The Effect of Medication Timing on Anticoagulation Stability in Users of Warfarin: The "INRange" RCT

Background

WARFARIN AND THE MONITORING OF ANTICOAGULATION THERAPY: Warfarin, an oral anticoagulant medication which disrupts vitamin K's ability to activate clotting factors in the liver, has strong evidence for efficacy in the prevention of thromboembolic disease (most notably in the setting of atrial fibrillation, mechanical heart valves or venous thromboembolism). However, it also has a very narrow therapeutic range and many drug / disease interactions - with risk of the clotting disorders it is meant to prevent (e.g. stroke, pulmonary embolism) in those who are "under anticoagulated" and risk of major bleeding as a side effect in those who are "over anticoagulated". For this reason warfarin users routinely undergo regular blood tests which measure the time it takes their blood to clot in a standardized setting compared to the population average. This test is commonly referred to as the "INR" (International Normalized Ratio) and patients will commonly have this blood test measured every 1 to 4 weeks with same-day decisions on warfarin dose adjustments made by their general practitioner (GP) based on that day's INR result. For atrial fibrillation and venous thromboembolism the target INR range is usually 2.0 to 3.0 (i.e. it takes 2 to 3 times longer for the patient's blood to clot while on warfarin). For mechanical heart valves, where the risk of stroke is higher, this target range is usually 2.5 to 3.5. Unfortunately some patients can have highly variable INR tests and the proportion of time they spend in the therapeutic range ("TTR") will be relatively low. This reduces the effectiveness of the therapy (if the INR is too low) and puts the patient at risk of potentially life threatening bleeding (if the INR is too high). The overall average for TTR in community settings in the published literature is 56.7% [95%CI 51.5-62.0%], which is considered suboptimal. "Good control" has been described as TTR > 75% and newer alternative anticoagulation strategies have been recommended when TTR is < 60%.

THE INFLUENCE OF VITAMIN K ON WARFARIN THERAPY: Vitamin K, which plants use in the process of photosynthesis, is found largely in green leafy vegetables (most notably kale, spinach, broccoli, lettuce and Brussels sprouts). It is an essential cofactor used by the liver to "activate" the clotting factors it releases into the blood. Vitamin K has a very short half-life in the body (approximately 2 ½ hours) and cycles through an active (so far as its ability to "activate" clotting factors) and inactive form in the liver. Warfarin acts by preventing one of the intermediary steps necessary to convert the inactive form of vitamin K back to the active form and hence it reduces the amount of available activated clotting factor. When vitamin K is first ingested, however, it is in an easily activated form upon which warfarin has little effect. As a result, consumption of high vitamin K containing foods can counteract the effect of warfarin and highly variable consumption of these foods may cause clinically important INR variability in some individuals. Most typically, low vitamin K containing foods are consumed in the morning, and highly variable vitamin K containing meals are consumed at dinner.

THE POTENTIAL INFLUENCE OF THE TIMING OF WARFARIN INGESTION ON ANTICOAGULATION THERAPY: Warfarin has a long half-life (36 to 42 hours) and, as a result, it is generally assumed by clinicians that the time of day it is ingested will not impact its efficacy. Although some patients will take it in the morning, most commonly warfarin is taken at dinnertime (or later in the evening) so that the results from each INR test can be communicated to the patient in time for dose adjustments (if any) the same day. However warfarin acts on the liver and immediately following ingestion (as is the case with all ingested substances) ALL of the ingested warfarin is delivered directly to the liver by the portal blood. Conceivably, because warfarin transitions through the liver first, there could be an important "first pass effect" that would make warfarin activity greatest around the time it is ingested (in comparison to a few hours later when the drug is distributed throughout the rest of the body). To date, this possibility has never been explored. If a first pass effect did exist, it is conceivable that ingesting warfarin at dinner (the same time as ingestion of the day's largest and most highly variable amount of easily activated vitamin K), may induce undesirable variability in what might be the period of greatest warfarin activity. Given vitamin K has such a short half-life, taking warfarin in the morning (when one would expect consistently low vitamin K activity) may produce a more consistent first pass effect (if one exists), and a more stable INR as a result.

Hypotheses

1. Morning, as compared to evening, administration of warfarin will produce a more consistent anticoagulant effect and improve the proportion of time a patient spends in the target therapeutic INR range.
2. A less variable daily (as opposed to weekly) dietary vitamin K consumption at dinner, whether consistently high or consistently low, will reduce variability in the anticoagulant effect of warfarin and improve the proportion of time a patient spends in the target therapeutic INR range.

Objectives

1. To determine (by RCT) whether switching current warfarin users from evening to morning dosing will alter the proportion of time spent in therapeutic INR range.
2. To determine (by cross-sectional analysis of baseline data) whether evening warfarin users with greater variability in daily dinner-time vitamin K ingestion have a lower time in therapeutic INR range.
3. To determine (by prospective subgroup analysis of RCT data) whether the effect of warfarin timing on TTR (i.e. the effect of changing to morning dosing) is influenced by day-to-day variability in vitamin K consumption.

Methods

Design: Prospective Randomized Open Blinded End-Point (PROBE) Study

Population: Community family physicians will be recruited from across two Canadian provinces, British Columbia and Alberta. These physicians will send a letter (from them) to their warfarin using patients which: a) Describes the project; b) Lets patients know their physician is participating; and c) Provides a central contact number (which reaches our study co-ordinator) for more information. The study co-ordinator dialogs with interested patients, ensures they are eligible, obtains written informed consent from all patients willing to be randomized and obtains baseline information presumed to be predictive of TTR [including age ≥ 80 years, hospitalization in the last 6 months, temporary planned discontinuation of warfarin in the last 6 months (e.g. for elective surgery), number of daily prescription medications, ≥ 6 months of warfarin use, the self-reported average number of days per week in which high vitamin K containing foods are consumed, and how variable this dietary consumption is]. For consenting patients who meet the interview criteria for eligibility, the family physician will be asked to provide investigators with the patient's target INR range and their last 6 months of INR results and test dates. This will provide baseline data for the cross-sectional analysis and allow determination of the last RCT eligibility criteria (which is availability of complete and adequate baseline INR data).

Intervention: Morning warfarin ingestion versus continued evening ingestion. Upon completing the determination of eligibility (i.e. upon receiving 6 months of baseline INR data from each patient's family physician) consented eligible patients will be randomized with stratification by baseline % of INR tests in range (<50%, 50-80%, >80%), using variable blocks of 2 or 4 (by a study co-ordinator with no clinical patient interactions), to either morning (active arm) or continued evening (control arm) ingestion of their warfarin. REDCap software will be used to perform randomization and ensure allocation concealment. Patients (and hence their providers) will not be blinded to the intervention.

Seven months after enrolment of each patient the investigators will ask their family physician to provide us all INR results for that period in order to determine the outcomes specified below.

Sample size: We wish to be able to demonstrate a 20% reduction in time out of therapeutic range and will conservatively estimate (since there are no prior studies exploring this outcome to guide us) that the standard deviation of this measure is twice the mean effect (i.e. SD = 40%). For a t-test with 1:1 allocation to control and experimental groups, power = 0.9, alpha = 0.05, minimum difference = 20% and SD = 40% the required sample size per group is 85 (i.e. 170 subjects in total). Providing for potential dropouts we will increase our target enrolment to 200 subjects.

Statistical analysis:

Calculating TTR:

The therapeutic INR range for each patient will vary and be determined by the treating physician. Typically the therapeutic range is 1 unit wide (often 2 - 3, or 2.5 - 3.5) but some physicians will choose narrower or wider ranges (e.g. 3.0 - 3.5, or 2 - 3.5). We will standardize the width of all target ranges by determining the midpoint of each patient's individual target range and use upper and lower limits that are 0.5 units above and below this midpoint. For example, if a physician is targeting a narrower than normal 3.0 to 3.5 range, we will use a midpoint of 3.25 and assume a (standardized width) target therapeutic range of 2.75 to 3.75. The proportion of time both in and out of therapeutic range will be determined using the linear interpolation method of Rosendaal which (conceptually) draws a line between sequential INR values no more than 8 weeks apart and assigns a projected INR value to every day in that interval.

RCT:

All analyses will be by intention to treat. The primary analysis of percentage change in time outside of therapeutic range will be by Student's t-test if the data appears normally distributed or by Mann Whitney U test if it is not. The secondary analyses will be by Student's t or Mann Whitney U (for percentage change in TTR and for maximum and minimum INR values), and by the Student's t or Fisher's exact test (for both percentages of patients with TTR > 75% \& < 60%, and major warfarin related cardiovascular events). A subgroup analysis of the influence on the intervention of the numbers of days per week that high vitamin K foods are ingested will be carried out looking at an ANOVA analysis of % change in time outside of therapeutic range according to 3 possible categories for the number of days per week of consumption of high vitamin K containing foods (these being less than 2, 2 to 5, and greater than 5 days per week). In the same analysis we will also examine the effect on the intervention of the patient's global assessment of how variable their consumption of high vitamin K containing foods is. To do this we will convert the 4-point scale of possible responses into a dichotomous variable that combines the two options indicating the most variable diet, and the two options indicating the least variable diet.

Baseline cross-sectional analysis:

The effect on baseline TTR of the number of days per week that high vitamin K containing foods are consumed will be analyzed with multiple linear regression using baseline covariates which include: gender, age ≥ 80 years, hospitalization in the last 6 months, temporary planned discontinuation of warfarin in the last 6 months, number of daily prescription medications, < 6 months of warfarin use, each of the 3 possible categories for number of days per week consuming high vitamin K containing foods, and the patient's dichotomous global assessment of how variable their pattern of vitamin K consumption is.