Dual antiplatelet therapy in patients with coronary artery disease after percutaneous coronary intervention with drug-eluting stents: A systematic review and network meta-analysis

Background: This network meta-analysis was committed to evaluating the ecacy and safety of different dual antiplatelet therapies (DAPTs) after percutaneous coronary intervention (PCI) with drug-eluting stents (DESs). Methods: Randomized controlled trials (RCTs) comparing two of the following DAPT strategies: long-term (>12 months) DAPT (L-DAPT), 12-months DAPT (DAPT 12Mo), short-term ( ≤ 6 months) DAPT followed by aspirin monotherapy (S-DAPT+ASA), short-term DAPT followed by a P2Y12 receptor inhibitor monotherapy (S-DAPT+P2Y12) were searched. Primary outcomes were all-cause mortality, cardiac death, myocardial infarction (MI), stroke, major bleeding, any bleeding, denite or probable stent thrombosis (ST). This Bayesian network meta-analysis was performed with the random-effects model. Results: Twenty-four RCTs (n=81,376) were included. L-DAPT increased the risk of major bleeding (OR 2.37, 95%CI 1.32-5.03 compared with S-DAPT+P2Y12) and any bleeding (OR 2.95, 95%CI 1.91-4.34 compared with S-DAPT+P2Y12). When compared with L-DAPT, DAPT 12Mo (OR 1.54, 95%CI 1.13-2.02) and DAPT+ASA (OR 1.67, 95%CI 1.22-2.19) were associated with higher rates of MI, but S-DAPT+P2Y12 obtained no statistical difference. The sensitivity analysis revealed that the risks of major bleeding and any bleeding further increased for ≥ 18 months of DAPT. In the subgroup analysis, short-term DAPT (S-DAPT) presented similar ecacy and safety to DAPT 12Mo for patients with the acute coronary syndrome (ACS), and lower risks of major bleeding and all-cause mortality were observed in S-DAPT+P2Y12 among patients with newer-generation DES. Conclusions: S-DAPT+P2Y12 presented superiority in after S-DAPT, which was a newer regimen under exploration. With a combination of direct and indirect comparisons, subsequently, on account of bleeding and ischemic risks, our study makes recommendations for patients with individualized DAPT. Our results suggest that S-DAPT especially continuing P2Y12 monotherapy, admires a lower rate of bleeding events under any clinical presentations. S-DAPT + P2Y12 is safe compared with L-DAPT without signicant ischemic harm for patients with ACS after DES implantation. Furthermore, S-DAPT + P2Y12 presents superiority compared with other therapies in patients with all clinical presentations, and in patients with ACS or at high ischemic risk L-DAPT is more admirable than DAPT 12Mo. More broadly, our ndings highlight the superiority of P2Y12 monotherapy after short-term DAPT in the precision medical practice and lay the foundation for further research between aspirin monotherapy and P2Y12 monotherapy.

ischemic events could be safely considered among ACS patients, which highlighted the feasibility of S-DAPT in patients enduring higher ischemia hazards [11].
Post-discharge bleeding was con rmed to be a signi cant contributing factor to cardiac and all-cause mortality, which greater than that associated with post-discharge MI [12]. Therefore, S-DAPT + P2Y12 was proposed lately to balancing thrombotic with hemorrhagic risks and has shown its superiority in several studies [13][14][15][16][17].
Thus, we conducted a network meta-analysis to narrow the gap where nite head-to-head experiment conclusions exist up to now and took full advantage of both direct and indirect evidence to comprehensively evaluate the safety and e cacy of various kinds of DAPT, eventually making recommendations for DAPT among patients after DES implantation.

Methods:
This network meta-analysis was performed in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines [18].

Search Strategy and Selection Criteria:
Two investigators (Y.S. and Y.H.) systematically and independently searched ve databases (PubMed, Embase, the Cochrane Library, Web of Science, and ClinicalTrials.gov) without language restrictions from their inception to 24 May 2020. The following keywords and controlled vocabularies: platelet aggregation inhibitors, aspirin, drug-eluting stent, percutaneous coronary intervention, randomized controlled trial, were used in different combinations (Additional le 1). The data of the TICO trial has published on 16 June 2020 was added to this study. Presentations about unpublished relevant RCTs of important international conferences were veri ed simultaneously for additional literature to guarantee a comprehensive search.
Eligible RCTs were supposed to meet the following inclusion criteria: 1) participants were adults with ACD who received DAPT after undergoing PCI with DES implantation; 2) the comparison interventions corresponded with more than two different arms of following candidate durations of DAPT: longer than 12 months (L-DAPT), 12 months (DAPT 12Mo), shorter than 6-month therapy followed by aspirin monotherapy (S-DAPT+ASA), shorter than 6-month therapy followed by a P2Y12 receptor inhibitor monotherapy (S-DAPT+P2Y12); 3) at least one of the following outcomes: all-cause mortality, cardiac death, MI, stroke, major bleeding, any bleeding, de nite or probable ST, were accessible; 4) a 12-month follow-up or longer.
Exclusion criteria included the following: 1) non-randomized trials; 2) bare-metal stents (BMSs) were predominantly used in trials; 3) uncompleted trials or irretrievable data (both baseline characteristics and outcomes); 4) trials were designed to analyze different kinds of stents, pharmacokinetics, platelet reactivity, or those with a cross-over design.
Data extraction and quality assessment: Two investigators (Y.X. and P.Z.) independently screened the titles, abstracts, and sequentially full articles under the guidance of inclusion criteria and then extracted data on the study design, baseline characteristics, and outcomes from full texts or published appendixes using pre-speci ed forms. The e cacy endpoints of interest included all-cause mortality, cardiac death, MI, stroke, major bleeding, any bleeding, de nite or probable ST. We gave preference to data from long term or mature follow-up of premier trials. And data extraction was under the instruction of the intention-to-treat principle. We appraised the quality of eligible studies according to the Cochrane Risk of Bias Tool based on six dimensions [19]. Moreover, a third investigator (J.J.) identi ed the accuracy of the information and handled contradictions by consensus.
Data synthesis and statistical analysis: We applied odds ratios (OR) with corresponding 95% con dence intervals (CI) to demonstrate time-toevent data quantifying the contributions of each duration, and 95% CI that do not cross one was considered statistically signi cant. We pooled evidence within the Bayesian framework for its superiority in reconciling complex conditions by using the Aggregate Data Drug Information System (ADDIS version 1.16.8). The network plots were drawn by Stata version 15.1 using the "networkplot" command.
First, we conducted a pair-wise meta-analysis to aggregate data of different treatments in terms of the aforementioned endpoints with a random-effects model (Additional le 2 g. 1). The heterogeneity of direct comparisons was assessed by the I 2 statistic and the p-value [19]. An I 2 value of smaller than 25% indicated low heterogeneity, 25-20% indicated moderate heterogeneity, above 50% indicated high heterogeneity, and the larger values represented the higher heterogeneity [19]. The p-value of 0.05 was identi ed as statistical signi cance [19]. Subsequently, we tted a Bayesian random-effects network meta-analysis model using Markov chain Monte Carlo (MCMC) algorithms to compare multiple therapies simultaneously. Models were calculated with MCMC simulations, using 4 chains with over-dispersed values, with Gibbs sampling basing on 50,000 simulation iterations after 20,000 tuning iterations.
Convergence was assessed by calculating the Potential Scale Reduction Factor (PSRF, the value <1.2 was considered as acceptable) according to the Brooks-Gelman-Rubin method [20]. The ranking probabilities of each intervention were calculated, and probabilities were positively related to values. Synthetic outcomes honored consistency when values of inconsistency factors were close to 0, outcomes of variance calculation were approximately equal, moreover, outcomes of inconsistency and consistency Search results: The systematic search outputted 16,042 citations and 42 articles were subsequently scrutinized. Then, the TICO trial was incorporated. Eventually, 24 randomized controlled trials with aggregately 81,376 participants were deemed eligible for inclusion ( Fig. 1) [7-10, 13-17, 23-37].
Characteristics of included studies and bias assessment: The fundamental features of enrolled studies, baseline characteristics of participants, de nitions of endpoints, outcomes of enrolled trials, and the network plots are summarized in additional le 2. Generally, 15,158 participants were randomly assigned to the L-DAPT arm, 35,743 to the DAPT 12Mo arm, 14,418 to the S-DAPT+ASA arm, and 16,057 to the S-DAPT+P2Y12 arm. The median follow-up duration was 18 months, with an interquartile range of 12 to 24 months. The distribution of baseline characteristics of patients was balance across comparisons.
An overwhelming majority of RCTs were judged to be low-risk categories according to the Cochrane Risk of Bias Tool, and other sources of bias made a huge contribution to the risk of bias (Additional le 2 table 5).
Outcomes of network meta-analysis: All-cause mortality and cardiac death: We synthesized 24 studies reporting all-cause mortality and 21 studies with 60,112 patients altogether reporting cardiac death. Although S-DAPT+P2Y12 resulted in diminished all-cause mortality and cardiac death compared with other DAPT durations, they demonstrated no signi cant differences (Fig. 2).
Hemorrhagic endpoints: Results of 18 studies with 54,746 participants about any bleeding and 22 studies with 79,009 participants about major bleeding were pooled. Short term DAPT, especially S-DAPT+P2Y12, was proved to honor overwhelming superiority (Fig. 2). The mate-analysis showed that L-DAPT resulted in an increased risk of major bleeding (OR 2.37, 95%CI 1.32-5.03) and any bleeding (OR 2.95, 95%CI 1.91-4.34) when compared with S-DAPT+P2Y12.

Stroke:
A total of 23 studies with 79,896 patients reported the data of stroke. We observed that 4 different therapies presented similar rates, and the differences were not statistically signi cant among them (Fig.   2).

Rank probabilities:
The rank probabilities were in accordance with the pooled results quanti ed by OR ( Fig. 3). S-DAPT+P2Y12 was ranked the best therapy for reducing all-cause mortality, cardiac death, major bleeding, and any bleeding. Besides, L-DAPT honored admiration in limiting myocardial infarction and de nite or probable ST.
Network coherence: The node-splitting analysis con rmed there was no signi cant difference between direct and indirect effects in closed loops, which veri ed favorable coherence in all endpoints (Additional le 2 table7).

Sensitivity analysis:
The SMART-DATE trial which de ned duration of long term DAPT as 12-month or longer might weaken the contradistinction among different interventions. We excluded it and rede ned long-term arm as received DAPT 18 months or longer. Results of sensitivity analysis did not in contradiction compared with the original meta-analysis but had more prominent differences (Additional le 2  DAPT 12Mo). The robustness of outcomes could also be con rmed in the pooled analysis of the remaining treatments.

Subgroup analysis:
We conducted subgroup analyses among patients with ACS for their higher ischemic risks relative to those with stable CAD, and among patients after newer-generation DES implantation owning to its superiority in ischemic outcomes [38,39].

Discussion:
Summary of evidence: We applied a Bayesian network meta-analysis, which included 24 RCTs with 81,376 participants to synthetically evaluate the e cacy and safety of four therapeutic regimens of DAPT in patients undergoing PCI with DES.
The principal ndings were summarized as follows:1) S-DAPT, especially when followed by a P2Y12 receptor inhibitor monotherapy, was associated with a signi cant reduction in hemorrhagic endpoints, and the discrimination became more remarkable when de ned long term arm as received DAPT longer than 18 months. 2) S-DAPT led to increasing rates of MI and de nite or probable ST, but no obvious difference was obtained when S-DAPT + P2Y12 compared with L-DAPT or DAPT 12Mo. 3) The four regimens had similar possibilities for all-cause mortality, cardiac death, or stroke overall S-DAPT + P2Y12 was relative optimum. 4) In patients with ACS, L-DAPT was observed a decreased risk of MI, besides short term DAPT was con rmed to has equivalent e cacy and safety to DAPT 12Mo. 5) In terms of newer-generation DES, S-DAPT + P2Y12 was associated with lower risks of major bleeding, any bleeding, and all-cause mortality.
Optimal DAPT is controversial concerning constantly updated DESs, clinical presentations, types of pathological coronary vessels, and constantly advanced antiplatelet drugs. On the strength of observational studies, previous guidelines recommended at least a 12-month duration of DAPT in patients receiving DES. Then, when concerning emerging RCTs that gave admiration to S-DAPT primarily enrolled non-ACS patients and prolonged DAPT proved to associated with lower ischemic complications rate without signi cant bleeding hazards in studies focusing on patients with increasing ischemia risk, recent guidelines gave weight to S-DAPT for patients at lower ischemic risk [1,2,29,30,32,[34][35][36][37].
Shortly afterward, the DAPT-STEMI trial, the REDUCE trial, and successive meta-analyses proved the validity and feasibility of S-DAPT for those with ACS, which contradicted with the guidelines [11,22,25,47].
Recent studies emphasized the correlation between bleeding events and mortality [12]. Based on the former conclusion, current researches were devoted to exploring the safety and e cacy of S-DAPT + P2Y12 to maximize the bene ts in terms of bleeding and ischemia events. The SMART-CHOICE, STOPDAOT-2, GLASSY, and TWILIGHT trials conformably honored S-DAPT + P2Y12 for signi cant reductions in hemorrhagic endpoints without increasing ischemic harm [14][15][16]48]. In line with the previous studies, our meta-analysis con rmed that S-DAPT + P2Y12 inhibitor monotherapy dramatically reduced the risk of bleeding events without a signi cant increase in ischemic complications [13][14][15][16]48]. And in patients implanted with newer-generation DES, which was much less thrombogenic owing to considerably thinner and more biocompatible struts, S-DAPT + P2Y12 was associated with enlarging bene ts according to our subgroup analysis.
Though without statistically signi cant differences, our outcomes revealed non-inferiority for cardiac death and all-cause mortality of S-DAPT when compared with L-DAPT. The thrombotic bene t of L-DAPT in our study, which may have been neutralized by signi cant bleeding harm, did not transform into a decline in all-cause or cardiovascular mortality. That is to say, bleeding makes a contribution to mortality risk in patients after PCI, and with suggestions from some studies, the hazard conveys risk comparable to that of an MI during follow-up [12]. It also endorsed the application of S-DAPT from another aspect.
As for patients with ACS, our subgroup analysis further con rmed the safety and e cacy of the use of a P2Y12 inhibitor monotherapy strategy, which was in accordance with the TICO trial [13]. However, although without a statistical difference in terms of the S-DAPT + P2Y12 arm, a decreased risk of MI veri ed in L-DAPT among ACS patients when compared with shorter-term DAPT was still unignorable. Besides, current study also favoring prolonged DAPT among patients who were at a higher ischemic risk, such as those with ACS, diabetes, and patients with previous stenting [49,50]. Taking all these into consideration, this kind of patient was inclined to gain more bene t from prolonged DAPT or potent antiplatelet agents compared with patients with stable ischemic heart disease. Thus, cautious interpretation was still required before applying S-DAPT + P2Y12 to patients confronting high ischemic risk in the absence of clinical data.
Our study is the most comprehensive network meta-analysis up to now. We committed to evaluating the e cacy and safety of P2Y12 monotherapy after S-DAPT, which was a newer regimen under exploration. With a combination of direct and indirect comparisons, subsequently, on account of bleeding and ischemic risks, our study makes recommendations for patients with individualized DAPT. Our results suggest that S-DAPT especially continuing P2Y12 monotherapy, admires a lower rate of bleeding events under any clinical presentations. S-DAPT + P2Y12 is safe compared with L-DAPT without signi cant ischemic harm for patients with ACS after DES implantation. Furthermore, S-DAPT + P2Y12 presents superiority compared with other therapies in patients with all clinical presentations, and in patients with ACS or at high ischemic risk L-DAPT is more admirable than DAPT 12Mo. More broadly, our ndings highlight the superiority of P2Y12 monotherapy after short-term DAPT in the precision medical practice and lay the foundation for further research between aspirin monotherapy and P2Y12 monotherapy.

Limitations:
Our study suffers from several limitations. First, a large majority of eligible RCTs pooled in our metaanalysis precluded patients related to recurrent ischemia and bleeding events; clopidogrel was predominantly used in enrolled trials, which weaken the universality of our conclusion. Second, a majority of trials were designed to test the non-inferiority of different DAPT strategies. Third, heterogeneity could be witnessed among the enrolled trials in terms of design and endpoints de nitions. Forth, results for ACS patients relied on subgroup or post hoc analyses of original RCTs, so the results should apply modestly in these patients. Finally, the comparison between S-DAPT + ASA and S-DAPT + P2Y12 was merely based on indirect evidence, so the conclusions must be considered exploratory on account of limited statistical power and should be interpreted with caution.

Conclusion:
S-DAPT associated with reduced risk of bleeding without a signi cant increase in ischemic complications showed non-inferiority compared with other therapies for patients undergoing PCI. In population with ACS or high risk of ischemic, L-DAPT is still honored. Further pre-speci ed randomized controlled trials comparing S-DAPT followed by P2Y12 monotherapy with aspirin monotherapy should be conducted to promote the individualization of DAPT. Availability of data and materials: The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.
Competing interests: The authors declare that they have no competing interests.
Authors' contributions: All authors contributed to the study conception and design. Y.S. and Y.H. searched the literature. Y.X. and P.Z. extracted the data. Y.X. and Y.S. analyzed data. Y.X. and P.Z written the rst draft of this manuscript. J.J contributed to the revision of manuscript. All authors read and approved the nal manuscript.  Forest plots illustrating network meta-analysis results of all endpoints