Oral Vitamin K for Warfarin Associated Coagulopathy

What is the problem to be addressed ? Although warfarin is a highly effective anticoagulant, it causes bleeding. The risk of bleeding in an individual is increased by a number of factors, including age, gender and previous hemorrhage. However, the single strongest predictor of hemorrhage is excessive prolongation of the international normalized ratio (INR). We have shown that low-dose oral vitamin K reliably reduces an excessively prolonged INR; we hypothesize this treatment will also reduce the risk of hemorrhage without increasing the risk of thrombosis. The objective of this study is to demonstrate that low dose oral vitamin K improves the safety of oral anticoagulant therapy, by reducing the risk of bleeding without causing thrombosis, in patients presenting with INR values of more than 4.5.

Background and rationale for the study (Efficacy and toxicity of warfarin): Warfarin reduces the risk of first or recurrent venous thrombosis by 80 to 90 percent, and reduces the risk of first or recurrent arterial thrombosis by 50 to 65 percent (1;2). In North America more than 1 million patients receive warfarin on a daily basis and in Ontario more than 300,000 prescriptions for warfarin are filled annually. The major toxicity of warfarin is bleeding, which may be fatal. Although bleeding may occur when the international normalized ratio (INR) is in the "therapeutic range" (2.0 to 3.0 for most indications), many patients who bleed have supratherapeutic INR values. Indeed, large studies suggest patients spend 10 to 20 percent of their time with excessively prolonged INR values and one anticoagulation clinic has reported that out of 7,279 patients managed over a 5 year interval, 1,995 had at least one INR greater than 6.0 (3). Furthermore, the risk of hemorrhage approximately doubles for each increase in the INR of 1.0 unit (4-7).

Background and rationale for the study (Risks of an excessively prolonged INR) The risk of hemorrhage in patients with asymptomatic warfarin-associated coagulopathy has been highlighted by Hylek and colleagues (8). In this study, 114 patients who presented with INR values of more than 6.0 simply had their warfarin withheld until the INR had declined to the desired range. In the two-week follow-up period, 10 of the 114 (8.8%, 95% confidence interval (CI) 3.6 to 14.0) patients experienced bleeding including 5 major and two fatal bleeds. Thus the risk of major hemorrhage and fatal hemorrhage was 4.4% (95% CI 0.6, 8.1), and 1.7% (95% CI 0, 4.2), respectively. Similarly, Oden and colleagues (4) demonstrated the mortality rate attributable to bleeding per 100 patient years of follow-up was 22 fold higher in patients with INR values between 6.0 and 6.9, compared with patients with an INR between 2.0 and 2.4. The observation that excessive elevation of the INR is associated with bleeding suggests that identifying an effective technique to lower the INR will reduce bleeding.

Background and Rationale (Treatment of warfarin associated coagulopathy) Other than simply withholding warfarin and allowing the INR to fall to the desired range, two "active" approaches are used to lower the INR in excessively anticoagulated patients. Transfusion of coagulation factor concentrates replaces missing coagulation factors. Supplemental vitamin K antagonizes the effect of warfarin, thus increasing endogenous synthesis of these same factors.

Transfusion with human derived or recombinant clotting factors can rapidly normalize the INR. However, this treatment is expensive, inconvenient, associated with infectious or other complications such as allergy or thrombosis, and has never been prospectively studied in non-bleeding patients (9). Thus, transfusion therapy should not be used to correct an excessively prolonged INR value in non-bleeding patients.

What is this study designed to prove ?

We anticipate that this study will prove that:

1. Oral vitamin K reduces the risk of bleeding in patients taking warfarin who present with INR values between 4.5 and 10.0.
2. That oral vitamin K does not increase the risk of thrombosis when given to such patients.
3. That low dose oral vitamin K will reduce the INR to an acceptable range in a high proportion of patients presenting with an INR of more than 10.0

Why is this trial needed now ? This trial is needed now because there is new evidence confirming the association between prolonged INR values and the risk of bleeding (4;8), because new evidence suggests that oral vitamin K effectively lowers the INR (12-15;17) and thus may reduce bleeding (13), and because the risk:benefit ratio of oral vitamin K remains unclear.

Studies which have been previously published by our group suggests that oral vitamin K effectively reduces the INR value. However, despite their rigour, and their publication in high-impact journals, our studies have not yet influenced clinical practice. A recent US survey confirmed that most clinicians faced with an excessively anticoagulated, non-bleeding patient would simply withhold warfarin and not give oral vitamin K (18). The lack of use of vitamin K is attributable to at least four factors; first, no study has unequivocally shown that oral vitamin K reduces the risk of hemorrhage. Although our first randomized trial (13) did find less bleeding in patients who received vitamin K, the bleeding in the placebo group occurred more than one week after randomization, making it possible that this observation was attributable to the play of chance. Second, there is concern that, irrespective of its route of administration, vitamin K may place patients at risk of thrombosis by "over-correcting" the INR. Although our two randomized trials showed no evidence that low dose vitamin K causes either warfarin resistance or thrombosis, at least one other publication has suggested this might occur (19). Third, the "passive" strategy of merely withholding warfarin is simple, inexpensive and easily applied. Finally, access to a convenient form of vitamin K is a major barrier as there is no convenient and easily administered form of oral vitamin K in many jurisdictions.

Study design and implementation

Trial design: This study is a randomized, double-blind, placebo-controlled, multicentre clinical trial with a concurrent cohort study (Appendix 1 - Study Flow Diagram). The randomized trial will enrol patients with INR values of 4.5 to 10.0. The cohort study will enrol patients with INR values of more than 10.0.

Study interventions:

RANDOMIZED TRIAL (INR 4.5 to 10.0): Eligible, consenting patients will temporarily discontinue warfarin and will be randomized to receive a single capsule containing either 1.25 mg of oral vitamin K or no active drug.() COHORT STUDY (INR more than 10.0): All eligible consenting patients will receive 2.5 mg of vitamin K.

Allocation to study groups:

Three types of anticoagulant clinics exist. In the first type, patients present to the clinic for blood-drawing, and wait for their INR. Their dose of warfarin is then prescribed based on these results. Within this type of clinic, patients will be identified at the time that their INR prolongation is discovered. Consent for the study will be obtained, and the patient will be provided with the next numbered study kit, from which a single capsule containing either vitamin K or placebo will be administered. In clinics in which a point-of-care device is used for INR determinations the INR will be confirmed with a venous sample before the patient is enrolled. For the purposes of this study, such centres will be called "group 1" sites.

The more common anticoagulant clinic is one within which patients have their INR determined by an outside laboratory, with the results called to a clinic physician, nurse or pharmacist. Within such a clinic, it is impractical to require patients to present to the hospital for assessment and randomization on the day that their INR is prolonged. For clinics using this model, we propose that when patients are found to have an INR of more than 4.5 they be assessed over the telephone for inclusion and exclusion criteria for the study. Patients meeting inclusion criteria, with no exclusion criteria, will be asked to provide consent over the telephone for the study. Once consent is obtained, and study forms are completed using information obtained over the telephone, the study centre will arrange to have the next numbered study drug allocated to the patient. The study drug will then be shipped to the patient using taxi cabs or a dedicated courier service. Receipt and consumption of the drug will then be confirmed at the time of the mandatory in centre assessment the following day.

For the purposes of this study, all centres using this form of allocation to study drug will be called "group 2" centres.

In both group 1 and group 2 centres patients with INR values of more than 10.0 will be given two capsules each containing 1.25 mg of oral vitamin K. A supply of such capsules will be provided to all participating centres in a clearly marked container. This drug will be provided to the patient in the same fashion as drug is provided to patients in the cohort study.

Timely return of INR values is a problem when INR values are obtained outside the hospital. In some cases, for example, the INR may not be available until late in the evening or the day following the blood draw. If more than 24 hours has elapsed since the blood draw and the time at which a patient would be eligible for randomization the patient will be deemed ineligible for the study. In such cases however, the INR can be repeated (for example at the hospital or clinic laboratory) and if the INR remains above 4.5, the patient can then be enrolled.

Patients who self-administer study drug will be asked to bring their unused capsule to the clinic on the day following randomization. This will ensure that they took the correct capsule.

The least frequently used mechanism for monitoring warfarin therapy is patient self-monitoring using a point of care testing (POCT) device. Patients monitored with POCT are responsible for determining their INR and adjusting their warfarin dose based on their INR result. Patients who self-monitor using point of care testing and who find themselves to have an INR of more than 4.5 will be asked to contact their clinic staff. Prior to being approached for consent to enter this study the patient's INR will be confirmed on a venous blood sample organized by their host anticoagulant clinic. Given that patients in this group will be randomized while at the study centre, for the purposes of this study such patients will be considered to be in group 1.

Group 1 and group 2 defines arbitrary groups. Clinics may have a mixture of group 1 and group 2 patients; such clinics will have some patients who have signed consents and who will be randomized "at home" (group 2 patient) and others who are randomized and receive study drug "in centre". Irrespective of where patients are at the time of randomization they will be allocated the next available study drug supply at the time of randomization.

Procedure for dealing with patients who are eligible for the study on more than one occasion: Patients can only be enrolled in either the randomized trial or the cohort study once. However, a patient previously enrolled in the cohort study who presented on a later occasion with an INR of 4.5 to 10 would be eligible for the randomized trial. Similarly, patients previously enrolled in the randomized trial are eligible for the cohort study. To ensure that the results of the randomized trial are not confounded by participation in the cohort study, a minimum of one month must pass between the time a patient is enrolled in the cohort study and their subsequent enrollment in the randomized clinical trial.

Method of randomization: Randomization will be performed using a computer-generated randomization sequence. This sequence will guide drug packaging. Study drug supply will be packaged and dispensed to each study centre in lots of sequentially numbered appropriately labeled pill bottles. An additional, clearly labeled, supply of capsules containing 1.25 mg of oral vitamin K will be dispensed to each participating centre. Patients will be randomized by allocating the next, sequentially numbered pill bottle.

What is the frequency of study follow-up ? Patients will be seen at each participating clinical centre by study staff on days 1 and 14 (+/- 2 days) after randomization. All patients will be contacted by telephone on days 3, 7, 28 and 90. Additional contact and INR sampling necessary to manage the patient's anticoagulant therapy will be made as required.

What are the study outcome measures ? The primary outcome measure is "all clinically overt bleeding". Secondary outcome measures are "all adjudication-confirmed major hemorrhage", "all adjudication-confirmed thrombotic events", "changes in INR values" and "cost effectiveness"

Why include all bleeding, and how is this endpoint defined ? Previous studies of both antithrombotic treatment and the treatment of warfarin-associated coagulopathy have used only major hemorrhage as the measure of bleeding. In those studies, "major hemorrhage" has been chosen as the primary measure because it is both detected and independently adjudicated more easily than "all hemorrhage". In addition, in most such studies of antithrombotic therapy, patients with acute thrombosis (or those at high risk of acute thrombosis) are enrolled. In the setting of acute thrombosis it is likely that physicians and patients are willing to tolerate minor bleeding as one of the "unavoidable" complications of antithrombotic therapy. Neither patients nor physicians would be willing to stop an effective treatment for acute, life threatening thrombosis because of minor bleeding. In the setting of warfarin maintenance therapy, however, minor bleeding such as unanticipated bruising, epistaxis or excessive menstrual blood loss in the setting of an excessively prolonged INR value is disconcerting, disruptive, costly if patients seek medical attention and likely to reduce compliance. Since we believe that reducing all bleeding complications (major and minor) is necessary to maximize the utility of warfarin therapy, we have chosen to measure all bleeding as our primary outcome measure of efficacy. We feel this is appropriate since the study intervention is masked; thus although minor bleeding is subjective, patients should be unaware of their treatment allocation and thus the treatment they receive will not influence their reporting of clinical events.

Secondary outcome measures:

Major Hemorrhage: Since we anticipate that many clinicians will wish to know the impact of vitamin K on "major bleeding", we will also compare the rates of all adjudication-confirmed major bleeds between groups. All bleeding events will be submitted to the central adjudication committee and major bleeding will be confirmed by standardized, masked central adjudication (see appendix 2).

Thrombotic Events: At each clinic visit and telephone follow-up signs and symptoms of thrombotic events will be sought. If a thrombotic event is suspected, appropriate objective documentation will be obtained and will be forwarded to the Central Adjudication Committee. The committee will adjudicate these thrombotic events using validated, pre-specified criteria (see appendix 2).

Cost-effectiveness: Details on the number and frequency of interventions, outcomes and their associated morbidity will be collected. This material will form the basis for a subsequent cost-effectiveness study. Given that vitamin K is very inexpensive (oral vitamin K costs less than 5$ Canadian per dose in most jurisdictions) and the expense of bleeding is high (previous cost effectiveness studies have estimated that an episode of major bleeding costs more than 5000 $ Canadian per event (20)) it is very likely that the use of oral vitamin K will be highly cost-effective.

Plan for ascertaining outcome events: Signs or symptoms of all (major and minor) bleeding and thrombosis will be sought at each clinic visit or telephone call. If bleeding or thrombosis is reported, all available objective documentation such as hospital records, transfusion records or results of diagnostic or therapeutic investigations (such as endoscopy) will be obtained. Minor bleeding will frequently not have led to medical evaluation; in cases where minor bleeding is reported the specific setting, nature, duration and required actions will be sought for each event reported by the patient.

Determination of the sample size (INR 4.5 to 10.0): The estimated rate of all bleeding in patients who present with excessive, warfarin-associated prolongation of the INR, treated with simple warfarin withdrawal, is 8% over the ensuing 30 days (8). We believe that clinicians would be willing to treat a minimum of 20 patients with oral vitamin K to prevent one episode of bleeding, thus the "minimally clinically important difference" that we wish to reliably detect is 5% (21;22). This reduction in the risk of bleeding (from 8% in those who receive placebo to 3% in those who receive vitamin K) represents about a 60% risk reduction. It is reasonable to expect that 1.25 mg of oral vitamin K will produce at least a 60% reduction in the risk of bleeding based on the following assumptions: (a) The INR in the placebo arm is likely to decline by about one INR point over the first 24 hours while the INR in the vitamin K group is likely to decline by about 3.9 INR points. These declines are based on the results achieved in our 3 published studies of oral vitamin K (10;12;13). (b) A 2.9 point reduction in the INR should be associated with a 6 fold reduction in the risk of hemorrhage (based on the assumption that a 1 point increase in the INR doubles the risk of bleeding (4-7)). (c) The 60 percent risk reduction is a conservative interpretation of the effect of a 6 fold decrease in the risk of hemorrhage.

To accomplish our goal of reliably detecting a 5% reduction in the risk of hemorrhage we will need to enrol 330 patients into each arm of the study to have 80% power with a two sided alpha of 0.05. This sample size calculation is based on the following assumptions:

Figure 4: Sample size calculation

2N = 4(Zα + Z)2 * (1-p)/(pc-pi)2 where Zα = 1.96, Z = 0.84

pc is the proportion with success in the control group (no bleed) and is equal to 0.92 pi is the proportion with success in the intervention group (no bleed) and is equal to 0.97

and p = (pi + pc)/2

thus the total sample size is 652

To accommodate for anticipated losses to follow up (which have been minimal in our previous studies) we will increase the sample size by 5% (to 690 patients)

Sample size determination (cohort study): The sample size for the cohort study is entirely one of convenience; since there is no good quality data on the efficacy of oral vitamin K in patients with INR values of more than 10.0 we will simply enrol and follow such patients as they become available. Data derived from this cohort might then be used as the basis for future investigations.

Study administration:

Accounting for non-compliance: The sample size has not been adjusted to reflect non-compliance. It is unlikely that patients will be non-compliant with a single dose of study drug administered at the time of randomization. For example, in our previous studies none of the 134 patients receiving oral vitamin K have been non-compliant or lost to follow-up.

Accounting for "loss to follow-up": We have not previously had patients lost to follow-up in our vitamin K studies, and we anticipate the rate in this study will be negligible (i.e. less than 5%). The low rate of loss to follow-up is due, in large part, to the fact that (a) patients will have only 2 clinic visits over the course of the study (day 1 and 14) and (b) all other follow-up will be performed over the telephone, Further, patients receiving warfarin are aware of the need for monitoring of their warfarin and thus accept the need for periodic medical and laboratory re-assessments. This increases the likelihood that they will be willing and able to participate in study follow-up.

Coordination of the study: The study will be coordinated by the Thromboembolism Research Unit at St Joseph's Hospital in Hamilton, Ontario. Data submission to the coordinating centre will occur over a secure, web-based database system, which will be supplemented by paper copies of the case report forms which will be maintained at each clinical centre. Details of the web-based data-entry system are currently being determined; a revised study manual with full data entry details will be included in the subsequent versions of this study manual.

The study will be coordinated by a professional data coordinator, who is based at St Joseph's hospital. Data integrity will be ensured by manually reviewing the electronic database for 5% of submitted records, and comparing these with copies of the paper CRFs from the relevant centres.

The randomization sequence for all patients will be generated at the coordinating centre; the coordinator will remain masked to treatment allocation at all times.

Provision of study drug: There is no oral form of vitamin K currently marketed in Canada, thus our previous investigations have used the intravenous formulation administered by mouth. Demonstration that a tablet form is effective would further simplify (and encourage) the use of oral vitamin K. A 5 mg vitamin K tablet is available in both Canada and the United States (in Canada the drug is available only after Health Canada approval). For this study, a supply of the 5 mg vitamin K tablets will be obtained from Merck Frosst Incorporated. Study pharmacies will compound this drug to produce matching capsules containing either lactose or lactose and 1.25 mg of vitamin K. This pharmacy will also provide drug packaging and will be responsible for preparing study drug kits which consist of sequentially numbered drug supplies for each clinical centre (following a randomization list provided by the study coordinating centre) and a supply of labeled 1.25 mg capsules (for the cohort study). These supplies will then be distributed, by the coordinating centre to each participating clinical centre.

Role of the participating investigators: Dr Mark Crowther is the principal applicant and will be primarily responsible for all aspects of the study. Drs Ageno and Garcia, in addition to fulfilling the roles of co-investigators, will coordinate recruitment and management of centres in Italy, and the United States, respectively. Drs Crowther, Ageno and Garcia will comprise the executive steering committee which will meet by teleconference at least monthly to discuss specific trial implementation and management issues. The steering committee will consist of the members of the executive steering committee, as well as the Canadian participating investigators (Drs Blostein, Anderson, Wells, Kahn, Ginsberg, Kovacs, Kearon and Selby) and Dr Witt from the United States. Telephone Steering Committee meetings will occur six monthly, or more frequently as determined by the Principal Investigator. The Steering Committee is responsible for the overall conduct of the study, and will act on the recommendations of the study's subcommittees.

Participating centres represented by letters of support on this grant include; Duke University (Dr Ortel); University of Washington (Dr Wittkowsky); University of Massachusetts (Dr Becker); IRCCS Maggiore Hospital of Milano (Dr Moia); Centro Emostasi e Trombosi Reggio Emilia (Dr Silingardi); Instituti Ospitalieri di Cremona (Dr Testa and Alatri); Dr. Siragusa (Policlinico Palermo)

Adjudication committee: An Adjudication Committee of three thromboembolism experts not otherwise involved in this study will be convened.All reported hemorrhagic and thrombotic events will be reviewed by this committee, which will use all available source documentation to adjudicate clinical events. This committee will be convened when complete data are available on the first 20 bleeds or clots, and subsequently with each multiple of 20 events. To ensure uniformity, all episodes of hemorrhage will be adjudicated according to pre-specified criteria. The committee will report the results of their adjudication to the Coordinating Centre for inclusion in the study database. Although adjudication of all events may seem excessive it is possible that adjudication could "upgrade" events deemed minor by the site investigator, highlighting the need for central review of all bleeding events.

DSMB: An independent Data Safety Monitoring Board (DSMB) will be constituted. The committee will consist of 2 experts who are not otherwise involved in the study. Proposed members are Dr Martin O'Donnell (Thromboembolism expert, Henderson Hospital, Hamilton, Ontario) and Dr William Geerts (Thromboembolism expert, Sunnybrook Hospital, Toronto, Ontario). The number of outcome events will be recorded for the cohort study and the randomized trial separately. The DSMB will be convened when a total of 20 (or multiples thereof) adjudication-confirmed major bleeds or thrombi have occurred in either of the randomized trial or the cohort study. While remaining masked to treatment allocation, the DSMB will review the source material describing the events. If the DSMB is concerned about the type or number of events, they will have the right to ask the steering committee to allow them to contact the study statistician and have an unblinded, unplanned, interim analysis for safety performed. Only experienced thromboembolism experts have been invited onto this committee, to ensure that such an interim analysis is not performed in response to less significant observations. If the interim analysis suggests that there is a significant safety concern (for example, if there is evidence of excess major bleeding or thrombotic events, as indicated by a P value of less than 0.05) the DSMB will have the right to recommend to the Steering Committee that the study be closed.

Analysis plan:

Primary analysis (Randomized trial): The primary analysis will be a simple comparison of the proportion of patients in the placebo and vitamin K arm with adjudication-confirmed bleeding. The proportion in each group will be compared using a Chi square or Fisher's exact test, as dictated by the number of events.

Secondary analysis: The proportion of patients in each group with major hemorrhage, thrombosis and a composite of all adverse outcomes (thromboembolism, death and major hemorrhage) will be performed using a Chi square or Fisher's exact technique, as indicated. A regression analysis will be performed to determine if baseline variables including the receipt of drug or placebo, demographic characteristics, degree of prolongation of the international normalized ratio or other baseline variables predict adverse events. Finally, the mean INR values in the two groups will be compared using ANOVA.

Primary analysis (cohort study): The change in the INR, and the number of bleeds and clots will be presented graphically.

Frequency of analyses (Randomized trial): Three interim analyses are planned. The first will occur after 200 patients have been enrolled. In this analysis, the statistician will not be aware of treatment allocation; the overall rate of bleeding for the entire population will be computed. Assuming balanced allocation, we anticipate the event rate will be 5.7 percent. If the observed rate of hemorrhage is more than 2.5 percent above or below 5.7 percent, a revised sample size will be calculated. If the revised sample size suggests an unfeasibly large study (defined by the need for more than 500 patients per arm), the statistician will determine whether limiting enrollment to patients with INR values of more than 6.0 will enrich the event rate sufficiently to allow completion of the study with less than 500 patients per arm. In these analyses, blinding to treatment allocation will be maintained and the sample size will be revised assuming a 60 percent reduction in the risk of bleeding in the vitamin K arm.

Second and third interim analyses will be performed after 350 and 500 patients have been enrolled. If the sample size is adjusted at the first interim analysis, the second and third analyses will be performed after 60% and 75%, respectively, of the required sample size is enrolled. For these analyses, the study statistician will be unblinded to treatment allocation and the proportion of patients suffering adjudication confirmed major hemorrhage will be compared between the two groups. If there is a treatment advantage as measured by a two-tailed Chi-square or Fisher's exact test with a P value of less than 0.001 the statistician will recommend to the study Steering Committee that the study be stopped. Any other result will result in a recommendation to continue this study.

Frequency of analyses (Cohort study): No interim analyses are planned. The Data Safety Monitoring Board (DSMB, see below) will be instructed to monitor the frequency of hemorrhage in this group. If the upper limit of the 95% CI about the risk of hemorrhage exceeds 5% after a minimum of 20 patients have completed the 90 day outcome ascertainment, the DSMB will notify the Steering Committee that the risk of hemorrhage in the cohort study is excessive. The steering committee will then decide whether this arm of the study should be discontinued or the intervention modified.

Planned subgroup analyses The response to vitamin K or placebo is measured in two ways in this study. The first way is by observing the number of clinical events (thrombosis and hemorrhage) and the second is by observing changes in the INR values between the day of study drug and the next day. The number of clinical events in the study is likely to be relatively low, reducing the power of subgroup analyses to determine if baseline variables such as the degree of prolongation of the INR influence the response to oral vitamin K. However, since changes in the INR value will occur in all patients and since the INR is a continuous variable, a subgroup analysis which examines the influence of baseline variables (such as the degree of prolongation of the INR) on the response to vitamin K will have more power if performed with the change in the INR value as the dependant variable. This analysis will include: baseline demographic variables (such as gender, weight, presence of cancer, and age), the degree of prolongation of the INR, baseline maintenance warfarin dose and the indication for warfarin therapy. The ability of these variables to predict changes in the INR in both the vitamin K and placebo groups will be calculated. The ultimate goal of this line of investigations is to generate hypotheses about those factors (if any) that influence the response to vitamin K. These hypotheses could be tested in future studies within which the dose of vitamin K is modified based on the presence or absence of these risk factors.