Dominick J. Angiolillo, University of Florida College of Medicine-Jacksonville, 655 West 8th Street, Jacksonville, FL 32209, USA. Tel.: +1 904 244 3933; fax: +1 904 244 3102. E-mail: firstname.lastname@example.org
Summary. Background: Thrombin receptor antagonists blocking protease-activated receptor-1 (PAR-1) on platelets represent a new class of oral antiplatelet agents for patients with atherothrombotic disease manifestations.
Objectives: We investigated the safety and efficacy of PAR-1 antagonists in patients with coronary artery disease (CAD).
Patients/Methods: Randomized, placebo-controlled trials of the PAR-1 antagonists atopaxar or vorapaxar in CAD patients were identified. The primary safety endpoint was the composite of Thrombolysis In Myocardial Infarction (TIMI) clinically significant bleeding. The primary efficacy endpoint was the composite of death, myocardial infarction (MI) or stroke.
Results: A total of 41 647 patients from eight trials were included. PAR-1 antagonists were associated with higher risks of TIMI clinically significant (odds ratio [OR] 1.48, 95% confidence interval [CI] 1.39–1.57, P < 0.001), major (OR 1.46, 95% CI 1.28–1.67, P < 0.001) and minor (OR 1.67, 95% CI 1.40–2.00, P < 0.001) bleeding than placebo in the fixed-effects model. PAR-1 antagonists reduced the composite of death, MI or stroke as compared with placebo (OR 0.87, 95% CI 0.81–0.92, P < 0.001), driven by a lower risk of MI (OR 0.85, 95% CI 0.78–0.92, P < 0.001). Conversely, PAR-1 antagonists and placebo did not differ in terms of risk of death (OR 0.99, 95% CI 0.90–1.09, P = 0.81) or stroke (OR 0.96, 95% CI 0.84–1.10, P = 0.59).
Conclusions: PAR-1 antagonists decrease ischemic events in patients with CAD as compared with placebo, mainly driven by a reduction in MI, at the cost of an increased risk of clinically significant bleeding.
Platelet activation, which is a crucial step in the onset of pathologic thrombotic processes, may be mediated by different signaling pathways [1,2]. Current antiplatelet treatment options include drugs that inhibit platelet activation mediated by thromboxane A2 (e.g. aspirin) and ADP (e.g. P2Y12 receptor antagonists). Blocking these pathways, however, does not affect platelet activation induced by thrombin, which is the most potent effector of platelet-mediated thrombosis . Thrombin exerts its platelet-activating effects via protease-activated receptors (PARs), most importantly PAR-1 in human platelets . As the PAR-1-mediated signaling pathway continues to promote platelet activation and aggregation even in the presence of aspirin and P2Y12 receptor antagonists, patients with coronary artery disease (CAD) may continue to experience recurrent atherothrombotic events [1–3].
PAR-1 antagonists have emerged as promising novel oral antiplatelet agents that may reduce ischemic risk among patients with atherothrombotic processes . Although the PAR-1 platelet activation pathway stimulated by thrombin has a critical role in pathologic thrombosis, it may be not essential for hemostasis [4–8]. This is in contrast to aspirin and P2Y12 receptor antagonists, which interfere with the platelet signaling pathways essential for primary hemostasis, thus exposing treated patients to an increased risk of bleeding [9–12]. Therefore, inhibiting thrombin-induced platelet activation while sparing other functions of thrombin might provide for a larger therapeutic index of PAR-1 antagonists. Results from phase II trials of two PAR-1 antagonists currently under clinical development, vorapaxar (SCH530348) and atopaxar (E5555), have provided promising clinical data in support of this concept, outlining the potential of PAR-1 antagonists to reduce ischemic events without substantially increasing bleeding risk when used as an adjunct to standard antiplatelet treatment regimens [13–17]. However, safety concerns emerged in two phase III trials of vorapaxar [18,19]. These observations have raised questions concerning the appropriateness of relying on data from limited phase II testing to guide further trial design and clinical development of a drug.
The aim of the present study was to perform a meta-analysis of all the trials conducted that have compared one of these PAR-1 antagonists with placebo, to address the important question of whether there are risks or benefits with these new agents when used in patients with CAD.
The primary aim was to evaluate the safety and efficacy of novel PAR-1 antagonists as compared with placebo in patients at high risk for cardiovascular events. A full electronic search strategy was used, and the terms used for research were PAR-1, SCH530348, E5555, vorapaxar, and atopaxar. We restricted our analysis to prospective, randomized trials that met all of the following inclusion criteria: (i) study population of patients with CAD; (ii) assignment of participants to PAR-1 antagonist treatment or a placebo group; and (iii) the supply of data on both bleeding outcomes and major adverse cardiac and cerebrovascular events (MACCEs).
The quality of the identified studies was assessed to ensure minimization of bias. No formal scoring system was used. Reviewers were not blinded to journal, author, or institution of publication.
All data were extracted independently by two evaluators (D.C. and D.J.A.), and discrepancies were resolved by a third evaluator (S.G.). From each eligible study, we recorded information regarding the number of patients that were randomly assigned to PAR-1 antagonists or placebo and patient demographics (mean age, proportion of women, diabetes mellitus, non-ST elevation myocardial infarction [MI], percutaneous coronary intervention [PCI], and use of aspirin or P2Y12 inhibitors).
The following major outcomes were extracted: Thrombolysis In Myocardial Infarction (TIMI) bleeding (major, minor, or minimal), MACCE (defined according to each trial’s definition, as shown in Table 1), death (mortality from any cause or cardiovascular death), MI, and stroke. We considered the longest available follow-up period for all events in each eligible study. The clinical endpoint definitions were similar among the trials. Each trial had independent adjudication of clinical events. MI was defined by electrocardiogram (ECG) changes and elevation of myocardial enzymes. The primary safety endpoint of this meta-analysis was clinically significant bleeding according to the TIMI classification, defined as TIMI major or minor bleeding or bleeding that required unplanned medical or surgical treatment or laboratory evaluation. The primary efficacy endpoint was the composite of death, MI or stroke.
Table 1. Studies included in the meta-analysis
Key safety endpoints
Key efficacy endpoints
CAD, coronary artery disease; CURE, Clopidogrel in Unstable angina to prevent Recurrent Events; GUSTO, Global Use of Strategies to Open Occluded Coronary Arteries; LD, loading dose; MACCE, major adverse cardiac and cerebrovascular event; MD, maintenance dose; MI, myocardial infarction; NSTE-ACS, non-ST elevation acute coronary syndrome; PCI, percutaneous coronary intervention; TIMI, Thrombolysis In Myocardial Infarction.
MACCE (cardiovascular death, MI, stroke, or recurrent ischemia) and individual components
We abstracted the raw number of patients experiencing the outcomes of interest among all patients in each randomized treatment group from each of the publications of the selected clinical trials. Data were analyzed according to the intention-to-treat principle [13,14] or a modified intention-to-treat principle, including only subjects who had taken at least one dose of study drug [15–17,19]. To give a global estimation of the treatment effect, the results of all studies were combined by use of a random-effects model to minimize heterogeneity between groups, and confirmed with a fixed-effects model to avoid small studies being overly weighted [20,21]. Odds ratio (ORs) with 95% confidence interval (CIs) were calculated. A two-tailed alpha risk of 5% was used for hypothesis testing.
Statistical heterogeneity was assessed with Cochran Q via a chi-squared test and quantified with the I2-test. The influence of individual studies was examined by excluding studies one at a time, and testing for systematic bias was performed with funnel plots and Begg’s test. Exploratory bivariate meta-regressions were performed to assess heterogeneous study effects, and included regressions of the log OR on a set of prespecified variables, including number of enrolled patients, percentage of women, and concomitant use of P2Y12 inhibitors. The statistical analysis was performed with Comprehensive Meta-Analysis Version 2.0 (Biostat, Englewood, NJ, USA).
A total of 1223 potentially eligible studies were identified, and 955 were excluded following title evaluation for not fulfilling inclusion criteria on the basis of intervention. On further review, 268 studies were excluded for other reasons. A flow diagram of the trial selection process is given in Fig. 1.
Eight trials that enrolled a total of 41 647 patients met the inclusion criteria and were included in the meta-analysis [13–19]. The main features of the included studies are listed in Tables 1 and 2. Six studies were randomized, double-blind phase II trials comparing oral PAR-1 inhibitors with matching placebo, two of which tested vorapaxar and four of which tested atopaxar. Two studies were large, double-blind phase III trials of vorapaxar vs. placebo. A spectrum of CAD manifestations was represented by the eight trials. The Thrombin Receptor Antagonist–Percutaneous Coronary Intervention (TRA-PCI) trial and its Japanese counterpart (J-TRA-PCI) evaluated the safety and tolerability of vorapaxar in patients with symptoms of CAD in whom catheterization with possible PCI was planned [13,14]. Both the TRA-PCI and the J-TRA-PCI trials subdivided the results into those of patients who actually underwent PCI (PCI cohort) and those who did not (no PCI cohort), being treated medically or surgically. The Lessons from Antagonizing the Cellular Effects of Thrombin in Acute Coronary Syndromes (LANCELOT-ACS) and the Japanese J-LANCELOT-ACS trials evaluated the safety and tolerability of atopaxar after a non-ST-segment elevation acute coronary syndrome (NSTE-ACS) [15,16]. These trials were conducted in parallel with the LANCELOT-CAD and the J-LANCELOT-CAD trials, evaluating the safety and tolerability of atopaxar in patients with stable CAD [15,17]. The Thrombin Receptor Antagonist for Clinical Event Reduction in Acute Coronary Syndrome (TRA-CER) trial evaluated the efficacy and safety of vorapaxar in patients with NSTE-ACS . Finally, the Thrombin Receptor Antagonist in Secondary Prevention of Atherothrombotic Ischemic Events (TRA-2P) trial evaluated the efficacy and safety of vorapaxar for secondary prevention in patients with established atherosclerosis, including those with a history of MI, ischemic stroke, or peripheral artery disease. Given the design of this meta-analysis, patients with qualifying CAD, such as those with a history of MI (N = 17 779), were abstracted and analyzed separately from the TRA-2P trial patients for the primary safety and efficacy endpoints.
Table 2. Patients and procedural characteristics of included studies
Overall, the risk of clinically significant bleeding was significantly higher with PAR-1 antagonists than with placebo (OR 1.48, 95% CI 1.39–1.57, P < 0.001) in both the fixed-effects (Fig. 2) and random-effects models, with no observed heterogeneity (I2 = 0%, P = 0.89) or apparent systematic bias across trials (Begg’s test, P = 0.53). These results did not change significantly for either model when only patients with qualifying CAD from the TRA-2P trial were included (OR 1.49, 95% CI 1.39–1.59, P < 0.001). No trial was found to unduly influence the primary safety effects, as pooling only data from the phase II trials after excluding the TRA-CER and TRA-2P trials resulted in a consistent increase, albeit slightly lower in magnitude, in the endpoint of clinically significant bleeding with PAR-1 antagonists (OR 1.32, 95% CI 1.03–1.69, P = 0.03). Meta-regressions conducted to investigate the impact of preselected variables on heterogeneity among studies demonstrated no variability in the log OR based on percentage of women (P = 0.13), but a significant interaction with the rate of P2Y12 inhibitor usage (P = 0.02) and the number of patients enrolled in each trial (P < 0.0001).
A total of 912 TIMI major and 537 TIMI minor bleeds occurred. The TIMI major bleeding OR for PAR-1 antagonists vs. placebo was 1.46 (95% CI 1.28–1.67, P < 0.001) in both the fixed-effects and random-effects models, with no observed heterogeneity (I2 = 0%, P = 0.92). The TIMI minor bleeding ORs for PAR-1 antagonists vs. placebo were 1.67 (95% CI 1.40–2.00, P < 0.001) in the fixed-effects model and 1.49 (95% CI 1.10–2.02, P = 0.011) in the random-effects models, with no significant heterogeneity (I2 = 28%, P = 0.19) (Fig. 3). Analyses restricted to phase II trials showed no statistically significant differences in the risks of TIMI major (OR 1.13, 95% CI 0.59–2.14, P = 0.72) and TIMI minor (OR 0.79, 95% CI 0.44–1.40, P = 0.42) bleeding. We were not able to separate intracranial hemorrhages from TIMI major bleeding from the phase II trials included in the meta-analysis. Thus, the intracranial hemorrhage estimate is drawn from the TRA-CER and TRA-2P trials of vorapaxar, the two largest studies. Vorapaxar was associated with a significantly increased risk of intracranial hemorrhage as compared with placebo (OR 2.17, 95% CI 1.61–2.91, P < 0.001), with no significant heterogeneity (I2 = 54%, P = 0.14).
The composite of death, MI or stroke occurred in 4023 patients. Each trial reported a decreased risk of death, MI or stroke associated with PAR-1 antagonist treatment. Overall, PAR-1 antagonists significantly reduced the risk of the composite of death, MI or stroke as compared with placebo in both the fixed-effects (OR 0.81, 95% CI 0.87–0.92, P < 0.001; Fig. 4) and random-effects (OR 0.86, 95% CI 0.79–0.94, P = 0.001) models, with no significant heterogeneity (I2 = 11%, P = 0.34) or apparent systematic bias among the trials (Begg’s test, P = 0.29). These results did not change significantly for either model when only patients with qualifying CAD from the TRA-2P trial were included (OR 0.84, 95% CI 0.78–0.91, P < 0.001). The magnitude of the treatment effect of PAR-1 antagonists for the composite of death, MI or stroke increased after exclusion of the TRA-CER and TRA-2P trials from the meta-analysis (OR 0.52, 95% CI 0.33–0.83, P = 0.006). None of the other studies was found to unduly influence the estimate of the composite of death, MI or stroke. The risk of the primary efficacy endpoint was significantly reduced with PAR-1 antagonists vs. placebo in analyses restricted to patients who received vorapaxar (OR 0.87, 95% CI 0.81–0.93, P < 0.001), and non-significantly reduced in those who received atopaxar (OR 0.59, 95% CI 0.29–1.21, P = 0.15), although the lower number of patients may account for this lack of difference. Patients treated in the setting of NSTE-ACS showed a significant benefit from PAR-1 antagonists over placebo in reducing the composite of death, MI or stroke (OR 0.87, 95% CI 0.79–0.97, P = 0.008). Meta-regression analysis did not show statistically significant interactions between PAR-1 antagonists vs. placebo and female gender (P = 0.64), use of P2Y12 inhibitors (P = 0.41) or number of patients enrolled in each study (P = 0.65).
A total of 4956 MACCEs (primary efficacy endpoint plus recurrent ischemia) occurred. The MACCE ORs for PAR-1 antagonists vs. placebo were 0.89 (95% CI 0.84–0.94, P < 0.001) in the fixed-effects model (Fig. 5) and 0.87 (95% CI 0.78–0.97, P = 0.02) in the random-effects model, with no significant heterogeneity (I2 = 29%, P = 0.20) or apparent systematic bias among the trials (Begg’s test, P = 0.14). The magnitude of the treatment effect of PAR-1 antagonists for MACCEs increased after exclusion of the TRA-CER and TRA-2P trials from the meta-analysis (OR 0.63, 95% CI 0.44–0.92, P = 0.02). None of the other included studies was found to unduly influence the MACCE estimate. Overall, death and stroke occurred in 1765 (OR 0.99, 95% CI 0.90–1.09, P = 0.81 for PAR-1 antagonists vs. placebo) and 843 patients (OR 0.96, 95% CI 0.84–1.10, P = 0.59), respectively. Each trial reported a decreased risk of MI associated with PAR-1 antagonist treatment. Cumulatively, PAR-1 antagonist vs. placebo use was associated with a significant reduction in MI in the fixed-effects (OR 0.84, 95% CI 0.76–0.91, P < 0.001; Fig. 5) and random-effects (0.81, 95% CI 0.70–0.94, P = 0.004) models, with non-significant heterogeneity (I2 = 31%, P = 0.19). The relative benefit of PAR-1 antagonists vs. placebo in reducing MI was consistent both with vorapaxar (OR 0.85, 95% CI 0.78–0.92, P < 0.001) and with atopaxar (OR 0.44, 95% CI 0.19–0.99, P = 0.048), and in the analysis restricted to patients presenting with NSTE-ACS (OR 0.86, 95% CI 0.77–0.96, P = 0.007). The magnitude of the treatment effect of PAR-1 antagonists for MI increased after exclusion of the TRA-CER and TRA-2P trials from the meta-analysis (OR 0.44, 95% CI 0.27–0.72, P = 0.001). There was no apparent systematic bias as assessed by funnel plots among the trials (Begg’s test, P = 0.29).
The primary findings from this meta-analysis of PAR-1 antagonists vs. placebo are as follows. First, in a pooled analysis of randomized clinical trials, treatment with PAR-1 antagonists was associated with a significant safety hazard in terms of TIMI clinically significant, major and minor bleeding. Intracranial hemorrhages occurred more frequently with vorapaxar than with placebo. Second, the randomized use of PAR-1 antagonists vs. placebo was associated with a significant relative reduction in ischemic events, which was primarily driven by decreased rates of MI. Deaths and strokes were not significantly different between PAR-1 antagonists and placebo. Third, a meta-regression analysis showed statistically significant interactions between PAR-1 antagonists vs. placebo and percentage of patients receiving P2Y12 inhibitors for the primary safety outcome.
Vorapaxar and atopaxar are reversible and orally active small-molecule inhibitors that interfere with thrombin-mediated platelet activation via the PAR-1 receptor . In contrast to the 5′-ADP and thromboxane A2 platelet activation pathways, which are crucial for both primary hemostasis and pathologic thrombosis, the PAR-1-mediated platelet activation pathway via thrombin is a key contributor to platelet-centered positive feedback of activation of coagulation and platelet activation, but has been shown to play a relatively weak role in hemostasis [4–8]. The potential to reduce recurrent ischemic events without exposing patients to increased bleeding risk was outlined by dose-finding phase II trials of both vorapaxar and atopaxar [13–17], but not confirmed by the phase III trials of vorapaxar [18,19]. In fact, the TRA-CER trial was terminated early because of the observed significant increase in major bleeding, which was not overshadowed by a significant decrease in the primary composite ischemic endpoint , and in the TRA-2P trial the data and safety monitoring board recommended the discontinuation of vorapaxar in patients with a history of stroke on the basis of an excess of intracranial hemorrhage in such patients . In this pooled analysis, we observed that the signals of clinical efficacy observed in phase II trials of PAR-1 antagonists translated into a significant reduction in ischemic events, but this was driven only by reduced MI rates, regardless of the impact of the TRA-CER and TRA-2P trials (representing 95% of the overall cohort) on the overall treatment effect. Importantly, in the TRA-CER and TRA-2P trials, the magnitude of the treatment effect of vorapaxar in reducing ischemic events was smaller than that observed in previous, small-sized phase II trials. However, the 95% CIs excluded a null effect for the combined endpoint of cardiovascular mortality, MI and stroke and other component secondary endpoints, including spontaneous recurrent MI. The observed beneficial reduction in ischemic events observed in this pooled analysis was offset by a significant increased risk of clinically significant major and minor bleeding (including intracranial hemorrhages), mostly attributable to patients enrolled in the TRA-CER and TRA-2P trials. Importantly, a meta-regression analysis showed that a significant interaction might be found between bleeding and the use of P2Y12 inhibitors. This is consistent with previous evidence indicating that more potent antithrombotic therapy incrementally increases the risk of bleeding [22,23] and with the exploratory observation that patients who were not receiving a P2Y12 inhibitor at randomization in the TRA-CER trial did not experience a higher risk of bleeding . This circumstance could be more relevant in acute coronary syndrome patients, as evidence of significant heterogeneity of the effect of vorapaxar on the basis of treatment with clopidogrel was not observed in the TRA-2P trial . Therefore, future studies of vorapaxar (or atopaxar) in patients not receiving a P2Y12 inhibitor or studies of vorapaxar (or atopaxar) vs. P2Y12 inhibitors on a background of aspirin therapy might be eventually considered. In the design of these trials, special attention should be given to the duration and dosing of PAR-1 antagonist therapy in conjunction with other antiplatelet drugs, because, in the TRA-CER and TRA-2P trials, the rate of bleeding was found to increase over time, possibly influencing the cumulative risk–benefit profile. The results of these studies could provide insights into the safety and efficacy of more comprehensive platelet inhibition with PAR-1 antagonism as an adjunct to standard of care treatment, which includes aspirin and thienopyridines, in selected patients with atherothrombotic disease. Indeed, the results of dedicated studies of vorapaxar and, eventually, the further clinical development of atopaxar are expected to provide important information not only on patients who may benefit, but also on patients in whom there is a potential for harm with more aggressive antiplatelet treatment regimens.
Other potential side effects of PAR-1 antagonists were not consistently reported across trials, and therefore do not constitute a focus of this pooled analysis. In the LANCELOT-ACS and LANCELOT-CAD trials, there were no significant differences in the proportions of patients who discontinued the study drug between the combined active atopaxar treatment group and the placebo group [16,17]. In these studies, atopaxar was generally well tolerated, especially at the 50-mg treatment dose. However, relative QTcF prolongation and liver function abnormalities were observed more frequently with the highest doses of atopaxar, a finding that is consistent with the results of the J-LANCELOT-ACS and J-LANCELOT-CAD trials [15–17]. No differences in vorapaxar or placebo discontinuation were seen in the TRA-CER and TRA-2P trials [18,19].
The inherent limitations of study-level meta-analyses must be acknowledged, such as variations in study design, variations in endpoint definitions, and publication bias. This study included phase II trials, which can be characterized by limitations such as small sample size, patient selection, differences in concomitant medications, and dose ranging. Importantly, although these trials were designed to provide important insights into safety, they are largely underpowered with respect to fully addressing safety concerns. Importantly, the statistically significant increase in intracranial hemorrhages with PAR-1 antagonists observed in the TRA-CER and TRA-2P trials was not captured by phase II trials, and the results of the TRA-CER and TRA-2P trials largely dominated the results of the present meta-analysis, particularly regarding the bleeding endpoint. It is known that the impact of novel antiplatelet therapies on bleeding mostly emerges from larger-scale phase III clinical trials, as observed from recent studies with novel P2Y12 receptor antagonists, which, albeit showing favorable safety profiles in phase II testing, showed an increased risk of major bleeding, including fatal intracranial hemorrhages, in phase III studies [22,23]. This has shown to be also the case with PAR-1 antagonists, thereby raising a note of caution on failure to recognize the unpredictability of random variation leading to premature acceptance of safety results from small, underpowered trials.
PAR-1 antagonists have emerged as promising novel oral antiplatelet agents that inhibit thrombin-mediated platelet activation processes. In this meta-analysis conducted in a total of 41 647 patients from eight trials, PAR-1 antagonists decreased ischemic events in patients with CAD as compared with placebo, the decrease being mainly driven by a reduction in MI, at the cost of increased risks for clinically significant, major and minor bleeding, including intracranial hemorrhages. Further clinical testing is needed in order to define the best niche for PAR-1 antagonist use in patients with CAD.
Disclosure of Conflict of Interests
D. J. Angiolillo (corresponding author) reports honoraria for lectures from Bristol Myers Squibb, Sanofi-Aventis, Eli Lilly, Daiichi Sankyo, and Astra Zeneca, consulting fees from Bristol-Myers Squibb, Sanofi-Aventis, Eli Lilly, Daiichi Sankyo, The Medicines Company, Portola, Novartis, Medicure, Accumetrics, Arena Pharmaceuticals, Merck, and AstraZeneca, and research grants from Bristol Myers Squibb, Sanofi-Aventis, GlaxoSmithKline, Otsuka, Boston Scientific, Eli Lilly, Daiichi Sankyo, The Medicines Company, Portola, Accumetrics, Merck-Schering-Plough, Astra Zeneca, Eisai, and Johnson & Johnson. D. L. Bhatt discloses the following relationships: Advisory Board, Medscape Cardiology; Board of Directors, Boston VA Research Institute, Society of Chest Pain Centers; Chair, American Heart Association Get With The Guidelines Science Subcommittee; Honoraria from the American College of Cardiology (Editor, Clinical Trials, Cardiosource), Duke Clinical Research Institute (clinical trial steering committees), Slack Publications (Chief Medical Editor, Cardiology Today Intervention), and WebMD (CME steering committees); Senior Associate Editor, Journal of Invasive Cardiology; research grants from Amarin, AstraZeneca, Bristol-Myers Squibb, Eisai, Ethicon, Medtronic, Sanofi Aventis, The Medicines Company; and unfunded research for FlowCo, PLx Pharma, Takeda. S. Goto has received research grants and honoraria from Sanofi-aventis, Eisai, and Otsuka, and consulting fees from Merck and Eisai. M. L. O’Donoghue receives research grants from GlaxoSmithKline and Eisai, and honoraria from Eli Lilly and Daiichi Sankyo (continuing medical education only). D. J. Moliterno has received honoraria, consulting fees and research grant funding from Merck-Schering-Plough. The other authors state that they have no conflict of interest.