IEEE TRANSACTIONS ON INFORMATION TECHNOLOGY IN BIOMEDICINE, VOL. 16, NO. 6, NOVEMBER 2012 WiiPD—Objective Home Assessment of Parkinson's Disease Using the Nintendo Wii Remote Jonathan Synnott, Student Member, IEEE, Liming Chen, Member, IEEE, Chris. D. Nugent, Member, IEEE, and George Moore, Member, IEEE Abstract—Current clinical methods for the assessment of
Fiix-prothrombin time versus standard prothrombin time for monitoring of warfarin anticoagulation: a single centre, double-blind, randomised, non-inferiority trialFiix-prothrombin time versus standard prothrombin time
for monitoring of warfarin anticoagulation: a single centre,
double-blind, randomised, non-inferiority trial
Páll T Onundarson*, Charles W Francis, Olafur S Indridason, David O Arnar, Einar S Bjornsson, Magnus K Magnusson, Sigurdur J Juliusson, Hulda M Jensdottir, Brynjar Vidarsson, Petur S Gunnarsson, Sigrun H Lund, Brynja R Gudmundsdottir* Summary
Background Rapid ﬂ uctuations in factor VII during warfarin anticoagulation change the international normalised Lancet Haematol 2015
ratio (INR) but contribute little to the antithrombotic eﬀ ect. We aimed to assess non-inferiority of anticoagulation Published Online
stabilisation with a warfarin monitoring method aﬀ ected only by factors II and X (Fiix-prothrombin time [Fiix-PT])
compared with standard PT-INR monitoring that includes factor VII measurement as well.
Methods The Fiix trial was a single centre, double-blind, prospective, non-inferiority, randomised controlled clinical http://dx.doi.org/10.1016/
trial. Ambulatory adults on warfarin with an INR target of 2–3 managed by an anticoagulation dosing service using S2352-3026(15)00075-7
software-assisted dosing at the National University Hospital of Iceland, Reykjavik, Iceland, were eligible for inclusion See Online for and audio
in this study. We excluded patients undergoing electroconversion and nursing home residents. Patients were interview with Páll Onundarson
randomly assigned (1:1) to either the Fiix-PT monitoring group or the PT monitoring group by block randomisation. *Contributed equally
A blinded research INR (R-INR) based on results of the respective test was reported to the dosing staﬀ . Participants The National University
Hospital of Iceland, Reykjavik,
were contacted by a study nurse at 4-week intervals to elicit information about thromboembolism or bleeding Iceland
otherwise unknown to the anticoagulation management centre. The primary eﬃ
cacy outcome was a composite of (Prof P T Onundarson MD,
objectively diagnosed non-fatal and fatal arterial or venous thromboembolism, including myocardial infarction and O S Indridason MD,
transient ischaemic attacks, assessed in all eligible patients who were randomised (intention-to-monitor population). D O Arnar MD,
Prof E S Bjornsson MD, The safety endpoint was major bleeding or other clinically relevant bleeding, assessed in the per-protocol population. Prof M K Magnusson MD,
We assumed a 3% annual thromboembolism incidence and a non-inferiority margin of 2·5%. This trial is registered S J Juliusson BS,
with ClinicalTrials.gov, number NCT01565239.
H M Jensdottir BS, B Vidarsson MD, P S Gunnarsson PharmD, Findings Between March 1, 2012, and Feb 28, 2014, we enrolled 1156 patients. 573 patients were assigned to Fiix-PT B R Gudmundsdottir MS);
and 575 to PT-INR monitoring after exclusion of four patients from each group for various reasons. Median follow-up University of Iceland School of
was 1·7 years (IQR 1·1–1·9). During days 1–720, ten (1·2% per patient year) thromboembolic events occurred in the Health Sciences, Faculty of
Fiix-PT group versus 19 (2·3% per patient year) in the PT group (relative risk [RR] 0·52, 95% CI 0·25–1·13; Medicine, Reykjavik, Iceland
(Prof P T Onundarson, <0·0001). Major bleeding occurred in 17 of 571 patients in the Fiix group (2·2% per patient year) versus 20 of
Prof E S Bjornsson, 573 patients in the PT group (2·5% per patient year; RR 0·85, 0·45–1·61; p
=0·0034). Anticoagulation stability
Prof M K Magnusson, was improved with Fiix-PT monitoring as manifested by fewer tests, fewer dose adjustments, increased time in range S H Lund PhD); and University
and less INR variability than reported with standard PT monitoring.
of Rochester Medical Center,
Rochester, New York, USA
(Prof C W Francis MD)
Interpretation Monitoring of warfarin with Fiix-PT improved anticoagulation and dosing stability and was clinically Correspondence to:
non-inferior to PT monitoring. Results from this trial suggest that during vitamin K antagonist treatment INR Prof Páll T Onundarson,
monitoring could be replaced by Fiix-PT and that this would lead to at least a non-inferior clinical outcome compared Haematology Laboratory and
with monitoring with PT-INR.
Coagulation Disorders, The National University Hospital of Iceland, Reykjavik 104, Iceland Funding Innovation Center Iceland, University of Iceland Science Fund, Landspitali Science Fund and Actavis.
normalised ratio (INR), which adjusts for the diﬀ erent Vitamin K antagonists have a narrow therapeutic sensitivities of thromboplastins used in the test.4 window and must be monitored at variable intervals to Additionally, software-assisted dosing5,6 and centralised ensure eﬃ cacy and minimise bleeding complications.1 anticoagulation dose-management centres1,7 have led to This monitoring is done by measurement of the major improvements in vitamin K antagonist treatment.
prothrombin time (PT),2,3 a test that is equally sensitive Inability to obtain stable anticoagulation with vitamin K to warfarin induced reductions in each of the coagulation antagonists leads to a variable need for dose adjustments factors II, VII, and X. Reductions in factor IX also occur and an increased risk of thrombotic and bleeding during vitamin K antagonist treatment but are not complications. Fluctuations in INR are often ascribed to detected by measurement of the PT. To manage food and drug interactions or patient non-compliance.1 vitamin K antagonist therapy, calibrators are used to However, experimental data8 suggest that INR ﬂ uctuations standardise the PT by calculation of an inter national might also be caused by rapidly changing factor VII www.thelancet.com/haematology Published online May 26, 2015 http://dx.doi.org/10.1016/S2352-3026(15)00073-3
Research in context
Evidence before this study
database review of the scientiﬁ c literature did not identify a We searched Medline between Jan 1, 1950, and Jan 1, 2012, for similar research idea. Subsequently, we did this single centre all studies in English using the search terms "anticoagulation", double-blinded, randomised controlled trial to test the "monitoring", "coumarin", "warfarin", "vitamin K antagonist", hypothesis, with centralised modern software-assisted dosing "prothrombin time", and "international normalised ratio". No of warfarin.
studies describing the research idea of the current study by Added value of this study
other scientists were identiﬁ ed. Warfarin and other vitamin K An improvement in the regulation of anticoagulation was antagonists have been dosed for more than 60 years on the reported and a clinically non-inferior reduction in basis of the prothrombin time (PT). The prothrombin time is thromboembolism in patients that were dosed based on the equally sensitive to a reduction in each of coagulation factors II, Fiix-PT (Fiix-INR) compared with high-quality dosing based on VII, and X, but not to reductions in factor IX that occur during PT (INR) in controls. Bleeding was not increased. A post-hoc vitamin K antagonist treatment. As factor VII has a much analysis suggests a possible improved long-term outcome of shorter half-life than factor II and factor X, changes in factor VII patients on warfarin monitored with the Fiix-PT. The new data cause a fast corresponding ﬂ uctuation of the PT-based suggest that anticoagulation variability during warfarin international normalised ratio (INR). However, experimental management is not only caused by food and drug interactions data suggests that the in-vitro anti-clot forming and in-vivo or patient non-compliance but partly by a confounding antithrombotic eﬀ ect of vitamin K antagonist drugs is mainly clinically irrelevant ﬂ uctuation of factor VII. Replacement of the caused by lowering factor II and factor X activity and not by PT with the Fiix-PT (Fiix-INR) eliminated this confounding factor VII. Such studies include thrombin generation studies, eﬀ ect of factor VII on the INR.
thromboelastometric studies, small clinical studies monitoring warfarin with the native prothrombin antigen, and a rabbit Implications of all the available evidence
study showing that disseminated intravascular coagulation Although a multicentre study ideally should be done to conﬁ rm induced by infusion of tissue factor was prevented by reduction the results of the Fiix-PT trial in diﬀ erent management settings, of factor II and possibly factor X, but not factor VII. On the basis our results suggest that INR variability during vitamin K of these studies, we postulated that the combined eﬀ ect of antagonist treatment could be reduced in practice by factor II and factor X should only be monitored during replacement of the PT with Fiix-PT and that this would lead to vitamin K antagonist anticoagulation and that elimination of at least a non-inferior clinical outcome. Finally, the high quality the eﬀ ect of factor VII in the test sample would increase stability of warfarin treatment in the PT control group might suggest of anticoagulation with at least a non-inferior clinical outcome. that vitamin K antagonists should be centrally managed by We invented a modiﬁ ed PT, the Fiix-PT, that measures only the dedicated staﬀ using software-assisted dosing.
eﬀ ect of factor II and factor X in test samples. Our extensive activity. Although severe reductions in factor VII can lead with fewer dose adjustments and at least equivalent to a risk of bleeding, the experiments suggest that the clinical outcomes.
main antithrombotic eﬀ ect of vitamin K antagonists is
due to stable reductions in factor II and possibly in Methods
factor X, which have much longer half-lives than Study design and patients
factor VII, whereas the contributions of reduced factor VII
We undertook an investigator-initiated, double-blind, non- activity and that of the undetected factor IX are minor.8–10 inferiority, randomised controlled trial in our We have developed a modiﬁ ed PT assay, Fiix-PT, that is anticoagulation management centre at The National sensitive to the combined reductions of factor II and University Hospital of Iceland, Reykjavik, Iceland. We factor X and is unaﬀ ected by the activity of factor VII in invited all community-dwelling participants, aged 18 years the test sample.8 This assay was based on our rotational or older, who were taking or starting short-term or long-thromboelastometric experiments,8 which showed that term warfarin during the study period with an INR target in a dilute thromboplastin model, the initiation, of 2–3 and being managed at our anticoagulation propagation, and stabilisation phases of clotting were management centre to participate in this study, irrespective equally aﬀ ected by reduced concentrations of factor II or of indication for anticoagulation. We excluded nursing factor X, but much less by factor VII or factor IX. home residents and participants being monitored at ﬁ xed Furthermore, thrombin generation in samples from weekly intervals before electroconversion of atrial patients receiving warfarin was detected equally well ﬁ brillation. We recruited patients consecutively as they with Fiix-PT or PT-INR, suggesting that measurement of came for their monitoring test appointments in the factor VIIc is not needed (Onundarson PT, unpublished). outpatient phlebotomy unit or in hospital wards during We therefore postulated that monitoring of warfarin inpatient warfarin initiation. This study was done in with Fiix-PT would lead to more stable anticoagulation, accordance with the principles of the Declaration of www.thelancet.com/haematology Published online May 26, 2015 http://dx.doi.org/10.1016/S2352-3026(15)00073-3
Helsinki at The National University Hospital of Iceland, Reykjavik, Iceland. All patients provided written informed 3234 patients screened consent. The National Bioethics Committee of Iceland (VSNb2011040019/03.15) and the Data Protection Agency of Iceland (2011040560AMK/-) approved the protocol.
For the protocol see
162 in nursing homes 169 refused to participate Randomisation and masking
1747 not oﬀered participation The study nurse recruited patients and obtained informed segavarnir/ﬁ ix- consent and then randomly assigned patients (in a 1:1 1156 enrolled in study ratio) to either the experimental monitoring Fiix-PT group or the PT monitoring control group. Block randomisation without a stratiﬁ cation procedure (with 48 participants in each block) was applied by each patient 577 randomly assigned to 579 randomly assigned Fiix-INR monitoring to INR monitoring drawing a colour-coded card from a closed box; each box contained 24 orange cards and 24 green cards that were also numbered with an anonymous code (a case report form number) that was recorded. Each colour directed 1 not on anticoagulants 1 with INR target 2·5–3·5 the patient to either the active Fiix-PT group or the 2 on electroconversion list 2 on electroconversion list standard PT control group. The patient then presented the colour card during phlebotomies and corresponding coloured stickers were placed on the blood sample tubes 573 included in analysis 575 included in analysis by phlebotomists. Once received in the coagulation laboratory, laboratory staﬀ directed the sample tube to the 573 assessed for 575 assessed for appropriate test on the basis of the sticker colour. Results of Fiix-PT or PT were reported in the laboratory information system in a masked manner as research INR (R-INR). Only the electronically reported R-INR was 2 discontinued warfarin 2 discontinued warfarin available to the dosing staﬀ , who were not involved in anticoagulation was anticoagulation was phlebotomies and, therefore, had no way of knowing which test had generated the R-INR and had no knowledge of each patient's colour card. Patients, dosing 571 assessed for 573 assessed for staﬀ , and event outcome assessors were masked to monitoring method assignment.
Figure 1: Trial proﬁ le
We classiﬁ ed patients according to indications for tests. PT-INR was calculated on the basis of Quick PT2 and
anticoagulation. More than one indication was present Fiix-INR was calculated on the basis of the new Fiix-PT, a
in some individuals. We also recorded associated modiﬁ ed PT that is only sensitive to factor II and factor X
conditions and selected medication use. We calculated due to mixing factor II and factor X double-deﬁ cient
a CHA DS -VASC risk score for thromboembolism for plasma into the test sample.8 Both tests used in-house
patients with atrial ﬁ brilliation.11 Patients were standardisation of the thromboplastin sensitivity index registered as short term when referred as such and if (ISI) with ISI calibrators and control plasma (Danish treatment lasted less than 6 months and were regarded Institute for External Quality Assurance in Health Care, as naive to warfarin if enrolled during the ﬁ rst 60 days Glostrup, Denmark). The calibrator is designed for PT of warfarin intake.12 standardisation but not for Fiix-PT standardisation.
An executive committee did the trial, managed the data, Standardised PT ratios and Fiix-PT ratios were reported and analysed the data. An independent adjudication electronically as R-INR to dosing staﬀ (nurses, biomedical committee consisting of three experienced physicians (a scientists, and physicians) who adjusted doses according haematologist, a cardiologist, and a gastroenterologist) to usual practice, aiming for an R-INR of 2–3, with the adjudicated all eﬃ cacy and safety endpoints. An DAWN anticoagulation software (4-S, Milnthorpe, UK)5 independent safety monitoring committee reviewed and in-house protocols designed for monitoring with study outcomes at 3-month intervals from the entry of the traditional PT based on the American College of Chest ﬁ rst patient.
Physicians guidelines.6,10 The maximum recommended Both Fiix-PT and PT tests were done at the centralised interval between monitoring tests was 6 weeks.
coagulation laboratory on citrated venous blood samples. All thromboembolism and bleeding events reported to The automated STA-R Evolution coagulation analyser the anticoagulation management centre by patients, (Diagnostica Stago, Asnieres, France) was used for both health-care workers, and hospital units were recorded. www.thelancet.com/haematology Published online May 26, 2015 http://dx.doi.org/10.1016/S2352-3026(15)00073-3
(Continued from previous column) Arterial thromboembolism without known atrial ﬁ brillation Cerebral thromboembolism or transient ischaemic attack Peripheral arterial Years of warfarin treatment before 3·7 (0·9–8·2) 3·4 (0·8–7·8) Venous thromboembolism Deep vein thrombosis alone Warfarin experienced (>60 days on warfarin when enrolled) Pulmonary embolism Short-term warfarin treatment* Pulmonary hypertension Indications for warfarin Associated conditions Atrial ﬁ brillation High blood pressure Atrial ﬁ brillation without previous 307 (75%) Ischaemic heart disease arterial thromboembolic events Peripheral vascular disease Atrial ﬁ brillation with previous History of congestive heart failure cerebral thromboembolic events or transient ischaemic attack Atrial ﬁ brillation with previous peripheral arterial embolism Active cancer chemotherapy CHA DS -VASC risk score in patients with atrial ﬁ brillation Acetylsalicylic acid Score 0 (low thromboembolic Non-steroidal anti-inﬂ ammatory Score 1 (moderate thromboembolic risk) Score ≥2 (high Hydrogen blockers and proton thromboembolic risk) Ischaemic heart disease Acute myocardial infarction Data are median (IQR) or n (%). Percentages may not total 100% owing to Other ischaemic heart disease presence of more than one indication in some patients or rounding of numbers. Congestive heart failure as only *Short-term warfarin is deﬁ ned as less than 6 months receiving warfarin treatment. Fiix-PT=Fiix-prothrombin time. PT=prothrombin time. Atrial septal defect Table 1: Baseline characteristics
Artiﬁ cial heart valves Rheumatic mitral valve disease (Table 1 continues in next column) objectively diagnosed non-fatal and fatal arterial or venous thromboembolism, including myocardial infarction during the whole 720 day study period. We included transient All participants were instructed to report any new health- ischaemic attacks of any kind if they had been diagnosed by related events as soon as possible. The study nurse a treating physician, but imaging studies were not contacted all participants at 4-week intervals using a mandatory. We did not include superﬁ cial thrombo-checklist to elicit information about thromboembolism phlebitis. Secondary eﬃ cacy outcomes included non- or bleeding otherwise unknown to the anticoagulation thromboembolism and non-haemorrhagic related deaths. management centre. After a recorded event, the next Safety outcomes were major bleeding and a composite of follow-up call was 4 weeks later. We assessed causes of major bleeding and other clinically relevant non-major death on the basis of information obtained from hospital bleeding occurring in the per-protocol population. Major records, physicians, autopsy reports, and death bleeding was deﬁ ned according to the International Society certiﬁ cates. Deaths were coded as caused by bleeding, on Thrombosis and Haemostasis criteria—ie, bleeding thrombo embolism, or other causes when neither could leading to hospital admission, transfusion of at least two be conﬁ rmed.
units of packed red cells, or bleeding into a closed compartment such as the cranium or pericardium.13 We deﬁ ned other clinically relevant non-major bleeding as cacy outcome was calculated in the overt bleeding not meeting the criteria for major bleeding intention-to-monitor population and was a composite of but associated with medical intervention, unscheduled www.thelancet.com/haematology Published online May 26, 2015 http://dx.doi.org/10.1016/S2352-3026(15)00073-3
p value for
patient observation patient observation Primary outcome population Total observation years Fatal and ﬁ rst non-fatal thromboembolism 0·52 (0·25–1·13) including myocardial infarction Cerebral infarction or transient ischaemic attacks 0·65 (0·28–1·48) Cerebral infarction 0·64 (0·25–1·64) Transient ischaemic attack 0·67 (0·11–3·99) Myocardial infarction 0·33 (0·03–3·21) Peripheral arterial occlusion Venous thromboembolism Per-protocol population Total observation years First major bleeding 0·85 (0·45–1·61) 1·2 (0·52–2·76) 0·4 (0·08–2·06) 0·33 (0·03–3·21) Other major bleeding 0·75 (0·17–3·35) Non-major clinically relevant bleeding All (including repeated) 0·88 (0·71–1·09) First non-major clinically relevant bleeding 0·92 (0·7–1·2) 0·93 (0·83–1·04) Death from any cause 0·75 (0·36–1·58) Non-vascular death 1 (0·42–2·39) Composite major vascular events 0·69 (0·43–1·12) Non-inferiority analysis of total major events occurring during days 1–720 from randomisation. Fatal events are shown in parentheses. Eﬃ cacy of the monitoring method is assessed based on intention-to-monitor analysis, but safety of monitoring method is based on actual time on warfarin including a 5-day washout period after warfarin discontinuation (per-protocol population). Fiix-PT=Fiix-prothrombin time. PT=prothrombin time. *Percentage with event per patient observation year. †p value by Farrington–Manning test of non-inferiority with a non-inferiority margin of 0·025. Table 2: Primary analysis of clinical outcome
physician contact, temporary cessation of treatment, mixed indications for anticoagulation, with an INR target discomfort such as pain, or impairment of activities of daily of 2–3, monitored and dosed by our anticoagulation life. We classiﬁ ed other bleeding as minor. We calculated management centre. On the basis of these studies, we composite major vascular events as the total incidence of used an expected 3% annual thromboembolism incidence both fatal and non-fatal thromboembolism events and and a non-inferiority margin of 2·5%. We regarded this major bleeding. Surrogate eﬃ cacy variables calculated in 2·5% diﬀ erence as clinically signiﬁ cant, yet within the the per-protocol population included the number of tests in 95% CI for event incidence in previous studies.20 On the each study group, dose adjustment frequency, and the basis of these considerations, with two separate online percentage of tests at deﬁ ned ranges. We calculated the calculators, we calculated the number of participants percentage of time that each individual patient spent within needed to show statistical non-inferiority of clinically the INR target range using the Rosendaal formula.14 We important events for the test method with an 80% also calculated the variance growth rate as an indicator of certainty per year of observation as being 576 patients in INR variability between tests (B2 method).15 each group. A study time of 24 months was expected to allow for a delay in recruitment and potential dropouts.
The prespeciﬁ ed primary and post-hoc secondary To calculate the non-inferiority margin, we used data analyses of eﬃ cacy were assessed in an intention-to- from previous prospective studies16–19 in patients with monitor population—ie, all events were counted from www.thelancet.com/haematology Published online May 26, 2015 http://dx.doi.org/10.1016/S2352-3026(15)00073-3
We compared continuous data with the Mann–Whitney test and categorical data with χ² or Fisher's exact tests. We Primary analysis 1–720 days: RR 0·52 (95% CI 0·25–1·13),p<0·0001 for non-inferiority, p=0·0890 for superiority used the Farrington–Manning test of non-inferiority to Post-hoc analysis 181–720 days: calculate non-inferiority of the eﬃ cacy and safety RR 0·41 (95% CI 0·16–1·05), p=0·0307 for superiority endpoints21,22 and calculated relative risk (RR) with a 95% CI. We generated Kaplan-Meier curves to show events occurring over time and compared them with the Breslow–Gehan–Wilcoxon test for superiority assessment. We did a subgroup analysis for the largest subgroup of non-valvular atrial ﬁ brillation. We regarded p of less than 0·05 as Cumulative thromboembolism signiﬁ cant. We did statistical analyses with GraphPad Prism version 6.0 and R version 3.1.1 (July 10, 2014).
This study is registered with ClinicalTrials.gov, number Number at risk
Role of the funding source
The funders of the study had no role in study design, Primary analysis 1–720 days: RR 0·85 (95% CI 0·45–1·61), data collection, data analysis, data interpretation, or p<0·0034 for non-inferiority, p=0·8261 for superiorityPost-hoc analysis 181–720 days: writing of the report. PTO, CWF, OSI, SJJ, HMJ, SHL, RR 0·71 (95% CI 0·31–1·65), and BRG had access to the data. All authors were p=0·5362 for superiority responsible for the ﬁ nal decision to submit this report for publication. Results
Between March 1, 2012, and Feb 28, 2014, we
enrolled 1156 patients (ﬁ gure 1). After exclusion of four
Cumulative major bleeding (%) patients in each group, 573 patients were monitored with Fiix-INR and 575 with PT-INR. The total intention-to- monitor analysis observation time was 828 patient-years in the Fiix-PT group and 835 patient-years in the PT group Time from randomisation (days) Number at risk
with a median follow-up of 1·7 years (IQR 1·1–1·9) per person in both groups. In the Fiix-PT group, 440 patients were followed up for more than 1 year, comprising 766 of Figure 2: Cumulative event incidences during days 1–720 for primary eﬃ
the observation years; in the PT group, 450 patients were outcome thromboembolism (A) and major bleeding (B) in all patients
followed up for more than 1 year, comprising monitored with Fiix-prothrombin time (Fiix-PT) versus quick prothrombin
773 observation years. Baseline characteristics did not time (PT)
Non-fatal and fatal events are shown. RR=relative risk. PT=prothrombin time.
diﬀ er between the groups (table 1), although minor numerical diﬀ erences were reported. The average CHA DS -VASC scores were similar in patients with atrial the day of enrolment until 5 days after ﬁ nal ﬁ brillation in both groups, mean 2·9 (SD 1·6) and median discontinuation of warfarin or study completion. Patients 3·0 (IQR 2–4) in the Fiix-PT group and 3·0 (SD 1·6) and with thromboembolism events were censored after the 3·0 (IQR 2–4) in the control group, supporting a similar occurrence. We analysed safety outcomes in the thromboembolism risk in both study groups. The number per-protocol population—ie, we included bleeding events of participants who discontinued warfarin, switched to occurring from enrolment of warfarin-experienced direct oral anticoagulants, or discontinued from the study patients or after two INRs fell within the target range for other reasons did not diﬀ er between groups. No See Online for appendix
after warfarin initiation in new patients and until 5 days patients were lost to follow-up (appendix p 3).
after discontinuation or study completion, but we In the primary analysis of thromboembolism events excluded periods of temporary discontinuation of occurring during days 1–720, thromboembolism occurred warfarin or dose management outside the anticoagulation in ten patients in the Fiix-PT group (incidence of 1·2% per management centre from analysis (eg, during surgery, patient-year) versus 19 (2·3% per patient-year) in the PT hospital admission, or precardioversion for atrial group (RR 0·52, 95% CI 0·25–1·13, p ﬁ brillation with monitoring at ﬁ xed weekly intervals) p =0·0890; 17 vs 20 events, p=0·9285 for superiority; until two INRs fell within the target range after restarting table 2, ﬁ gure 2). The occurrence of ischaemic strokes warfarin. Patients were censored after the ﬁ rst major (including transient ischaemic attack) alone or myocardial bleed. We did calculations of surrogate variables in the infarction alone was also non-inferior (table 2). In a per-protocol population.
secondary analysis, that was not prespeciﬁ ed but based on www.thelancet.com/haematology Published online May 26, 2015 http://dx.doi.org/10.1016/S2352-3026(15)00073-3
Primary analysis 1–720 days: RR 0·69 (95% CI 0·43–1·12),p=0·0006 for non-inferiority, p=0·2107 for superiority Primary analysis 1–720 days: RR 0·61 (95% CI 0·28–1·31), Post-hoc analysis 181–720 days: p=0·0010 for non-inferiority, RR 0·62 (95% CI 0·38–1·30), p=0·1736 for superiority p=0·0492 for superiority Post-hoc analysis 181–720 days: RR 0·45 (95% CI 0·18–1·17), p=0·0536 for superiority with atrial ﬁbrillation (%) Cumulative major vascular events (%) Cumulative arterial Time from randomisation (days) Number at risk
Number at risk
Figure 3: Composite fatal and non-fatal major vascular events in all patients
Primary analysis 1–720 days: RR 0·98 (95% CI 0·48–1·20), monitored with either Fiix-prothrombin time (Fiix-PT) or quick prothrombin
p=0·0244 for non-inferiority, p=0·9114 for superiority time (PT; control group)
Post-hoc analysis 181–720 days: Thromboembolic events are assessed based on intention-to -monitor analysis and RR 0·82 (95% CI 0·31–2·19), major bleeding based on per-protocol (on-treatment) analysis. The primary p=0·8344 for superiority analysis is for days 1–720 and a secondary analysis was done for days 181–720 when only warfarin-experienced long-term patients are observed. RR=relative risk. PT=prothrombin time. INR=international normalised ratio.
the observation that a diﬀ erence emerged after 6 months, with atrial ﬁbrillation (%) we excluded the ﬁ rst 6 months after randomisation from Cumulative major bleeding in patients analysis. In this analysis, Fiix-PT monitoring led to a signiﬁ cant long-term reduction in thromboembolism (ﬁ gure 2A), six versus 15 cases (1·1% vs 2·2% per patient- Time from randomisation (days) Number at risk
year; RR 0·41, 0·16–1·05, p=0·0307 for superiority).
Major bleeding was analysed in the per-protocol population and was similar between the Fiix-PT group and the PT group (table 2, ﬁ gure 2B). Unlike thrombo- Figure 4: Cumulative event incidences during days 1–720 for primary eﬃ
outcome (A) and major bleeding (B) in the subgroup of patients with atrial
embolism, the incidence was consistent and showed no ﬁ brillation monitored with Fiix-prothrombin time (Fiix-PT) versus quick
divergence after 6 months. An intention-to-monitor prothrombin time (PT)
analysis did not change the outcome (19 vs 21 events, Non-fatal and fatal events are shown. RR=relative risk. PT=prothrombin time.
p=0·9090 for superiority; data not shown). Despite the INR=international normalised ratio.
small number of events, Fiix-PT was also non-inferior to
PT in the frequency of gastrointestinal bleeding, intra-
major vascular events and deaths from any cause did not cranial haemorrhage, intracerebral haemorrhage, and diﬀ er signiﬁ cantly between groups (p=0·0537 for other major bleeding (table 2). Incidences of clinically superiority; appendix p 6).
relevant non-major bleeding events and minor bleeding In a non-prespeciﬁ ed subgroup analysis of patients events with Fiix-PT were similar to the incidences with atrial ﬁ brillation, ten arterial thromboembolism observed with PT (table 2). The incidence of composite events (1·63% per patient-year) occurred in the major bleeding and non-major clinically relevant Fiix-PT group versus 17 (2·7% events per patient-year) in bleeding was also non-inferior in the Fiix-PT group the PT group (RR 0·62, 95% CI 0·29–1·33, (p=0·0351; appendix p 5).
=0·1763; ﬁ gure 4A). In the 12 patients died in the Fiix-PT group (1·45% per secondary analysis of thromboembolism beyond the patient-year) versus 16 (1·92% per patient-year) in the ﬁ rst 6 months in participants with atrial ﬁ brillation, six control group (table 2). Composite major vascular events events occurred in the Fiix-PT group (1·4% per patient- occurred in 27 patients in the Fiix group compared with year) versus 14 (3·2% per patient year) in the PT group 39 controls (table 2, ﬁ gure 3). In the secondary post-hoc (RR 0·45, 95% CI 0·18–1·17, p =0·0536). Major
analysis beyond 6 months of Fiix-PT monitoring, bleeding occurred in 14 participants with atrial signiﬁ cantly fewer major vascular events occurred in the ﬁ brillation in the Fiix-PT group (2·5% per patient-year) Fiix-PT than in the PT group (p=0·0492 for superiority; and 15 (2·5% per patient-year) in the PT group (RR 0·98, ﬁ gure 3). However, incidence of combined composite 0·48–2·0, p =0·0244; ﬁ gure 4B).
www.thelancet.com/haematology Published online May 26, 2015 http://dx.doi.org/10.1016/S2352-3026(15)00073-3
in the PT group than in the Fiix-PT group (table 3). Fiix-PT group (n=571),
PT group (n=573),
Patients with major events had a higher variance growth rate than did those without major events (table 3).
Number of monitoring tests 16 months into the study, an ISI calibration issue occurred, causing the median Fiix-INR to be reported 0·2 points higher than previously despite control samples being within limits, whereas the PT-INR did not Number of tests within deﬁ ned INR ranges change. This issue would have mainly aﬀ ected the analysis during days 450–630 of observation. A corresponding temporary lowering of times in range in the Fiix group but not the PT group was observed and Days between monitoring tests might have led to unneeded and aberrant dose reductions in the Fiix-PT group. The diﬃ culty was identiﬁ ed to coincide with the receipt of a new order of the DEKS ISI Dose changes per monitoring test in each calibrator. When the old batch DEKS calibrator could be obtained again and used to recalibrate ISI, the median Fiix-INR returned to the previous levels, and a higher time in the range returned. The appendix (p 4) shows reanalysed data from table 3 after exclusion of R-INRs done during this calibration period.
Number of dose changes per patient per year 5·3 (2·2–10·9) 6·5 (2·3–11·7) Warfarin monitoring with Fiix-PT (Fiix-INR) stabilised and 4·1 (2·0–7·1) 5·0 (2·2–8·0) increased the intensity of anticoagulation in a well TTR of each patient managed typical anticoagulation management centre patient population with an INR target of 2–3. This resulted in fewer thromboembolisms per year in patients monitored with Fiix-PT than in patients monitored with standard PT, without an increase in bleeding. Additionally, although not a prespeciﬁ ed analysis, we noted signiﬁ cantly VGR, patients with major events fewer thromboembolisms, with a low bleeding incidence VGR, patients without major events in patients who were monitored long term with Fiix-PT, Daily warfarin dose (mg) 4·7 (3·3–6·3) 4·7 (3·3–6·3) compared with those monitored by PT-INR, although this was based on a small number of events.
Data are median (IQR), unless otherwise stated. Per-protocol (on-treatment) analysis of the safety of monitoring Warfarin anticoagulation instability from food and method. Only patients that have three or more tests were included in the TTR interval calculation and warfarin initiation periods were excluded from the TTR analysis until two INRs fell within target range. Fiix-PT=Fiix- drug interactions and patient non-compliance has long prothrombin time. PT=prothrombin time. NA=not applicable. INR=international normalised ratio. TTR=time within been regarded to contribute to adverse events during target range by Rosendaal method. VGR=variance growth rate by B2 method—ie, between test variance in INR.
warfarin management.1 On the basis of our results, Table 3: Surrogate outcome variables
however, an additional source of INR ﬂ uctuation emerges—namely, a confounding side-eﬀ ect of the PT. The PT is equally sensitive to reduction in the activity of Dose change frequency was reduced with Fiix-PT factor II, factor VII, or factor X.8 However, the long half- monitoring, particularly long term (ie, after day 180, life factor II and factor X and not the short half-life p=0·0101), with no signiﬁ cant diﬀ erences in the ﬁ rst factor VII mainly aﬀ ect clot and thrombus formation.8,9,23 6 months; however, the median daily warfarin dose was By replacement of PT with Fiix-PT, the confounding identical in both groups, 4·7 mg (IQR 3·3–6·3; table 3). eﬀ ect of the ﬂ uctuating factor VII in the test sample on With long-term Fiix-PT monitoring (180–720 days), INR is eliminated, which improves stability. However, as monitoring tests were reduced by 5·8%. More monitoring evident from our study, a special Fiix-INR calibrator will tests were within the target range in the Fiix-PT group probably be needed.
than in the control group and fewer tests with an INR less An increased time in range was evident in the Fiix-PT than 2·0 occurred in the Fiix-PT group (table 3). The group during the ﬁ rst 6 months, but the long-term median percent time in range in the control group was clinical beneﬁ t only emerged subsequently. Reduced 81%, 80%, 81%, and 79% during four consecutive thromboembolism in the Fiix-PT group was associated 6-month observation periods, whereas in the Fiix- with proportionally more tests in range and fewer tests monitoring group the median percent time in range was below range (table 3), whereas the proportion with an 85%, 85%, 80%, and 87%, respectively. INR ﬂ uctuation increased INR with Fiix-PT was similar to PT-INR, the measured as variance growth rate was signiﬁ cantly higher similar proportion above target range possibly partly www.thelancet.com/haematology Published online May 26, 2015 http://dx.doi.org/10.1016/S2352-3026(15)00073-3
explaining similar, albeit low, bleeding incidence in both not yet have antidotes for emergency reversal and might study groups. The signiﬁ cantly lower variance growth not be straightforward in elderly patients and in those rate in the Fiix-PT group than the PT group suggests less with reduced kidney function.
variability of anticoagulation monitored with the Fiix- Our trial compared in a randomised and masked INR. Although why the clinical eﬀ ect of improved manner the incidence of thromboembolism and bleeding anticoagulation stability only becomes evident after in 1148 patients treated with warfarin monitored by two 6 months of Fiix-PT monitoring is unclear, we suggest diﬀ erent tests at a single medical centre. We regard the that this could be explained by the lower variance growth masking to have been successful because the dosing staﬀ rate in the Fiix-PT group because a high variance growth could not decipher which monitoring group a patient rate (unstable INR) has been retrospectively shown to be was assigned to. A larger conﬁ rmatory multicentre study predictive of clinical events 3–6 months later.15 would, however, be preferred because such a trial could Alternatively, other long-term eﬀ ects such as eﬀ ects on provide increased statistical power and the ability to do the vessel wall might be implicated. Our trial might be further subgroup analyses that would increase the the ﬁ rst prospective study to show the eﬀ ect of variance generalisability of the results. The setting of our trial, growth rate on clinical events in patients receiving nevertheless, was a typical anticoagulated population at wafarin.
one specialised academic medical centre applying Large multicentre clinical trials have suggested that specialised anticoagulation software that ensured that unmonitored direct oral anticoagulants are clinically non- treatment and follow-up between the two groups was inferior or even superior to warfarin in both non-valvular identical. The only diﬀ erence was the absence of the atrial ﬁ brillation12,24–26 and in venous thromboembolism.27–29 eﬀ ect of factor VII in the test sample on monitoring in However, the beneﬁ t of direct oral anticoagulants in atrial the Fiix-PT group. Despite not detecting factor VII in the ﬁ brillation has been reported mainly at study sites24,30,31 Fiix-PT group, anticoagulation stability was improved, that did not maintain a high time within target range in thromboembolism was reduced in the long term and patients in the warfarin control group and this inevitably bleeding incidence remained low. Because we exaggerates the reported beneﬁ t of direct oral investigated only few warfarin-naive patients, conclusions anticoagulants. Our subgroup analysis of patients with should not be made from this study on clinical outcome non-valvular atrial ﬁ brillation (75% of the study diﬀ erences in this group. Patients in the Fiix-PT group population) raises the question of whether management were dosed by a protocol that was designed for the of patients with atrial ﬁ brillation on warfarin with Fiix-PT ﬂ uctuating PT-INR. Therefore, dosing staﬀ , on the basis instead of PT has the potential to improve long-term of their PT-INR management experience, might have outcomes in atrial ﬁ brillation even further than that responded to the masked Fiix-INR with unnecessary achievable with unmonitored direct oral anticoagulants. dose adjustments that might have caused a bias in favour In the patients with atrial ﬁ brillation in this study, Fiix-PT of the PT group. Hence, even further improvements led to a long-term reduction after 6 months in arterial could possibly be achieved with the Fiix-INR than those thromboembolism (combined ischaemic strokes, reported in this trial. Although minor diﬀ erences were transient ischaemic attacks, other arterial embolism, and noted in patient characteristics, mainly indications for myocardial infarction). In the direct oral anticoagulant warfarin, they are probably to be expected in a study of trials,12,24–26 the change in arterial thromboembolism this size. Also, although slightly fewer patients had atrial (including myocardial infarction but excluding transient ﬁ brillation in the Fiix-PT group than in the PT-INR ischaemic attacks) in the active study groups ranged from group, somewhat more had a high-risk CHA DS -VASC +11% to –19% compared with outcome during only score in the Fiix-PT group. moderately well managed warfarin monitored with PT Finally, an unexpected Fiix-INR calibration issue with time within target range in the 58–65% range. occurred 16 months into the study that might have Notably, the absolute incidence of total arterial caused an erroneous dose reduction that would not have thromboembolism in our patients with atrial ﬁ brillation been in favour of an improved outcome in the active Fiix-monitored with Fiix-PT was lower than in the direct oral PT group. Figure 3 might indicate that the composite anticoagulation studies and the low major bleeding major vascular event incidence increased in the Fiix-PT incidence, including intracranial haemorrhage incidence, group during this period (mainly days 450–630) because compared favourably with direct oral anticoagulants. the event curves seem to converge then diverge again Improved stability of warfarin management with Fiix-PT, once the issue was corrected. The issue was not reported with resulting improved long-term outcome and a low with PT-INR calibration, which the calibrator is designed bleeding incidence could therefore lead to less impetus to for. Hence, diﬀ erent calibrators will probably be needed switch patients to the unmonitored and more expensive to standardise the Fiix-INR.
direct oral anticoagulants. Although unmonitored direct We conclude that monitoring warfarin with Fiix-PT oral anticoagulants might be convenient to use, the improves the stability of warfarin management and is eﬃ cacy, safety, reversibility, titratability, and aﬀ ordability clinically at least non-inferior to PT-INR. Furthermore, a must all be taken into account. These new drugs also do post-hoc analysis suggests that thromboembolism is www.thelancet.com/haematology Published online May 26, 2015 http://dx.doi.org/10.1016/S2352-3026(15)00073-3
reduced during long-term treatment when standard INR 13 Schulman S, Kearon C, Subcommittee on Control of Anticoagulation of the Scientiﬁ c and Standardization Committee of monitoring is replaced with Fiix-INR. the International Society on Thrombosis and Haemostasis. Deﬁ nition of major bleeding in clinical investigations of PTO chaired the study. PTO, CWF, OSI, and BRG designed the trial. antihemostatic medicinal products in non-surgical patients.
J Thromb Haemost 2005; 3: 692–94.
PTO, CWF, OSI, SJJ, SHL, HMJ, and BRG analysed the data. PTO, CWF, OSI, and BRG interpreted the data. PTO wrote the report draft. 14 Rosendaal FR, Cannegieter SC, van der Meer FJ, Briet E. A method to determine the optimal intensity of oral anticoagulant therapy. CWF, OSI, DOA, ESB, MKM, SJJ, HMJ, BV, PSG, and BRG edited the Thromb Haemost 1993; 69: 236–39.
report. OSI, BV, and PSG were members of the data safety and 15 Ibrahim S, Jespersen J, Poller L, European Action monitoring board. DOA, ESB, and MKM were adjudication committee on Anticoagulation. The clinical evaluation of International members. SHL was the statistician, consultant, and responsible for Normalized Ratio variability and control in conventional oral statistical calculations. OSI was also responsible for statistical anticoagulant administration by use of the variance growth rate. J Thromb Haemost 2013; 11: 1540–46.
Declaration of interests
16 Abdelhaﬁ z AH, Wheeldon NM. Results of an open-label, PTO and BRG together with the Landspitali and University of Iceland prospective study of anticoagulant therapy for atrial ﬁ brillation in have applied for a patent (patent pending) for the Fiix-prothrombin time an outpatient anticoagulation clinic. Clin Ther 2004; 26: 1470–78.
invention. All other authors declare no competing interests.
17 Menendez-Jandula B, Souto JC, Oliver A, et al. Comparing self- management of oral anticoagulant therapy with clinic management: a randomized trial. Ann Intern Med 2005; 142: 1–10.
We thank Brynja Laxdal, Gunnhildur Magnusdottir, 18 Palareti G, Leali N, Coccheri S, et al. Bleeding complications of oral Tinna Halldorsdottir, Hallgerdur Bjarnhedinsdottir, Halla Arnardottir, anticoagulant treatment: an inception-cohort, prospective Kristin A Einarsdottir, Oddny Olafsdottir, and Loic Letertre for their help.
collaborative study (ISCOAT). Italian Study on Complications of
Oral Anticoagulant Therapy. Lancet 1996; 348: 423–28.
19 Wilson SJ, Wells PS, Kovacs MJ, et al. Comparing the quality of oral Ageno W, Gallus AS, Wittkowsky A, et al. Oral anticoagulant anticoagulant management by anticoagulation clinics and by family therapy: antithrombotic therapy and prevention of thrombosis, physicians: a randomized controlled trial. CMAJ 2003; 169: 293–98.
9th edn: American College of Chest Physicians Evidence-Based
Clinical Practice Guidelines. Chest 2012; 141: e44S–88S.
20 Julious S. Sample sizes for clinical trials. Florida: Chapman & Hall/ Quick AJ, Stanley-Brown M, Bancroft FW. A study of the coagulation
defect in hemophilia and in jaundice. Am J Med Sci 1935; 190: 501–11.
21 Miettinen O, Nurminen M. Comparative analysis of two rates. Stat Med 1985; 4: 213–26.
Owren PA, Aas K. The control of dicumarol therapy and the quantitative determination of prothrombin and proconvertin. 22 Farrington CP, Manning G. Test statistics and sample size Scand J Clin Lab Invest 1951; 3: 201–08.
formulae for comparative binomial trials with null hypothesis of non-zero risk diﬀ erence or non-unity relative risk. Stat Med 1990; Kirkwood TB. Calibration of reference thromboplastins and standardisation of the prothrombin time ratio. Thromb Haemost
1983; 49: 238–44.
23 Xi M, Beguin S, Hemker HC. The relative importance of the factors II, VII, IX and X for the prothrombinase activity in plasma Poller L, Shiach CR, MacCallum PK, et al. Multicentre randomised of orally anticoagulated patients. Thromb Haemost 1989; 62: 788–91.
study of computerised anticoagulant dosage. European Concerted
Action on Anticoagulation. Lancet 1998; 352: 1505–09.
24 Granger CB, Alexander JH, McMurray JJ, et al. Apixaban versus warfarin in patients with atrial ﬁ brillation. N Engl J Med 2011; Onundarson PT, Einarsdottir KA, Gudmundsdottir BR. Warfarin anticoagulation intensity in specialist-based and in computer-
assisted dosing practice. Int J Lab Hematol 2008; 30: 382–89.
25 Giugliano RP, Ruﬀ CT, Braunwald E, et al. Edoxaban versus warfarin in patients with atrial ﬁ brillation. N Engl J Med 2013; Wieloch M, Sjalander A, Frykman V, Rosenqvist M, Eriksson N, Svensson PJ. Anticoagulation control in Sweden: reports of time in therapeutic range, major bleeding, and thrombo-embolic 26 Patel MR, Mahaﬀ ey KW, Garg J, et al. Rivaroxaban versus warfarin complications from the national quality registry AuriculA. in nonvalvular atrial ﬁ brillation. N Engl J Med 2011; 365: 883–91.
Eur Heart J 2011; 32: 2282–89.
27 Bauersachs R, Berkowitz SD, Brenner B, et al. Oral rivaroxaban for Gudmundsdottir BR, Francis CW, Bjornsdottir AM, Nellbring M, symptomatic venous thromboembolism. N Engl J Med 2010; Onundarson PT. Critical role of factors II and X during coumarin anticoagulation and their combined measurement with a new 28 Schulman S, Kearon C, Kakkar AK, et al. Dabigatran versus Fiix-prothrombin time. Thromb Res 2012; 130: 674–81.
warfarin in the treatment of acute venous thromboembolism. Zivelin A, Rao LV, Rapaport SI. Mechanism of the anticoagulant N Engl J Med 2009; 361: 2342–52.
eﬀ ect of warfarin as evaluated in rabbits by selective depression of 29 Agnelli G, Buller HR, Cohen A, et al. Oral apixaban for the individual procoagulant vitamin K-dependent clotting factors. treatment of acute venous thromboembolism. N Engl J Med 2013; J Clin Invest 1993; 92: 2131–40.
10 Ansell J, Hirsh J, Hylek E, Jacobson A, Crowther M, Palareti G. 30 Wallentin L, Yusuf S, Ezekowitz MD, et al. Eﬃ cacy and safety of Pharmacology and management of the vitamin K antagonists: dabigatran compared with warfarin at diﬀ erent levels of American College of Chest Physicians Evidence-Based Clinical international normalised ratio control for stroke prevention in atrial Practice Guidelines, 8th edn. Chest 2008; 133: 160S–98S.
ﬁ brillation: an analysis of the RE-LY trial. Lancet 2010; 376: 975–83.
11 Lip GY, Nieuwlaat R, Pisters R, Lane DA, Crijns HJ. Reﬁ ning 31 Singer DE, Hellkamp AS, Piccini JP, et al. Impact of global clinical risk stratiﬁ cation for predicting stroke and geographic region on time in therapeutic range on warfarin thromboembolism in atrial ﬁ brillation using a novel risk factor- anticoagulant therapy: data from the ROCKET AF clinical trial. based approach: the euro heart survey on atrial ﬁ brillation. J Am Heart Assoc 2013; 2: e000067.
Chest 2010; 137: 263–72.
12 Connolly SJ, Ezekowitz MD, Yusuf S, et al. Dabigatran versus warfarin in patients with atrial ﬁ brillation. N Engl J Med 2009;
www.thelancet.com/haematology Published online May 26, 2015 http://dx.doi.org/10.1016/S2352-3026(15)00073-3
Race and Medicine Genetic studies of population differences, although controversial, promise David Goldstein of University College in clues to disease as well as new drug targets, scientists believe London agrees: "If you say on average the difference between West Africans and Eu- Mention race and medicine in a group of racial identity biologically irrelevant. But ropeans is slight, that does not rule out a