SEARCH

SEARCH BY CITATION

Keywords:

  • anticoagulants;
  • atrial fibrillation;
  • bias;
  • meta-analysis;
  • stroke

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Addendum
  8. Acknowledgements
  9. Funding
  10. Disclosure of Conflict of Interest
  11. References

Background

The prospective, randomized, open, blinded endpoint evaluation (PROBE) design has been proposed as a valid alternative to the double-blind (DB) design for trials comparing new oral anticoagulants (NOAs) with INR-adjusted vitamin K antagonists in patients with non-valvular atrial fibrillation (NVAF).

Objectives

To determine whether the observed treatment effects of NOAs in patients with NVAF differ between PROBE/open-label trials and DB trials.

Methods

All phase II or III trials were eligible. The main efficacy and safety outcomes were stroke/systemic embolism (SSE) and major bleeding, respectively. Other outcomes included ischemic SSE, hemorrhagic stroke, intracranial and extracranial bleeding, myocardial infarction, and all-cause and cardiovascular mortality. Interaction (Cochran's chi-squared test) between PROBE and DB designs was tested.

Results

Thirteen studies (61 620 patients) were included. For SSE, a greater treatment effect of NOAs vs. INR-adjusted warfarin was observed in PROBE trials (RR 0.76, CI 0.65–0.89) compared with DB trials (RR 0.88, CI 0.78–0.98), but the interaction test was non-significant (P = 0.16). A significant 67% enhancement of treatment effect was found with PROBE/open-label trials compared with DB trials (interaction test, P = 0.05) for hemorrhagic stroke. No other interaction was significant. A non-significant interaction (P = 0.07) between oral direct thrombin inhibitors (RR 0.33; 0.22–0.51) and factor Xa inhibitors (RR 0.54; 0.40–0.72) was seen. No heterogeneity was found for any outcome.

Conclusions

Our meta-analysis showed no significant interaction of study design for the main efficacy and safety outcomes. However, the non-significantly exaggerated reduction in SSE suggests interdependence of treatment effect and PROBE design, especially for hemorrhagic stroke.


Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Addendum
  8. Acknowledgements
  9. Funding
  10. Disclosure of Conflict of Interest
  11. References

Non-valvular atrial fibrillation (NVAF) is a major cause of ischemic stroke and systemic embolism and is consequently characterized by increased mortality and morbidity and higher costs of medical care [1, 2]. Vitamin K antagonists (VKAs), principally warfarin, have been proven to be highly effective in preventing thromboembolic events in patients with paroxysmal, persistent or permanent NVAF [3]. In 29 randomized trials involving more than 28 000 patients pooled according to meta-analytic methods, adjusted-dose warfarin reduced the risk of stroke by 64% compared with the control and by 37% compared with aspirin, but at the cost of an increased risk of bleeding [3]. Furthermore, warfarin was associated with a 26% reduction in all-cause mortality in randomized, controlled trials when compared with no anticoagulation therapy in patients with NVAF [3].

New oral anticoagulants (NOAs), directly inhibiting thrombin or factor (F) Xa, have recently been developed. Their wide therapeutic windows permit the use of fixed doses without any need for monitoring [4, 5]. These new drugs could potentially overcome the well-known limitations of VKAs, such as slow onset of action, need for regular monitoring of the international normalized ratio (INR) on blood samples, narrow therapeutic windows, marked inter-individual variations in drug metabolism, and multiple drug-drug and drug-food interactions, all of which lead to an increased risk of bleeding [6-8].

Different designs with respect to blinding of participants and researchers have been used to assess the efficacy and safety of NOAs, namely double-blind (DB), open-label, and prospective, randomized, open, blinded endpoint evaluation (PROBE) designs [9, 10]. Several authors have presented evidence indicating an association between lack of blinding and over-optimistic estimates of treatment effects for subjectively assessed outcomes such as cardiovascular events [11, 12]. However, the PROBE design has been debated as a valid alternative to the development of NOAs, as it improves the external validity of trials through the presence of an event adjudication committee, which guarantees the absence of information bias and the facilitation of patient recruitment, thereby enhancing sample representivity [13, 14]. As a DB design of studies comparing VKAs with other anticoagulants necessitates a double-dummy strategy for INR monitoring, which is very difficult to perform, a demonstration that this design is not essential would be of great help in the development of NOAs. We performed a meta-analysis of all randomized NVAF trials to determine whether PROBE and open designs were associated with an enhancement of treatment effects in this clinical setting.

Methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Addendum
  8. Acknowledgements
  9. Funding
  10. Disclosure of Conflict of Interest
  11. References

Inclusion criteria

The meta-analysis was performed according to a prospectively developed protocol (available upon request from the corresponding author), which prespecified the research objective, the search strategy, the study eligibility criteria, and the methods of data extraction and statistical analysis. All subgroup variables were defined before analysis.

All randomized, controlled trials conducted in patients with NVAF were eligible for inclusion in the present meta-analysis. Patients in the control group had to receive VKA and patients in the treated group had to receive an oral FXa or thrombin inhibitor. Double-blind and open-label trial designs with or without a blinded evaluation of outcomes were eligible. Studies were excluded if they concerned duplicate cohorts. Trials with short-term follow-ups (< 12 weeks) were excluded.

Data sources and searches

We searched Medline (PubMed) and Embase up to October 2012 using sensitive methods and employing the keywords: rivaroxaban, apixaban, betrixaban, edoxaban (DU-176b), eribaxaban, ximelagatran, dabigatran (BIBR1048), LY 517717, darexaban (YM150), letaxaban, AZD0837, TTP889, RB006, MCC977 and TAK442 [15, 16]. Search terms included combinations of free text and medical subject headings (MeSH or Emtree). The complete search strategies may be requested from the authors. We also reviewed the citations of the retrieved studies, reviews and meta-analyses obtained by searches of PubMed and Embase. Unpublished and ongoing trials were sought in clinical trial repositories, including those of the National Institute of Health, the National Research Register, Current Controlled Trials, Meta-Embol and Trials Central. We also searched the Internet using the keywords listed above, including websites dedicated to the dissemination of clinical trial results, such as TheHeart.org and the US Food and Drug Administration web site, as well as the web sites maintained by the drug manufacturers and product information sheets.

Unpublished studies were included in the meta-analysis if the design had been previously published in detail and the patient characteristics, follow-up and main results had been presented at international congresses. No restrictions with regard to language or small population size were applied. All qualifying studies were assessed for adequate blinding of randomization, completeness of follow-up, and objectivity of the outcome assessment. Phase II trials were included if an arm corresponding to the dose of interest was selected in a subsequent phase III trial. When an abstract in congress proceedings and a full paper referred to the same trial, only the full article was included in the analysis. When two or more papers reported the same study (for example one for the protocol, a second for the results and a third for additional safety data), we included all that were useful for our purpose.

Outcomes

The main outcomes were the composite of stroke and systemic embolism (SSE; main efficacy outcome), major bleeding (main safety outcome) as defined by the International Society on Thrombosis and Haemostasis [17], intracranial bleeding, and all-cause and cardiovascular mortality. Other outcomes included ischemic stroke and systemic embolism separately, hemorrhagic stroke, myocardial infarction, and extracranial bleeding (including hemorrhagic stroke).

Data extraction

Studies were selected and data extracted from publications by two reviewers independently (JCL and CC). A 2 × 2 table was constructed for each study to compute the relative risk. If the data required to complete this table were missing or incomplete, the hazard ratio and its confidence interval were extracted and directly included in the pooled results [18]. Data regarding inclusion criteria, proportion of VKA-naive patients, proportion of patients with a CHADS2 score < 2, treatments, pharmacological class of study drugs, duration of follow-up, type of design (DB vs. PROBE or open-label), study phase (II or III) and randomization procedure were extracted from each individual study. The risk of bias was assessed by the Cochrane Collaboration's tool [19]. The results obtained for the intention-to-treat population were used for the main analyses. Disagreements were resolved by consensus. When the trial evaluated different dosages of the NOA and therefore allowed more than one comparison with the VKA arm, we divided the number of events and the number of participants in the VKA arm accordingly so that each was counted only once.

Statistical analysis

The relative risks (RRs) or hazard ratios were weighted by the inverse of their variance and combined using the logarithm of RR method according to a fixed-effects model with EasyMA 2.0 [20, 21]. The heterogeneity between studies was assessed using Cochran's chi-squared test and I2 [22]. We retained heterogeneity at P < 0.10. In the event of unexplained heterogeneity, the results were pooled according to a random-effects model. Interaction was systematically tested for PROBE or open-label vs. DB design and pharmacological class and was considered as significant at P < 0.10. Combined effect estimates were also calculated separately for trials with and without a DB design. We used the logistic regression models described previously to estimate RR ratios, comparing treatment effects in trials with and without a DB design [23]. Enhancement of the treatment effect, in terms of RR ratio, was determined by dividing the RR values obtained in PROBE trials by those obtained in DB trials. For example, a RR ratio of 1.3 would imply that the estimates of treatment effect were enhanced by 30% in trials with a PROBE or open-label design compared with trials with a DB design. We derived 95% confidence intervals (CIs) using robust standard errors allowing for heterogeneity between meta-analyses [23].

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Addendum
  8. Acknowledgements
  9. Funding
  10. Disclosure of Conflict of Interest
  11. References

Literature search and study selection

We identified 1151 references through electronic searches and 17 references by manual searches and contacts with experts (Fig. 1). Among these, 13 studies (including 61 620 patients) were eligible for analysis [9, 10, 24-35]. The funnel plot showed an asymmetry around the point estimate due to the lack of expected phase II studies in the bottom right-hand quadrant (Fig. 2). This asymmetry was due to the intrinsic properties of phase II studies, which are designed with the aim of selecting the dose with the best safety and efficacy profile and excluding doses showing adverse profiles. Patient characteristics are shown in Table 1 and study designs, methodological features and patient characteristics in Tables 2 and 3. The risk of bias according to the Cochrane Collaboration's tool mainly reflected the lack of blinding (Table 3). Random sequence generation was not reported in one small phase III study [34] and two phase II studies [31, 32], and allocation concealment was not reported in one small phase III study [34] and four phase II studies [27, 31-33]. Patient inclusion criteria were based on various combinations of the known risk factors for thromboembolism included in the CHADS2 score and consequently the proportion of patients with a CHADS2 score < 2 differed greatly from study to study, ranging from 0 to 47%. The proportion of VKA-naive patients ranged from 10 to 64%. Four trials had a DB design [9, 25, 26, 34], three were open-label studies [28, 31, 32] and the other six had a PROBE design (Table 2).

Table 1. Patient characteristics in each trial included in the meta-analysis
Trial, yearMean age (yr)Men (%)VKA-naive patients (%)Patients with CHADS2 < 2 (%)Lost to follow-up (%)TTR (%)
  1. Pts, patients; NR, not reported; TTR, time during which the INR was in the therapeutic range; yr, years.

SPORTIF III, 2003706927301.366
SPORTIF V, 2005726916260.668
NCT 01136408, 20076888NRNRNRNR
PETRO, 20077082NRNR057
Lip et al. [28]69682930058
RE-LY, 2009716450320.164
Weitz et al. [29]65626400NR
ROCKET, 201173603700.0155
ARISTOTLE, 2011706543342.162
ARISTOTLE-J, 2011708315430NR
Chung et al. [24]656548470.845
J-ROCKET-AF, 201271811001.565
Yamashita et al. [32]698215NRNR73 for pts aged < 70 yrs; 83 for pts aged ≥ 70 yrs
Table 2. Characteristics of the studies included
Trial, yearDesignNOA classNOA treatmentRandomization methodITTMean follow-up (months)Timing of endpoint (ITT)
  1. Bid, twice daily; ITT, intention-to-treat analysis; IVRS, interactive voice-response system; NR, not reported; od, once daily; PROBE, prospective, randomized, open, blinded-endpoint; SSE, stroke and systemic embolism. *Additional data were provided by the trial investigators. Doses were reduced in patients with renal failure.

SPORTIF III, 2003Phase III PROBEThrombin inhibitorXimelagatran 36 mg bidComputer-generated and centralized (IVRS)Death, SSE, myocardial infarction17.43 months after MI, bleeding, SSE
SPORTIF V, 2005Phase III Double-blindThrombin inhibitorXimelagatran 36 mg bidComputer-generated and centralized (IVRS)Death, SSE20.03 months after MI, bleeding, SSE
NCT 01136408, 2007*Phase II OpenThrombin inhibitorDabigatran 110 mg bid Dabigatran 150 mg bidNRNR3NR
PETRO, 2007Phase II PROBEThrombin inhibitorDabigatran 150 mg bidDose stratificationPer protocol3.012-week follow-up visit
Lip et al. [28]Phase II OpenThrombin inhibitorAZD0737 300 mg odComputer-generated and centralized (IVRS) – stratification according to prior VKA statusNR5.0NR
RE-LY, 2009Phase III PROBEThrombin inhibitorDabigatran 110 mg bid Dabigatran 150 mg bidComputer-generated and centralized (IVRS)All outcomes24.0NR
Weitz et al. [29]Phase II PROBEFactor Xa inhibitorEdoxaban 30 mg od Edoxaban 60 mg odComputer-generated and centralized (IVRS)NR4.0End of follow-up
ROCKET, 2011Phase III Double-blindFactor Xa inhibitorRivaroxaban 20 mg odComputer-generated and centralized (IVRS)SSE23.2Switch to open label-treatment with conventional anticoagulants
ARISTOTLE, 2011Phase III Double-blindFactor Xa inhibitorApixaban 5 mg bidComputer-generated and centralized (IVRS) – stratification according to investigative site and prior VKA statusAll outcomes21.6Efficacy outcomes: cut-off date Safety outcomes: 2 days after last dose received
ARISTOTLE-J, 2011Phase II PROBEFactor Xa inhibitorApixaban 2.5 or 5 mg bidNRPer protocol3End of follow-up
Chung et al. [24]Phase II PROBEFactor Xa inhibitorEdoxaban 30 mg od Edoxaban 60 mg odComputer-generated and centralized (IVRS)All outcomes3End of follow-up
J-ROCKET-AF, 2012Phase III Double-blindFactor Xa inhibitorRivaroxaban 15 mg odNRSSE5.812-week follow-up visit
Yamashita et al. [32]Phase II OpenFactor Xa inhibitorEdoxaban 30 mg od Edoxaban 60 mg odStratification according to prior VKA statusPer protocol38-week follow-up visit
Table 3. Assessment of the risk of bias according to the Cochrane Collaboration's tool
 Random sequence generationAllocation concealmentBlinding of participants and personnelBlinding of outcome assessmentIncomplete outcome dataSelective reporting
  1. +, low risk of bias; −, high risk of bias; ?, unclear risk of bias.

SPORTIF III, 2003+++++
SPORTIF V, 2005++++++
NCT 01136408, 2007???+
PETRO, 2007+?+++
Lip et al. [28]++++
RE-LY, 2009+++++
Weitz et al. [29]+++++
ROCKET, 2011++++++
ARISTOTLE, 2011++++++
ARISTOTLE-J, 2011+?+++
Chung et al. [24]+++++
J-ROCKET-AF, 2012??++++
Yamashita et al. [32]??+
image

Figure 1. Flow chart of trial selection.

Download figure to PowerPoint

image

Figure 2. Funnel plot of primary efficacy outcome (stroke or systemic embolism). Grey squares indicate phase II study arms and black squares phase III study arms. RR, relative risk; SE (log RR), standard error of relative risk.

Download figure to PowerPoint

Oral direct antithrombin inhibitors were assessed in six studies [10, 26-28, 30, 31] and oral direct FXa inhibitors in seven studies [9, 24, 25, 29, 32-34] (Tables 2 and 4). A reduced dose of the NOA was used in patients with renal failure in two trials (apixaban 2.5 mg and rivaroxaban 15 mg) and in Japanese patients in one trial (rivaroxaban 15 mg) [9, 25, 34]. All studies used adjusted-dose warfarin (target INR, 2.0–3.0) as the control, except for three trials in Japan in which the targeted INR was lower in patients aged > 70 years [32-34]. Depending on the study, the INR was in the therapeutic range for 45–83% of the treatment period (Table 1).

Table 4. Trial results
Trial name, YearGroupPatients (n)Stroke/SEIschemic stroke/SEHemorrhagic strokeMBIntracranial bleedingAll-cause mortalityCardiovascular mortalityAMI
  1. A, apixaban; AMI, acute myocardial infarction; bid, twice daily; D, dabigatran; E, edoxaban; MB, major bleeding; NOA, new oral anticoagulant; NR, not reported; od, once daily; R, rivaroxaban; SE, systemic embolism; VKA, vitamin K antagonist; X, ximelagatran.

SPORTIF III, 2003X 36 mg bid1704403642910784024
VKA1703564894113793313
SPORTIF V, 2005X 36 mg bid196051512707116NR26
VKA196237372939123NR37
NCT 01136408, 2007D 110 bid4600000000
D 150 bid5800010000
VKA6211010000
PETRO, 2007D 150 mg bid16600000NRNRNR
VKA7000000NRNRNR
Lip et al. [28]AZD073715100000001
VKA31811020201
RE-LY, 2009D 110 mg6015183170143222744628998
D 150 mg6076134124123993643827497
VKA6022202156454218748731775
Weitz et al. [29]E 30 mg od2351000NRNR22
E 60 mg od2341001NRNR02
VKA2503301NRNR20
ROCKET, 2010R 15 or 20 mg od70812691542939555208170101
VKA70903061835038684250193126
ARISTOTLE, 2011A 2.5 or 5 mg912021217740327526030.89 (0.76–1.04)90
VKA908126519278462122669 102
ARISTOTLE-J 20112.5 mg bid7400000000
5 mg bid7400000000
VKA7432111000
Chung et al., [24]E 30 mg od7900000000
E 60 mg od8000000110
VKA7500020100
J-ROCKET-AF, 2012R 10 or 15 mg od6402283264763
VKA640261843010521
Yamashita et al., [32]E 30 mg od13000000000
E 60 mg od13000021100
VKA12500000100

Assessment of the impact of study design on the results

Stroke and systemic embolism

A greater treatment effect was observed in trials with a PROBE/open-label design, showing a 24% relative risk reduction (nine trials including 37 665 patients; RR, 0.76; CI, 0.65–0.89), than in trials with a DB design, indicating a 12% relative risk reduction (four trials including 23 955 patients; RR, 0.88; CI, 0.78–0.98). The interaction test was not significant (P = 0.16), indicating a non-significant enhancement of the treatment effect by 16% (CI −5; + 41%; Fig. 3).

image

Figure 3. Analysis of interaction between study design and treatment effect.

Download figure to PowerPoint

Major bleeding and intracranial bleeding

We found no interaction with respect to major bleeding. The interaction tests showed a non-significant enhancement of the treatment effect (+33%, CI −9; + 92%) in trials with a PROBE/open-label design (RR, 0.40; CI, 0.30; 0.54) compared with those with a DB design (RR, 0.52; CI, 0.43; 0.66; interaction test; P = 0.16; Fig. 3) for intracranial bleeding. This result was not associated with an interaction concerning pharmacological class (Fig. 4).

image

Figure 4. Analysis of interaction between pharmacological class and treatment effect.

Download figure to PowerPoint

Cardiovascular mortality and all-cause mortality

No interaction was concluded for these outcomes.

Other outcomes

Interaction was evident with respect to the risk of hemorrhagic stroke, which was reduced to a greater extent in trials with a PROBE/open-label design (RR, 0.33; CI, 0.21–0.50) than in those with a DB design (RR, 0.55; CI, 0.41–0.73; interaction test; P = 0.05; Fig. 3). The test for interaction was also significant for pharmacological class (P = 0.07; Fig. 4). With respect to myocardial infarction, the subgroup analysis according to study design (PROBE/open-label vs. DB) revealed a significant interaction (P < 0.0001), DB trials showing a reduction in the risk of myocardial infarction (RR, 0.82; CI, 0.69–0.98) whereas PROBE/open-label trials indicated a significant increase (RR, 1.37; CI, 1.07–1.75). The test of interaction between these two trial subsets was significant (P < 0.0001) and corresponded to a RR enhancement of −35% (CI, −15; −49%; Fig. 3). In addition, the risk of myocardial infarction was reduced by oral direct FXa inhibitors (RR, 0.85; CI, 0.70–1.03) compared with VKA, the decrease in risk being less pronounced in the case of thrombin inhibitors (RR, 1.20; CI, 0.96–1.50) with a significant interaction effect (P = 0.02; Fig. 4). No interaction was found for ischemic stroke/systemic embolism or extracranial bleeding.

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Addendum
  8. Acknowledgements
  9. Funding
  10. Disclosure of Conflict of Interest
  11. References

The aim of this meta-analysis was to determine whether open designs with or without blinded outcome assessments were associated with an enhancement of the treatment effects compared with double-blind designs in randomized trials evaluating NOA in patients with NVAF. Our meta-analysis showed a non-significant enhancement with respect to the main efficacy outcome. The PROBE design may influence the results by a 16% enhancement of the observed reduction in the risk of SSE compared with the DB design. This point estimate is consistent with a previous estimate of 17% derived from a meta-analysis of blinded/unblinded randomized trials in pediatric and obstetrical settings [36]. The lack of statistical significance may be due to a lack of power in detecting a significant interdependence of treatment effect and PROBE study design. The interaction test is known to suffer from lack of power even in a meta-analysis combining the results from many studies, as previously reported [37]. Only meta-epidemiological approaches encompassing more than 1000 trials can achieve statistical significance [11].

Other authors have already highlighted the potential biases introduced by the PROBE design [11, 38]. In a trial comparing rosiglitazone with a combination of metformin and sulphonylurea in type 2 diabetes, an FDA review revealed a differential assessment of outcomes, resulting from under-reporting of events occurring in the active arm to the blinded endpoint committee, compared with those occurring in the control arm, leading to spurious underestimation of the rate of myocardial infarction [38]. Robust outcomes, such as all-cause mortality, were less likely to be influenced by reporting bias [11]. Our meta-analysis showed no impact of the PROBE study design on the reported risk of cardiovascular and all-cause mortality. Nevertheless, the other potential biases that could lead to enhancement of the treatment effect with the PROBE design in comparison to the double-blind design are less well known. Apart from reporting bias, the lack of blinding of investigators might potentially lead to an imbalance between patients receiving VKAs or NOAs with regard to medical interventions such as invasive investigations or prescriptions (e.g. co-administration of antiplatelet agents) after randomization and result in confusion bias. The enrichment of our meta-analysis by publication of the results of ongoing trials may provide additional arguments to confirm or refute our assumptions.

Regarding safety, the PROBE design seems to influence the results by overestimating the observed reduction in the risk of intracranial hemorrhage by 33% and that of hemorrhagic stroke by 67% compared with the DB design, resulting in the consistent core benefit of these new drugs across all trials. Even though this observation concerning safety may result from a detection bias due to the lack of blinding of suspected outcome occurrence in PROBE/open studies (i.e. observer bias), we cannot rule out the impact of a confounding factor. Indeed, the large PROBE/open studies mainly assessed oral direct thrombin inhibitors whereas double-blind studies mainly assessed FXa inhibitors, and it is conceivable that the risk of intracranial bleeding differs between the two pharmacological classes. If the site of bleeding is related to the pharmacological class, it is impossible to know if the interaction is related to the study design or the type of drug evaluated. As for myocardial infarction [39], we cannot preclude a difference between thrombin and FXa inhibitors as regards the reduction of hemorrhagic stroke risk, with a greater risk reduction by oral direct thrombin inhibitors.

In conclusion, our meta-analysis did not show a significant interaction of study design on the main efficacy and safety outcomes. Nevertheless, the non-significant interaction seen on analysis of the effect of NOAs in reducing the risk of SSE suggests an interdependence of treatment effect and PROBE study design, especially with regard to hemorrhagic stroke risk, necessitating careful interpretation of trials using this design. However, we cannot rule out a difference in intracranial bleeding risk between the two pharmacological classes of NOAs.

Addendum

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Addendum
  8. Acknowledgements
  9. Funding
  10. Disclosure of Conflict of Interest
  11. References

P. Mismetti, S. Laporte and M. Cucherat designed the study. J-C. Lega, C. Chapelle and T. Fassier contributed to data acquisition. J-C. Lega, S. Laporte and M. Cucherat were responsible for the statistical analyses. J-C. Lega, S. Laporte, L. Bertoletti and P. Mismetti interpreted data. J-C. Lega, S. Laporte and P. Mismetti planned and wrote the first draft of the paper, which was subsequently revised by all authors. All authors read and approved the final manuscript. J-C. Lega is guarantor.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Addendum
  8. Acknowledgements
  9. Funding
  10. Disclosure of Conflict of Interest
  11. References

We are grateful to P. Harry for her valuable help in the preparation of this manuscript.

Funding

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Addendum
  8. Acknowledgements
  9. Funding
  10. Disclosure of Conflict of Interest
  11. References

This work was part of the META EMBOL project, supported by the Programme Hospitalier de Recherche Clinique 2008, Ministère de la Santé, France.

Disclosure of Conflict of Interest

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Addendum
  8. Acknowledgements
  9. Funding
  10. Disclosure of Conflict of Interest
  11. References

J-C. Lega, P. Mismetti, M. Cucherat, T. Fassier, L. Bertoletti and C. Chapelle received no direct support for this study. S. Laporte received support from the Ministère de la Recherche through a grant from the Programme Hospitalier de Recherche Clinique (META EMBOL) in 2008. P. Mismetti and S. Laporte sit on advisory boards for Boehringer Ingelheim, BMS/Pfizer and Bayer, as well as Daichii Sankyo in the case of P. Mismetti. P. Mismetti has received honoraria from Sanofi-Aventis, GSK, Astra Zeneca, Merck Serono, Boehringer Ingelheim and Bayer. S. Laporte has received honoraria from Sanofi-Aventis, Merck Serono, Boehringer Ingelheim and Bayer. No other relationships or activities that could appear to have influenced the submitted work are declared. M. Cucherat has received research funding and speaking fees from, or acted as a consultant for, GSK, Bayer, Pfizer and BMS.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Addendum
  8. Acknowledgements
  9. Funding
  10. Disclosure of Conflict of Interest
  11. References
  • 1
    Dulli DA, Stanko H, Levine RL. Atrial fibrillation is associated with severe acute ischemic stroke. Neuroepidemiology 2003; 22: 11823.
  • 2
    Slot KB, Berge E, Dorman P, Lewis S, Dennis M, Sandercock P. Impact of functional status at six months on long term survival in patients with ischaemic stroke: prospective cohort studies. BMJ 2008; 336: 3769.
  • 3
    Camm AJ, Kirchhof P, Lip GY, Schotten U, Savelieva I, Ernst S, van Gelder IC, Al-Attar N, Hindricks G, Prendergast B, Heidbuchel H, Alfieri O, Angelini A, Atar D, Colonna P, De Caterina R, De Sutter J, Goette A, Gorenek B, Heldal M, et al. Guidelines for the management of atrial fibrillation: the Task Force for the Management of Atrial Fibrillation of the European Society of Cardiology (ESC). Eur Heart J 2010; 31: 2369429.
  • 4
    Rupprecht HJ, Blank R. Clinical pharmacology of direct and indirect factor Xa inhibitors. Drugs 2010; 70: 215370.
  • 5
    Garcia D, Libby E, Crowther MA. The new oral anticoagulants. Blood 2010; 115: 1520.
  • 6
    Fuster V, Ryden LE, Cannom DS, Crijns HJ, Curtis AB, Ellenbogen KA, Halperin JL, Le Heuzey JY, Kay GN, Lowe JE, Olsson SB, Prystowsky EN, Tamargo JL, Wann S, Smith SC Jr, Jacobs AK, Adams CD, Anderson JL, Antman EM, Hunt SA, et al. ACC/AHA/ESC 2006 guidelines for the management of patients with atrial fibrillation–executive summary: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines and the European Society of Cardiology Committee for Practice Guidelines (Writing Committee to Revise the 2001 Guidelines for the Management of Patients With Atrial Fibrillation). J Am Coll Cardiol 2006; 48: 854906.
  • 7
    Singer DE, Albers GW, Dalen JE, Fang MC, Go AS, Halperin JL, Lip GY, Manning WJ. Antithrombotic therapy in atrial fibrillation: American College of Chest Physicians Evidence-Based Clinical Practice Guidelines (8th Edition). Chest 2008; 133: 546S92S.
  • 8
    Hylek EM, Go AS, Chang Y, Jensvold NG, Henault LE, Selby JV, Singer DE. Effect of intensity of oral anticoagulation on stroke severity and mortality in atrial fibrillation. N Engl J Med 2003; 349: 101926.
  • 9
    Patel MR, Mahaffey KW, Garg J, Pan G, Singer DE, Hacke W, Breithardt G, Halperin JL, Hankey GJ, Piccini JP, Becker RC, Nessel CC, Paolini JF, Berkowitz SD, Fox KA, Califf RM. Rivaroxaban versus warfarin in nonvalvular atrial fibrillation. N Engl J Med 2011; 365: 88391.
  • 10
    Connolly SJ, Ezekowitz MD, Yusuf S, Eikelboom J, Oldgren J, Parekh A, Pogue J, Reilly PA, Themeles E, Varrone J, Wang S, Alings M, Xavier D, Zhu J, Diaz R, Lewis BS, Darius H, Diener HC, Joyner CD, Wallentin L. Dabigatran versus warfarin in patients with atrial fibrillation. N Engl J Med 2009; 361: 113951.
  • 11
    Wood L, Egger M, Gluud LL, Schulz KF, Juni P, Altman DG, Gluud C, Martin RM, Wood AJ, Sterne JA. Empirical evidence of bias in treatment effect estimates in controlled trials with different interventions and outcomes: meta-epidemiological study. BMJ 2008; 336: 6015.
  • 12
    Schulz KF, Grimes DA. Blinding in randomised trials: hiding who got what. Lancet 2002; 359: 696700.
  • 13
    Casteels M, Flamion B. Open-label trials and drug registration: a European regulator's view. J Thromb Haemost 2008; 6: 2324.
  • 14
    Buller HR, Halperin JL, Bounameaux H, Prins M. Double-blind studies are not always optimum for evaluation of a novel therapy: the case of new anticoagulants. J Thromb Haemost 2008; 6: 2279.
  • 15
    Wong SS, Wilczynski NL, Haynes RB. Developing optimal search strategies for detecting clinically sound treatment studies in EMBASE. J Med Libr Assoc. 2006; 94: 417.
  • 16
    Haynes RB, McKibbon KA, Wilczynski NL, Walter SD, Werre SR. Optimal search strategies for retrieving scientifically strong studies of treatment from Medline: analytical survey. BMJ 2005; 330: 1179.
  • 17
    Schulman S, Kearon C. Definition of major bleeding in clinical investigations of antihemostatic medicinal products in non-surgical patients. J Thromb Haemost 2005; 3: 6924.
  • 18
    Parmar MK, Torri V, Stewart L. Extracting summary statistics to perform meta-analyses of the published literature for survival endpoints. Stat Med 1998; 17: 281534.
  • 19
    Higgins JP, Altman DG, Gotzsche PC, Juni P, Moher D, Oxman AD, Savovic J, Schulz KF, Weeks L, Sterne JA. The Cochrane Collaboration's tool for assessing risk of bias in randomised trials. BMJ 2011; 343: d5928.
  • 20
    DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials 1986; 7: 17788.
  • 21
    Cucherat M, Boissel JP, Leizorovicz A, Haugh MC. EasyMA: a program for the meta-analysis of clinical trials. Comput Methods Programs Biomed 1997; 53: 18790.
  • 22
    Higgins JP, Thompson SG. Quantifying heterogeneity in a meta-analysis. Stat Med 2002; 21: 153958.
  • 23
    Sterne JA, Juni P, Schulz KF, Altman DG, Bartlett C, Egger M. Statistical methods for assessing the influence of study characteristics on treatment effects in ‘meta-epidemiological’ research. Stat Med 2002; 21: 151324.
  • 24
    Chung N, Jeon HK, Lien LM, Lai WT, Tse HF, Chung WS, Lee TH, Chen SA. Safety of edoxaban, an oral factor Xa inhibitor, in Asian patients with non-valvular atrial fibrillation. Thromb Haemost 2011; 105: 53544.
  • 25
    Granger CB, Alexander JH, McMurray JJ, Lopes RD, Hylek EM, Hanna M, Al-Khalidi HR, Ansell J, Atar D, Avezum A, Bahit MC, Diaz R, Easton JD, Ezekowitz JA, Flaker G, Garcia D, Geraldes M, Gersh BJ, Golitsyn S, Goto S, et al. Apixaban versus warfarin in patients with atrial fibrillation. N Engl J Med 2011; 365: 98192.
  • 26
    Albers GW, Diener HC, Frison L, Grind M, Nevinson M, Partridge S, Halperin JL, Horrow J, Olsson SB, Petersen P, Vahanian A. Ximelagatran vs warfarin for stroke prevention in patients with nonvalvular atrial fibrillation: a randomized trial. JAMA 2005; 293: 6908.
  • 27
    Ezekowitz MD, Reilly PA, Nehmiz G, Simmers TA, Nagarakanti R, Parcham-Azad K, Pedersen KE, Lionetti DA, Stangier J, Wallentin L. Dabigatran with or without concomitant aspirin compared with warfarin alone in patients with nonvalvular atrial fibrillation (PETRO Study). Am J Cardiol 2007; 100: 141926.
  • 28
    Lip GY, Rasmussen LH, Olsson SB, Jensen EC, Persson AL, Eriksson U, Wahlander KF. Oral direct thrombin inhibitor AZD0837 for the prevention of stroke and systemic embolism in patients with non-valvular atrial fibrillation: a randomized dose-guiding, safety, and tolerability study of four doses of AZD0837 vs. vitamin K antagonists. Eur Heart J 2009; 30: 2897907.
  • 29
    Weitz JI, Connolly SJ, Patel I, Salazar D, Rohatagi S, Mendell J, Kastrissios H, Jin J, Kunitada S. Randomised, parallel-group, multicentre, multinational phase 2 study comparing edoxaban, an oral factor Xa inhibitor, with warfarin for stroke prevention in patients with atrial fibrillation. Thromb Haemost 2010; 104: 63341.
  • 30
    Olsson SB. Stroke prevention with the oral direct thrombin inhibitor ximelagatran compared with warfarin in patients with non-valvular atrial fibrillation (SPORTIF III): randomised controlled trial. Lancet 2003; 362: 16918.
  • 31
    A dose response study of dabigatran exilate (BIBR 1048) in pharmacodynamics and safety in patients with non-valvular atrial fibrillation in comparison to warfarin. Available at http://clinicaltrials.gov/ct2/show/NCT01136408. Accessed 23 May 2013.
  • 32
    Yamashita T, Koretsune Y, Yasaka M, Inoue H, Kawai Y, Yamaguchi T, Uchiyama S, Matsumoto M, Ogawa S. Randomized, multicenter, warfarin-controlled phase II study of edoxaban in Japanese patients with non-valvular atrial fibrillation. Circ J 2012; 76: 18407.
  • 33
    Ogawa S, Shinohara Y, Kanmuri K. Safety and efficacy of the oral direct factor xa inhibitor apixaban in Japanese patients with non-valvular atrial fibrillation. -The ARISTOTLE-J study. Circ J 2011; 75: 18529.
  • 34
    Hori M, Matsumoto M, Tanahashi N, Momomura S, Uchiyama S, Goto S, Izumi T, Koretsune Y, Kajikawa M, Kato M, Ueda H, Iwamoto K, Tajiri M. Rivaroxaban vs. warfarin in Japanese patients with atrial fibrillation – the J-ROCKET AF study. Circ J 2012; 76: 210411.
  • 35
    Connolly SJ, Ezekowitz MD, Yusuf S, Reilly PA, Wallentin L. Newly identified events in the RE-LY trial. N Engl J Med 2010; 363: 18756.
  • 36
    Schulz KF, Chalmers I, Hayes RJ, Altman DG. Empirical evidence of bias. Dimensions of methodological quality associated with estimates of treatment effects in controlled trials. JAMA 1995; 273: 40812.
  • 37
    Altman DG, Bland JM. Interaction revisited: the difference between two estimates. BMJ 2003; 326: 219.
  • 38
    Psaty BM, Prentice RL. Minimizing bias in randomized trials: the importance of blinding. JAMA 2010; 304: 7934.
  • 39
    Mak KH. Coronary and mortality risk of novel oral antithrombotic agents: a meta-analysis of large randomised trials. BMJ Open 2012; 2. pii: e001592.