Author + information
- Received March 21, 2016
- Revision received June 3, 2016
- Accepted July 7, 2016
- Published online November 1, 2016.
- Gianluigi Savarese, MDa,b,
- Robert P. Giugliano, MD, SMc,
- Giuseppe M.C. Rosano, MD, PhDd,e,
- John McMurray, MDf,
- Giulia Magnani, MDc,
- Gerasimos Filippatos, MD, PhDg,
- Santo Dellegrottaglie, MD, PhDh,i,
- Lars H. Lund, MD, PhDb,
- Bruno Trimarco, MD, PhDa and
- Pasquale Perrone-Filardi, MD, PhDa,∗ ()
- aDepartment of Advanced Biomedical Sciences, Federico II University, Naples, Italy
- bCardiology Unit, Department of Medicine, Karolinska Institutet, Stockholm, Sweden
- cTIMI Study Group, Cardiovascular Division, Brigham and Women's Hospital and Harvard Medical School, Boston, Massachusetts
- dCardiovascular and Cell Sciences Research Institute, St. George’s University, London, United Kingdom
- eIRCCS San Raffaele Pisana, Rome, Italy
- fInstitute of Cardiovascular and Medical Sciences, BHF Glasgow Cardiovascular Research Centre, University of Glasgow, Glasgow, United Kingdom
- gAthens University Hospital Attikon, Athens, Greece
- hVilla dei Fiori, Acerra, Napoli, Italy
- iMount Sinai Medical School, New York City, New York
- ↵∗Reprint requests and correspondence:
Dr. Pasquale Perrone-Filardi, Department of Advanced Biomedical Sciences, Federico II University, Via Pansini, 5–I-80131–Naples, Italy.
Objectives This study investigated the efficacy and safety of novel oral anticoagulants (NOACs) in patients with atrial fibrillation (AF) and heart failure (HF) by a meta-analysis.
Background AF is quite prevalent in patients with HF.
Methods Four phase III clinical trials comparing NOACs to warfarin in patients with AF were included. Each patient was defined as affected by HF according to the criteria of the trial in which the patient was enrolled. Pre-specified outcomes were the composite of stroke/systemic embolism (SSE); major, intracranial, and any bleeding; and cardiovascular (CV) and all-cause death.
Results A total of 55,011 patients were enrolled, 26,384 (48%) with HF, and 28,627 (52%) without HF; 27,518 receiving NOACs and 27,493 receiving warfarin (median, 70 years of age; 36% females; follow-up: 1.5 to 2.8 years). Rates of SSE (relative risk [RR]: 0.98; 95% confidence interval [CI]: 0.90 to 1.07]; p = 0.68) and major bleeding (RR: 0.95; 95% CI: 0.88 to 1.03; p = 0.21) were comparable in patients with and without HF. HF patients had reduced rates of any (RR: 0.86; 95% CI: 0.81 to 0.91; p < 0.01) and intracranial (RR: 0.74 95% CI: 0.63 to 0.88; p < 0.01) bleeding but increased rates of all-cause (RR: 1.70 95% CI: 1.31 to 2.19; p < 0.01) and CV death (RR: 2.05 95% CI: 1.66 to 2.55; p < 0.01). NOACs, compared with warfarin significantly reduced SSE and major, intracranial, and any bleeding, regardless of the presence or absence of HF (pinteraction > 0.05 for each).
Conclusions Patients with AF and HF had increased mortality but reduced rates of intracranial and any bleeding compared with the no-HF patients, with no differences in rates of SSE and major bleeding. NOACs significantly reduced SSE, major bleeding, and intracranial hemorrhage in HF patients. No interactions in efficacy and safety of NOACs were observed between AF patients with and without HF.
HF is considered a major risk factor for thromboembolic (3,4) and bleeding risk (5,6) in patients with AF and is included in the CHA2DS2-VASc score (7). In addition, HF is widely recognized as a risk factor for reduced time in therapeutic range (TTR) in AF patients taking warfarin-based anticoagulation therapy (8,9), which may contribute to excess thromboembolic and bleeding risk in AF patients with HF.
Novel oral anticoagulants (NOACs), due to their more predictable therapeutic effect and more favorable hemorrhagic risk profile, represent a particularly attractive therapeutic option in AF patients with concomitant HF. However, currently, no randomized clinical trials assessing the efficacy and safety of NOACs in HF patients with AF have been conducted, and only subgroup analyses from the 4 major NOAC phase III clinical trials reporting data for HF patients that have not enough statistical power to investigate less frequent events (i.e., intracranial bleeding) are available (10–13). Notably, Ruff et al. (14), in a recent meta-analysis, evaluated the relative efficacy and safety of NOACs across clinically relevant patient subgroups, but no data for patients with AF and concomitant HF were reported. Xiong et al. (15) reported in another meta-analysis a significant reduction of thromboembolic events and major bleeding with high-dose NOACs compared with warfarin in HF patients. There was a similar incidence of both major bleeding and thromboembolic events in AF patients with and without HF treated with NOACs, but in this study, data from the ENGAGE AF-TIMI 48 (Effective Anticoagulation with Factor Xa Next Generation in Atrial Fibrillation–Thrombolysis In Myocardial Infarction 48) trial (13) were not available for most of the outcomes and subanalyses.
Thus, the aim of the present analysis was to assess the efficacy and safety of NOACs in patients with AF and HF enrolled in all phase III NOACs clinical trials and to detect whether the efficacy and safety of NOACs differed between AF patients with and without HF.
The current analysis included the 4 major phase III randomized clinical trials comparing NOACs to warfarin in patients with AF: ARISTOTLE (Apixaban for Reduction in Stroke and Other Thromboembolic Events in Atrial Fibrillation) trial, comparing apixaban, a direct factor Xa inhibitor, with warfarin (10); the RE-LY (Randomized Evaluation of Long Term Anticoagulation Therapy) trial, comparing dabigatran, a direct inhibitor of thrombin, with warfarin (11); the ROCKET AF (Rivaroxaban Once Daily Oral Direct Factor Xa Inhibition Compared with Vitamin K Antagonism for Prevention of Stroke and Embolism Trial in Atrial Fibrillation) trial (12); and the ENGAGE AF–TIMI 48 study, comparing edoxaban, a direct factor Xa inhibitor, with warfarin (13).
Data extraction and quality assessment
Two reviewers (G.S. and P.P.F.) independently selected potentially eligible trials. Discrepancies were resolved by consensus. From each study, information about the inclusion criteria, year of publication, number of patients in treatment and control arms, duration of follow-up, age, sex, cardiovascular (CV) risk factors, prior myocardial infarction or coronary artery disease, CHADS2 score, New York Heart Association (NYHA) functional class and baseline medications were abstracted by one author (G.S.) and checked by another author (P.P.F.). Pre-specified outcomes for the current analysis were the composite of stroke and systemic embolism (SSE), major bleeding, intracranial bleeding, any bleeding, CV mortality, and all-cause death. These data were abstracted for subgroups of patients with and without HF and randomized to NOACs or warfarin and were available in published articles for apixaban (10), dabigatran (11), and rivaroxaban (12), whereas they were obtained directly from authors for edoxaban, as the analyses of patients stratified by HF in ENGAGE AF-TIMI 48 were still not published at the time of the current meta-analysis. Patients were reported to be affected or not by HF according to the different definitions used by the trials in which they were enrolled (Table 1).
In 2 of 4 trials, 2 different dosages of NOACs (dabigatran, 110 and 150 mg twice daily, in RE-LY; and edoxaban, 30 or 60 mg once daily, in ENGAGE AF-TIMI 48) were compared to those of warfarin (11,13). In order not to merge the effects of different doses, we performed a meta-analysis including the higher dose arms of the RE-LY (11) and ENGAGE AF-TIMI 48 (13) studies together with the single-dosage arms of the ARISTOTLE (apixaban, 5 mg twice daily) (10) and ROCKET AF (rivaroxaban, 20 mg once daily) (12) trials.
Some outcome data were not available for the secondary analyses: all-cause and CV mortality were not reported in the ARISTOTLE and RE-LY HF subgroups (10,11), respectively. In the ROCKET AF HF substudy, major bleeding was not reported, and major or nonmajor clinically relevant bleeding was used instead of any bleeding (12).
Methodological quality of trials was assessed by the Detsky method, scoring the following items: method of randomization: 1 point; adequate description of method of randomization: 2 points; blindness: 2 points; adequate description of outcome: 1 point, and of outcome assessment: 2 points; inclusion/exclusion criteria: 2 points; number of patients excluded and reasons: 2 points; description of therapy in treatment and control groups: 4 points; and appropriateness of statistical analysis: up to 5 points (16).
Relative risks (RR) of the outcomes in HF versus no-HF patients and of effects of randomized treatments were calculated using the metan routine (version 12.0 software, StataCorp LLC, College Station, Texas) to account for the probability of events occurring in the treatment group versus the control group (17). RR and 95% confidence interval (CI) for each outcome were separately calculated for each trial, with grouped data, using the intention-to-treat principle (18). Pooled RRs were logarithmically transformed and weighted for the inverse of variance. Overall estimates of effect were calculated with a fixed-effects model or with a random effects model when heterogeneity could not be explained. The assumption of homogeneity between the treatment effects in different trials was tested using Q statistic and further quantified by I2 statistic. A significant heterogeneity was defined by a p value of <0.05 at Q statistic; i2 values ranging from 0% to 40% indicated unimportant heterogeneity, 30% to 60% represented moderate heterogeneity, 50% to 90% indicated substantial heterogeneity, and 75% to 100% represented considerable heterogeneity (19). The significance level for all outcome and heterogeneity analyses was set at a p value ≤0.05.
To explore the influence of potential effect modifiers on outcomes, weighted random effects metaregression analysis was performed with the metareg command (version 12.0, Statacorp) to test demographic characteristics of the study population, duration of follow-up, CV risk factors, New York Heart Association (NYHA) functional class, CHADS2 score, and concomitant medications (20). For all metaregression analyses, the weight used for each trial was the inverse of the sum of the within-trial variance and the residual between trial variance. Additionally, the residual maximum likelihood methods were used to explain residual heterogeneity not explained by potential effect modifiers, including an additive between-study variance component Tau2 (21).
Because the main analysis was performed by pooling the trials with a single-dose arm (ARISTOTLE and ROCKET AF) together with the higher dose arm of the trials assessing 2 different doses of the same drug (RE-LY and ENGAGE AF-TIMI 48), further analysis was performed in HF patients, also including the lower dose arms.
Characteristics of included trials
The 4 trials included in analyses of 55,011 subjects, 26,384 with HF, of whom 13,251 were treated with NOACs and 13,133 with warfarin, and 28,627 without HF, of whom 14,267 were treated with NOACs and 14,360 with warfarin (Figure 1). The median follow-up ranged from 1.5 to 2.8 years across the 4 trials. Characteristics of patients with and without HF are reported in Table 1. Patients with HF on average were younger and less likely to be female and hypertensive but had higher rates of diabetes and prior myocardial infarction or coronary artery disease.
The median Detsky score was 100%, meaning that the quality of the trials included in the meta-analysis was very high. All trials satisfied all Detsky method items, except for RE-LY, which was not double-blinded.
Outcome analysis: HF versus no HF subgroup
A total of 1.74 SSEs per 100 patient-years were reported in HF patients compared with 1.67 in subjects without HF; thus, the rates of SSE in patients with HF were comparable to those in patients without HF (p = 0.68) (Figure 1). Similarly, 2.70 major bleedings per 100 patient-years occurred in HF patients and 3.02 in subjects without HF, with no differences between the subgroups (p = 0.21).
Both the rates of intracranial bleeding (0.45 vs. 0.60 per 100 patient-years; p < 0.01) and any bleeding (11.80 versus 15.62 per 100 patient-years; p < 0.01) were significantly lower in patients with HF than in patients without HF. Rates of CV death (3.62 vs. 1.84 per 100 patient-years; p < 0.01) and all-cause death (5.21 vs. 3.20 per 100 patient-years; p < 0.01) were approximately 2-fold significantly higher in HF patients than in patients without HF.
Treatment analysis: Effects of NOACs in HF versus no HF subgroups
No differences in treatment effect were detected between HF and no-HF patients for SSE (p = 0.23), major bleeding (p = 0.09), intracranial bleeding (p = 0.32), any bleeding (p = 0.99), CV death (p = 0.11), and all-cause mortality (p = 0.13) (Figure 2).
Effects of NOACs in HF patients
In patients with HF, NOACs significantly reduced the rate of SSE by 14.4% (p = 0.01), major bleeding by 23.2% (p < 0.01), intracranial bleeding by 57.3% (p < 0.01), and any bleeding by 11.6% (p = 0.02) compared with warfarin. CV and all-cause death rates were not significantly different between NOAC- and warfarin-treated patients (p = 0.11 and p = 0.08, respectively) (Figures 2 and 3).
Effects of NOACs in patients without HF
NOACs significantly reduced the rate of SSE by 23.1% (p < 0.01), major bleeding by 12.2% (p = 0.01), and intracranial bleeding by 49.0% (p < 0.01) compared with warfarin. Rates of any bleeding were numerically lower with NOACs, but this did not meet statistical significance (p = 0.054). NOACs significantly reduced both CV and all-cause death by 19.2% (p = 0.02) and 14.7% (p = 0.04), respectively, compared with warfarin (Figures 2 and 4).
In ROCKET AF (12), major or nonmajor clinical relevant bleeding was used instead of any bleeding, as the latter was not available. For this reason, we also excluded data from ROCKET AF in a sensitivity analyses of the outcome of any bleeding, and results were confirmed in HF versus no-HF analysis and within the HF subgroup. Within the no-HF subgroup, after the removal of data from ROCKET AF, NOACs significantly reduced the risk of any bleeding by 14.1% (RR: 0.86; 95% confidence interval [CI]: 0.77 to 0.96; comparison p = 0.01; heterogeneity p < 0.01; i2 = 88.3%).
In a metaregression analysis, no relevant effect modifiers were identified that influenced the findings of the meta-analysis (Online Table 1).
When lower dose arms of RE-LY and ENGAGE AF-TIMI 48 trials were also included in the analyses assessing the effects of NOACs on outcomes in HF population (3,979 patients in edoxaban, 30 mg once daily, arm of ENGAGE AF-TIMI 48 trial and 1,641 patients in dabigatran, 110 mg twice daily, arm of RE-LY trial), NOACs significantly reduced the risk of major bleeding by 27.3% (RR: 0.73; 95% CI: 0.62 to 0.85]; comparison p < 0.01; heterogeneity p = 0.04; i2 = 61.1%), total bleeding by 12.8% (RR: 0.87; 95% CI: 0.80 to 0.95; comparison p < 0.01; heterogeneity p < 0.01; i2 = 78.3%), and intracranial bleeding by 58.1% (RR: 0.42; 95% CI: 0.32 to 0.54; comparison p < 0.01; heterogeneity p = 0.257; i2 = 24.7%), without any effect on risk of SSE (RR: 0.94; 95% CI: 0.85 to 1.04; p comparison = 0.241; p heterogeneity = 0.120; i2 = 42.8%) and CV death (RR: 0.94; 95% CI: 0.86 to 1.02; p comparison = 0.123; p heterogeneity = 0.824; i2 = 0.0% [data for edoxaban, 30 mg once daily, were missing]). Data for all-cause mortality in RE-LY and for edoxaban, 30 mg once daily, were not available; therefore, this outcome analysis could not be performed.
The findings of the present study indicate that NOACs, compared with warfarin, significantly reduce the risk of SSE as well as the risk of major, intracranial, and any bleeding in AF patients regardless of the presence or absence of HF.
Thromboembolic and bleeding risk in patients with AF and HF
Current guidelines include HF among the factors to be summed in the CHADS2 and CHA2DS2-VASc risk scores (7), even though the evidence that HF is associated with a higher thromboembolic risk in patients with AF is reported in some studies (3,4) but not in others (6,22,23). In fact, this apparent discrepancy needs to be interpreted with caution, taking into account the different enrollment criteria adopted in the studies. In ARISTOTLE and ROCKET AF, no differences in cardioembolic or bleeding risk were observed between patients with and without HF, although a higher mortality was reported (10,12). In addition, in both of those trials (10,12), no interaction was observed between ejection fraction and SSE risk. However, patients enrolled in the ROCKET AF trial were at substantially higher stroke risk than patients in the other NOAC trials, which might have diluted the prognostic impact of HF (12). In the RE-LY trial, although the rate of SSE was numerically higher in patients with HF than in those without HF, after multivariate adjustment for baseline characteristics, the risk of SSE was similar among patients with and without HF, whereas in the same analysis, HF was an independent risk factor for CV death (11). In contrast to other NOAC phase III trials, in ENGAGE AF-TIMI 48, significantly higher SSE and bleeding risks were observed in patients with HF, after adjustment for other confounders (13). In our analysis summarizing all the data from the 4 major NOAC clinical trials, we found no differences in the risk of either thromboembolic or major bleeding events between patients with versus those without a history of HF, but an increased mortality was present in patients with HF.
It is important to recognize that HF is associated with an increased likelihood of a suboptimal TTR in AF patients treated with vitamin K antagonists (8,9). Because inadequate control of the level of anticoagulation increases the risk of thromboembolic events, bleeding, and death, this might by itself contribute to increased risk in HF patients with AF (9,13). In fact, a recent meta-analysis of all 4 major NOAC versus warfarin trials reported a greater reduction in bleeding risk with NOACs in patients enrolled at centers with a median TTR <66% (14). In RE-LY, patients with HF, compared with those without HF, showed a significantly lower rate of any bleeding, whereas no significantly reduced rate of intracranial bleeding in patients with HF was observed in RE-LY, ARISTOTLE, and ROCKET AF (10–12). In our analysis, the rates of intracranial and of any bleeding were significantly lower in patients with HF than in patients without. This counterintuitive finding might be explained in part by the enrollment criteria of NOAC clinical trials, in which patients with HF, compared with those without HF, were younger, had a lower prevalence of hypertension, and, consequently, had lower blood pressure values that might explain in particular the reduced rates of intracranial bleeding in these patients.
Finally, the absence of differences in the risk of either thromboembolic or bleeding events shown in our meta-analysis between patients with and those without a history of HF does not deny the unfavorable impact of HF and could be explained by the enrollment criteria used in the 4 NOAC trials. In fact, to be enrolled in these trials, patients needed to have AF plus only 1 additional risk factor, including HF. This resulted in a selection bias among patients without HF, as other major risk factors for SSE were needed for participation in the trials, which might have obscured the impact of HF on SSE and bleeding risk. In addition, all patients in these trials received anticoagulation therapy, whether they had HF or not, and that might have attenuated any difference in thromboembolic and bleeding related to HF status.
Efficacy and safety of NOACs in HF versus No-HF patients
This meta-analysis of HF patients enrolled in the 4 major trials on NOACs allowed us to investigate the overall effects of these drugs on major efficacy and safety end points.
In these 4 trials, NOACs achieved efficacy and safety outcomes that were at least as good as warfarin in patients with HF.
In the main ARISTOTLE trial, apixaban significantly reduced the risk of SSE compared with warfarin, and the subgroup analysis comparing patients with HF with reduced ejection fraction to patients without HF demonstrated no significant interaction between treatments and subgroups (10). Risks of intracranial, major, and total bleeding were significantly reduced by apixaban compared with warfarin in the main trial and to a similar degree in subgroups stratified by HF, with no evidence of subgroup by treatment interaction for these safety endpoints (10).
In RE-LY, no differences in treatment efficacy were reported between HF and no-HF subgroups (11). In both the patients with and those without HF, intracranial and any bleeding were significantly reduced by dabigatran, 150 mg twice daily, whereas no differences were observed between dabigatran and warfarin for major bleeding (11).
In ROCKET AF, rivaroxaban compared with warfarin did not significantly reduce the risks of SSE, bleeding, CV death, or overall mortality in patients with or in patients without HF (12).
In ENGAGE AF-TIMI 48, edoxaban, 60 mg, was similar to warfarin for the prevention of thromboembolism, regardless of the presence or absence of prior HF. In addition, edoxaban significantly reduced CV death and CV hospitalization even in patients with severe HF. Regardless of HF severity, edoxaban significantly reduced the risk of major bleeding and intracranial hemorrhage compared with warfarin (13).
Our analysis shows that NOACs, compared with warfarin, significantly reduced the risks of SSE, major and intracranial bleeding, and any bleeding in patients with HF, with no significant interactions with patients without HF. Furthermore, when the lower dosage arms of RE-LY (11) and ENGAGE AF-TIMI 48 (13) were also included in the analysis, NOACs were still shown not to be inferior to warfarin in reducing the risk of SSE but superior in reducing the risk of bleeding in AF patients with HF. Thus, NOACs represent a valuable and preferable alternative to warfarin in this clinical setting. The current meta-analysis confirms and expands the previous study performed by Xiong et al. (15). The inclusion of data from ENGAGE AF-TIMI 48 and the evaluation of other outcome data not reported in the previous meta-analysis complement the previous analysis and further strengthen the efficacy and safety of NOACS in HF patients.
The current meta-analysis has some limitations to acknowledge. First, the studies included in the analysis were not prospectively designed to assess the efficacy and safety of NOACs in patients with HF versus without HF. Consequently, the definitions of HF varied across trials, and this could have led to consider patients with similar characteristics as affected by HF in some trials but not in others, with the possibility of introducing a bias in the analysis. Additionally, our analyses were based on aggregate trial-level data and not on patient-level data. However, we performed meta-regression to assess for potential confounding by differences in baseline characteristics in patients with versus without HF, and the results were similar. Because we did not have patient-level data, we could not perform subgroup analyses to assess the impact of treatment in relation to relevant clinical characteristics, including TTR, level of left ventricular dysfunction, renal impairment, or other comorbid conditions commonly observed in HF patients. It is also important to acknowledge that our results cannot be generalized to the overall AF population with HF but may be applied only to patients showing similar characteristics to those enrolled in NOACs randomized clinical trials. Finally, the J-ROCKET study (24), even if it was a phase III trial randomizing AF patients to NOAC (rivaroxaban) versus warfarin, was not included in the analysis because it would have been the only trial with a smaller sample size and enrolling patients in just 1 country, leading to the possibility of including a bias and generating heterogeneity in our analysis.
In HF patients NOACs significantly reduced the risks of SSE and bleeding events, including major bleeding and intracranial hemorrhage, compared with warfarin. No interaction in the efficacy or safety profile of NOACs compared with warfarin was present in patients with versus those without HF.
COMPETENCY IN MEDICAL KNOWLEDGE: AF patients with HF enrolled in NOAC clinical trials have reduced rates of any and intracranial bleeding, but increased rates of all-cause and CV death. NOACs, compared with warfarin, significantly reduce SSE and major, intracranial, and any bleeding in AF patients regardless of the presence or absence of HF.
TRANSLATIONAL OUTLOOK: NOACs represent a valuable and preferable alternative to warfarin in HF clinical setting.
For a supplemental table, please see the online version of this article.
Daiichi-Sankyo provided support for the ENGAGE AF-TIMI 48 trial with a grant to the Brigham and Women's Hospital. Dr. Giugliano has received honoraria for CME lectures from Daiichi Sankyo and Merck; honoraria for consulting from the American College of Cardiology, Boehringer Ingelheim, Bristol-Myers Squibb, Daiichi-Sankyo, Pfizer, and Portola. Dr. Filippatos is a member of the steering committee of trials sponsored by Bayer, Novartis, and Vifor. Dr. Lund has received speaker or consulting honoraria from St Jude, Novartis, Bayer, ViforPharma, and HeartWare; and research grants to his institution from Boston Scientific, Medtronic and AstraZeneca.
All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- atrial fibrillation
- confidence interval
- heart failure
- novel oral anticoagulants
- New York Heart Association
- relative risk
- stroke and systemic embolism
- time in therapeutic range
- Received March 21, 2016.
- Revision received June 3, 2016.
- Accepted July 7, 2016.
- American College of Cardiology Foundation
- Friberg L.,
- Rosenqvist M.,
- Lip G.Y.
- Camm A.J.,
- Lip G.Y.,
- De Caterina R.,
- et al.
- McMurray J.J.,
- Ezekowitz J.A.,
- Lewis B.S.,
- et al.
- van Diepen S.,
- Hellkamp A.S.,
- Patel M.R.,
- et al.
- Magnani G.,
- Giugliano R.P.,
- Ruff C.T.,
- et al.
- ↵Sharp S, Sterne J. Meta-analysis. Stata Technical Bulletin Reprints 1998;7:100–108. Available at: http://www.stata.com/products/stb/journals/stb16.pdf. Accessed August 11, 2016.
- Whitehead A.
- ↵Cochrane Handbook for Systematic Reviews of Interventions. Version 5.1.0. Identifying and measuring heterogeneity. Available at: http://handbook.cochrane.org/. Accessed March 11, 2016.
- Sharp S.J.
- Olesen J.B.,
- Lip G.Y.,
- Hansen M.L.,
- et al.