Preventing warfarin-related bleeding.

Abstract: Warfarin is one of the most common anticoagulants in current use. It maintains a narrow therapeutic index resulting in a significant amount of bleeding. Changes in concomitant medication use and acute illnesses are common causes of overanticoagulation. Monitoring regimens based on patient characteristics and the intensity and duration of anticoagulant therapy, along with simple prediction rules, can reduce the risk of warfarin-related bleeding. Supratherapeutic international normalized ratios are usually best treated by holding a dose and/or reducing the weekly dose of warfarin. Patients with acute bleeding require more urgent warfarin reversal.

Key Words: anticoagulation, bleeding, Coumadin, drug interactions, over-anticoagulation, warfarin

**********

Case Presentation I

An 82-year-old white female presents to the emergency department with a 1-hour history of epistaxis. She noted the onset of bleeding after a brief episode of sneezing while in her garden. She reports no other sites or sources of bleeding and has never had nosebleeds in the past. She takes a daily aspirin for coronary artery disease prevention and warfarin for a history of intermittent atrial fibrillation, but otherwise reports no significant medical history. She was recently discharged from the hospital after a 4-day stay for community-acquired pneumonia. She has completed a 5-day course of 500 mg azithromycin daily and is on the ninth day of a 10-day course of 500 mg levofloxacin daily.

Her vital signs reveal a temperature of 37.2[degrees]C, blood pressure of 145/73 mm Hg, heart rate of 85 beats per minute, and respiratory rate of 16 breaths per minute. On examination, her left nare shows active bleeding with surrounding clot. The rest of the examination is normal. Laboratory results show a hematocrit of 41.5%, a platelet count of 376, and an international normalized ratio (INR) of 9.4.

Scope of the Problem

Warfarin has been used for many decades for its anticoagulant effect. It is the most commonly used long-term anticoagulant in the United States, with annual retail sales exceeding $400 million, and is one of the most commonly used cardiovascular agents. The drug exerts its effect by antagonizing vitamin K, which is responsible for the activation of the clotting factors II, VII, IX, and X, and the natural anticoagulants proteins C and S. These factors are biologically inactive until vitamin K-dependent carboxylation renders them functional. Antagonism of this process by warfarin leads to a relative reduction in the activity of these factors and ultimately a state of anticoagulation. When dosed properly, warfarin has been shown to decrease morbidity and mortality rates in several common ailments including deep venous thrombosis, pulmonary embolism, atrial fibrillation, and cardiac valve replacement. Unfortunately, this benefit must be weighed against warfarin's significant annual risk of minor or major bleeding (10%), intracranial hemorrhage (1%), and death (0.6%). (1)

How Should Warfarin Be Monitored?

The primary goal of anticoagulation therapy is to find the lowest possible dose that prevents propagation or recurrence of clot formation. A secondary goal is to find a monitoring schedule that balances patient safety against the discomfort, cost, and inconvenience of monitoring. In doing this, the clinician must recognize that the level of vigilance in monitoring must be adaptable and flexible enough to change in concert with changes in the patient's risk of overanticoagulation. For example, the risk of bleeding is dependent on factors such as the duration of warfarin use, dietary intake, the level of anticoagulation sought, and the use of interacting medications.

To better understand the clinically important aspects of warfarin monitoring, it is important to have a working knowledge of the pharmacokinetics and pharmacodynamics of warfarin. The drug exists most commonly in an oral form as a 50-to-50 mixture of the more potent S-enantiomer and less potent R-enantiomer. Once ingested, it is rapidly and extensively absorbed from the gastrointestinal (GI) tract to the plasma, where it is 99% albumin bound. It reaches its maximal blood level within 90 minutes and enjoys a mean plasma half-life of 40 hours and overall effect of approximately 5 days. The drug concentrates in the liver and is renally and hepatically metabolized, with its inactive metabolites cleared in the urine and stool. (1,2) Pharmacodynamically, it is important to recognize the difference between the anticoagulant and antithrombotic effects of warfarin. The drug's anticoagulant effect is determined by its ability to reduce the circulating levels of functional clotting factors and is reflected in changes in the INR. This effect is most dependent upon the level of factor VII, which has the shortest half-life of all the clotting factors (approximately 6 hours). The antithrombotic effect of warfarin is a reflection of its ability to inhibit the patient's capability to generate or expand clots, and is most dependent on the inhibition of factor II (thrombin), which has a half-life of 50 hours. Therefore, it is imperative to recognize that early increases in INR (within 2 to 3 days) are most consistent with depletion of factor VII levels, and are not indicative of reductions in factor II levels, which are most predictive of the patient's inability to form clots (Table 1). Consequently, because the maximum antithrombotic effect of warfarin will not be seen for 4 to 5 days, other anticoagulants such as unfractionated or low-molecular-weight heparin should not be discontinued until the INR has been therapeutic for at least 1 or 2 days. (1,2)

Warfarin is most commonly monitored by following the INR, which is the patient's prothrombin time (PT) divided by the standard PT and adjusted for the sensitivity of the individual laboratory's assay. The therapeutic level of INR is most commonly 2 to 3 but varies, depending on the patient and their indication for warfarin use. The risk of warfarin-related bleeding is highest at the start of anticoagulant therapy, with one study suggesting this risk to be 10 times higher at the outset of therapy as opposed to at the 1-year mark of use. (1) Therefore, when commencing warfarin, the INR should be monitored daily until it is therapeutic on two occasions. Afterward, it should be checked 2 to 3 times weekly for 1 to 2 weeks, followed by periodic checks every 4 to 6 weeks, depending on the individual patient's stability. More frequent measurements should be made if dose adjustments occur or if the patient becomes acutely ill or commences a potentially interacting medication (Table 2). (2,3) An evolving body of literature shows that the use of intensive multidisciplinary coagulation clinics, in combination with home self-monitoring, can be beneficial in maximizing a patient's time in therapeutic range. In addition, home INR monitoring has been shown to reduce the adverse event rate and has proven especially important to patients who require frequent testing or live far from a testing facility.

Case Presentation II

A review of the patient's records show that her INR on the day before discharge from the hospital was 2.9 and that she was scheduled for a follow-up appointment 10 days later. However, no interim appointment was made to check the level of her anticoagulation. The patient asks if anything could have been done to prevent her supratherapeutic INR and ensuing bleed. You tell her that it is generally recommended that she have her INR checked 2 to 3 times a week after any change from her baseline health or addition of potentially interacting medications. During this discussion she asks about other risk factors that may increase her risk of bleeding.

What Are the Risk Factors for Warfarin-related Bleeding?

A warfarin-related bleed that requires a blood transfusion or occurs intracranially or retroperitoneally is defined as a major bleed, with all other bleeding referred to as minor. The major predisposing factors for warfarin-related bleeding can be classified into four broad categories: intensity of anticoagulant therapy, patient characteristics, duration of therapy, and the concomitant use of interacting medications (Table 3).

Intensity of anticoagulant therapy

Although bleeding can occur with any level of anticoagulation, studies evaluating the use of warfarin for various indications have consistently shown that higher levels of anticoagulation are associated with higher rates of bleeding. (5,6,9) In fact, several studies and many experts believe intensity of anticoagulation to be the single biggest determinant of a patient's bleeding risk. (4,7,8) Landefeld et al (9) showed that for every increase in the INR of 1.0, the odds ratio for major bleeding was 1.8. Indeed, patients treated with more aggressive levels of anticoagulation (goal INR > 3) have at least twice as many bleeds than those treated less intensely (INR 2-3). (4) Yet, the overall bleeding risk is not necessarily linear. In fact, the bleeding risk increases exponentially as the INR surpasses 5.0 in otherwise healthy persons and 3.0 in older patients and those with a history of cerebrovascular disease. (17)

Although some of the excess bleeding at higher INR levels is attributable to the incremental increase in the INR (ie, 3.5 versus 2.5), much of the increased risk of bleeding is explained by the fact that the higher INR goal predicts those who will achieve significant levels of overanticoagulation. Panneerselvam et al (8) compared 31 INR measurements above 7.0 with a cohort of 100 consecutive patients with a therapeutic INR to determine risk factors for this extreme level of over-anticoagulation. The group reported that the most important risk factor for overanticoagulation (odds ratio, 7.3) was a target INR of 3.5 (Table 4).

Patient characteristics

Bleeding that occurs within the therapeutic range (INR 2-3) is much more likely to be associated with a pathologic lesion than a bleed that occurs when the INR is elevated beyond the therapeutic range. Furthermore, those patients who bleed from remediable lesions are more likely to have a stable PT at the time of bleeding than patients who bleed from nonremediable lesions. This latter group is much more likely to see a sharp spike in their INR before their bleeding episode. (9)

Two other important predictors of bleeding risk are increasing age and a history of GI bleeding. Age is a specifically strong predictor of intracranial hemorrhage, with the risk rising significantly after 75 years of age. Interestingly, a history of peptic ulcer disease does not seem to reliably predict future bleeds. Other comorbid states that are associated with bleeding during anticoagulation include hypertension, cerebrovascular disease, heart disease, renal insufficiency, and cancer. (2,4)

Duration of therapy

As mentioned earlier, the bleeding risk is highest during initiation of warfarin and decreases substantially after the first month of therapy. Landefeld et al retrospectively described a cohort of patients commencing outpatient therapy with warfarin. The risk of major bleeding, which was 3% during the first month of therapy, decreased to 0.8% during months 2 to 12 and 0.3% for each month after the 12th. The risk of fatal hemorrhage followed a similar trajectory with the highest frequency seen in the first few months. (10)

Which Medications Can Potentiate the Warfarin Effect?

The Panneerselvam et al (8) study found that a recent medication change was one of the most common precipitants associated with overanticoagulation (Table 4). Many mechanisms for this association exist. Most often, the bleeding results from a potentiation of the effect of warfarin. This effect is often variable between drug classes but is nearly uniformly predictable for a particular drug. In other words, although different medications can either inhibit or induce the drug's effect, the effect is usually consistent in one direction for a particular medication. In some instances, notably corticosteroids and nonsteroidal anti-inflammatory (NSAID) drugs, the risk is not necessarily from a drug interaction but rather from the drug's potential to induce peptic ulcer disease. Since numerous medications can have at least some effect on warfarin's anticoagulant effect, a reasonable approach is to cast a cautious eye at any drug before introducing it into a warfarin regimen.

In 1994, Wells et al (11) attempted to better categorize which medications interact with warfarin. Recognizing that extensive clinical trials into drug interactions are rarely conducted, they performed a systematic review of several databases, searching for articles reporting original data on drug interactions with warfarin. The authors found 120 references outlining 186 reports of food or drug interactions and independently rated each article on strength of evidence for the interaction, using a complex model incorporating information such as the objectivity/purity of the data, pharmacologic plausibility, and reproducibility of the effect. Antibiotics, cardiac medications, and drugs active on the central nervous system (eg, anti-epileptic agents) were the most common drug groups reported to potentiate the effect of warfarin (Table 5).

Antibiotics are generally thought to be the most common group of drugs to interact with warfarin. (2,8,23) Indeed, antiinfectives comprise more then 25% of the medications thought to be at least possible of potentiating the effect of warfarin (Table 5). (11) Penning-van Beest et al compared 300 patients with an INR greater than 6.0 with 302 case-matched controls, looking for the use of 87 potentially interacting drugs in the 4 weeks before study involvement. Forty-two of the medications in the preselected group of potentially interacting medications were used in the 4-week study period, with 15 medications being used by more than 10 patients. Antibiotics were the only medications that were significantly associated with overanticoagulation in this study. The most incriminating evidence was found for trimethoprim-sulfamethoxazole (adjusted OR, 24.2 [2.6-144.5]) and the penicillins (adjusted OR, 2.4 [1.00-5.5]), with the majority of the latter group associated with the use of amoxicillin/clavulonate. (12) More recent data implicating the newer generation fluoroquinolones have appeared in the form of case reports and a controlled prospective trial. Despite the fact that there have been 8 cases reporting levofloxacin potentiation of warfarin, this latter publication reported no significant change in the INR (2.61 before levofloxacin versus 2.74 after levofloxacin) after institution of levofloxacin for clinical indications. (13) However, as our group pointed out in a follow-up alternative viewpoint, 44% of the patients saw their INR increase beyond therapeutic levels, with 17% increasing by more than one point and 11% surpassing an INR of 4.0. (14) In our opinion, these numbers, combined with good data to support an interaction with other quinolones, would suggest that a clinically significant interaction exists between levofloxacin and warfarin.

Table 6 outlines the current understanding of warfarin's interaction with a selection of more commonly used antibiotics. It is important to note that in an acutely ill patient, it is nearly impossible to differentiate the effect of the antibiotic from the effect of the patient's illness that requires the antibiotic. Many of the factors associated with acute infections, such as fever, diarrhea, anorexia, and changes in hepatic and renal function, are understood to affect the body's response to warfarin. Regardless, as stated in the monitoring section, it is imperative that warfarin be monitored more vigilantly in any acutely ill patient, especially if an antibiotic is introduced into the mix.

Case Presentation III

After discussing her risk factors for bleeding, which include chronic aspirin use, advanced age, acute illness, and the addition of potentially interacting medications (levofloxacin and azithromycin), she asks if you think she should remain on warfarin. You explain the potential benefits of warfarin in reducing stroke risk in atrial fibrillation and state that ultimately the risk of stroke must be weighed against the risk of bleeding. She asks if there is any way to gauge her future risk of bleeding.

Are There Prediction Rules for Who May Bleed?

As outlined above, several risk factors for warfarin-related bleeding have been uncovered, thereby allowing for an estimation of a patient's risk for bleeding at the outset of therapy. Beyth et al retrospectively developed an outpatient bleeding risk index based on 556 patients. The prediction rule produced a score based on the presence of various risk factors. One point was given to those patients aged 65 or older, those with a history of stroke or those with a history of GI bleed. Likewise, a point was given if any of the following were present: recent myocardial infarction, anemia (defined as hematocrit < 30%), chronic renal failure (defined as creatinine > 1.5 mg/dL), or diabetes mellitus. The group then applied the prediction rule prospectively to 264 outpatients commencing warfarin therapy. In the validation cohort, this prediction rule ably predicted those at the highest risk of bleeding events at 6, 12, and 48 months (Table 7). For example, at 4 years, bleeding occurred in 3% of those at the lowest risk, 12% of those at moderate risk, and in 53% of those at the highest risk. Furthermore, this objective measurement performed better than the prescribing physicians' opinion of the patient's bleeding risk. (16)

A bleeding risk index, such as the one outlined here, can be used prospectively to aid in decisions regarding which patients should not commence warfarin therapy. However, in many patients, the decision to start warfarin therapy may be unalterable (eg, pulmonary embolism or cardiac valve replacement). In these patients, the use of such a prediction rule can still be helpful in determining a monitoring regimen or whether or not to use medications that increase bleeding risk such as NSAIDs or corticosteroids.

How Should Supratherapeutic INRs Be Managed?

Three methods can be used to lower a supratherapeutic INR: withholding warfarin, administering vitamin K, and transfusion of fresh frozen plasma (FFP). The former method, which will result in complete correction of the INR in approximately 4 to 5 days, carries the advantage of being the least invasive and is unlikely to result in warfarin resistance when the anticoagulant is re-instituted. Administering FFP affords the most rapid correction (within hours) of hypoprothrombinemia but is more expensive, exposes the patient to potential infectious agents, and achieves an effect that is not durable (less than 24 hours). The use of vitamin K will result in a significant reversal of INR within 24 hours, offering a balance between the first two options in level of invasiveness as well as onset and duration of effect but is associated with warfarin resistance. Vitamin K should nearly always be administered orally because this method is more rapid than the subcutaneous route and associated with less anaphylaxis risk than the use of intravenous dosing. (21)

The goal of treatment is to apply an intervention that safely lowers the INR to a therapeutic level or one associated with hemostasis without causing warfarin resistance or requiring the use of bridge heparin therapy. Generally, the decision of which modality to use is based on clinical judgment, since no prospective trials comparing the effect of these three approaches on clinical outcomes beyond INR level exist. The presence or absence of bleeding complications and the height of INR elevation should be used to determine the treatment course. Patients with higher INRs and/or active bleeding should be treated more aggressively, whereas those with lower INRs or the absence of bleeding can be treated more conservatively.

Acute hemorrhage

Most patients with an elevated (or even therapeutic) INR associated with acute hemorrhage should be reversed rapidly. This typically involves administration of FFP to rapidly lower the INR. The dose is usually 800 mL of FFP but may be adjusted up or down, depending on patient size and the desired amount of correction. Meanwhile, a dose of vitamin K should be given to ensure more long-term reversal after the effect of FFP has waned. The dose of vitamin K should be tailored to the level of INR, as outlined below. If the bleeding episode is thought to be life-threatening, then an intravenous dose of 10 mg of vitamin K may be given slowly while watching closely for signs of anaphylaxis. (3) Again, the key point is to find the dose of vitamin K that will lower the INR into an acceptable range (usually < 4.0) without overcorrecting the INR or inducing warfarin resistance once the drug is restarted.

INR 3.0 to 4.0

Assuming a goal INR of 2.0 to 3.0, these levels of modest overanticoagulation are associated with minimal increased risk of bleeding over therapeutic levels in most patients and can generally be managed conservatively, assuming the absence of acute hemorrhage. If the INR is between 3.0 and 3.5, a reduction in the weekly warfarin dose by 5 to 15% is warranted. If the INR is greater than 3.5 but less than 4.0, then a dose reduction of 10 to 15%, along with holding a dose, should be considered (Table 8). (2)

INR 4.0 to 10.0

Since the risk of hemorrhage increases dramatically as the INR increases beyond 4.0, most experts agree that actions should be taken to reduce the INR to less than 4.0 as quickly as possible. Still, the absolute daily risk of hemorrhage is low, even when a patient's INR is severely prolonged and the indiscriminate use of vitamin K is associated with resistance when the warfarin is restarted. Consequently, one approach is to treat patients with INRs between 4 and 10 by withholding warfarin and monitoring daily INRs. A recent study evaluated 633 patients with INRs greater than 6.0 and found that with-holding two doses of coumadin lead to an INR less than 4.0 at 2 days in 63% of patients. The implication of this study is that most patients in this range will not require more aggressive therapy. However, the investigators did note that patients with lower baseline maintenance doses, advanced age, decompensated heart failure, and active cancer were at risk for slower return to therapeutic levels. (19) In these cases, more aggressive measures may be warranted, such as the addition of low dose vitamin K. Crowther et al investigated the use of low-dose vitamin K (1 mg orally) in combination with with-holding one dose of warfarin in patients with INR values between 4.0 and 10.0. This method resulted in an INR reduction in 95% of patients at 24 hours, with 85% achieving an INR less than 4.0. (20) No data exist as to when the warfarin should be restarted, but re-dosing the warfarin as the INR approaches the therapeutic level will minimize the risk of overcorrecting the INR and the consequent need for heparin bridge therapy.

INR greater than 10.0

Although withholding warfarin is essential to the treatment of patients with an INR greater than 10, most authorities recommend vitamin K reversal therapy for these patients. The dose is generally higher than that used for INRs between 4.0 and 10.0. Weibert et al retrospectively analyzed 81 patients with an INR greater than 5.0 and found that a 5 mg oral dose of vitamin K more reliably reversed the INR in the cohort with an INR greater than 9.0 than a dose of 2.5 mg. Therefore, a reasonable approach is to withhold at least 2 doses of warfarin while administering a single dose of 5 mg of vitamin K orally. For extreme INR elevations greater than 20, the American College of Chest Physicians recommends a slow intravenous infusion of 10 mg of vitamin K with simultaneous dosing of FFP. The vitamin K dosing may be repeated every 12 hours as needed. (3)

Case Presentation IV

After applying gentle pressure to her nose, the bleeding stops. She is dosed with 1 mg of vitamin K orally and monitored for 4 hours in the emergency department without further bleeding. Upon discharge from the emergency department, she is told to hold her nightly warfarin dose and return to her primary care physician's clinic in the morning for another INR level. On recheck the next morning, she reports no bleeding and her INR is 4.39. She is instructed to hold another dose of her warfarin and return to clinic for another recheck in the morning. The following day her INR is 3.4, the warfarin is restarted at her prior dose (she is now off antibiotics), and she is told to follow-up in the morning for another INR level. At this visit her INR is 3.1 and she again reports no signs of recurrent bleeding. She is asked to follow-up in 2 days, and her INR is 2.9. At this point she resumes her routine follow-up schedule and continues on warfarin without incident.

Summary

Warfarin-related bleeding is the principal complication of long-term anticoagulation and can be life threatening. This risk increases with the prescribed intensity of anticoagulation as well as with several patient characteristics such as history of GI bleeding, age, and other comorbidities such as hypertension, cerebrovascular disease, heart disease, renal failure, and active cancer. Another ominous risk factor for bleeding during warfarin therapy is the introduction, removal, or alteration of medications. This peril exists for most medications but is most predictably witnessed with anti-epileptic, antibiotic, and antiarrhythmic medications. In addition, there is intrinsic risk to combining warfarin with medications that increase bleeding risk such as NSAIDs and corticosteroids. These risks can be reduced by patient awareness and appropriate monitoring that takes into account the myriad health, dietary, and social changes the patient may have during their course of warfarin therapy. Moreover, validated prediction models can be used to better assess an individual patient's risk for subsequent bleeding. Supratherapeutic INRs are common and often treated by withholding the anticoagulant for 1 or 2 days or reducing the overall weekly warfarin dose. Acute bleeding or severely elevated INRs should be treated more aggressively with vitamin K alone or in combination with fresh frozen plasma.

It is dangerous to be sincere unless you are also stupid.
--George Bernard Shaw

Table 1. Warfarin pharmacodynamics and importance to treatment (a)

Pharmacodynamic effect  Dependent on                 Timing

Anticoagulant effect    Factor VII levels            24-72 hr
Antithrombotic effect   Factor II levels (thrombin)  96-120 hr

Pharmacodynamic effect  Clinical importance

Anticoagulant effect    Clinically insignificant elevation of INR, still
                        able to form thrombin clot
Antithrombotic effect   Clinically significant elevation of INR, less
                        likely to form clot

(a) INR, international normalized ratio.

Table 2. Monitoring frequency based on treatment phase

Patient treatment phase                Monitoring frequency

Warfarin commencement                  Daily until two consecutive
                                         therapeutic measurements
Initial therapeutic range              2-3 times/wk for 1-2 wk
Stable patient                         Every 4-6 wk
Potentially unstable/dosing change     2-3 times/wk until therapeutic
                                         on two measurements

Table 3. Factors associated with bleeding complications during
anticoagulation (a)

Goal INR > 3.0
Age > 65 yr
Use of interacting medications
Comorbid disease states
  History of gastrointestinal bleed
  Hypertension
  Cerebrovascular disease
  Heart disease
  Chronic renal insufficiency
  Diabetes mellitus
  Anemia
  Cancer
Acute illness
  Diarrhea
  Fever
  Stress
  Anorexia
Duration of therapy (rate of bleeding per month)
  1st mo                                                   3%
  2-12 mo                                                  0.8%
  >12 mo                                                   0.3%

(a) INR, international normalized ratio.

Table 4. Odds ratios for severe overanticoagulation (INR > 7.0) (a)

Target INR 3.5                                             7.26
Antibiotic use                                             6.22
Intercurrent illness                                       4.48
Recent drug change                                         4.02

(a) INR, international normalized ratio.

Table 5. Medication and food interactions that potentiate warfarin by
level of supporting evidence

Level of evidence                                  Medication/food

Highly probable          Alcohol (a)               Metronidazole
                         Amiodarone                Miconazole
                         Cimetidine                Omeprazole
                         Clofibrate                Phenylbutazone
                         Cotrimoxazole             Piroxicam
                         Erythromycin              Propafenone
                         Fluconazole               Propranolol
                         Isoniazid                 Sulfinpyrazone
Probable                 Acetaminophen             Itraconazole
                         Anabolic steroids         Quinidine
                         Aspirin                   Phenytoin
                         Chloral hydrate           Simvastatin
                         Ciprofloxacin             Tamoxifen
                         Dextropropoxyphene        Tetracycline
                         Disulfiram                Influenza vaccine
Possible                 Disopyramide              Norfloxacin
                         5-Fluorouracil            Ofloxacin
                         Ifosphamide               Topical salicylates
                         Lovastatin                Sulindac
                         Metolazone                Tometin
                         Nalidixic acid

(a) When associated with underlying liver disease.
Adapted from Wells et al. (11)

Table 6. Selected warfarin-antibiotic interactions and level of
supporting evidence (a)

Antibiotic        Strength of evidence  Type of evidence/comment

Sulfonamides              ++++          TMP-SMX probably has largest
                                        effect of all antibiotics
Fluoroquinolones          +++           More than 65 case reports of
                                        potentiation with ciprofloxacin
                                        Increasing evidence of
                                        levofloxacin interaction
Penicillins/                            Minimal evidence for penicillin
cephalosporins            +++           or cephalosporins
                                        Stronger evidence for
                                        potentiation by
                                        aminopenicillins (b)
                                        Strongest evidence for reduction
                                        in INR by synthetic
                                        penicillins (c)
Macrolides                ++            Strongest for erythromycin,
                                        which changes warfarin
                                        metabolism
                                        Retrospective analysis of
                                        azithromycin did not show
                                        clinically significant
                                        interaction

(a) TMP-SMX, trimethoprim-sulfamethoxazole; INR, international
normalized ratio.
(b) Amoxicillin and ampicillin.
(c) Dicloxacillin and nafcillin especially.

Table 7. Outpatient bleeding risk index

Sum of risk factors      0                 1-2              3-4
                         [down arrow]     [down arrow]     [down arrow]
Risk stratification      Low              Intermediate     High
                         [down arrow]     [down arrow]     [down arrow]
Estimated risk for
  major bleeding (a)     [down arrow]     [down arrow]     [down arrow]
                         [down arrow]     [down arrow]     [down arrow]
  @ 6 mo                 3%                8%              16%
  @ 12 mo                3%                8%              30%
  @ 48 mo                3%               12%              53%

Risk factors were given 1 point for each: age [greater than or equal to]
65 yr, history of cerebrovascular disease, history of gastrointestinal
bleed; 1 point if any present: recent myocardial infarction, anemia
(hematocrit < 30%), chronic renal failure (creatinine > 1.5 mg/dL),
diabetes mellitus.
(a) Intracranial bleed or need for transfusion.

Table 8. Altering weekly warfarin dosage to achieve INR of 2 to 3 (a b)

                     Withhold                  Change in weekly
                       dose                     warfarin dose

INR < 2          No                       Increase weekly dose 5-10%
INR 3.0-3.5      No                       Decrease weekly dose 5-15%
INR 3.5-4.0      None to one dose         Decrease weekly dose 10-15%
INR > 4.0        None to one dose         Decrease weekly dose 10-20%

(a) INR, international normalized ratio.
(b) Will require more urgent treatment if acutely bleeding.
Adapted from Horton and Bushwick. (2)

Accepted September 2, 2004.

References

1. Landefeld CS, Beyth RJ. Anticoagulation-related bleeding: clinical epidemiology, prediction, and prevention. Am J Med 1993;95:315-327. This extremely detailed review outlines much of this group's earlier work in predicting anticoagulant-related bleeding but also covers the full body of literature published before 1993. Includes an essential section reviewing the level and strength of evidence for various patient-specific risk factors.

2. Horton JD, Bushwick BM. Warfarin therapy: evolving strategies in anticoagulation. Am Fam Physician 1999;59:635-646. A well-written general review of warfarin with an especially strong and thorough section covering medication interactions based on the severity, onset, and strength of evidence for more than 30 medications.

3. Ansell J, Hirsh J, Dalen J, et al. Managing oral anticoagulant therapy. Chest 2001;119:22S-38S. This brilliant and timely piece covers a spectrum of issues, from dosing to peri-operative anticoagulant management to management of acute hemorrhage. Also reviews the literature for models of anticoagulant management including the utility of home anticoagulant monitoring. Includes a wonderful section on management of nontherapeutic INRs.

4. Levine MN, Raskob G, Landefeld S, et al. Hemorrhagic complications of anticoagulant treatment. Chest 2001;119:108S-121S. One of the best overviews of the risks of anticoagulant-related bleeding with great sections on bleeding rates by anticoagulant indication.

5. Hull R, Hirsh J, Jay R, et al. Different intensities of oral anticoagulant therapy in the treatment of proximal-vein thrombosis. N Engl J Med 1982;307:1676-1681. A seminal article aimed at documenting the most efficacious level of anticoagulant therapy for this common disorder. Secondarily, this group reported the risk of bleeding increases along with the level of anticoagulation.

6. Turpie AG, Gunstensen J, Hirsh J, et al. Randomized comparison of two intensities of oral anticoagulant therapy after tissue heart valve replacement. Lancet 1988;1:1242-1245. This group randomized 108 patients status post tissue heart valve replacement to receive standard therapy with warfarin to a goal INR of 2.5 to 4.0 or 102 patients to a less intense regimen. The investigators reported no significant change in

embolism at 3 months but did discover a significant increase in bleeding in the standard INR group.

7. Van der Meer FJ, Rosendaal FR, Vandebroucke JP, et al. Assessment of a bleeding risk index in two cohorts of patients treated with oral anticoagulants Thromb Haemost 1996;76:12-16. This study, conducted on the Leiden Thrombosis Service in 1988 and 1991, identified age and level of achieved INR as the two most important predictors of bleeding.

8. Panneerselvam S, Baglin C, Lefort W, Baglin T. Analysis of risk factors of over-anticoagulation in patients receiving long-term warfarin. Br J Haematol 1998;103:422-424. This group compared 31 patients with an INR greater than 7.0 with 100 patients with therapeutic INRs and identified several key risks for anticoagulant-related bleeding: height of target INR, recent drug addition (especially antibiotics), and intercurrent illness.

9. Landefeld CS, Rosenblatt MW, Goldman L. Bleeding in outpatients treated with warfarin: relation to the prothrombin time and important remediable lesions. Am J Med 1989;87:153-159. This useful study evaluated the association of bleeding to prothrombin time and presence of a remediable lesion. The authors report that every increase in prothrombin time-to-control ratio of 1.0 was associated with an 80% increase in risk of bleeding. They also noted that the prothrombin time was stable in those patients who bled from pathologic lesions but that it tended to rise sharply before bleeding in patients without remediable lesions.

10. Landefeld CS, Goldman L. Major bleeding in outpatients treated with warfarin: incidence and prediction by factors known at the start of outpatient therapy. Am J Med 1989;87:144-152. The first of many papers looking at the risk factors for bleeding in patients commencing warfarin therapy as an outpatient. The authors reviewed the charts of 565 patients and established a history of GI bleed, stroke, atrial fibrillation, age 65 years or older, and serious comorbid diseases as major risk factors for future bleeding.

11. Wells PS, Holbrook AM, Crowther NR, Hirsh J. Interactions of warfarin with drugs and food. Ann Intern Med 1994;121:676-683. This offering represents the best article outlining the various medication interactions that interact with warfarin. The authors systematically reviewed 120 reports of 186 medication interactions and then listed the medications based on the strength of evidence for the interaction. This article implicated antibiotics, anti-epileptics, and cardiac medications most often.

12. Penning-van Beest FJ, Van Meegen E, Rosendaal FR, et al. Drug interactions as a cause of overanticoagulation on phenprocoumon or acenocoumarol predominantly concern antibacterial drugs. Clin Pharmacol Ther 2001;69:451-457. These authors compared 302 controls with 300 patients with an INR greater than or equal to 6.0, looking for drugs that appeared to contribute to overanticoagulation. They found that antibiotics were the most commonly implicated drug class, with trimethoprim-sulfamethoxazole and amoxicillin plus clavulonate leading the way.

13. Yamreudeewong W, Lower DL, Kipatrick DM, et al. Effect of levofloxacin coadministration on the international normalized ratios during warfarin therapy. Pharmacotherapy 2003;23:333-338. These authors followed 18 warfarin-using patients after the addition of the levofloxacin for clinical indications. They report that no change was seen from the baseline INR of 2.61 to the follow-up INR of 2.74. This is one of the few prospective analyses of a warfarin-drug interaction.

14. Glasheen JJ, Fugit RV, Prochazka AV. Effect of levofloxacin coadministration on the international normalized ratios during warfarin therapy: a comment. Pharmacotherapy 2003;23:1079-1080. This alternative viewpoint, based on a reanalysis of Yamreudeewong's data, provided evidence that perhaps an interaction does exist between warfarin and this very commonly used antibiotic. This comment pointed out that 44% of patients in the study had an increase in INR to supratherapeutic levels, 17% had an increase of greater than one point, and 11% saw their INR rise to a level of 4.0 or beyond.

15. Beyth RJ, Quinn LM, Landefeld CS. Prospective evaluation of an index for predicting the risk of major bleeding in outpatients treated with warfarin. Am J Med 1998;105:91-99. This article, published by some of the leaders in the warfarin research field, validated their prior work. This outpatient bleeding risk index consists of four major risk factors and is easy to use in the clinical setting.

17. Bussey HI. Managing excessive warfarin anticoagulation. Ann Intern Med 2001;135:460-462. This wonderful editorial on Hylek's paper (Reference 19) adroitly reviews the literature and offers a roadmap for how to manage overanticoagulation in the warfarin-using patient.

19. Hylek EM, Regan S, Go AS, et al. Clinical predictors of prolonged delay in return of the international normalized ratio to within the therapeutic range after excessive anticoagulation with warfarin. Ann Intern Med 2001;135:393-400. This group studied 633 patients who had an INR greater than 6.0 while using warfarin. Thirty-seven percent had an INR greater than or equal to 4.0 after withholding two doses of warfarin (without receiving vitamin K reversal). Those patients who were older, had decompensated congestive heart failure or active cancer, or required higher doses of warfarin were more likely to have an INR above 4.0 after 2 days of withholding warfarin.

20. Crowther MA, Donovan D, Harrison L, et al. Low-dose oral vitamin K reliably reverses over-anticoagulation due to warfarin. Thromb Haemost 1998;79:1116-1118. This prospective study evaluated 62 patients with INR levels between 4.0 and 10.0. After withholding one dose of warfarin and administering 1 mg of vitamin K, the INR fell to less than 4.0 in 85% of patients within 24 hours. Only 35% fell to an INR below 1.9.

21. Crowther MA, Douketis JD, Schnurr T, et al. Oral vitamin K lowers the international normalized ratio more rapidly than subcutaneous vitamin K in the treatment of warfarin-associated coagulopathy. Ann Intern Med 2002;137:251-254. After studying 51 patients with an INR between 4.5 and 10.0 randomly assigned to either subcutaneous or oral vitamin K, these authors report that the latter is more efficacious at rapidly lowering supratherapeutic INRs than the former. Fifty-seven percent versus 24% had therapeutic INRs the day after the drug was administered in the oral versus subcutaneous vitamin K, respectively.

22. Gage BF, Fihn SD, White RH. Management and dosing of warfarin therapy. Am J Med 2000;109:481-488. This comprehensive review of warfarin therapy includes excellent sections on the pharmacology, initiation, and maintenance of warfarin therapy. Also contains a concise review of the literature surrounding home anticoagulant self-monitoring.

23. Glasheen JJ, Fugit RV. How Warfarin interacts with common antibiotics. Emerg Med 2004;36:30C-30I. A concise, informative review of this common drug-drug interaction.

RELATED ARTICLE: Key Points

* Patient characteristics such as age and history of bleeding as well as the intensity and duration of warfarin therapy are the strongest predictors of warfarin-related bleeding.

* Antibiotics, antiepileptics, and cardiac medications are the most common drugs to interact with warfarin.

* Simple rules can reliably predict a patient's risk of bleeding while using warfarin.

* Supratherapeutic international normalized ratios in a nonbleeding patient are usually best treated by holding the warfarin and subsequently reducing the weekly dose.

Jeffrey James Glasheen, MD

From the Hospital Medicine Section, Division of General Internal Medicine, and Department of Medicine, University of Colorado Health Sciences Center, Denver, CO.

Reprint requests to Dr. Jeffrey Glasheen, Campus Box B178, University Hospital, Room 6201, University of Colorado Health Sciences Center, Denver, CO 80262-0002. Email: [e-mail]