Aspiration thrombectomy prior to percutaneous coronary intervention in ST-elevation myocardial infarction: a systematic review and meta-analysis

Background Trials of aspiration thrombectomy (AT) prior to primary percutaneous intervention (PCI) in patients with ST-segment elevation MI (STEMI) have shown apparently inconsistent results and therefore generated uncertainty and controversy. To summarize the effects of AT prior to PCI versus conventional PCI in STEMI patients. Methods Searches of MEDLINE, EMBASE and CENTRAL to June 2015 and review of reference lists of previous reviews. We included randomized controlled trials (RCTs) comparing AT prior to PCI with conventional PCI alone. Pairs of reviewers independently screened eligible articles; extracted data; and assessed risk of bias. We used the GRADE approach to rate overall certainty of the evidence. Results Among 73 potential articles identified, 20 trials including 21,660 patients were eligible; data were complete for 20,866 patients. Moderate-certainty evidence suggested a non statistically significant decrease in overall mortality (risk ratio (RR) 0.89, 95 % confidence interval, 0.78 to 1.01, risk difference (RD) 4/1,000 over 6 months), no impact on recurrent MI (RR 0.94, 95 % CI, 0.79 to 1.12) or major bleeding (RR 1.02, 95 % CI, 0.78 to 1.35), and an increase in stroke (RR 1.56, 95 % CI, 1.09 to 2.24, RD 3/1,000 over 6 months). Conclusions Moderate certainty evidence suggests aspiration thrombectomy is associated with a possible small decrease in mortality (4 less deaths/1000 over 6 months) and a small increase in stroke (3 more strokes/1000 over 6 months). Because absolute effects are very small and closely balanced, thrombectomy prior to primary PCI should not be used as a routine strategy.


Background
In patients with ST-segment elevation myocardial infarction (STEMI), primary percutaneous coronary intervention (PCI) rapidly restores myocardial flow resulting in decreased infarct size and decreased mortality compared to thrombolysis or conservative medical management [1]. Some patients may, however, experience distal embolization of thrombus and plaque debris with failure to adequately restore distal microcirculatory flow. This "no reflow" phenomenon is associated with an increase in infarct size and lower survival [2].
Randomized clinical trials (RCTs) comparing aspiration or mechanical thrombectomy prior to primary PCI to PCI alone have shown improvement in markers of myocardial reperfusion (e.g. "myocardial blush", ST-segment resolution post procedure) [3]. A recent meta-analysis of 20 RCTs addressing patient-important outcomes and including over 11,000 patients reported that aspiration thrombectomy prior to primary PCI was associated with a reduction in major coronary adverse events and 1-year mortality [4]. A more recent meta-analysis including 26 RCTs, reported a different conclusion: aspiration thrombectomy did not improve clinical outcomes [5]. Neither of these metaanalyses included the recently published Trial of Routine Aspiration Thrombectomy with PCI versus PCI Alone in Patients with STEMI (TOTAL), which randomized over 10,000 patients [6].
We therefore undertook a systematic review of all RCTs comparing aspiration thrombectomy prior to PCI versus PCI alone in patients with STEMI, focusing on patient-important outcomes. As composite endpoints varied between trials and can produce misleading results [7,8], we focused on individual endpoints of overall mortality, recurrent MI, stroke, and major bleeding.

Methods
This review adheres to the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) Statement [9]; the Quality of Reporting of Meta-analyses QUOROM [10]; and the Cochrane Handbook for Systematic Reviews of Interventions [11].

Eligibility criteria
We included RCTs that compared aspiration thrombectomy prior to PCI with conventional PCI in patients with STEMI, included any one of the following patientimportant outcomes: overall mortality, cardiovascular (CV) mortality, myocardial infarction (MI), stroke (including ischemic and hemorrhagic stroke) and, non-fatal extracranial major bleeding, and followed patients for at least 30 days. We excluded studies reported only as conference abstracts.

Data source and searches
A previous review with similar inclusion criteria identified studies up to December 2013 [5]. Using Medical Subject Headings (MeSH) based on the terms "thrombectomy," "thrombus aspiration," "thromboaspiration," "infarction," and "myocardial infarction" (Appendix Table 1) we replicated the search strategy of that review [5] for Medline, EMBASE, and Cochrane Controlled Trials Register (CENTRAL) from January 1, 2014 to June 26, 2015. We also reviewed reference lists of relevant review articles [4,5,12] and primary studies.

Selection of studies
Teams of two reviewers independently screened all titles and abstracts identified by the literature search, obtained full-text articles of all potentially eligible studies, and evaluated these studies for eligibility criteria.

Data extraction and risk of bias assessment
Three pairs of reviewers independently extracted the following data using a pre-standardized data extraction form: characteristics of the study design; participants; interventions; outcomes event rates and follow-up.
Reviewers independently assessed risk of bias by using a modified version of the Cochrane Collaboration's tool for assessing risk for bias tool [13] (http:/distillercer.com/resources/) [14] that includes nine domains: adequacy of sequence generation, allocation sequence concealment, blinding of participants and caregivers, blinding of data collectors, blinding for outcome assessors, blinding of data analysts, incomplete outcome data, selective outcome reporting, and the presence of other potential sources of bias not accounted for in the previously cited domains [14]. For incomplete outcome data we stipulated as low risk of bias loss to follow-up of less than 10 % and a difference of less than 5 % in missing data in intervention and control groups.

Certainty of evidence
The reviewers used the Grading of Recommendations Assessment, Development and Evaluation (GRADE) methodology to rate certainty of the evidence for each outcome as high, moderate, low, or very low [15]. Detailed GRADE guidance was used to assess overall risk of bias [16], imprecision [17], inconsistency [18], indirectness [19] and publication bias [20], and summarized results in an evidence profile. We assessed publication bias through visual inspection of funnel plots for 10 or more studies.
For decisions regarding eligibility, risk of bias assessment, and data abstraction, reviewers resolved disagreement through discussion with third party adjudication if necessary.

Data synthesis and statistical analysis
We chose six months as a follow-up time that represented duration important to patients, sufficient to include most events that would likely be influenced by thrombectomy, and would include relatively few events that would not be potentially influenced by thrombectomy. For meta-analyses we used six months data if available; and otherwise we chose the time point closest to six months, but preferring 1-year over 30 days.
We calculated pooled risk ratios (RRs) and associated 95 % confidential intervals (CIs) using random-effects models with statistical method of Mantel-Haenszel. Absolute effects and 95 % CI were calculated by multiplying pooled RRs and 95 % CI by baseline risk estimates derived from the TOTAL study (the most recent and largest of the included RCTs) [6]. We addressed variability in results across studies by using I 2 statistic and the P value obtained from the Cochran chi square test. Our primary analyses were based on eligible patients who had reported outcomes for each study (complete case analysis). For overall mortality we used all-cause mortality when available. For studies that did not present all-cause mortality we used cardiovascular mortality. We assessed publication bias through visual inspection of funnel plots for outcomes addressed in 10 or more studies. Review Manager (RevMan) provided the software for all analyses (version 5.3; Nordic Cochrane Centre, Cochrane) [21].
We also performed a meta-regression with a fixedeffect model using restricted estimated maximum likelihood with an observed log-odds ratio to predict whether mortality and recurrent myocardial infarction rates changed significantly by mean age. Meta-regression analysis was performed using Stata-13 (StataCorp LP, College Station, TX).

Selection of titles
Our search strategy focusing on publications since the last review identified 103 unique citations (Fig. 1).
After title and abstract screening, we assessed the full-text version of 38 relevant citations. In addition, we identified 42 potentially eligible publications included in previous systematic reviews, six [6,[22][23][24][25][26] of which were also identified in our search strategy. Thereafter, we assessed eligibility of 74 unique publications and excluded 49 studies (Fig. 1). As a result, we included 25 publications documenting 20 randomized controlled trials [6, involving 21,660 participants. Two studies [28,35] and one updated follow-up [46] were not included in any of the previous reviews.
EXPORT [34] The choice of medication during the procedure such as aspirin, heparin, clopidogrel, and glycoprotein IIb/IIIa inhibitors was also at the investigator's discretion, and were administrated according to standard hospital procedure.
IMPACT [35] Aspirin 300 mg and clopidogrel 600 mg preloading in the ambulance and anticoagulated with a heparin bolus (70-100 U/kg) after arterial sheath insertion to achieve an activated clotting time (ACT) >250 s. Adjunctive pharmacotherapy, including abciximab and bivalirudin, was given at the operator's discretion.

Liistro 2009 [40]
Aspirin (a loading dose of 500 mg), heparin (70 IU/kg), and clopidogrel (a loading dose of 600 mg). All patients also received the glycoprotein IIb/IIIa inhibitor abciximab with an intravenous procedural bolus of 0.25 mg/kg followed by a continuous intravenous infusion of 0.125 μg/kg/min for 12 hours and postprocedural infusion without heparin.
REMEDIA [41] Heparin (initial weight-adjusted IV bolus then further boluses administered with the aim of obtaining an activated clotting time of 250 to 300 s in patients treated with abciximab and > 300 s in the remaining subjects) and with double antiplatelet therapy with aspirin and clopidogrel (loading dose of 300 mg followed by 75 mg/day) for at least four weeks. Unless contraindicated, abciximab (0.25 mg/kg bolus plus infusion of 0.125 μg/kg/min for 12 h) was intravenously administered in all patients undergoing primary PCI, whereas in those with failed thrombolysis, abciximab use was left to the operator's discretion.

Shehata 2014 [25]
Aspirin (a loading dose of 500 mg), heparin (70 IU/kg), and clopidogrel (a loading dose of 600 mg). All patients also received the glycoprotein IIb/IIIa inhibitor abciximab with an intravenous procedural bolus of 0.25 mg/kg followed by a continuous intravenous infusion of 0.125 g/kg/min for 12 hours and postprocedural infusion without heparin.
TAPAS [43,44] Aspirin (a loading dose of 500 mg), heparin (5000 IU), and clopidogrel (a loading dose of 600 mg). Patients also received the glycoprotein IIb/IIIa inhibitor abciximab, with the dose based on body weight, unless contra-indicated, and additional heparin, with the dose based on the activated clotting time.
TASTE [26,27] Patients received the following procedure-related medication: bivalirudin, clopidogrel or ticlopidine, acetylsalicylic acid, ticagrelor, prasugrel, heparin, low-molecular-weight heparin, and glycoprotein IIb/IIIa blocker. The use of platelet inhibitors or anticoagulants was left to the discretion of the treating physician.
VAMPIRE [47] Aspirin and intravenous heparin boluses were administered during the procedure to maintain an activated clotting time ≥ 300 s.
The choice of medication during the procedure such as aspirin, heparin, clopidogrel, and glycoprotein IIb/IIIa inhibitors was at the investigator's discretion in one of the included studies [34]. The patients in one further trial [26,27] received the following procedure-related medication: bivalirudin, clopidogrel or ticlopidine, acetylsalicylic acid, ticagrelor, prasugrel, heparin, low-molecular-weight heparin, and glycoprotein IIb/IIIa blocker, while in other one [6] patients received unfractionated heparin; bivalirudin; enoxaparin and; glycoprotein IIb/IIa inhibitor (Table 2). Patients in TROFI trial [45,46] received only heparin in ambulance and, in VAMPIRE trial [47] aspirin and intravenous heparin boluses were administered during the procedure to maintain an activated clotting time ≥ 300 s.

Risk of bias assessment
A possibly important limitation with respect to risk of bias was lack of blinding for caregivers. A number of studies, including the larger ones, blinded the adjudicators of outcome. Follow-up was largely satisfactory: 14 trials lost less than 10 % of patients to follow-up (Table 3 and Fig. 2). Defined as less than 10 % loss to outcome data or difference between groups less than 5 % and those excluded are not likely to have made a material difference in the effect observed All answers as: yes (low risk of bias), probably yes, probably no, no (high risk of bias)

Outcomes
Appendix Table 2 presents the mortality data by individual study and Appendix Table 3 presents individual study outcome data for recurrent MI, stroke, and bleeding.

Fig. 2 Risk of bias assessment
More than 10 studies addressed overall mortality and recurrent MI; for both, funnel plots did not suggest publication bias (Appendix: Figures 1 and 2).

Meta-Regression analysis
Data from studies assessed in a meta-regression showed that the relationship between mortality rates decreased with increasing mean age; however was not significant (slope: −0.011; 95 % confidence interval: −.0980 to .0765; P = 0.784; Fig. 7). Similarly, the relationship between recurrent myocardial infarction rates decreased with increasing mean age; however was not significant (slope: −0.011; 95 % confidence interval: −.1175 to .0944; P = 0.811; Fig. 8).

Main findings
Based on pooled data from 20 randomized trials with more than 20,000 patients, we found moderate quality evidence for a non-statistically significant reduction in overall mortality (4 fewer deaths/1000 treated over 6 months) ( Table 4) and a small potential increase in stroke (3 additional strokes/1000 treated over 6 months) ( Table 4) in patients treated with thrombectomy. Moderate quality evidence suggests no impact of thrombectomy on either recurrent MI or major bleeding (Table 4).
A number of factors decreased our certainty in the estimates for overall mortality. In particular, the confidence interval included both no reduction in deaths and a mortality reduction that although small (8 fewer deaths in 1,000 over six months), many would consider important. Similarly with stroke: the confidence interval includes no increase in stroke and an increase of 6 more strokes in 1,000 patients over 6 months with thrombectomy, which many would consider an important risk. Other issues decreasing confidence in our estimates included potential risk of bias imposed by lack of blinding of patients and health care providers in all studies, and lack of blinding of outcome adjudicators in some studies.
The meta-regression analyses showed that both mortality and recurrent myocardial infarction rates decreased with increasing mean age. However, there was a non-significant difference between these two variables and the mean age of participants in both studied groups. A study [49] evaluated through a meta-regression whether there is an association between age, gender, diabetes mellitus, previous myocardial infarction and ejection fraction, and the choice of revascularization, focusing on death, myocardial infarction, repeat revascularization and stroke. The authors found that the reduction in stroke was significantly higher in females, and that women and patients with diabetes mellitus were at increased risk of subsequent revascularization after PCI [49].

Strengths and limitations
Strengths of our review include a comprehensive search; assessment of eligibility, risk of bias, and data abstraction independently and in duplicate; use of the GRADE  No studies were blinded to patient or caregiver. Some studies (minority of subjects enrolled) did not indicate blinded adjudication. While not specifically rating down for risk of bias, these additional concerns plus borderline clinically important imprecision led to downgrading of certainty in estimates for all outcomes 2 Some studies only report cardiovascular and not all cause mortality. However cardiovascular mortality constituted significant proportion of overall mortality in studies reporting both types of mortality. Therefore we opted against rating down for indirectness 3 95% CI for absolute effects include clinically important benefit and no benefit 4 Baseline risk estimates for mortality, recurrent MI, stroke, and major bleeds come from control arm of TOTAL study (largest and most recent randomized trial) 5 95% CI for absolute effects include benefit and harm 6 95% CI for absolute effects include clinically important harm and no harm approach in rating the quality of evidence for each outcome; and focus on absolute as well as relative effects of the intervention on patient-important outcomes. In this case, the small and more or less equivalent number of possible deaths prevented and strokes caused by thrombectomy, and the uncertainty consequent on the imprecision and risk of bias issues, are crucial in considering patient management (Table 4).
Potential limitations are related to the available data. Trials often suffered from incomplete outcome reporting, and lack of blinding consequent on the nature of the intervention, but for some studies also avoidable lack of blinding (outcome adjudication).

Relation to prior work
Recently published results from another metaanalysis [50] as well as data from a limited metaanalysis conducted as part of an evaluation of the outcome of stroke in the TOTAL study [12] are in general consistent with our findings. Results from all three analyses are in general consistent with our findings. Our systematic review and meta-analysis nevertheless adds important information as a result of our comprehensive assessment of risk of bias issues, our use of a complete case analysis that avoids assumptions regarding patients lost to follow-up, our use of the GRADE approach to rate quality of evidence, and our focus on absolute effects of thrombectomy required for optimal decisionmaking.
Furthermore, another review compared the effects of thrombectomy as an adjunct to PCI in the management of acute myocardial infarction in 20,853 patients [51]. The authors concluded that mortality; reinfarction and; stent thrombosis rates did not differ significantly between patients treated with or without AT; but stroke rates were increased with AT [51].   Meta-regression of mortality rates by mean age. Each circle represents a study highlighted by its weight in the analysis. The relationship between mortality and mean age in both groups was not significant (slope: -0.011; 95 % confidence interval: -.0980 to .0765; P = 0.784) Fig. 6 Meta-analysis comparing aspiration PCI versus conventional PCI on major bleeding Fig. 8 Meta-regression of recurrent myocardial infarction rates by mean age. Each circle represents a study highlighted by its weight in the analysis. The relationship between recurrent myocardial infarction and mean age in both groups was not significant (slope: -0.011; 95 % confidence interval: -.1175 to .0944; P = 0.811)

Implications
The possible magnitude of benefit with respect to mortality and magnitude of harm with respect to stroke are smallsome might say very smalland similar both with respect to magnitude and likelihood that the effects are real. With respect to mortality, the most likely mechanism of benefit would be a reduction in recurrent MI; the data, however, provide no support for an impact of thrombectomy on MI.
Similarly the mechanism of an increase in stroke is not immediately apparent. In a recent analysis of data from the TOTAL study, thrombectomy was associated with a small increase in procedure time as well as increased use of larger catheters (99.2 % vs. 97.5 % > 5 French) [12]. One could postulate this could lead to an increase in embolization of aortic atherosclerotic plaque leading to increased early ischemic events. More frequent development of subsequent atrial fibrillation would constitute another possible mechanism; no study reported this outcome.
Initial enthusiasm for thrombectomy was motivated by evidence of improvement in markers of myocardial tissue reperfusion. Our findings emphasize the need for caution with respect to surrogates, and the desirability of focus on outcomes important to patients. While it is not routinely justified there may be individual cases in which an operator may feel the potential benefit of the procedure outweighs potential risks.
The absolute effects of thrombectomy prior to primary PCI are very small and still associated with uncertainty. Given the best estimates of effect and associated quality of evidence, fully informed risk adverse patients -and particularly those who are highly stroke risk averse -would likely decline thrombectomy. Patients who place high value on an uncertain mortality reduction and have limited concern regarding a possible stroke increase would be more likely to choose to undergo the procedure. Given current concerns regarding overtreatment and efficient use of health care resources, a policy decision to not use thrombectomy in a particular catheterization laboratory is defensible.

Conclusions
Moderate certainty evidence suggests aspiration thrombectomy is associated with a possible small decrease in mortality (4 less deaths/1000 over 6 months) and a small increase in stroke (3 more strokes/1000 over 6 months). Because absolute effects are very small and closely balanced, thrombectomy prior to primary PCI should not be used as a routine strategy.