The Prognostic Impact of High On-Treatment Platelet Reactivity with Aspirin or ADP Receptor Antagonists: Systematic Review and Meta-Analysis

Objective. Negative results of recent randomized clinical trials testing the hypothesis of target therapy for patients with high on-treatment platelet reactivity (HOPR) have questioned its independent impact on clinical outcomes. 26 studies with 28.178 patients were included, with a median age of 66.8 (64–68) and 22.7% (22.4–27.8), of female gender. After a median follow-up of 1 year (0.1–1), cardiac adverse events occurred in 8.3% (3–11; all results are reported as median and interquartile range) of patients. Pooling all studies together, on-treatment platelet reactivity significantly increased the risk of adverse events (OR 1.33 [1.09, 1.64], I 2 = 0%). However, a sensitivity analysis showed that HOPR did not increase the risk of adverse events for patients with ACS, AMI, or stable angina as well as patients resistant to aspirin, ADP antagonists, or both. For all studies, publication bias was formally evident; after adjusting for this, HOPR did not significantly increase adverse cardiac events (OR 1.1 : 0.89–1.22, I 2 0%). Conclusions. After adjusting for clinical confounders (like risk factors and clinical presentation) and for relevant publication bias, HOPR was not an independent prognostic indicator in unselected patients with both stable and unstable coronary disease for an adverse cardiac event. The clinical importance of HOPR for high-risk populations remains to be assessed.


Introduction
Aspirin and ADP receptor antagonists represent an unquestionable strategy for patients undergoing percutaneous coronary intervention (PCI), both for stable and unstable coronary disease [1]. High on-treatment platelet reactivity (HOPR), variously defined and analyzed, has been reported in up to 30% of these patients [2] and has been linked to adverse cardiac events at follow-up [3][4][5][6].
Due to the high prevalence of HOPR and the assumption that HOPR increases the risk of adverse cardiac events, randomized clinical trials were performed to test the safety and efficacy of a tailored strategy (defined as an increase in dose or a switch to another ADP receptor antagonist) in patients undergoing PCI. When appraised separately, most of these studies were negative, without achieving the expected reduction in recurrent thrombotic events [7][8][9].
Prognostic impact of HOPR was assessed by at least two meta-analyses, although limited from methodological flaws [3,4], due to lack of adjustement for baseline differences in burden of traditional risk factors and clinical presentation, which may explain themselves the increased risk of adverse cardiac events in selected patients. These two studies, however, have not tested the independent clinical effect of inadequate platelet inhibition on outcomes; moreover they evaluated patients with different risk profiles (ACS and stable angina) and different treatments (aspirin together with ADP antagonists or periprocedural glycoprotein inhibitors [10,11]).
Randomisation of patients to HOPR and non-HOPR groups is obviously not feasible; consequently a bias analysis may help to elucidate the impact of HOPR on clinical prognosis independently from cardiovascular risk factors and clinical presentations.

Methods
The recent Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) amendment to the Quality of Reporting of Meta-analyses (QUOROM) statement, and recommendations from The Cochrane Collaboration and Meta-analysis of Observational Studies in Epidemiology (MOOSE) were followed during the development of the present systematic review [11][12][13][14][15][16].

Search Strategy and Study
Selection. Pertinent articles were searched in Medline, Cochrane Library, Biomed Central, and Google Scholar in keeping with established methods with MESH strategy and with the following terms: (Prognosis/Broad[filter]) AND (platelet * AND (reactivity OR aggregation OR activation OR response * ) AND (death OR (myocardial AND infarction))). Three independent reviewers (Fabrizio D' Ascenzo, Umberto Barbero, and Marta Bisi) screened the retrieved citations via the title and/or abstract; divergences were resolved via consensus. If potentially pertinent, studies were then appraised as complete reports according to the following explicit selection criteria. Studies were included if (i) reporting more than 50 patients (ii) independent prognostic impact of HOPR evaluated through multivariate analysis, while exclusion criteria were (i) nonhuman setting, (ii) duplicate reporting (in which case the manuscript reporting the largest sample of patients was selected), and (iii) interventional studies.

Data Extraction, End Points, and Sensitivity Analysis.
Three unblinded independent reviewers (Fabrizio D' Ascenzo, Umberto Barbero, and Marta Bisi) abstracted the following data on prespecified forms: authors, journal, year of publication, location of the study group, and baseline clinical and interventional features. Data extraction was conducted by mutual agreement and all potential disagreement was solved by consensus. Incidence of adverse cardiac events (all-cause mortality and cardiovascular mortality, nonfatal myocardial infarction and stroke, and revascularization and stent thrombosis) was the primary end point. Sensitivity analyses were performed appraising aspirin and ADP receptor antagonists separately. Similarly we appraise indications for PCI in stable and unstable disease (i.e., either unstable angina, ST and non-ST segment elevation myocardial infarction). Finally, we analyze all-cause death, stent thrombosis and major bleedings.

Internal Validity and Quality
Appraisal. Unblinded independent reviewers (Fabrizio D' Ascenzo, Umberto Barbero, and Marta Bisi) evaluated quality of included studies on prespecified forms. Modifying the MOOSE items to take into account the specific features of included studies [11], we separately abstracted and appraised study design, setting, and data source, as well as risk of analytical, selection, adjudication, detection, and attrition bias (expressed as low, moderate, or high risk of bias, as well as incomplete reporting leading to inability to ascertain the underlying risk of bias).

Data Analysis and Synthesis.
Continuous variables are reported as mean (standard deviation) or median (interquartile). Categorical variables are expressed as / (%). Statistical pooling was performed according to a random-effect model with generic inverse-variance weighting, computing risk estimates with 95% confidence intervals, using RevMan 5 (The Cochrane Collaboration, The Nordic Cochrane Centre, Copenhagen, Denmark), and Comprehensive Meta-Analysis. Metaregression analysis was performed to identify impact of length of follow-up on results. Small study bias was appraised by graphical inspection of funnel plots and formally through Begg and Mazumdar rank correlation, Egger's regression intercept, and Duval and Tweedie trim and fill [14].
For all studies, publication bias was graphically evident ( Figure 6) and formally assessed with Begg and Mazumdar rank correlation (with a positive Tau of 0.31) and with Egger's regression intercept (Intercept 0.42 : 0.11-0.69; -value 2.81). After adjusting for this bias with Duval and Tweedie trim and fill, HOPR was not a significant prognostic indicator for all studies (OR 1.1 : 0.89-1.22, 2 0%; trim and fill methods evaluate publication bias by evaluating number of "asymmetric" trials on the right side, removing and replacing them with  Angiolillo et al., 2007 [17] 173 67 ±   Multiple electrode aggregometry (AA and ADP) Sibbing et al., 2012 [37] Siller   missing counterparts at the pooled estimate, and evaluating the adjusted confidence interval [14]).

Discussion
The main results of the present meta-analysis, investigating incidence and impact of HOPR on prognosis, are as follows: (a) HOPR represents a frequent finding for patients with coronary artery disease, both in chronic and acute settings; (b) current evidence is limited from relevant publication bias; (c) after adjustment for clinical and methodological confounders HOPR appraised for "all comers" with CAD does not significantly increase the hazard of adverse cardiac events; and (d) usefulness in high-risk patients may not be excluded and remains to be assessed. Many reasons can explain nonresponsiveness to antiplatelet medications, such as interindividual variability in the metabolism of clopidogrel (which is a prodrug activated by CYP-3A4, CYP-2C19, and CYP1A2), drug-drug interactions (i.e., interaction on the same metabolic pathway for clopidogrel, but also competition for binding sites on COX-1 by nonsteroidal anti-inflammatory medications and aspirin), P2Y12 receptor polymorphisms and increased platelet turnover during inflammation, acute coronary events, and diabetes mellitus. Interestingly, conventional cardiovascular risk factors themselves (smoking, diabetes, and hyperlipidemia) and also the same clinical pattern of unstable angina, increasing macrophage's thromboxane synthesis, enhance resistance to aspirin [40].
Previously, numerous observational studies have demonstrated the causal relationship between laboratory evidence of nonresponsiveness to aspirin or clopidogrel and an increase hazard of death, myocardial reinfarction, and stent thrombosis during secondary prevention for coronary disease [18,19,8 BioMed Research International Collet et al., 2012 [7] 23,[41][42][43]. The obvious induction was that individualization of antiplatelet therapy based on laboratory tests should improve outcomes, even if most of these studies were limited by absence of multivariate adjustments, that is, without a global assessment of potential clinical confounders [19], for example, the presence of diabetes, which increases both HOPR and recurrent cardiac events after ACS.
However, subsequent randomized controlled trials questioned this hypothesis. In the ARMYDA-2 study, pretreatment with a 600 mg loading dose of clopidogrel given before PCI was demonstrated to be safe and, as compared with the 300-mg dose, reduced periprocedural MI without increased bleeding [44]. On the other hand, the GRAVITAS and the ARCTIC trials, which randomized patients with HOPR after PCI with drug eluting stents to high-dose clopidogrel compared with standard-dose, did not showe significant improvements in clinical outcomes [22,33]. Later, new evidence suggested that a more tailored therapy could be attained by switching to newer drugs [9,45,46]. Similarly, randomized evidence failed to demonstrate a clinical impact. The TRIGGER-PCI study showed that HOPR after elective PCI with DES implantation, if detected, can be reliably corrected by switching from clopidogrel to prasugrel but again failed to demonstrate an improvement in clinical outcomes [47]. A similar result emerged from the TRILOGY-ACS trial, randomizing patients with NSTE-ACS who were medically managed [48]. More recently, switching to ticagrelor seems to be associated to an effective reduction in HOPR but studies about the effective clinical impact are still lacking [47,49].
This meta-analysis indicates that HOPR does not seem to be a useful predictor of outcomes in an "all comers" CAD population. These results hold true both for overall studies, and, after appraisal for diagnosis, types of antiplatelet medication analysed and assays were exploited. These findings may be explained because they derive from data drawn from multivariate analysis, with a critical adjustment (even though limited by absence of randomization itself) for clinical features both increasing platelet resistance and risk of adverse events (like diabetes mellitus, smoking, or renal disease).
While HOPR should not totally be disregarded, a focus on high-risk patients seems more appropriate [49][50][51][52][53], for example, those with recurrent stent thrombosis in the absence of periprocedural or adherence problems or in diabetic or in HIV populations who have a well-known increased risk of recurrent events.
Current evidence remains burdened from relevant publication bias, which deeply affects clinical interpretation of HOPR. This phenomenon was described by psychologist Robert Rosenthal as the "file drawer problem"; he wrote that "journals are filled with the 5% of the studies that show Type I errors, while the file drawers are filled with the 95% of the studies that show nonsignificant results" [54]. In the cardiovascular field, this problem was recently demonstrated by Ioannidis and colleagues [55], who stated that, among 56 meta-analyses reporting relationships between biomarkers and cardiovascular events, only 13 were not affected by selection bias. However, most of current guidelines do not include this kind of evaluation, which may deeply influence every day clinical decisions.
Our analysis has some limitations, including a great number of observational studies, which brings incomplete data around follow-up and about the correct reporting of adverse Parodi et al., 2011 [32]  Heterogeneity: 2 = 0.00; 2 = 4.23, df = 17 (P = 1.00); I 2 = 0% Test for overall effect: Z = 1.04 (P = 0.30) Figure 5: Pooled analysis of odds ratio according to reactivity (aspirin: 10066 patients; ADR receptor antagonists: 6750 patients; both: 17436 patients, from above to below in Figure 5). effects, different definitions, and outcomes. Moreover, for each sensitivity analysis, the number of patients was inferior to that of overall population, although superior or similar to that of previous meta-analysis on this topic [3,4]. Again, just a small number of studies could reliably monitor compliance. Platelet reactivity tests differed in each study, which also limits the HOPR definition. Because of the selection criteria, no studies selected use the Platelet Vasodilator-Stimulated Phosphorylation test (PLT-VASP test), a flow cytometry test that is today the most specific test to assess the effect of the platelet P2Y12 antagonists (clopidogrel, ticlopidine, and prasugrel) [51]. Thus, the included studies' quality was evaluated according to standardized criteria and we separately abstracted and appraised study design, setting, and data source, as well as risk of analytical, selection, adjudication, detection, and attrition bias. For all studies, publication bias was formally assessed. After adjusting for this bias, HOPR did not significantly increase adverse cardiac events for all studies.
We therefore conclude that routine assessment of HOPR is not useful, but high-risk subsets of patients (i.e., diabetics, multiple cardiovascular risk factors, and important  Figure 6: Funnel plot of standard error (a) and of precision (b). White box: observed studies. Black box: imputed study (trim and fill methods evaluates publication bias by evaluating number of "asymmetric" trials on the right side, removing and replacing them with missing counterparts at the pooled estimate and evaluating the adjusted confidence interval).
comorbidities, especially if they need therapies potentially interacting with antiplatelet drugs) may potentially benefit from its assessment and interpretations remain to be assessed.

Abbreviations
ACS: Acute coronary syndrome AMI: Acute myocardial infarction HOPR: High on-treatment platelet reactivity PCI: Percutaneous coronary intervention.