Felix J. M. van der Meer, Department of Thrombosis and Haemostasis, C2-R, Leiden University Medical Center, PO Box 9600, 2300 RC, Leiden, The Netherlands. Tel.: +31 71 526 2267; fax: +31 71 526 6755; e-mail: email@example.com
Summary. Background: One of the causes of unstable anticoagulant control in patients using vitamin K antagonists is a fluctuating intake of vitamin K. Research suggests that patients with a low dietary intake of vitamin K have a less stable anticoagulant control than patients with a higher intake. Objectives: To study whether supplementation with a low daily dose of vitamin K improves anticoagulant control. Methods: We performed a double-blind, randomized, placebo-controlled trial. 200 patients of the Leiden anticoagulation clinic, who used the vitamin K antagonist phenprocoumon, were randomized to receive either adjusted-dose phenprocoumon and 100 μg vitamin K once daily or adjusted-dose phenprocoumon and a placebo. Treatment duration was 24 weeks. The primary outcome was the percentage of time the International Normalized Ratio was within the therapeutic range. Results: The time in the therapeutic range was 85.5% in the placebo group and 89.5% in the vitamin K group (adjusted difference 3.6%; 95% CI −0.8% to 8.0%). The time below the therapeutic range was 3.1% in the placebo group and 2.1% in the vitamin K group (adjusted difference −0.7%; 95% CI −2.5% to 1.1%) and the time above the therapeutic range was 11.4% in the placebo group and 8.5% in the vitamin K group (adjusted difference −2.9%; 95% CI −6.9% to 1.1%). The relative risk (RR) of a maximal stability in the vitamin K group compared to the placebo group was 1.8 (95%, CI 1.1–2.7). Conclusion: Supplementation of vitamin K antagonists with 100 μg vitamin K improves stability of anticoagulant therapy. Because the risk of side effects is inversely related to anticoagulant stability, such an improvement is likely to reduce the number of bleeding and thrombotic events.
Oral anticoagulant treatment with vitamin K antagonists is indicated for the primary and secondary prevention of both arterial and venous thrombosis . A major disadvantage of vitamin K antagonists is their narrow therapeutic window and the large inter- and intra-individual variability in anticoagulant response. Despite intensive monitoring, the intensity of anticoagulation, expressed as the International Normalized Ratio (INR), is within the target range only approximately 60% of the time . The impact of under- and overanticoagulation is high, with a sharp increase of severe hemorrhage when the INR rises above the therapeutic range and a rise in thrombotic risk as the INR falls below the therapeutic range [2–4]. An improvement of the quality of anticoagulant treatment will reduce the number of these adverse events.
One of the causes of the variability in anticoagulant response is a fluctuating vitamin K intake [5,6]. Studies have shown that the INR is especially sensitive to vitamin K changes when vitamin K intake is low [6,7]. In patients with unstable anticoagulant control, the daily intake of vitamin K has been shown to be lower than in stably anticoagulated patients . We hypothesized that supplementation with a low daily dose of vitamin K results in an increased stability of anticoagulant control. The objective of this study was to test this hypothesis clinically, by assessing the effect of vitamin K supplementation on anticoagulant stability in patients treated with vitamin K antagonists.
The study was a double blind, randomized, placebo controlled trial. 200 patients were randomized in blocks of eight patients into two equal groups. All patients were treated with adjusted-dose phenprocoumon, a vitamin K antagonist with a long half-life of 140 h. In addition, patients in the treatment group received 100 μg vitamin K once daily, and patients in the other group received a placebo once daily. The study medication was used for 24 weeks. After this, patients were followed for an additional four weeks, to observe possible side effects after stopping the study medication. The study was approved by the local Medical Ethics Committee. Written informed consent was obtained from all participants prior to enrolment.
Participants were recruited from the Leiden anticoagulation clinic. Patients were considered for enrolment if they were between 18 and 80 years of age, had an indication for long-term oral anticoagulant therapy and had been using phenprocoumon for at least one year. Exclusion criteria were: treatment by a medical specialist for liver failure; hemodialysis or peritoneal dialysis; pregnancy or planned pregnancy; puerperium; any chronic condition with an expected survival of less than six months; an expected interruption of oral anticoagulant treatment of more than one week; self-management of oral anticoagulant treatment and non-compliance, based on information from the anticoagulation clinic computer records.
The required sample size was calculated using data of the Leiden anticoagulation clinic. We expected an improvement of the time in the therapeutic zone from 75% to 85%. The standard deviation of the time in the therapeutic zone was estimated at 23%. To achieve 80% power for detecting this difference at a significance level of 5%, 84 subjects per group were required. Allowing for a loss of patients of 15%, the total number of patients was rounded to 200.
The study was performed at the Leiden anticoagulation clinic, where approximately 7000 patients are treated each year. During the first four weeks after starting the study medication, the INR was measured weekly to be able to adjust the dosage of phenprocoumon in case of INR changes as a result of the vitamin K. Subsequently, the interval between visits depended on anticoagulant stability, with a maximum of four weeks. After stopping the study medication, patients were again seen at weekly intervals for four weeks. INR results and prescribed dosages were recorded in a central database, along with relevant history details, changes in medication, information on bleeding and thrombotic complications, admissions and surgical interventions. After inclusion, patients were asked to list their current co-medications. No specific dietary recommendations were given. Compliance with the study medication was verified by pill-counting. We classified patients as compliant when they had used over 90% of the prescribed capsules, and non-compliant otherwise.
The therapeutic ranges were according to the guidelines of the Federation of Dutch Anticoagulation Clinics: INR 2.0–3.5 with a target INR of 3.0 for low-intensity treatment and INR 2.5–4.0 with a target INR of 3.5 for high-intensity treatment.
Capsules containing 100 μg of vitamin K or placebo were manufactured by Numard Pharmaceuticals (Lelystad, the Netherlands) from 5% dry vitamin K (Acatris, Londerzeel, Belgium). The dose of 100 μg vitamin K was chosen because of the results of a pilot study, in which we found that this dose had only a minor effect on the INR and subsequent dosage adjustments, while being close to the recommended daily intake (90 μg for women and 120 μg for men ) .
The primary study outcome was the quality of anticoagulant treatment, expressed as percentage of time of the INR in the therapeutic range. The time in range was calculated using the linear interpolation method, as described by Rosendaal et al. . In this method, the INR is modeled to increase or decrease linearly between two consecutive measurements. Each time period between two INR measurements can then be divided into time in, above or below the therapeutic range. For each patient, the time per category is summed and divided by the total follow-up time. Because the research question was mechanistic rather than a question of efficacy, we only calculated the time in range over the period in which the study medication was used. Also, the first four weeks after starting the study medication were excluded from the analysis, to allow some time to find a new dose balance. Secondary outcome measures were the number of bleeding and thrombotic complications. Major hemorrhage was defined as fatal or intracranial hemorrhage and any hemorrhage that required blood transfusion, hospitalization or surgery, as well as muscle and joint bleeding.
Finally, we classified patients in categories of stability and contrasted patients with maximal stability (time in range of 100% during follow-up) between the treatment and placebo group, which allowed the calculation of a relative risk (RR).
Age, sex, target range, use of co-medication and use of medication interacting with the anticoagulant drug were considered possible confounders and were adjusted for in the analysis.
Patients were enrolled between December 2004 and January 2006 (Fig. 1). Of 1053 patients, 200 consented to be randomized. Data on 18 patients were not analyzed or only partly analyzed, because they did not complete follow-up. Follow-up time was 34.2 patient years (py) in the placebo group and 33.7 py in the vitamin K group.
Patient characteristics are shown in Table 1. Age, sex, duration of anticoagulant treatment, previous stability and indication for anticoagulant treatment were similar in both groups. In the placebo group, there were slightly more patients in the high target range. Patients in the placebo group used more co-medications, including drugs interacting with anticoagulants.
Table 1. Baseline characteristics of the study population. Data are median (interquartile range) or number (%)
Placebo (n = 95)
Vitamin K (n = 94)
Female sex (%)
Duration of anticoagulant treatment (years)
Previous stability (time in therapeutic range)
Secondary prevention venous thrombosis
Mechanical heart valve
Therapeutic range (%)
Number of co-medications
Interacting medication (%)
The primary study outcome is shown in Table 2. The time in the therapeutic range was 85.5% in the placebo group (95% CI 82.3–88.6%) and 89.5% in the vitamin K group (95% CI 86.4–92.5%). The time in range was 4.0% higher in the vitamin K group than in the placebo group (95% CI −0.3 to 8.3%). After adjusting for age, sex, target range, number of co-medications and use of interacting medication, this difference was 3.6% (95% CI −0.8 to 8.0%). Both the time below the therapeutic range and the time above the therapeutic range were slightly lower in the vitamin K group.
Table 2. Time in, below and above the therapeutic range. Data provided are mean percentages (95% CI) of time in, below and above the therapeutic range
Placebo (n = 95) (%)
Vitamin K (n = 94) (%)
Adjusted difference (%)*
*Adjusted for age, sex, target range, number of co-medications and use of interacting medication.
Time in the therapeutic range
4.0 (−0.3 to 8.3)
3.6 (−0.8 to 8.0)
Time below the therapeutic range
−1.0 (−2.8 to 0.7)
−0.7 (−2.5 to 1.1)
Time above the therapeutic range
−3.0 (−6.9 to 0.9)
−2.9 (−6.9 to 1.1)
Fig. 2 shows the percentage of patients classified in categories of time in the therapeutic range. Of the patients in the vitamin K group, 43% were in the therapeutic range 100% of the time vs. 24% in the placebo group. The RR of a maximal stability in the vitamin K group compared to the placebo group was 1.8 (95% CI 1.1–2.7).
Table 3 shows the time in the therapeutic range in the vitamin K and placebo groups for different patient and treatment characteristics. The results are similar for all categories. Patients in the vitamin K group were more stable in all but three subgroups, although the confidence intervals are wide. Previous stability was the best predictor of time in the therapeutic range during the trial.
Table 3. Percentage of time in the therapeutic range in various patient categories. Data provided are mean percentages (95% CI) of time in the therapeutic range
Placebo group (n = 95)
Vitamin K group (n = 94)
Number of patients
Percentage time in the therapeutic range
Number of patients
Percentage time in the therapeutic range
3.7 (−4.3 to 11.7)
2.5 (−5.0 to 10.0)
5.6 (−1.9 to 13.1)
4.5 (−0.6 to 9.6)
2.6 (−6.2 to 11.4)
Duration of anticoagulant treatment (years)
−1.5 (−11.9 to 8.7)
7.5 (−3.6 to 18.6)
3.1 (−3.4 to 9.6)
Previous stability (time in the therapeutic range)
5.8 (−30.5 to 42.2)
0.1 (−5.9 to 6.1)
4.6 (−3.2 to 12.5)
Secondary prevention venous thrombosis
11.5 (−0.7 to 23.7)
Mechanical heart valve
5.0 (−6.5 to 16.6)
0.4 (−6.3 to 7.0)
5.9 (−0.4 to 12.2)
1.7 (−4.1 to 7.4)
Compliance study medication
2.2 (−2.3 to 6.7)
13.0 (−21.8 to 47.8)
12.6 (−0.4 to 25.7)
−3.8 (−18.5 to 10.9)
−5% to 5%
6.6 (−0.2 to 12.9)
1.6 (−6.1 to 9.3)
Number of co-medications
3.1 (–4.1 to 10.2)
–2.3 (–12.7 to 8.1)
0.8 (−6.5 to 8.1)
Compared to the 24 weeks before the start of the study medication, the dose was increased by an average of 0.8% in the placebo group (range −16% to 57%) and 7.0% in the vitamin K group (range −57% to 39%). The difference in time in range between the placebo group and the vitamin K group was largest in patients with a dose change between −5% and 5% (Table 3).
Three patients in the placebo group and five in the vitamin K group were classified as non-compliant with the study medication (more than 10% of the prescribed capsules not taken). Both compliant and non-compliant patients were more stable in the vitamin K group than in the placebo group (Table 3).
The beneficial effect of vitamin K was strongest (7.2% time in range) in those with no or few co-medications and absent in those with six or more co-medications. When we looked specifically at drugs that potentially interact with vitamin K antagonists, we found that the effect of vitamin K supplementation appeared restricted to those who did not use interacting agents (Table 3).
Two patients died, one in the vitamin K group (intracranial hemorrhage as a result of cerebral contusion) and one in the placebo group (cause of death unknown). Two patients in the vitamin K group had a major bleeding event; the before-mentioned fatal bleed (last known INR 2.8) and a muscular bleed (INR 5.1). There were 17 minor bleeding episodes, 10 in the placebo and seven in the vitamin K group (RR for any bleeding event, vitamin K group vs. placebo group, 0.9; 95% CI 0.3–2.3). No thrombotic events were reported. There were five self-reported putative side effects, four of which were in the placebo group. These included weight gain (two patients in the placebo group, one in the vitamin K group), flatulence and indigestion.
We performed a randomized placebo-controlled trial in 200 outpatients of a Dutch anticoagulation clinic, to study the effect of vitamin K supplementation in patients treated with oral anticoagulants. We observed a small increase in the stability of anticoagulant treatment in patients receiving vitamin K supplementation. There was a 2-fold increase in the percentage of patients with a maximal stability of 100% of time within the therapeutic range, from 24% to 43% of patients.
The need to investigate whether vitamin K supplementation leads to an improved anticoagulant stability has been identified by several authors [12,13]. The question was raised based on recent studies, which have shown that the INR is more sensitive to vitamin K changes in patients with a low vitamin K status than in those with a normal or high vitamin K status [6,7] and that dietary vitamin K intake in unstable patients is considerably lower than in stable patients . Recently, Sconce et al.  published their results of a study investigating vitamin K supplementation in 70 patients receiving warfarin with an unstable anticoagulant control. They found that supplementation with 150 μg of vitamin K increased stability of anticoagulation, expressed as the standard deviation of the INR and the percentage of time in the target range. Our results confirm that this strategy increases the time in the therapeutic range in an unselected population of patients using vitamin K antagonists.
The observed improvement in time in the therapeutic range of 4% was smaller than the improvement of 10% on which the study was powered. This explains why the results do not reach the significance level of 5%; the size of our study was too small to detect this smaller difference. Taking into consideration the biologic plausibility of the results, previous studies showing low vitamin K status in unstable patients and the previous trial by Sconce et al. , we believe that the observed difference is real. The key question, of course, is whether this difference is clinically relevant. The 4% improvement of time in the therapeutic range will theoretically prevent one to two serious side effects per 1000 treatment years . It must be kept in mind that time in the therapeutic range in this study improved from 85% to 89%. So, even in the placebo group, the time in range was unexpectedly high. In a less stable population, the benefit of vitamin K supplementation is likely to be higher than the observed 4% difference.
There are several reasons why the time in the therapeutic range was higher in this study than is usually reported. One reason is the definition of therapeutic range, which is wider in the Netherlands than in other countries, where a target range of 2.0–3.0 is frequently employed. Furthermore, only patients who had been using anticoagulants for more than a year were included. These patients are often more stable than short-term patients. Finally, even in the placebo group, the time in range was 7% higher than before the trial. There are two possibilities why the time in the therapeutic range was so high in the placebo group. First, the study population (placebo and treatment groups) differed from the routinely treated patients. This is a common phenomenon in randomized trials, because of exclusion criteria, self-selection of consenting patients or because patients changed their behavior during the trial. Another possibility is that the placebo group differed from the treatment group by chance. If this were the case, the true effect was seemingly reduced, and in fact higher.
Despite randomization, the groups differed in use of co-medication. Patients in the placebo group used more co-medications in general and also more drugs interacting with vitamin K antagonists. Because the time in the therapeutic range was associated with both variables, they were adjusted for in the analysis. An interesting finding was that vitamin K supplementation especially improved anticoagulant control in patients using no or few co-medications and in patients not using interacting medication. However, because this was a post hoc subgroup analysis with wide CI, these results should be interpreted with caution.
A theoretical concern is that increasing the dose of the vitamin K antagonist increases the risk of side effects unrelated to the INR. Other proteins than those involved in hemostasis depend on the enzyme system affected by vitamin K antagonists . However, in the past decades of oral anticoagulant use, the number of reported side effects other than hemorrhage has been remarkably low.
A limitation of this study was that blinding may not have been completely maintained because of dose adjustments in the treatment group. Because patients in the vitamin K group did better when there were no large dose adjustments, it is unlikely that this affected the results.
Oral anticoagulants are among the most frequently prescribed drugs. In Western countries they are used by over 1% of the adult population [16,17]. 10% of hospital admissions as a result of adverse drug events are caused by oral anticoagulants, often as a result of overanticoagulation . An improvement of the quality of oral anticoagulant therapy will result in less thrombosis and fewer bleeding events, and thus in a major reduction of burden of disease. Our results show that supplementation with a low dose of oral vitamin K contributes to improved anticoagulant stability. Further research on the optimal dosage of vitamin K supplementation in various patients groups is necessary to optimize anticoagulant control.
Disclosure of Conflict of Interests
The authors state that they have no conflict of interest.