ratio was in therapeutic range was statistically higher in the pharmacogenetic-guided group (67.4% vs 60.3%), but not after week 8.
• In another trial, pharmacogenetic-guided algorithm using only CYP2C9 was compared to a local algorithm. The length of the trial was not explicitly stated but may have been until stabilization of the INR. The proportion of time that the International Normalized Ratio was in therapeutic range was statistically higher in the pharmacogenetic-guided group (80.4% vs 63.4%, respectively). The pharmacogenetic-guided algorithm led to less minor bleeding (3.2 vs 12.5%, P<0.02).
• The Kovacs 10 mg algorithm performed similarly to a pharmacogenetic-guided algorithm using 3CYP2C9 and VKORC1. The proportion of patients who did not have an out-of-range International Normalized Ratio statistically insignificantly 1.04 times higher in the pharmacogenetic group (69% vs 67%).

Observational studies:

• A pharmacogenetic-guided algorithm using 3CYP2C9 and VKORC1 outperformed an unvalidated, clinical algorithm in modeled results. In this cohort study of 5052 patients, the pharmacogenetic estimated a dose within 20% of the actual dose 1.2 times more often than the clinical algorithm (46% vs 38%).
• A pharmacogenetic-based model from a cohort of orthopedic patients using 3CYP2C9 and VKORC1 genotype results predicted 80% of the variation in warfarin doses. It is awaiting validation in larger populations and has not been reproduced in those who require warfarin for other indications.

Cost-benefit analysis and decision analysis concluded that patients at average risk do not benefit from testing for CYP2C9.

Adjusting the maintenance dose

See also atrial fibrillation

The goal is INR is 2 to 3, so then adjustments should be made when the INR is 1.7 or less or when 3.3 or greater.A meta-analysis observed that the time in therapeutic range varies by setting:

• Anticoagulation clinics: 66%
• Other clinics: 57%

The protocol for the RELY trial, which yielded 64% TTR is online. Increased use of the RELY protocol is associated with increased time in thereapeutic range.

The protocol for the CoumaGen-II, yielded 58% TTR as is online.

The protocol for the ARISTOTLE trial, yielded 62% TTR.

The optimal INRis unclear.

While usually warfarin is monitored by measing the INR once a month at a healthcare provider's office, other alternatives are:

• Some patients may use home monitoring.
• Patients whose dose is stable for at least 6 months, may not require monthly monitoring.

Discontinuation before procedures

Details have been provided by the American College of Chest Physicians at The Perioperative Management of Antithrombotic Therapy.

Management of warfarin around invasive procedures has been reviewed and new bridges have been studied.

For procedures that need discontinuation of anticoagulation, a cohort study that found that interruption for 5 days or less was generally safe.

For patients needing a bridge, low molecular weight heparin is preferable to intravenous infusion of unfractionated heparin. For patients needing a low molecular weight heparin bridge, a protocol is available. One bridge before coronary artery bypass grafting is to stop 6 days before and give 5 mg/day of vitamin K on the day of warfarin cessation and use low molecular weight heparin with last dose the night before surgery.However, bridging with low molecular weight heparin increases bleeding as compared to no bridging.

Effectiveness

The effect of warfarin is measured by the prothrombin time (or the International Normalized Ratio derived from the prothrombin time) although warfarin can also affect the partial thromboplastin time.

Warfarin's effectiveness depends the proportion of time that a patient's spends in the therapeutic range of anticoagulation, time in the optimum therapeutic range - the time in therapeutic range (TTR). In carefully conducted and monitored randomized controlled trials, the TTR ranges from 55% to 65%. However, the TTR is lower in the community practice. The TTR may be calculated with linear interpolation between INR values, or may be more simply reported as the number of INR values out of range. The highest reported TTR is approximately 75%.

Interactions and contraindications

Warfarin interacts with many medications. A proposed classifications of mechanisms is:

• Interference with platelet function
• Injury to gastrointestinal mucosa
• Reduced synthesis of vitamin K by intestinal flora
• Interference with warfarin metabolism by cytochrome P-450 CYP2C9 isoenzyme.
• Interruption of the vitamin K cycle

Some foodstuffs have also been reported to interact with warfarin.

Adverse effects

The most accurate clinical prediction rule for estimating the risk of bleeding is the HAS-BLED score.

The risk of bleeding can be predicted with the ATRIA score based on:

• anemia (3 points)
• glomerular filtration rate <30 ml/min or receiving dialysis
• age 75 years or more (2 points)
• prior bleeding (1 point)
• hypertension (1 point)

The annual risk of bleeding was:

• Low risk (0 to 3 points) - 0.8%
• Intermediate risk (4 points) - 2.6%
• High risk (5 to 10 points) - 5.8%

Elderly patients

Patients aged 80 years or more may be especially susceptible to bleeding complications with a rate of 13 bleeds per 100 person-years.

Patients with prior intracranial hemorrhage from warfarin

These patients are at high risk of bad outcomes regardless of whether anticoagulation is resumed.

Patients with cancer

Rates of major bleeding and recurrent venous thromboembolism (VTE)
using warfarin with a target INR of 2.0 and 3.0
among patients with cancer.

		No cancer	Cancer
			Stage I or II	Stage III	Stage IV
Major bleeding	Events per 100 patient-years	8.6	3.4	19.1	42.8
	Hazard ratio	1	0.5	2.15	4.8
Recurrent VTE	Events per 100 patient-years	12.8	14.5	44.1	54.1
	Hazard ratio	1	1.9	5.3	4.6
Adapted from Table 3 of Prandoni et al.

Patients with cancer are more likely to have bleeding complications, especially if they have Stage III (regionally extensive) or IV (metastatic) cancer. Regardless of the extent of cancer, the risk of bleeding was less than the risk of recurrent embolism and thromboembolism:

Unstable anticoagualation

Supplementing warfarin with 150 micrograms of vitamin k may reduce the frequency of unstable anticoaguation among patients who are difficult to anticoagulate. This amount of vitamin k led to 16% increase in the warfarin dose from 3.8 mg per day to 4.4 mg per day after one week.

Antagonism and reversal

A detailed table on reversing warfarin are provided in clinical practice guidelines from the American College of Chest Physicians. For patients who are not bleeding and have an International Normalized Ratio (INR) between 4.5 and 10.0, either withholding warfarin or 1 mg of oral vitamin K is effective. Vitamin K should not be given subcutaneously, even at low doses then can lower the INR too much.

Drug interactions

Acetaminophen

Acetaminophen in doses of 2 grams per day or more for several consecutive days may interact with warfarin.

Anti-platelet agents

Patients taking aspirin, clopidogrel, or dipyridamole may be at higher risk of hemorrhage.

Antilipemic agents

Medications, such as the antilipemic agents atorvastatin, simvastatin, or gemfibrozil that are metabolized by cytochrome P-450 CYP3A4 may increase bleeding when added to patients taking warfarin.

Antibiotics

Antibiotics metabolized by cytochrome P-450 CYP2C9, such as trimethoprim-sulfamethoxazole and ciprofloxacin may increase risk of hospitalization for gastrointestinal hemorrhage.