Effects of clopidogrel, prasugrel and ticagrelor on prevention of stent thrombosis in patients underwent percutaneous coronary intervention: A network meta‐analysis

Abstract Background Clopidogrel, prasugrel and ticagrelor, acting on platelet P2Y12 receptor, are commonly used for prevention of stent thrombosis (ST) among patients who underwent percutaneous coronary intervention (PCI). This study aimed to compare the effects of these drugs by a systematic review and network meta‐analysis. Hypothesis Efficacies of clopidogrel, prasugrel and ticagrelor on preventing ST are not the same. Methods PubMed, Embase and Cochrane Library were searched for randomized controlled trials (RCTs) that investigated the effect of clopidogrel, prasugrel, or ticagrelor on prevention of ST in patients who underwent PCI. The efficacies between groups were compared by a Bayesian network meta‐analysis, by which the pooled odds ratios (ORs) and 95% confidence intervals (CIs) was calculated. Results Fourteen studies and 46 983 participants were included in this study. The pooled results illustrated that clopidogrel, prasugrel and ticagrelor were effective on prevention of ST. Patients treated with prasugrel (OR = 0.30, 95% CI = 0.052 ~ 0.73, P < 0.05) and ticagrelor (OR = 0.25, 95% CI = 0.035 ~ 0.65, P < 0.05) had lower incidence of ST compared to those treated with clopidogrel. Patients treated with ticagrelor showed similar frequency with those in prasugrel group (OR = 0.86, 95% CI = 0.22 ~ 2.3, P > 0.05). No significant heterogeneity was observed across included studies. Conclusions Our findings suggest that prasugrel and ticagrelor are more effective than clopidogrel on prevention of ST among patients underwent PCI. Simultaneously, there is no significant difference in the prevention of ST between prasugrel and ticagrelor.

controlled trials (RCTs) that investigated the effect of clopidogrel, prasugrel, or ticagrelor on prevention of ST in patients who underwent PCI. The efficacies between groups were compared by a Bayesian network meta-analysis, by which the pooled odds ratios (ORs) and 95% confidence intervals (CIs) was calculated.   4 Therefore, it is increasingly concerned to reduce the incidence of ST after PCI treatment.
As is known, the occurrence of ST is associated with patient-, lesion-, procedure-, and post-procedure-related risk factors. 5 Platelets play an important role in the pathophysiology of ST. Implantation of stent struts initiates platelet activation and adhesion, followed by thrombus formation, which results in early ST. 5 Therefore, inhibition of platelet function is considered effective for preventing the occurrence of ST. Currently, one of the clinical approaches for inhibition of platelet aggregation is dual antiplatelet therapy (DAPT), which is based on the combination of aspirin with one of P2Y12 inhibitors (i.e., clopidogrel, prasugrel and ticagrelor). 6 Studies have focused on the differences in the efficacies of P2Y12 inhibitors. Pharmacodynamic studies have suggested that the effect of clopidogrel is weaker than prasugrel or ticagrelor. 7 Clinical trials and real-world studies discovered that patients using clopidogrel suffered from higher incidence of ST compared to those treated with prasugrel or ticagrelor. [8][9][10][11] However, up to date, the difference in the incidences of ST between prasugrel and ticagrelor has not been clearly elucidated. Two studies reported that prasugrel was more effective than ticagrelor for preventing ST 12,13 ; however, other studies released insignificant results. [14][15][16] Here, we conducted a systematic review and network meta-analysis to compare the efficacies of three P2Y12 inhibitors (i.e., clopidogrel, prasugrel and ticagrelor).

| Search strategy
PubMed, Embase and Cochrane Library were searched for randomized controlled trials (RCTs) that compared the effect of clopidogrel, prasugrel, and/or ticagrelor in patients who underwent PCI. The search terms were "ticagrelor or brilinta", "clopidogrel or plavix", "pra-

| Data extraction and quality assessment
The following data were collected from eligible studies: first author name, year of publication, region, type of study, duration of follow-up, number of participants, gender of the subjects, and treatment regimens. The data were independently collected by two authors. The discrepancies were resolved by discussion with a third author.
Quality of included studies was assessed in accordance with the Cochrane Risk of Bias Tool, which contains seven domains: (a) random sequence generation, (b) allocation concealment, (c) blinding of participants and personnel, (d) blinding of outcome assessment, (e) incomplete outcome data, (f) selective reporting, and (g) other bias. Two authors completed quality assessment separately, and any discrepancies would be resolved by a third author.

| Statistical analysis
We performed statistical analyses with Stata Version 14.0 (Stata Corp, College Station, TX, USA) and R Version 3.4.2 (R Development Core Team, Vienna, Austria). We carried out a Bayesian network metaanalysis using the "gemtc" package for R. We calculated odds ratios (ORs) and 95% confidence intervals (CIs) for comparisons between groups. We analyzed the convergence of Markov chain Monte Carlo chains for all model parameters using trace plots and Gelman-Rubin diagnostic statistics. With regard to heterogeneity test, a P < 0.10 indicated statistically significant. The potential publication bias was estimated by funnel plot asymmetry.

| Study quality
The Cochrane Risk of Bias tool was applied for assessment of methodological quality. 27 The quality evaluation of included studies was shown in Figures 2 and 3. Of the 14 studies included in this metaanalysis, only four studies applied blinding method, while six studies had potential risk of bias due to lack of blinding for participants. In addition, the lack of clear statement of allocation concealment (in nine studies) and blinding of outcome assessment (in 10 studies) might result in potential risk of selection bias and detection bias, respectively.

| Results of network meta-analysis
As shown in Figure 4

| Model convergence of our network metaanalysis
The model convergence was checked using Brooks-Gelman-Rubin diagnostic statistics ( Figure S1) and trace plots ( Figure S2) for all model parameters. The Brooks-Gelman-Rubin statistics got close to one fast, which revealed that the three Markov chain Monte Carlo chains mixed well regardless of their different initial starting points.
Meanwhile, the trace plots illustrated that every Markov chain Monte Carlo chains converged well.

| DISCUSSION
Our network meta-analysis indicated that prasugrel and ticagrelor had similar efficacy for preventing ST and both of them were more potent than clopidogrel. However, none of the three drugs could completely prevent ST.
As the most common complication, ST is a major clinical concern of post-PCI treatment. Studies showed that the occurrence of ST was significantly related to inadequate antiplatelet therapy. 28,29 According to clinical guidelines, DAPT, with a combination of aspirin and one P2Y12 inhibitor (i.e., clopidogrel, prasugrel or ticagrelor), has been recommended to be used for antiplatelet therapy. 6,30 TRITON-TIMI 38 study revealed that prasugrel caused less ST than clopidogrel (68/6745 vs. 142/6653). 8 Similarly, PLATO investigators reported that ticagrelor was more efficient on the prevention of ST events than clopidogrel (71/5569 vs. 106/5543). 17 Our network meta-analysis also obtained similar result. This might due to the different pharmacokinetics of these drugs. Prasugrel and ticagrelor generate active metabolite more rapidly than clopidogrel. [31][32][33] The conversion of clopidogrel is linked to ABCB1 which act as a transporting molecule mediating the uptake of drug by intestinal cells, CYP2C19 and paraoxonase 1 (PON1) which affect the biotransformation of clopidogrel. [34][35][36] This leads to a delayed anti-platelet action of clopidogrel. 29 By contrast, prasugrel requires only a single CYPdependent oxidative step during its metabolic transformation, 37 while the action of ticagrelor needs no transformation. 32,33 Although ticagrelor has pharmacokinetic advantages over prasugrel, two indirect comparison meta-analyses showed prasugrel might be more effective than ticagrelor for preventing ST. 12 More interestingly, several recent real-world studies actually found that ticagrelor was not superior to clopidogrel for preventing ST, [38][39][40] which was different from PLATO trial as well as our study. This might be explained by the following reasons. A potent mediator for efficacy attenuation of ticagrelor outside RCT might be decreased compliance because of the higher rate of adverse events (dyspnea, bleeding), administration of the drug twice daily and higher costs. 38

| CONCLUSION
In conclusion, our findings suggest that prasugrel and ticagrelor are more effective than clopidogrel on prevention of ST among patients underwent PCI. Additionally, there is no significant difference in the prevention of ST between prasugrel and ticagrelor.

CONFLICT OF INTEREST
The authors have no conflicts of interest to declare.

AUTHOR CONTRIBUTIONS
Conceived and designed the meta-analysis: Yuanmin Li and Haifeng Hou. Performed the study: Chen Zhang and Jian Zhao. Analyzed the data: Xiuxiu Xu and Qiang Xiao. Wrote the paper: Wenwen Chen.

DATA AVAILABILITY STATEMENT
The data supporting this network meta-analysis are from previously reported studies and datasets, which have been cited.