Multicenter Randomized Controlled Trial of Vitamin K Antagonist

CLINICAL RESEARCH www.jasn.org

Multicenter Randomized Controlled Trial of Vitamin K Antagonist Replacement by Rivaroxaban with or without Vitamin K2 in Hemodialysis Patients with Atrial Fibrillation: the Valkyrie Study

An S. De Vriese ,1,2 Rogier Caluwé ,3 Lotte Pyfferoen,4 Dirk De Bacquer,5 Koen De Boeck,6 Joost Delanote,4 Didier De Surgeloose,7 Piet Van Hoenacker,8 Bruno Van Vlem,3 and Francis Verbeke 9

1Division of Nephrology and Infectious Diseases and 4Department of Medical Imaging, AZ Sint-Jan Brugge, Brugge, Belgium; Departments of 2Internal Medicine and 5Public Health, Ghent University, Ghent, Belgium; 3Division of Nephrology and 8Department of Medical Imaging, Onze Lieve Vrouw Hospital, Aalst, Belgium; 6Division of Nephrology and 7Department of Medical Imaging, ZNA Middelheim, Antwerp, Belgium; and 9Division of Nephrology, University Hospital, Ghent, Belgium

ABSTRACT Background Vitamin K antagonists (VKAs), although commonly used to reduce thromboembolic risk in atrial fibrillation, have been incriminated as probable cause of accelerated vascular calcification (VC) in patients on hemodialysis. Functional vitamin K deficiency may further contribute to their susceptibility for VC. We investigated the effect of vitamin K status on VC progression in 132 patients on hemodialysis with atrial fibrillation treated with VKAs or qualifying for anticoagulation. Methods Patients were randomized to VKAs with target INR 2–3, rivaroxaban 10 mg daily, or rivaroxaban 10 mg daily plus vitamin K2 2000 mg thrice weekly during 18 months. Systemic dp-ucMGP levels were quantified to assess vascular vitamin K status. Cardiac and thoracic aorta calcium scores and pulse wave velocity were measured to evaluate VC progression. Results Baseline dp-ucMGP was severely elevated in all groups. Initiation or continuation of VKAs further increased dp-ucMGP, whereas levels decreased in the rivaroxaban group and to a larger extent in the rivaroxaban+vitamin K2 group, but remained nevertheless elevated. Changes in coronary artery, thoracic aorta, and cardiac valve calcium scores and pulse wave velocity were not significantly different among the treatment arms. All cause death, stroke, and cardiovascular event rates were similar between the groups. Bleeding outcomes were not significantly different, except for a lower number of life-threatening and major bleeding episodes in the rivaroxaban arms versus the VKA arm. Conclusions Withdrawal of VKAs and high-dose vitamin K2 improve vitamin K status in patients on hemodialysis, but have no significant favorable effect on VC progression. Severe bleeding complications may be lower with rivaroxaban than with VKAs.

JASN 31: ccc–ccc, 2020. doi: https://doi.org/10.1681/ASN.2019060579

Received June 8, 2019. Accepted September 11, 2019.

The use of vitamin K antagonists (VKAs) in patients Published online ahead of print. Publication date available at on hemodialysis with nonvalvular atrial ﬁbrillation www.jasn.org. (AF) is the subject of an ongoing debate. The ben- ﬁ Correspondence: Prof. An S. De Vriese, Division of Nephrology e cial effects of VKAs have been mooted, owing to and Infectious Diseases, AZ Sint-Jan Brugge, Brugge, and De- the lack of a straightforward relationship between partment of Internal Medicine, Ghent University, Ghent, Bel- AF and stroke in patients on dialysis and the ab- gium.Email:[email protected] sence of convincing evidence that VKAs reduce Copyright © 2020 by the American Society of Nephrology

JASN 31: ccc–ccc,2020 ISSN : 1046-6673/3101-ccc 1 CLINICAL RESEARCH www.jasn.org thromboembolic risk in this patient population.1,2 In Significance Statement addition, a disproportionately increased hazard of bleeding, in particular of hemorrhagic stroke, may tilt the benefit-to- Functional vitamin K deficiency, exacerbated by the use of vitamin K risk ratio of VKAs toward a net harm.1,2 Finally, ample cir- antagonists (VKAs), is thought to contribute to the rapid progression of vascular calcifications in patients on dialysis. We randomized cumstantial evidence implicates VKAs in the development of fi fi patients receiving chronic hemodialysis with atrial brillation to vascular calci cations (VCs), although high-quality clinical VKAs, rivaroxaban, or rivaroxaban with high-dose vitamin K2 sup- data are currently lacking.3,4 VKAs exert their anticoagulant plements. During 18 months of follow-up, vitamin K status improved effects by blocking the ɤ-carboxylation of coagulation factors. significantly by withdrawal of VKAs and vitamin K2 supplementation. Inevitably, they also prevent the activation of other vitamin Nevertheless, changes in coronary artery, thoracic aorta, and K–dependent proteins, some of which play a germane role in cardiac valve calcium scores and pulse wave velocity were not different among the treatment arms. Replacement of VKAs by the inhibition of VC, including matrix Gla protein (MGP), rivaroxaban was safe and potentially associated with less life- Gla-rich proteins, and growth arrest–specific protein 6. The threatening and major bleeding. Further studies should deter- procalcific effects of VKAs are thus intrinsic to their mode of mine whether earlier and multitargeted intervention can halt the action. Notwithstanding the controversy, current guidelines progression of vascular calcifications in dialysis. recommend the consideration of VKAs in patients with a $ 5,6 CHA2DS2-VASc score 2 and VKAs are still commonly calcification scores and also predicts cardiovascular disease used in the dialysis population with AF, albeit with a large and mortality.18,19 We therefore used these methods to eval- 7 practice variability and physician uncertainty. uate VC extent and progression in our study. Dephosphory- In the past few years, the use of direct oral anticoagulants lated uncarboxylated MGP (dp-ucMGP) is the inactive form (DOACs) in the hemodialysis population as an alternative for of MGP and is currently considered to be the most accurate VKAs has gathered momentum, despite a paucity of data on biomarker for vascular vitamin K stores when vascular end fi their safety and effectiveness and speci cdosingguidelines. points are studied.11 Hence, systemic dp-ucMGP levels were fi fi DOACs may have a better risk-bene tpro le, because they also measured to assess vitamin K status in each treatment provide more on-target anticoagulation and are associated group. with lower rates of intracerebral bleeding. Furthermore, they are not expected to accelerate the progression of VC, because they do not interfere with vitamin K metabolism. Selective METHODS inhibition of factor Xa may even have beneficial effects on – the development of atherosclerosis.8 10 Trial Design Patientswith CKD, andin particular those on dialysis, havea This study is an investigator-driven, randomized, prospec- high prevalence of subclinical vitamin K deficiency.11 Because tive, open-label interventional clinical trial, conducted at the propensity for VC may be a key corollary of vitamin K three sites in Belgium (AZ Sint-Jan Brugge, Onze Lieve deficiency, vitamin K supplements have garnered attention Vrouw Ziekenhuis Aalst, ZNA Middelheim Antwerpen). as a means to improve the dismal cardiovascular prognosis The study was approved by appropriately authorized ethics in the dialysis population. However, whether correction of committees in all participating sites and registered on Clin- vascular vitamin K status has a beneficial effect on the pro- icalTrials.gov (identifier NCT02610933). The design of the gression of VC in patients on dialysis is currently unknown.11 study has been described in detail previously.4 The study This study was designed to examine the effect of vitamin K has a three-arm parallel group design with a 1:1:1 allocation deficiency on the development of VC in the hemodialysis pop- ratio (Supplemental Figure 1). ulation. Patients on hemodialysis treated with VKAs epitomize a status of severe vitamin K deficiency, which we Participants sought to compare with replete vitamin K stores as attained by Adults on chronic hemodialysis with nonvalvular AF, with a withdrawal or avoidance of VKA plus administration of high- CHA2DS2-VASc score of $2, and therefore candidates for an- dose vitamin K supplements. To delineate the contribution of ticoagulation therapy or already receiving VKAs, were eligible each measure, an intermediate group was included in whom for inclusion. Inclusion and exclusion criteria are listed in the anticoagulation was achieved by a DOAC without additional Supplemental Material. All patients provided written, in- vitamin K supplements. formed consent. Computed tomography scan of the heart and thoracic aorta is a reliable technique to assess the extent of VC in patients on Interventions and Measurements hemodialysis.12 High calcification scores identify patients at Patients in the first treatment arm were started on a VKA or risk for cardiovascular events and death.13,14 Calcification continued the VKA, with dose adjustments to achieve an in- scores have been widely used as surrogate markers to assess ternational normalized ratio of 2–3 on the basis of weekly the effect of an intervention on cardiovascular risk in patients international normalized ratio measurements. Time in the with CKD and on hemodialysis.15–17 Arterial stiffness quan- therapeutic range was recorded. Patients in the second treat- tified by pulse wave velocity (PWV) correlates with ment arm received a daily dose of 10 mg rivaroxaban. The

2 JASN JASN 31: ccc–ccc,2020 www.jasn.org CLINICAL RESEARCH choice of this dose was on the basis of a comprehensive phar- experienced investigator (L.P.) reviewed all scans for con- macokinetic and pharmacodynamic analysis in patients on sistency of interpretation. For each subject, all imaging pro- hemodialysis, revealing that a dose of 10 mg provided a similar cedures were done on the same equipment using the same exposure as a dose of 20 mg in healthy individuals.20 Patients parameters at each session to permit valid longitudinal im- in the third treatment arm received a daily dose of 10 mg age comparisons. Scans started in the upper thorax above rivaroxaban and 2000 mg menaquinone-7 (MK-7) thrice the aortic arch, advancing caudally to the level of the dia- weekly after dialysis with directly observed therapy to ensure phragm to include the coronary arteries, the aortic and mi- adherence. MK-7 is a form of vitamin K2 that has a superior tral valve, and the ascending and descending thoracic aorta. absorption and bioavailability as compared with other me- Segments with stents were excluded from the analysis. In naquinones and vitamin K1.21,22 Thechoiceofthedosewas patients with coronary artery bypass grafts, the coronary on the basis of a short-term dose-finding study in patients arteries were excluded from the analysis in case of imaging receiving chronic hemodialysis.23 Incremental doses of 360, artifacts. 720, and 1080 mg MK-7 thrice weekly resulted in a dose- proportional decrease of the dp-ucMGP levels (17%, 33%, PWV and 46%, respectively), without achieving a plateau phase.23 Hemodynamic measurements were performed in the supine For this study, the dose of MK-7 was therefore increased to position by trained research nurses during the first hour of 2000 mg thrice weekly. the midweek dialysis session. Mean arterial pressure (MAP) Clinical data, biochemical data, imaging data, and PWV was calculated as diastolic BP+0.43(systolic BP–diastolic BP), data were collected at baseline, 6, 12, and 18 months. Subjective on the basis of the mean of two measurements. PWV was tolerability was evaluated by questioning the patients about obtained by sequential recording of electrocardiogram-gated adverse eventsor byspontaneous reporting of adverse events by carotid and femoral artery pressure waves using applanation the patients. Objective tolerability was evaluated by moni- tonometry (SphygmoCor version 7; AtCor Medical, Sydney, toring vital signs and routine clinical laboratory tests. Primary Australia), as described previously.26 The path length was cal- outcome measures were change of coronary artery calcifica- culated as 0.8 times the direct distance between the carotid and tion, thoracic aorta calcification, and PWV over 18 months femoral recording sites. PWV was calculated as the path length versus baseline. Secondary outcome measures are listed in the divided by transit time (m/s). Supplemental Material. Statistical Methods Biochemical Analysis Sample size calculation, randomization method, and blinding Blood samples were taken at the start of dialysis at the dialyzer considerations are described in the Supplemental Material. inlet line. Plasma was obtained by 15 minutes of centrifugation Descriptive statistics used were proportions, means, SDs, me- with a Rotina 38R and stored at 280°C in three aliquots of 1 ml dians, and interquartile ranges (IQRs). Baseline and end-of- until dry-ice shipment for central analysis. Circulating dp- study characteristics at 18 months were compared between the ucMGP was quantified using an automated assay (Ina K tif study arms according to Fisher’s exact test for categoric vari- MGP iSYS kit; IDS, Boldon, UK). ables and the Kruskal–Wallis test for continuous variables. Associations between baseline dp-ucMGP levels and PWV, Imaging calcification scores, and dialysis and warfarin vintages were For calcification assessment, an unenhanced electrocardio- evaluated using Spearman rho coefficients, partialized for graphically gated computed tomography of the heart and tho- age, sex, body mass index, systolic BP, presence of diabetes, racic aorta was performed at 120 kV on a Revolution parathyroid hormone, calcium, phosphate, cardiovascular (GE Healthcare), Aquillion One (Toshiba), or Somatom Def- history, and statin use. Percentage changes in calcification inition Flash (Siemens) scanner. Tolimit artifacts owing to high scores were calculated as the change in calcification scores heart rates (.70 Bpm), b-blockade (bisoprolol 2.5–5mg)was from baseline divided by the baseline value+1, the latter to administered orally 60 minutes before scanning when neces- account for zero values at baseline. Percentage changes at 18 sary. Calcium scores were calculated on 2.5-mm slices using months were annualized by dividing them by a factor of 1.5. Smartscore v.4.0 (GE Healthcare) by the Agatston method24 Proportions of “rapid progressors,” i.e.,patientswithanan- and by the volume method.25 All foci within the arteries with nualized percentage change of $15%, were compared be- attenuation .130 Hounsfield units (Hu) and a minimum area tween treatment groups according to Fisher’s exact test. To of 1 mm2 were considered significant and were counted into account for clustering of multiple observations over time the total score. An Agatston score for each calcific lesion was within patients, changes from baseline were analyzed accord- calculated by multiplying the density factor (1, 130–199 Hu; 2, ing to mixed linear regression modeling for continuous data 200–299 Hu; 3, 300–399 Hu; and 4, $400 Hu) by the area. A (dp-ucMGP levels, other laboratory parameters, calcification total Agatston score was obtained by adding the scores of all scores, PWV, MAP) and mixed logistic regression modeling individual lesions in all slice levels. The total volume score was for binary data (use of cinacalcet, sevelamer, calcium contain- obtained by adding the volumes of all .130 Hu lesions. One ing binders, and alfacalcidol). In these models, a random

JASN 31: ccc–ccc,2020 Vitamin K Status in Hemodialysis 3 CLINICAL RESEARCH www.jasn.org

A B 3000 3000 VKA Rivaroxaban 2000 Rivaroxaban + vit K2 2000

1000 1000

0 0

-1000 -1000

-2000 -2000 VKA dp-ucMGP (pmol/L) change from baseline dp-ucMGP (pmol/L) change from baseline Rivaroxaban Rivaroxaban + vit K2 -3000 -3000 0 6 12 18 061218 Follow-up time (months) Follow-up time (months)

Figure 1. Changes in dp-ucMGP levels over time in the VKA (beige squares), rivaroxaban (blue triangles), and rivaroxaban1vitamin K2 (red circles) groups. (A) Estimated marginal mean changes from baseline (95% confidence interval) in the entire study population (P,0.001). (B) Estimated marginal mean changes from baseline (95% confidence interval) in the patients with a warfarin vintage of 0(n534) (P,0.01). intercept at the patient level was used, time as a fixed factor, had a median dialysis vintage of 2.4 years with 16.7% patients and an unstructured covariance matrix. Differences in receiving incident dialysis, a median CHA2DS2-VASc score of changes over time across treatment groups were analyzed by 5.0 with a 30.3% history of stroke, a median warfarin vintage evaluating the treatment-by-time interaction in these mixed- of 1.1 years including 34.8% patients with a warfarin vintage effects models. To deal with the considerable skewness in ,3 months and 25.8% patients who were warfarin-naïve, both calcification scores and in their changes, Agatston and and a median HAS-BLED score of 5.0 with a 28.0% history of volume were entered in these models after square root gastrointestinal bleeding. Baseline demographic and clinical transformation in line with previous studies.27 Goodness- characteristics (Table 1), baseline biochemical characteristics of-fit statistics for all models demonstrated acceptable fitto (Supplemental Table 1), and baseline maintenance medication the data. Estimated marginal mean changes from baseline (Supplemental Table 2) were not different between the groups. and their 95% confidence intervals at 6, 12, and 18 months, Baseline calcification scores (Table 2) and baseline hemody- as depicted in Figures 1, A and B and 2, A–D, were obtained namic parameters and PWV (Table 3) were not significantly from the mixed models. Survival times, as well as times to different across the treatment arms. the first bleeding episode occurring during the total obser- The 132 patients represented 163.5 person-years of ob- vation period of 18 months, were compared between the servation over the course of the study. Follow-up was in- treatment arms using Cox proportional hazards models. complete in 55 patients, due to death (n=47) or withdrawal The total number of bleeding events across patients was of consent (n=8). In 14 patients, the study drug (VKA or compared using Poisson regression analysis. We rivaroxaban) was discontinued, in all but one case (pa- considered a two-sided P value ,0.05 to indicate statistical tient unwilling) as a consequence of a bleeding event. In significance. All analyses were on the basis of the intention- the rivaroxaban+vitamin K2 arm, only rivaroxaban was to-treat principle. withdrawn and vitamin K2 supplements were continued (Supplemental Figure 2). Median (IQR) dp-ucMGP at baseline was 1983 (1486–3087) RESULTS pmol/L in the VKA arm, 1632 (1083–2390) pmol/L in the rivaroxaban arm, and 1598 (1058–3324) pmol/L in the rivarox- Participants aban+vitamin K2 arm (P=0.12). Baseline dp-ucMGP levels Between February of 2015 and July of 2017, 143 patients on were strongly correlated with warfarin vintage (partial Spear- hemodialysis with documented AF were evaluated and 132 man rho, +0.44, P,0.001) after adjustment for age, sex, body enrolled (Supplemental Figure 2). The overall study popula- mass index, systolic BP, presence of diabetes, parathyroid hor- tion was almost exclusively of western European ancestry, and mone, calcium, phosphate, cardiovascular history, and statin

4 JASN JASN 31: ccc–ccc,2020 www.jasn.org CLINICAL RESEARCH

A B 10 5 VKA VKA Rivaroxaban Rivaroxaban 8 Rivaroxaban + vit K2 4 Rivaroxaban + vit K2

6 3

4 2 change from baseline change from baseline 2 1 sqrt(Total coronary artery volume score)

sqrt(Total coronary artery Agatston score) 0 0 0 6 12 18 0 6 12 18 Follow-up time (months) Follow-up time (months)

C D 20 10 VKA VKA Rivaroxaban Rivaroxaban 16 Rivaroxaban + vit K2 8 Rivaroxaban + vit K2

12 6

8 4 change from baseline change from baseline 4 2 sqrt(Thoracic aorta volume score) sqrt(Thoracic aorta Agatston score) 0 0 0 6 12 18 0 6 12 18 Follow-up time (months) Follow-up time (months)

Figure 2. Changes in coronary artery and thoracic aorta calcium scores over time in the VKA (beige squares), rivaroxaban (blue triangles), and rivaroxaban1vitamin K2 (red circles) groups. (A) Estimated marginal mean changes in total coronary artery Agatston scores from baseline (95% confidence interval) (P50.364). (B) Estimated marginal mean changes in total coronary artery volume scores from baseline (95% confidence interval) (P50.616). (C) Estimated marginal mean changes in thoracic aorta Agatston scores from baseline (95% confidence interval) (P50.210). (D) Estimated marginal mean changes in thoracic aorta volume scores from baseline (95% confidence interval) (P50.707). Sqrt, square root. use. No significant associations between baseline dp-ucMGP levels were significantly larger when vitamin K2 was added and baseline calcification scores, baseline PWV (partial Spear- to rivaroxaban (P=0.004). At 18 months, median (IQR) dp- man rho, 20.19, P=0.17), or dialysis vintage (partial Spearman ucMGP was 2967 (1982–4737) pmol/L in the VKA group, rho, 20.09, P=0.32) were observed. 981 (729–1453) pmol/L in the rivaroxaban group, and 853 Mixed modeling demonstrated that the change in dp- (707–1176) pmol/L in the rivaroxaban+vitamin K2 group ucMGP levels over time was significantly different across (P,0.001). Drops in dp-ucMGP levels induced by vitamin treatment arms (P,0.001) (Figure 1A). Dp-ucMGP levels K2 were not associated with diabetes status (P=0.48) or with increased significantly in the VKA arm (P=0.03), whereas the use of phosphate binders (P=0.42). In the subgroup of levels decreased significantlyintherivaroxabanarm patients who were warfarin-naïve (n=34, 25.8%), the (P=0.04) and the rivaroxaban+vitamin K2 arm (P=0.04). change in dp-ucMGP levels over time was also signifi- In the rivaroxaban and rivaroxaban+vitamin K2 groups, cantly different across treatment arms (P,0.01), with an dp-ucMGP levels dropped significantly after 6 months increase in the VKA group, stable levels in the rivaroxaban and remained stable from then on throughout the entire group, and a decrease in the rivaroxaban+vitamin K2 observation period (P=0.65). Decreases in dp-ucMGP group (Figure 1B).

JASN 31: ccc–ccc,2020 Vitamin K Status in Hemodialysis 5 CLINICAL RESEARCH www.jasn.org

Table 1. Baseline demographic and clinical characteristics Baseline Characteristic VKA (n=44) Rivaroxaban (n=46) Rivaroxaban+Vit K2 (n=42) Pa Age, yr 80.3 (71.5–84.3) 79.9 (74.4–83.9) 79.6 (73.2–83.1) 0.92 Men, % (no.) 56.8 (25 of 44) 76.1 (35 of 46) 66.7 (28 of 42) 0.16 Body mass index, kg/m2 25.6 (22.3–30.4) 24.7 (22.0–27.5) 24.4 (21.8–29.0) 0.51 History of stroke, % (no.) 36.4 (16 of 44) 32.6 (15 of 46) 21.4 (9 of 42) 0.30 History of acute myocardial infarction, % (no.) 47.7 (21 of 44) 45.7 (21 of 46) 45.2 (19 of 42) 0.98 History of gastrointestinal bleeding, % (no.) 27.3 (12 of 44) 19.6 (9 of 46) 38.1 (16 of 42) 0.16 Diabetes, % (no.) 45.5 (20 of 44) 43.5 (20 of 46) 52.4 (22 of 42) 0.74 Congestive heart failure, % (no.) 20.5 (9 of 44) 37.0 (17 of 46) 35.7 (15 of 42) 0.18 Preexisting vascular disease, % (no.) 63.6 (28 of 44) 54.3 (25 of 46) 40.5 (17 of 42) 0.10 Dialysis vintage, yr 1.8 (0.4–5.5) 2.7 (0.9–5.9) 2.7 (0.6–5.1) 0.64 Incident dialysis (,3 mo), % (no.) 20.5 (9 of 44) 8.7 (4 of 46) 21.4 (9 of 42) 0.18 Kt/V 1.77 (1.53–2.06) 1.68 (1.54–1.91) 1.72 (1.51–1.89) 0.68 Permanent AF, % (no.) 51.2 (21 of 41) 56.5 (26 of 46) 59.5 (25 of 42) 0.75 b CHAD2D2-VASc score Mean (SD) 4.8 (1.5) 4.7 (1.4) 4.5 (1.4) Median (IQR) 5 (4–6) 5 (4–5) 4 (4–5) 0.52 HAS-BLED scorec Mean (SD) 4.7 (0.9) 4.6 (0.8) 4.7 (0.9) Median (IQR) 5 (4–5) 4 (4–5) 5 (4–5) 0.61 VKA vintage, yr 0.9 (0.1–4.7) 1.1 (0.0–2.8) 1.5 (0.0–6.3) 0.46 VKA naïve (,3 mo), % (no.) 31.8% (14 of 44) 37.8% (17 of 45) 35.7% (15 of 42) 0.87 Pacemaker, % (no.) 18.2% (8 of 44) 23.9% (11 of 46) 9.5% (4 of 42) 0.21 Numbers displayed are median (IQR) unless otherwise speciﬁed. Vit K2, vitamin K2. aAccording to Fisher’s exact test or the Kruskal–Wallis test. b The CHAD2D2-VASc score ranges from 2 (minimum score to be included) to 9, with higher scores indicating an increased risk. cThe HAS-BLED score ranges from 1 (because patients by deﬁnition have renal failure) to 9, with higher scores indicating an increased risk.

Over the course of the 18 months of observation, changes Mixed modeling revealed that the change in PWVover time in calcium, phosphate, parathyroid hormone levels, and was not significantly different across treatment arms (P=0.56). HgA1c (in patients with diabetes); in MAP; and in use of When MAP was included as a time-varying covariate in the phosphate binders, cinacalcet, and active vitamin D were not model, the change in PWVover time was also not significantly significantly different between the study groups (Supple- different across treatment arms (P=0.12). mental Table 3). Secondary Efficacy Outcomes Primary Outcomes The rates of all cause deathwere 36.0/100 person-years in the Mixed model analyses revealed that longitudinal changes in cal- VKAgroup,26.0/100 person-years in the rivaroxaban group, cification were not significantly different between the treatment and 24.6/100 person-years in the rivaroxaban+vitamin groups (total coronary arteries Agatston score P=0.36, total cor- K2 group (Supplemental Figure 3; P=0.47). The main cau- onary arteries volume score P=0.62, thoracic aorta Agatston ses of death were withdrawal of dialysis, infectious disease, score P=0.21, thoracic aorta volume score P=0.71; Figure 2, and sudden death (Supplemental Table 4). An ischemic or A–D). The percentage changes from baseline values over 18 uncertain type of stroke occurred in eight of the 132 pa- months in Agatston calcification scores and calcium volume tients, corresponding with a stroke rate of 4.89/100 person- scores were not significantly different between the treatment years. A hemorrhagic stroke was diagnosed in two of the groups (Tables 4 and 5). After 18 months, the proportion of 132 patients, corresponding with a stroke rate of 1.22/100 patients with an annualized percentage change in Agatston cal- person-years. The number of strokes did not differ between cification scores of $15% was not different between the VKA, the treatment groups, although it is noticeable that both rivaroxaban, and rivaroxaban+vitamin K2 arms: 50.0%, 47.4%, hemorrhagic strokes occurred in the VKA group (Table and 40.0% (P=0.87) for the sum of the coronary arteries and 6). Mean (SD) CHA2DS2-VASc score was 5.13 (1.13) in 50%, 44.4%, and 50.0% (P=0.91) for the thoracic aorta, respec- the patients that developed an ischemic stroke versus 4.67 tively. Similarly, the proportion of patients with an annualized (1.44) in those that did not (P=0.26). A history of stroke was percentage change in volume scores of $15% was not different recorded in 62.5% (five of eight) of patients with an ische- between the VKA, rivaroxaban, and rivaroxaban+vitamin K2 mic stroke and in 28.2% (35 of 124) of those without a arms: 50.0%, 38.9%, and 50.0% (P=0.75) for the sum of the stroke during the study period (P=0.06). An exploratory coronary arteries and 50.0%, 38.5%, and 46.2% (P=0.76) for the analysis of stroke and systemic embolism events in the thoracic aorta, respectively. pooled rivaroxaban arms versus the VKA arm revealed no

6 JASN JASN 31: ccc–ccc,2020 www.jasn.org CLINICAL RESEARCH

ﬁ a signi cant difference between the groups (Supplemental P 0.45 0.50 Table 5).

=42) Adverse Events n No adverse events related to the intake of vitamin K2 ) 0.40 4541) 0.61 – 993) 0.54 thrice weekly after dialysis were reported. Thirty-six life- – 387) 0.84 430 490) 0.63 190) 0.36 – – – – 156) 186) – – threatening or major bleeding episodes occurred in the 132

3 patients, corresponding with a rate of 22.0/100 person- 90 (3 60 (14 37 (0 139 (37 237 (54 141 (18 years. Given multiplicity of bleeding episodes within patients, which may be more related to underlying patient characteristics (e.g., angiodysplasia of the bowel) than to the type of anticoagulant, bleeding rates were analyzed both

=46) Rivaroxaban+Vit K2 ( as times to ﬁrst occurrence of bleeding episode and as total n 6244) 2834 (1478 – 487) 1199) 548 (108 ﬁ – – 660) 233) 377) 145) number of bleeding episodes. No statistically signi cant – – – – 496)

– differences in the bleeding outcomes were found, except for the total number of combined life-threatening and major bleeding episodes being lower in both rivaroxaban arms as compared with the VKA arm (Table 7). An exploratory analysis of the bleeding rates in the pooled rivaroxaban

5432) 2930 (1352 ﬁ – 419) 200 (119 1215) 647 (195 446) 175 (43 arms versus the VKA arm revealed signi cantly fewer major =44) Rivaroxaban ( – – – 188) 101 (38 569) 174 (59 210) 83 (20 n – – – 243) 97 (7 bleedings as well as combined life-threatening and major – bleedings in the pooled rivaroxaban arms than in the VKA VKA ( arm (Supplemental Table 6). a P 0.52 117 (49 0.08 121 (61 0.70 103 (8 DISCUSSION =42) ex artery; RCA, right coronary artery. n ﬁ

fl This multicenter randomized, controlled trial is the rst to report the effects of vitamin K status on the progression of 14,019) 0.61 3636 (1672 – 1170) 0.80 264 (101 2755) 0.57 632 (249 VC in patients on chronic hemodialysis. We did not find a – – 1303) 0.66 161 (39 1028) 0.80 239 (111 – – 443) 489) 583) – – – differenceintheprogressionofVCoverthecourseof18 months among patients treated with VKA, patients in whom 72 (8 81 (0 207 (3 268 (11 339 (40 the VKAwas replaced by rivaroxaban, and patients treated with rivaroxaban that additionally received high-dose vitamin K2 supplements, despite significant differences in dp-ucMGP lev- cation Score Calcium Volume Score, mm

ﬁ els, considered to be the most accurate marker of vascular =46) Rivaroxaban+Vit K2 (

n vitamin K status. 22,185) 7864 (4135 – 1349) 675 (116 3078) 1465 (155 – – 588) 1002) 346) 2276) The conviction that correction of functional vitamin K de- – – – – 1595) – ﬁciency may retard the development of VC has led to the initiation of several randomized, controlled trials of vitamin Agatston Calci K supplements in prevalent or incident patients in hemodialysis,11 including this study. However, this conviction rests heavily on the assumption that vitamin K deﬁciency can be

18,193) 8991 (4165 corrected in patients on hemodialysis. Unfortunately, the re- – 1284) 664 (240 3000) 1611 (427 1391) 415 (71 =44) Rivaroxaban ( – – – 439) 260 (39 1226) 356 (76 609) 116 (22 816) 154 (4 n – – – – sults of our study reveal that withdrawal of VKAs and long- term vitamin K2 administration at pharmacologic doses does cation scores VKA (

ﬁ not normalize systemic dp-ucMGP levels. Initiation or con- 763 (211 263 (72 435 (53 264 (89 277 (17 689 (222

Wallis test. tinuation of VKAs led to a steady rise in dp-ucMGP concen- – trations. In the rivaroxaban group, dp-ucMGP decreased as a consequence of the wash-out of the VKA, but rivaroxaban had no intrinsic effect on dp-ucMGP levels, as revealed in Baseline calci the subgroup of patients that were VKA-naïve. Although an additional drop in dp-ucMGP was observed with vitamin K2 supplements, the median levels remained between 800 and According to the Kruskal Anatomic Location Table 2. LAD Numbers displayed area median (IQR). Vit K2, vitamin K2; LAD, left anterior descending; LCX, left circum LCX RCA Total coronary artery 1816 (461 Thoracic aorta 11,566 (5722 Aortic valve Mitral valve Total valves 900 pmol/L, which is still about two-fold the value observed

JASN 31: ccc–ccc,2020 Vitamin K Status in Hemodialysis 7 CLINICAL RESEARCH www.jasn.org

Table 3. Baseline hemodynamic parameters and PWV transdifferentiate into osteoblast-like cells dur- Rivaroxaban Rivaroxaban+Vit ing the course of VC. It is tempting to specu- Parameter VKA (n=44) Pa (n=46) K2 (n=42) late that the transdifferentiation process SBP, mm Hg 133 (116–153) 122 (112–145) 127 (119–146) 0.64 adversely affects the synthesis of active MGP. DBP, mm Hg 65 (56–71) 62 (53–68) 60 (52–70) 0.45 In patients on hemodialysis, the process MAP, mm Hg 91 (80–101) 86 (78–99) 88 (78–97) 0.58 of VC is governed by a mosaic of factors, Heart rate, bpm 69 (61–77) 64 (60–71) 67 (60–80) 0.41 including disruption of calcium and phos- PWV, m/s 12.7 (10.4–14.2) 14.6 (12.1–16.1) 13.1 (10.6–15.6) 0.19 phate metabolism and imbalance between Numbers displayed are median (IQR). Vit K2, vitamin K2; SBP, systolic blood pressure; DBP, diastolic calcification promotors and inhibitors.31 blood pressure. Vitamin K deficiency thus represents only aAccording to the Kruskal–Wallis test. one of many pathways by which VC is accelerated. So far, single therapeutic inter- in healthy volunteers of the same age group that do not receive ventions aiming at VC progression, including more frequent supplements.28,29 and extended hemodialysis, cinacalcet, phosphate binders, The question of why dp-ucMGP cannot be normalized in and cholecalciferol, have had limited success in the dialysis patients on hemodialysis is intriguing. Inactive uncarboxy- population.31 Recently, magnesium oxide supplementation latedMGPisconvertedtoactivecarboxylatedMGPbythe was found to retard the progression of calcification in the g-carboxylase enzyme that operates in the endoplasmic re- coronary arteries but not in the thoracic aorta of predialysis ticulum of vascular smooth muscle cells and requires vita- patients with CKD.17 Perhaps treatment should be directed at min K as an essential cofactor. High levels of the inactive multiple targets simultaneously; although a number of bio- dp-ucMGP can be taken as a prima facie confirmation of logic processes such as oxidative stress and chronic inflamma- vitamin K deficiency. Multiple issues contribute to vitamin tion currently remain beyond our therapeutic reach. As a final Kdeficiency in patients on dialysis.11 However, long-term point, once VCs are advanced past a certain stage, they may no high-dose supplementation can be expected to overrule longer be susceptible to reversal and opportunities to inter- most of these, including the effects of deficient dietary in- vene may have been missed. take, potential exhaustion of stores by high demands from Our study is the first randomized trial to report on the long- the procalcific uremic environment, and abnormalities of term use of DOACs to reduce thromboembolic risk in patients the gut microbiome. In addition, there was no demonstra- on hemodialysis. The overall stroke risk in our study was 4.89/ ble interference by phosphate binder intake in our study. 100 person-years, taking into account that patients had a me- Compliance with the supplements was ascertained by ad- dian CHA2DS2-VASc score of 5 and a 30% history of stroke. ministration after each dialysis session under supervision For comparison, stroke risk was 7.8/100 person-years in a of a dialysis nurse. Finally, vitamin K removal by dialysis can Taiwanese hemodialysis population with a CHA2DS2-VASc be expected to be negligible due to its lipophilic nature. An score of 5 that did not receive oral anticoagulation.32 A alternative potential explanation for the failure to normal- meta-analysis of 13 studies reported a stroke rate of 5.2/100 ize dp-ucMGP is that the g-carboxylase enzyme itself is de- person-years in patients on hemodialysis with AF, but details fective in uremia, even when there is an abundant supply of on anticoagulation coverage and CHA2DS2-VASc scores were its cofactor vitamin K. Indeed, in experimental uremia the not available.33 The incidence of ischemic or hemorrhagic activity of the g-carboxylase was impaired whereas gene stroke did not differ significantly between the rivaroxaban expression was normal.30 Finally, MGP is produced and se- and VKA groups. However, it is worth noting that hemor- creted by vascular smooth muscle cells that are known to rhagic strokes only occurred in the VKA group.

Table 4. Percentage changes of Agatston scores Anatomic Location Mo VKA (n=44) Rivaroxaban (n=46) Rivaroxaban+Vit K2 (n=42) Pa Total coronary artery 6 27.5% (4.0%–47.6%) 7.1% (0.6%–19.8%) 7.7% (20.4%–38.0%) 0.43 12 18.7% (8.1%–92.0%) 12.1% (4.3%–29.1%) 11.3% (1.9%–63.0%) 0.47 18 29.4% (17.2%–95.9%) 19.8% (7.2%–37.9%) 18.6% (7.2%–110.0%) 0.73 Thoracic aorta 6 6.3% (2.5%–14.6%) 6.5% (2.8%–19.0%) 6.6% (2.9%–20.2%) 0.94 12 13.5% (3.5%–42.4%) 11.6% (3.5%–22.0%) 12.6% (4.6%–29.1%) 0.96 18 19.9% (4.7%–56.9%) 18.7% (4.6%–48.8%) 25.6% (8.2%–61.4%) 0.79 Aortic & mitral valves 6 14.2% (2.5%–24.1%) 23.0% (4.9%–51.1%) 9.1% (0.0%–21.6%) 0.21 12 30.8% (7.8%–63.6%) 19.5% (0.0%–52.1%) 23.0% (0.0%–47.1%) 0.99 18 57.1% (23.0%–98.6%) 33.4% (5.8%–84.2%) 36.3% (3.1%–132.6%) 0.81 Numbers displayed are median (IQR). Vit K2, vitamin K2. aAccording to the Kruskal–Wallis test.

8 JASN JASN 31: ccc–ccc,2020 www.jasn.org CLINICAL RESEARCH

Table 5. Percentage changes of volume scores Anatomic Location Mo VKA (n=44) Rivaroxaban (n=46) Rivaroxaban+Vit K2 (n=42) Pa Total coronary artery 6 18.8% (5.9%–38.2%) 8.4% (1.0%–17.0%) 13.0% (3.6%–28.4%) 0.37 12 20.5% (6.6%–51.9%) 7.8% (2.3%–25.8%) 10.3% (2.5%–30.3%) 0.40 18 25.1% (10.6%–44.7%) 14.9% (2.7%–34.0%) 29.3% (9.7%–56.0%) 0.43 Thoracic aorta 6 6.0% (3.9%–12.6%) 8.3% (2.2%–18.3%) 6.8% (1.0%–17.8%) 0.81 12 13.8% (6.1%–37.5%) 12.2% (3.2%–24.9%) 8.9% (2.9%–26.8%) 0.97 18 24.9% (8.8%–46.6%) 15.6% (5.1%–35.0%) 19.5% (7.3%–56.0%) 0.62 Aortic & mitral valves 6 20.8% (26.8%–44.0%) 1.1% (29.1%–39.0%) 9.5% (23.5%–15.0%) 0.10 12 47.9% (5.2%–121.1%) 12.2% (23.3%–37.3%) 14.4% (0.0%–58.8%) 0.52 18 43.9% (15.6%–77.1%) 15.4% (25.6%–44.3%) 37.7% (12.1%–109.9%) 0.33 Numbers displayed are median (IQR). Vit K2, vitamin K2. aAccording to the Kruskal–Wallis test.

In our study, the incidence of life-threatening or major The main limitation of our study is the relatively small bleeding was 22/100 person-years, with patients having a me- sample size. Because it is the first clinical study to address dian HAS-BLED score of 5and a28% historyof gastrointestinal the effects of VKAs and vitamin K supplements on the pro- bleeding. In the Dialysis Outcomes and Practice Patterns Study, gression of VC in patients on dialysis, projections of effect size major bleeding rates were 7.8/100 person-years in patients on were on the basis of preliminary evidence and in hindsight oral anticoagulation and 20/100 person-years in patients with a appeared to have been too optimistic. Also, between-patient history of gastrointestinal bleeding.34 The results of our study variability in baseline calcification scores and progression of reveal that severe bleeding complications may occur less fre- VC was larger than anticipated. Therefore, our study has lim- quently with rivaroxaban than with VKA. ited ability to exclude a type 2 error. Another limitation is that Our study was not designed or powered to address the com- the population consisted mainly of prevalent patients receiving parative benefitsofDOACsversusVKAswithrespecttostroke dialysis, with a high burden of cardiovascular disease prevention and bleeding complications. However, because our as revealed by the elevated baseline calcification score and data on stroke and bleeding are the first to be generated in a PWV, of whom the majority was taking a VKA at inclusion. randomized, controlled trial setting, they may serve as a pilot As such, vascular damage may have progressed beyond a to inform the design of larger, definitive trials to determine the point of no return and any intervention at this stage may prove optimal anticoagulation strategy in patients on hemodialysis with to be futile. Conversely, our study population consists of an AF.Pending further evidence, our results suggest that rivaroxaban unselected real-life cohort of patients on dialysis from three 10 mg once daily can be used safely and effectively in patients on large European dialysis units. Our results therefore have direct hemodialysis. implications for clinical practice.

Table 6. Secondary efﬁcacy outcomes Outcome Parameter VKA (n=44) Rivaroxaban (n=46) Rivaroxaban+Vit K2 (n=42) Pa Death from any cause, no. (%) 19 (43.2) 15 (32.6) 13 (31.0) 0.46 Sudden death, no. 3 4 2 0.91 Stroke or systemic embolism, no. Ischemic or uncertain type of stroke 5 2 1 0.22 Hemorrhagic stroke or intracerebral bleeding 2 0 0 0.21 Systemic embolism 0 0 0 – Cardiac disease, no. Acute coronary syndrome 2 6 2 0.32 Symptom-driven revascularizationb 2 5 1 0.28 Hospitalization for heart failure 3 0 0 0.07 Symptomatic aortic valve stenosis 1 0 0 0.65 Death from cardiac cause 2 2 1 .0.99 Other vascular disease, no. Symptomatic lower limb ischemia 12 6 7 0.23 Calciphylaxis 1 1 0 .0.99 Bowel ischemia 1 0 0 0.65 Vit K2, vitamin K2; –, not applicable. aAccording to Fisher’s exact test. bIncluding acute coronary syndrome.

JASN 31: ccc–ccc,2020 Vitamin K Status in Hemodialysis 9 CLINICAL RESEARCH www.jasn.org

Table 7. Bleeding outcomes

Outcome Parameter VKA (n=44) Rivaroxaban (n=46) Rivaroxaban+Vit K2 (n=42) PCox PPoisson Life-threatening bleeding 6/7/13.3 2/2/3.5 4/5/9.5 0.66 0.25 Major bleeding 7/12/22.7 6/8/13.8 2/2/3.8 0.16 0.06 Life-threatening or major bleeding 13/19/36.0 7/10/17.3 5/7/13.2 0.09 0.04 Minor bleeding 10/14/26.5 14/18/31.1 14/16/30.3 0.68 0.90 Gastrointestinal bleeding 8/12/22.7 7/10/17.3 9/13/24.6 0.62 0.69 Cell entries are number of patients with at least one bleeding episode/total number of bleeding episodes/total number of bleeding episodes per 100 person-yr;

PCox,significance of differences in time to first bleeding episode according to Cox’s proportional hazard model analysis; PPoisson,significance of differences in total number of bleeding episodes according to Poisson model analysis. Life-threatening bleeding was defined as fatal bleeding, symptomatic intracranial bleeding, a decrease in hemoglobin of 5 g/dl or more, or a requirement for transfusion of four or more units of blood, inotropic agents, or surgery. Major bleeding was defined as a requirement for transfusion of two or more units of blood or a decrease in hemoglobin of 2 g/dl or more, and not fulfilling the criteria for life-threatening bleeding. All other bleedings were regarded as minor. Vit K2, vitamin K2.

In conclusion, we did not find an effect of vitamin K status on SUPPLEMENTAL MATERIAL VC progression in patients receiving chronic hemodialysis withAF. TheresultsofourstudyhighlightthecomplexityoftheVCprocess, This article contains the following supplemental material online which apparently cannot be redeemed by a single intervention. at http://jasn.asnjournals.org/lookup/suppl/doi:10.1681/ASN.2019060579/-/ Further research should be directed at optimizing vitamin K status, DCSupplemental. combining multiple interventions, and identifying a window of Patient inclusion and exclusion criteria. therapeutic opportunity perhaps long before stigmata of VC be- Secondary end points. come detectable with standard techniques. Sample size calculation. Randomization and blinding. Supplemental Table 1. Baseline biochemical characteristics. Supplemental Table 2. Baseline maintenance medication. ACKNOWLEDGMENTS Supplemental Table 3. Characteristics during follow-up. Supplemental Table 4. Causes of death. The authors are indebted to Bart Beauprez, David Bosman, Kaatje Supplemental Table 5. Stroke or systemic embolism in the VKA Bruggeman, Mieke Debou, Mirjam Demesmaecker, Xandra Smeulders, versus pooled rivaroxaban arms. Antje Straeten, Karel Van Hese, and Gwendolyn Verbeerst for their in- Supplemental Table 6.Bleeding outcomesin theVKAversuspooled valuable help in the collection of the patient data. rivaroxaban arms. KaydencePharma and Nattopharmahad no roleinthedesignofthe Supplemental Figure 1. Study design. trial; collection, analysis, and interpretation of the data; or in the Supplemental Figure 2. CONSORT diagram. submission of the results. Supplemental Figure 3. Kaplan–Meier curves for patient survival. Dr. Caluwé, Dr. De Vriese, and Dr. Verbeke designed the study. Dr. Caluwé, Dr. De Boeck, Dr. Delanote, Dr. De Surgeloose, Dr. De Vriese, Dr. Van Hoenacker, and Dr. Van Vlem collected the data. Dr. REFERENCES Pyfferoen read the computed tomography scans, Dr. Verbeke analyzed the PWV curves, Dr. De Bacquer analyzed the data, and Dr. De 1. De Vriese AS, Caluwé R, Raggi P: The atrial fibrillation conundrum in Vriese drafted the manuscript. All authors revised the manuscript and dialysis patients. Am Heart J 174: 111–119, 2016 approved the final version of the manuscript. 2. Van Der Meersch H, De Bacquer D, De Vriese AS: Vitamin K antagonists for stroke prevention in hemodialysis patients with atrial fibrillation: A systematic review and meta-analysis. Am Heart J 184: 37–46, 2017 DISCLOSURES 3. Chatrou ML, Winckers K, Hackeng TM, Reutelingsperger CP, Schurgers LJ: Vascular calcification: The price to pay for anticoagulation therapy Dr. Caluwé reports personal fees from Bayer, outside the submitted work. Dr. De with vitamin K-antagonists. Blood Rev 26: 155–166, 2012 Vriese reports personal fees from Ablynx, Achillion, Alexion, Amgen and Baxter 4. Caluwé R, Pyfferoen L, De Boeck K, De Vriese AS: The effects of vitamin and grants from Amgen, outside the submitted work. Dr. Van Vlem reports grants K supplementation and vitamin K antagonists on progression of vas- from Baxter, grants from Fresenius Medical Care, and grants from Amgen, outside cular calcification: Ongoing randomized controlled trials. Clin Kidney J the submitted work. All remaining authors have nothing to disclose. 9: 273–279, 2016 5. January CT, Wann LS, Alpert JS, Calkins H, Cigarroa JE, Cleveland JC Jr., et al.: American College of Cardiology/Ameri- can Heart Association Task Force on Practice Guidelines: 2014 FUNDING AHA/ACC/HRS guideline for the management of patients with atrial fibrillation: A report of the American College of Cardiology/ The conduction of this trial was financially supported by Kaydence Pharma American Heart Association Task Force on Practice Guidelines (New Brunswick, NJ, USA). Nattopharma (Lysaker, Norway) provided the and the Heart Rhythm Society. JAmCollCardiol64: e1–e76, MK-7 study medication. 2014

10 JASN JASN 31: ccc–ccc,2020 www.jasn.org CLINICAL RESEARCH

6. Potpara TS, Ferro CJ, Lip GYH: Use of oral anticoagulants in patients 21. Schurgers LJ, Teunissen KJ, Hamulyák K, Knapen MH, Vik H, Vermeer with atrial fibrillation and renal dysfunction. Nat Rev Nephrol 14: 337– C: Vitamin K-containing dietary supplements: Comparison of synthetic 351, 2018 vitamin K1 and natto-derived menaquinone-7. Blood 109: 3279–3283, 7. Königsbrügge O, Posch F, Antlanger M, Kovarik J, Klauser-Braun R, 2007 Kletzmayr J, et al.: Prevalence of atrial fibrillation and antithrombotic 22. Marles RJ, Roe AL, Oketch-Rabah HA: US Pharmacopeial Convention therapy in hemodialysis patients: Cross-sectional results of the Vienna safety evaluation of menaquinone-7, a form of vitamin K. Nutr Rev 75: InVestigation of AtriaL Fibrillation and Thromboembolism in Patients 553–578, 2017 on HemoDIalysis (VIVALDI). PLoS One 12: e0169400, 2017 23. Caluwé R, Vandecasteele S, Van Vlem B, Vermeer C, De Vriese AS: 8. Schurgers LJ, Spronk HM: Differential cellular effects of old and new Vitamin K2 supplementation in haemodialysis patients: A random- oral anticoagulants: Consequences to the genesis and progression of ized dose-finding study. Nephrol Dial Transplant 29: 1385–1390, atherosclerosis. Thromb Haemost 112: 909–917, 2014 2014 9. Hara T, Fukuda D, Tanaka K, Higashikuni Y, Hirata Y, Nishimoto S, et al.: 24. Agatston AS, Janowitz WR, Hildner FJ, Zusmer NR, Viamonte M Jr., Rivaroxaban, a novel oral anticoagulant, attenuates atherosclerotic Detrano R: Quantification of coronary artery calcium using ultrafast plaque progression and destabilization in ApoE-deficient mice. Ath- computed tomography. JAmCollCardiol15: 827–832, 1990 erosclerosis 242: 639–646, 2015 25. Callister TQ, Cooil B, Raya SP, Lippolis NJ, Russo DJ, Raggi P: Coro- 10. Rattazzi M, Faggin E, Bertacco E, Nardin C, Pagliani L, Plebani M, et al.: nary artery disease: Improved reproducibility of calcium scoring with Warfarin, but not rivaroxaban, promotes the calcification of the aortic an electron-beam CT volumetric method. Radiology 208: 807–814, valve in ApoE-/- mice. Cardiovasc Ther 36: e12438, 2018 1998 11. Caluwé R, Verbeke F, De Vriese AS: Evaluation of vitamin K status and 26. Caluwé R, De Vriese AS, Van Vlem B, Verbeke F: Measurement of pulse rationale for vitamin K supplementation in dialysis patients [published wave velocity, augmentation index, and central pulse pressure in atrial online ahead of print Dec 26, 2018]. Nephrol Dial Transplant doi: fibrillation: A proof of concept study. J Am Soc Hypertens 12: 627–632, 10.1093/ndt/gfy373 2018 12. Bellasi A, Raggi P: Vascular imaging in chronic kidney disease. Curr 27. Budoff MJ, Hokanson JE, Nasir K, Shaw LJ, Kinney GL, Chow D, et al.: Opin Nephrol Hypertens 21: 382–388, 2012 Progression of coronary artery calcium predicts all-cause mortality. 13. Bellasi A, Ferramosca E, Ratti C, Block G, Raggi P: The density of cal- JACC Cardiovasc Imaging 3: 1229–1236, 2010 cified plaques and the volume of calcium predict mortality in hemodi- 28. Cranenburg EC, Koos R, Schurgers LJ, Magdeleyns EJ, Schoonbrood alysis patients. Atherosclerosis 250: 166–171, 2016 TH, Landewé RB, et al.: Characterisation and potential diagnostic value 14. Raggi P: Cardiovascular disease: Coronary artery calcification predicts of circulating matrix Gla protein (MGP) species. Thromb Haemost 104: risk of CVD in patients with CKD. Nat Rev Nephrol 13: 324–326, 2017 811–822, 2010 15. Block GA, Spiegel DM, Ehrlich J, Mehta R, Lindbergh J, Dreisbach A, 29. Schlieper G, Westenfeld R, Krüger T, Cranenburg EC, Magdeleyns EJ, et al.: Effects of sevelamer and calcium on coronary artery calcification Brandenburg VM, et al.: Circulating nonphosphorylated carboxylated in patients new to hemodialysis. Kidney Int 68: 1815–1824, 2005 matrix gla protein predicts survival in ESRD. JAmSocNephrol22: 387– 16. Raggi P, Chertow GM, Torres PU, Csiky B, Naso A, Nossuli K, et al.: 395, 2011 ADVANCE Study Group: The ADVANCE study: A randomized study to 30. Kaesler N, Magdeleyns E, Herfs M, Schettgen T, Brandenburg V, Fliser evaluate the effects of cinacalcet plus low-dose vitamin D on vascular D, et al.: Impaired vitamin K recycling in uremia is rescued by vitamin K calcification in patients on hemodialysis. Nephrol Dial Transplant 26: supplementation. Kidney Int 86: 286–293, 2014 1327–1339, 2011 31. Ruderman I, Holt SG, Hewitson TD, Smith ER, Toussaint ND: Current 17. Sakaguchi Y, Hamano T, Obi Y, Monden C, Oka T, Yamaguchi S, et al.: A and potential therapeutic strategies for the management of vascular randomized trial of magnesium oxide and oral carbon adsorbent for calcification in patients with chronic kidney disease including those on coronary artery calcification in predialysis CKD. JAmSocNephrol30: dialysis. Semin Dial 31: 487–499, 2018 1073–1085, 2019 32. Chao TF, Liu CJ, Wang KL, Lin YJ, Chang SL, Lo LW, et al.: Incidence 18. Raggi P, Bellasi A, Ferramosca E, Islam T, Muntner P, Block GA: Asso- and prediction of ischemic stroke among atrial fibrillation patients with ciation of pulse wave velocity with vascular and valvular calcification in end-stage renal disease requiring dialysis. Heart Rhythm 11: 1752– hemodialysis patients. Kidney Int 71: 802– 807, 2007 1759, 2014 19. Verbeke F, Van Biesen W, Honkanen E, Wikström B, Jensen PB, 33. Zimmerman D, Sood MM, Rigatto C, Holden RM, Hiremath S, Clase Krzesinski JM, et al.: CORD Study Investigators: Prognostic value of CM: Systematic review and meta-analysis of incidence, prevalence and aortic stiffness and calcification for cardiovascular events and mortality outcomes of atrial fibrillation in patients on dialysis. Nephrol Dial in dialysis patients: Outcome of the calcification outcome in renal dis- Transplant 27: 3816–3822, 2012 ease (CORD) study. Clin J Am Soc Nephrol 6: 153–159, 2011 34. Sood MM, Larkina M, Thumma JR, Tentori F, Gillespie BW, Fukuhara S, 20. De Vriese AS, Caluwé R, Bailleul E, De Bacquer D, Borrey D, Van Vlem et al.: Major bleeding events and risk stratification of antithrombotic B, et al.: Dose-finding study of rivaroxaban in hemodialysis patients. agents in hemodialysis: Results from the DOPPS. Kidney Int 84: 600– Am J Kidney Dis 66: 91–98, 2015 608, 2013

JASN 31: ccc–ccc,2020 Vitamin K Status in Hemodialysis 11 Multicenter randomized controlled trial of vitamin K antagonist replacement by rivaroxaban with or without vitamin K2 in hemodialysis patients with atrial fibrillation Supplementary Appendix

Table of Contents Supplementary Methods

Patient inclusion and exclusion criteria

Secondary end-points

Sample size calculation

Randomization and blinding

Supplementary Tables

Supplementary Table 1: Baseline biochemical characteristics

Supplementary Table 2: Baseline maintenance medication.

Supplementary Table 3: Characteristics during follow-up

Supplementary Table 4: Causes of death

Supplementary Table 5: Stroke or systemic embolism in the VKA vs. pooled rivaroxaban arms

Supplementary Table 6: Bleeding outcomes in the VKA vs. pooled rivaroxaban arms

Supplementary Figures

Supplementary Figure 1: Study design

Supplementary Figure 2: CONSORT Diagram

Supplementary Figure 3: Kaplan-Meier curves for patient survival

Supplementary Methods

Patient inclusion and exclusion criteria Inclusion criteria:

• man or woman, age ≥18 years • signed informed consent • end-stage renal disease treated with chronic hemodialysis thrice weekly • non-valvular atrial fibrillation documented by electrocardiogram: paroxysmal (self- terminating) or permanent (accepted by the patient and physician); first-diagnosed (not diagnosed before, irrespective of the duration of the arrhythmia or related symptoms), new-onset or pre-existing

• CHA2DS2-VASc Score ≥2 • treatment with a vitamin K antagonist (VKA) or no treatment • female subjects should be postmenopausal, surgically sterilized or willing to use highly effective contraception during the study period

Exclusion criteria:

• known intestinal malabsorption or inability to take oral medication • inability to stop co-medication that causes major interactions with rivaroxaban (e.g. ketoconazole, itraconazole, voriconazole, posaconazole, ritonavir, rifampicin, phenytoin, carbamazepine, phenobarbital or St John’s wort) • investigator’s assessment that the subject’s life expectancy is less than 1 year • prosthetic mechanical heart valve • contraindication for anticoagulation • liver dysfunction Child-Pugh grade B-C • pregnancy, breastfeeding, inadequate contraception • inability to understand the nature, scope and possible consequences of the study • incompliance with medication and scheduled investigations, unlikelihood of completing the study

Secondary end-points 1) Change of aortic and mitral valve calcification over 18 months versus baseline.

2) Mortality from any cause.

3) Acute coronary syndrome, symptom-driven revascularization, hospitalization for heart failure and death from cardiac cause.

4) Stroke, defined as sudden onset of focal neurological deficit consistent with the territory of a major cerebral artery and categorized as ischemic, hemorrhagic, or uncertain type. Hemorrhagic transformation of ischemic stroke was not deemed to be hemorrhagic stroke. Systemic embolism, defined as an acute vascular occlusion of a limb or organ documented by imaging, surgery, or autopsy.

5) Life-threatening bleeding, defined as fatal bleeding, symptomatic intracranial bleeding, a decrease in hemoglobin of 5 g/dL or more, or a requirement for transfusion of four or more units of blood, inotropic agents, or surgery. Major bleeding, defined as a requirement for transfusion of two or more units of blood or a decrease in hemoglobin of 2 g/dL or more and not fulfilling the criteria for life-threatening bleeding. All other bleedings were regarded as minor.

Sample size calculation Assuming a mean CAC score of 800 at baseline, a standard deviation of annual CAC changes of 550 and accounting for an anticipated annual mean CAC progression of 60% (+480) in the VKA arm, 30% (+240) in the rivaroxaban arm and 0% (+0) in the rivaroxaban+vitamin K2 arm, power calculations revealed that 27 patients in each arm are required to provide 80% statistical power at the α=0.05 significance level (according to an ANOVA F-test). Taking into account an estimated drop-out rate of 30% during 18 months of follow-up and including a 10% security margin in view of the paucity of data to guide the power calculation, 132 patients were randomized.

Randomization and blinding The randomization schedule was computer-generated, stratified according to site, and concealed with the use of a web-based, locked central randomization system. Due to the requirement for frequent INR measurements to adjust the dose of warfarin, a complete blinding would be challenging. The patients and site investigators were aware of the treatment assignments. However, the primary endpoints were objectively measured by investigators (the investigator who reviewed all CT scans and the investigator who analyzed the pulse wave analysis curves) that were blinded to the treatment allocation.

Supplementary Tables

Supplementary Table 1: Baseline biochemical characteristics Baseline biochemical VKA Rivaroxaban Rivaroxaban+vit K2 P** characteristics (n=44) (n=46) (n=42) Calcium – mmol/L 2.19 (2.06-2.33) 2.24 (2.12-2.34) 2.18 (2.02-2.30) P=0.612 Phosphate – mmol/L 1.37 (1.17-1.64) 1.29 (1.09-1.70) 1.33 (1.18-1.62) P=0.870 Magnesium – mmol/L 0.86 (0.79-0.96) 0.88 (0.78-0.96) 0.85 (0.79-0.93) P=0.645 iPTH – ng/L 274 (192-475) 332 (181-433) 342 (174-503) P=0.971 25-hydroxy-vitamin D – µg/L 38.4 (29.7-47.5) 39.3 (30.3-49.3) 36.5 (23.4-48.4) P=0.731 Hemoglobin A1c – mmol/mol* 47 (41-60) 43 (39-51) 48 (42-55) P=0.504 LDL cholesterol – mg/dL 82 (56-117) 85 (57-113) 68 (59-97) P=0.539 HDL cholesterol – mg/dL 43 (35-56) 44 (38-55) 45 (38-54) P=0.869 Numbers displayed are median (interquartile range); VKA, vitamin K antagonist; *in patients with diabetes; **according to the Kruskal-Wallis test

Supplementary Table 2: Baseline maintenance medication. Baseline medication VKA Rivaroxaban Rivaroxaban+vit K2 P** (n=44) (n=46) (n=42) Cinacalcet % (no.) 11.4% (5/44) 8.7% (4/46) 4.8% (2/42) Mean (median) dose – mg/day* 48 (30) 45 (30) 60 (60) P=0.579 Sevelamer % (no.) 22.7% (10/44) 21.7% (10/46) 28.6% (12/42) Mean (median) dose – mg/day* 1680 (1200) 3022 (2400) 2800 (1600) P=0.740 Lanthanum % (no.) 2.3% (1/44) 0.0% (0/46) 4.8% (2/42) Mean (median) dose – mg/day* 750 (750) - 875 (875) P=0.208 Sucroferric oxyhydroxide % (no.) 0.0% (0/44) 0.0% (0/46) 2.4% (1/42) Mean (median) dose – mg/day* - - 1500 (1500) P=0.318 Calcium containing binders % (no.) 27.3% (12/44) 39.1% (18/46) 26.2% (11/42) Mean (median) dose – mg/day* 1675 (1334) 2321 (2001) 2272 (2000) P=0.363 Alfacalcidol % (no.) 20.5% (9/44) 30.4% (14/46) 35.7% (15/42) Mean (median) dose – mg/day* 1.69 (1) 2.33 (1.75) 2.12 (1) P=0.281 Aspirin – % (no.) 31.8% (14/44) 32.6% (15/46) 40.5% (17/42) P=0.665 ARB or ACE inhibitor – % (no.) 18.2% (8/44) 21.7% (10/46) 9.5% (4/42) P=0.301 Betablocker – % (no.) 52.3% (23/44) 52.2% (24/46) 57.1% (24/42) P=0.883 Amiodarone – % (no.) 20.5% (9/44) 10.9% (5/46) 19.0% (8/42) P=0.413 Statin – % (no.) 29.5% (13/44) 26.1% (12/46) 28.6% (12/42) P=0.941 Proton-pump inhibitor – % (no.) 50.0% (22/44) 56.5% (26/46) 69.0% (29/42) P=0.190 VKA, vitamin K antagonist; ARB, angiotensin receptor blocker; ACE, angiotensin-converting enzyme; *in patients taking this medication; **according Fisher’s exact test Supplementary Table 3: Characteristics during follow-up Baseline 6 months 12 months 18 months P** Calcium – mmol/L P=0.435 VKA 2.19 (2.06-2.33) 2.18 (2.09-2.28) 2.15 (2.08-2.25) 2.08 (2.00-2.24) Rivaroxaban 2.24 (2.12-2.34) 2.19 (2.11-2.29) 2.19 (2.10-2.33) 2.17 (2.10-2.29) Rivaroxaban+vitK2 2.18 (2.02-2.30) 2.19 (2.08-2.31) 2.15 (2.00-2.28) 2.22 (2.03-2.36) Phosphate – mmol/L P=0.572 VKA 1.37 (1.17-1.64) 1.41 (1.19-1.77) 1.44 (1.12-1.72) 1.33 (1.20-1.65)

Rivaroxaban 1.29 (1.09-1.70) 1.39 (1.04-1.66) 1.24 (1.10-1.71) 1.37 (1.22-1.64) Rivaroxaban+vitK2 1.33 (1.18-1.62) 1.43 (1.19-1.65) 1.41 (1.23-1.62) 1.37 (1.14-1.69) iPTH – ng/L P=0.093 VKA 274 (192-475) 385 (166-524) 438 (264-730) 473 (323-627)

Rivaroxaban 332 (181-433) 366 (198-589) 421 (229-625) 447 (188-584) Rivaroxaban+vitK2 342 (174-503) 322 (169-427) 373 (171-563) 384 (173-664) Hemoglobin A1c – mmol/mol* P=0.495 VKA 47 (41-60) 42 (39-50) 43 (37-52) 35 (33-52)

Rivaroxaban 43 (39-51) 45 (39-52) 42 (38-48) 43 (39-49) Rivaroxaban+vitK2 48 (42-55) 44 (40-50) 45 (36-52) 43 (38-51) Mean arterial pressure – mmHg P=0.370 VKA 91 (80-101) 86 (78-96) 92 (87-99) 99 (85-103)

Rivaroxaban 86 (78-99) 92 (80-99) 88 (81-107) 85 (70-95) Rivaroxaban+vitK2 88 (78-97) 88 (74-96) 85 (77-95) 86 (79-97) Cinacalcet – % P=0.955 VKA 11.4% (5/44) 10.5% (4/38) 17.9% (5/28) 28.6% (6/21)

Rivaroxaban 8.7% (4/46) 10.3% (4/39) 14.3% (5/35) 17.2% (5/29) Rivaroxaban+vitK2 4.8% (2/42) 5.4% (2/37) 8.8% (3/34) 11.1% (3/27) Sevelamer – % P=0.437 VKA 22.7% (10/44) 26.3% (10/38) 39.3% (11/28) 38.1% (8/21)

Rivaroxaban 21.7% (10/46) 25.6% (10/39) 20.0% (7/35) 24.1% (7/29) Rivaroxaban+vitK2 28.6% (12/42) 35.1% (13/37) 32.4% (11/34) 22.2% (6/27) Calcium containing binders – % P=0.541 VKA 27.3% (12/44) 26.3% (10/38) 17.9% (5/28) 9.5% (2/21)

Rivaroxaban 39.1% (18/46) 33.3% (13/39) 37.1% (13/35) 37.9% (11/29) Rivaroxaban+vitK2 26.2% (11/42) 32.4% (12/37) 29.4% (10/34) 37.0% (10/27) Alfacalcidol – µg/week, % P=0.517 VKA 20.5% (9/44) 15.8% (6/38) 14.3% (4/28) 33.3% (7/21)

Rivaroxaban 30.4% (14/46) 35.9% (14/39) 37.1% (13/35) 41.4% (12/29) Rivaroxaban+vitK2 35.7% (15/42) 40.5% (15/37) 41.2% (14/34) 44.4% (12/27) Numbers displayed are median (interquartile range) or % (n); VKA, vitamin K antagonist; *in patients with diabetes; **significance of treatment-by-time interaction according to mixed modeling Supplementary Table 4: Causes of death VKA Rivaroxaban Rivaroxaban+vit K2 P (n=44) (n=46) (n=42)

Death from any cause – no. (%) 19 (43.2%) 15 (32.6%) 13 (31.0%) P=0.457

Sudden death – no. 3 4 2 Cardiovascular disease Fatal acute coronary syndrome – no. 0 2 1 Terminal heart failure – no. 2 0 0 Limb ischemia – no. 1 0 1 Intracranial bleeding – no. 1 0 0 Infectious disease Endocarditis – no. 1 0 0 Sepsis/Septic shock – no. 1 3 1 Respiratory infection – no. 1 1 1 Influenza – no. 0 1 0 Malignancy – no. 1 2 1 MOF – no. 1 0 0 Lactic acidosis – no. 1 0 0 Accident – no. 0 0 2 Withdrawal of dialysis Palliative care – no. 5 1 2 Choice of the patient – no. 1 1 2

VKA, vitamin K antagonist Supplementary Table 5: Stroke or systemic embolism in the VKA vs. pooled rivaroxaban arms

VKA Rivaroxaban P* (n=44) (n=88)

Stroke or systemic embolism Ischemic or uncertain type of stroke – no. 5 3 P=0.116 Hemorrhagic stroke or intracerebral bleeding – no. 2 0 P=0.109

VKA, vitamin K antagonist; *according to Fisher’s exact test

Supplementary Table 6: Bleeding outcomes in the VKA vs. pooled rivaroxaban arms

VKA Rivaroxaban PCox PPoisson (n=44) (n=88) Life-threatening bleeding 6 / 7 / 13.3 6 / 7 / 6.3 P=0.207 P=0.166

Major bleeding 7 / 12 / 22.7 8 / 10 / 9.0 P=0.130 P=0.031

Life-threatening + major bleeding 13 / 19 / 36.0 12 / 17 / 15.4 P=0.034 P=0.011

Minor bleeding 10 / 14 / 26.5 28 / 34 / 30.7 P=0.284 P=0.644

Gastrointestinal bleeding 8 / 12 / 22.7 16 / 23 / 20.8 P=0.722 P=0.800

Cell entries are: number of patients with at least one bleeding episode / total number of bleeding episodes / total number of bleeding episodes per 100 person-years; PCox: significance of differences in time to first bleeding episode according to

Cox’s proportional hazard model analysis; PPoisson: significance of differences in total number of bleeding episodes according to Poisson model analysis; ATG, antagonist Life-threatening bleeding was defined as fatal bleeding, symptomatic intracranial bleeding, a decrease in hemoglobin of 5 g/dL or more, or a requirement for transfusion of four or more units of blood, inotropic agents, or surgery. Major bleeding was defined as a requirement for transfusion of two or more units of blood or a decrease in hemoglobin of 2 g/dL or more, and not fulfilling the criteria for life-threatening bleeding. All other bleedings were regarded as minor.

Supplementary Figures

Supplementary Figure 1: Study design

Hemodialysis patients with AF

CHAD2S2-VASc Score ≥2

On vitamin K antagonistO or no treatment

Assessment at baseline:

-Biochemical

-CT scan

-Pulse Wave Velocity

Vitamin K antagonist, thrice Rivaroxaban 10 mg od Rivaroxaban 10 mg od + weekly after dialysis, titrated vitamin K2 2000 µg thrice to INR 2-3 weekly after dialysis

Assessment at 6 months, 12 months, 18 months:

-Biochemical

-CT scan

-Pulse Wave Velocity

Supplementary Figure 2: CONSORT Diagram

Assessed (n=143)

Excluded (n=11) • Not meeting inclusion criteria Enrollment (n=0)

• Declined to participate (n=11) • Other reasons (n=)

Randomized (n=132)

Allocated to vitamin K Allocated to Rivaroxaban Allocated to Rivaroxaban + antagonist (n=44) (n=46) vitamin K2 (n=42) • Received allocated • Received allocated • Received allocated intervention (n=44) intervention (n=46) intervention (n=42) Allocation • Did not receive allocated • Did not receive allocated • Did not receive allocated intervention (n=0) intervention (n=0) intervention (n=0)

Follow-up incomplete (n=23) Follow-up incomplete (n=17) Follow-up incomplete (n=15) • Death (n=19) • Death (n=15) • Death (n=13) • Lost to follow-up (n=0) • Lost to follow-up (n=0) • Lost to follow-up (n=0) • Unwilling (n=4) • Unwilling (n=2) • Unwilling (n=2) Follow-Up Discontinued intervention Discontinued intervention Discontinued intervention (n=4) (n=5) (n=5)* • Unwilling (n=0) • Unwilling (n=0) • Unwilling (n=1) • Adverse event (n=4) • Adverse event (n=5) • Adverse event (n=4)

Analyzed (n=44) Analyzed (n=46) Analyzed (n=42) • Outcome data complete (n=43) • Outcome data complete (n=44) • Outcome data complete (n=41) • Outcome data incomplete • Outcome data incomplete • Outcome data incomplete Analysis (n=1)** (n=2)** (n=1)** Excluded from analysis (n=0) Excluded from analysis (n=0) Excluded from analysis (n=0)

*Patients discontinued rivaroxaban, but continued vitamin K2

**No reliable calcification score measurements could be made at one of the time points due to movement or other artefacts.

Supplementary Figure 3: Kaplan-Meier curves for patient survival

P=0.466 according to a Cox proportional hazards model.

Baseline 6 months 12 months 18 months Patients at risk, n VKA 44 38 28 21 Rivaroxaban 46 39 35 29 Rivaroxaban+vitK2 42 37 34 27 Patients taking the study drug, n VKA 44 35 27 20 Rivaroxaban 46 36 31 25 Rivaroxaban+vitK2 42 34 30 23