OUP user menu

Role of adjunctive thrombectomy and embolic protection devices in acute myocardial infarction: a comprehensive meta-analysis of randomized trials

Anthony A. Bavry, Dharam J. Kumbhani, Deepak L. Bhatt
DOI: http://dx.doi.org/10.1093/eurheartj/ehn421 2989-3001 First published online: 23 September 2008


Aims Adjunctive thrombectomy and embolic protection devices in acute myocardial infarction have been extensively studied, although outcomes have mainly focused on surrogate markers of reperfusion. Therefore, the effect of adjunctive devices on clinical outcomes is unknown. This study sought to determine whether the use of a thrombectomy or embolic protection device during revascularization for acute myocardial infarction reduces mortality compared with percutaneous coronary intervention (PCI) alone.

Methods and results The Cochrane and Medline databases were searched for clinical trials that randomized patients with ST-elevation acute myocardial infarction to an adjuvant device prior to PCI compared with PCI alone. Devices were grouped into catheter thrombus aspiration, mechanical thrombectomy, and embolic protection. There were a total of 30 studies with 6415 patients who met our selection criteria. Over a weighted mean follow-up of 5.0 months, the incidence of mortality among all studies was 3.2% for the adjunctive device group vs. 3.7% for PCI alone (relative risk, 0.87; 95% confidence interval, 0.67–1.13). Among thrombus aspiration studies, mortality was 2.7% for the adjunctive device group vs. 4.4% for PCI alone (P = 0.018), for mechanical thrombectomy, mortality was 5.3% for the adjunctive device group vs. 2.8% for PCI alone (P = 0.050), and for embolic protection, mortality was 3.1% for the adjunctive device group vs. 3.4% for PCI alone (P = 0.69).

Conclusion Catheter thrombus aspiration during acute myocardial infarction is beneficial in reducing mortality compared with PCI alone. Mechanical thrombectomy appears to increase mortality, whereas embolic protection appears to have a neutral effect.

  • Acute coronary syndrome
  • ST-elevation myocardial infarction
  • Meta-analysis
  • Mortality
  • Major adverse cardiac events
  • Thrombectomy
  • Embolic protection
  • Thrombus aspiration
See page 2953 for the editorial comment on this article (doi:10.1093/eurheartj/ehn488)


Percutaneous coronary intervention (PCI) during acute myocardial infarction usually achieves normalized coronary epicardial flow; however, distal embolization is common and it impairs microcirculation.1,2 Poor myocardial reperfusion is associated with adverse outcomes including reduced left ventricular function and survival.35 Numerous adjunctive coronary devices have been developed in an attempt to decrease or prevent embolization during revascularization and therefore to improve clinical outcomes.

The use of embolic protection devices to reduce adverse cardiac events in stable patients with saphenous vein graft lesions has been documented previously.6,7 Accordingly, embolic protection of saphenous vein grafts is considered class I therapy by the recent practice guidelines.8,9 Conversely, during acute myocardial infarction, the effect of adjunctive devices on clinical outcomes is unknown.1012 Moreover, some devices such as the AngioJet device (Possis Medical, Inc., Minneapolis, MN, USA) have been shown to increase the adverse cardiac events.13 One explanation why clinical outcome data with adjunctive devices is currently unknown is that individual trials have been powered to study only the surrogate markers of clinical outcomes, such as the Thrombolysis In Myocardial Infarction (TIMI) blush grade14 and ST-segment resolution.15 Therefore, the aim is this study was to examine clinical outcomes, most notably mortality with similar type thrombectomy or embolic protection devices.


Literature review

A computerized literature search of the Cochrane and MEDLINE (R) databases from January 1996 to June 2008 was conducted for randomized clinical trials using the following MeSH terms and keywords: myocardial infarction, thrombectomy, suction, embolization, microcirculation, thrombus aspiration, embolic protection, and balloon occlusion. There was no language restriction. Supplements from the Journal of the American College of Cardiology, Circulation, European Heart Journal, and American Journal of Cardiology were hand-searched, prior meta-analyses were reviewed, and ‘Thrombectomy’ and ‘Embolic Protection’ were used as a search terms in the websites http://clinicaltrials.gov and www.tctmd.com in an attempt to locate unpublished studies and increase the sensitivity of our search. We contacted selected study investigators to clarify outcome data. Eligible studies were finally cross-referenced with the Science Citation Index. Figure 1 displays the flow diagram of trial selection.

Figure 1

Flow diagram of the search strategy used to obtain the eligible studies.

Selection criteria

We selected studies that randomized patients within 12 h of acute myocardial infarction to one of the three strategies to protect against embolization: (i) thrombus aspiration prior to PCI vs. PCI alone, (ii) mechanical thrombectomy prior to PCI vs. PCI alone, and (iii) embolic protection prior to PCI vs. PCI alone. We required that studies reported clinical outcome data and/or markers of myocardial reperfusion. We excluded studies that performed thrombectomy on saphenous vein grafts or studies that compared one type of thrombectomy device against another.

Definition of adjunctive devices

Catheter thrombus aspiration is performed by a low-profile catheter (Export, Medtronic, Santa Rosa, CA, USA; Pronto, Vascular Solutions, Minneapolis, MN, USA; Diver, Invatec, Brescia, Italy; Rescue, Boston Scientific/Scimed, Maple Grove, MN, USA; TVAC, Nipro, Osaka, Japan) that is advanced to the thrombus over a guidewire and aspiration performed through syringe suction. Mechanical aspiration devices (AngioJet; X-sizer, ev-3, White Bear Lake, MN, USA) apply energy through saline jets or a rotating catheter head to facilitate break-up of the thrombus prior to aspiration. Embolic protection devices (Percusurge GuardWire, Medtronic; FilterWire, Boston Scientific, Natick, MA, USA; SpideRX, ev3, Minneapolis, MN, USA; Angioguard, Cordis, Miami, FL, USA) employ an occlusive balloon or filter that is advanced distal to the thrombus on it’s own guidewire prior to coronary revascularization. The GuardWire uses balloon occlusion, whereas the remaining devices use a filter.

Endpoints/data abstraction

The primary clinical endpoint was all-cause mortality. Secondary endpoints were major adverse cardiac events (MACEs) and the individual cardiac outcomes of myocardial infarction, target vessel revascularization, and stroke. MACE was defined as the composite of death, myocardial infarction, or target vessel revascularization. Myocardial infarction was defined as recurrent ischaemic symptoms with an elevation of creatine-kinase or troponin to at least twice the upper limit of normal. Stroke was defined as a major ischaemic or haemorrhagic cerebrovascular accident resulting in disabling neurological symptoms for at least 24 h. Target vessel revascularization was defined as recurrent ischaemic symptoms that resulted in the need for repeat PCI or surgical revascularization.

Markers of myocardial reperfusion included myocardial blush grade and ST-segment resolution. Optimal myocardial reperfusion was defined as TIMI blush grade of 3 (or at least 2 in some studies), whereas complete ST-segment resolution was defined as 70% (or at least 50% in some studies) resolution in peak ST-segments. We attempted to adjudicate this outcome at 60 min, although if this was unavailable, we recorded this outcome immediately after the procedure or up to 90 min after intervention. For studies with multiple time points, we used the time closest to 60 min. Outcomes were tabulated by two independent reviewers (A.A.B., D.J.K.) and the number of events that occurred in each arm of each trial was recorded. Discrepancies were resolved through discussion, and by a third reviewer (D.L.B.) if needed. Baseline information such as patient demographics, initial TIMI flow 0 or 1, ischaemic duration, and the extent of clinical follow-up was also tabulated.

Statistical analysis

For clinical outcomes, we used the intention-to-treat analysis, so that the denominator of the incidence was the total number of patients randomized to a given therapy. There were generally fewer patients available for the determination of myocardial blush grade or ST-segment analysis; therefore, to determine the incidence of these outcomes more accurately, we used the treatment-received (or protocol) analysis. Automatic ‘zero cell’ correction was used for studies with no events for a particular outcome.

A Mantel–Haenszel model was used to construct fixed effects summary risk ratios (RRs) and risk differences, whereas a DerSimonian–Laird model was used to construct random effects summary estimates. For all outcomes, we reported the fixed-effects model, unless there was significant heterogeneity, in which case we reported the random-effects model. Heterogeneity between studies was assessed by calculating I2 statistic and publication bias was assessed using the Begg’s method. Quality was assessed on the following components: adequate description of treatment allocation, blinded outcome assessment, and description of losses to follow-up.16 Since the experimental arm was an interventional device, concealment of allocation sequence was not possible. We followed the QUOROM statement for conducting high-quality meta-analysis.17

Clinical data were initially analysed among all the studies, and further analysed with similar type adjunctive devices grouped together. Since a proportion of the studies were abstract presentations, sensitivity analysis was performed on the published and non-published studies separately; however, the principal results were derived from all available studies. All outcomes were analysed at the maximal extent of clinical follow-up, while the mortality data were additionally analysed at 1 month or less. The weighted mean duration of follow-up was calculated according to an individual trial’s sample size and length of follow-up. Events reported to hospital discharge were assumed to occur at 3 days. All P-values were two-tailed, with statistical significance set at 0.05, and confidence intervals (CI) were calculated at the 95% level. All analyses were performed using STATA software v9.0. (STATA Corporation, College Station, TX, USA).


Baseline characteristics

A total of 30 studies with 6415 patients met our selection criteria (Figure 1, Tables 1, 2).13,1847 All patients received aspirin and intravenous heparin. At least 300 mg of clopidogrel (or an equivalent dose of ticlopidine) was given in most studies. The remaining studies either did not give a thienopyridine loading dose or no information was available.20,25,27,2934,4447 Glycoprotein IIb/IIIa inhibitors were used according to operator discretion. Thrombus aspiration devices represented 47% of the overall patient population, whereas embolic protection devices represented 38% and mechanical aspiration devices represented 15% of the study population. The proto-typical patient was male, with ST-elevation myocardial infarction (baseline TIMI 0/1 flow and frequently with visible thrombus) who presented <12 h from symptom onset. The duration of follow-up ranged from the time of hospital discharge to 12 months.

View this table:
Table 1

Baseline characteristics and duration of follow-up for the study participants

View this table:
Table 2

Quality assessment

Clinical events

Over a mean follow-up of 5.0 months, the incidence of mortality among the patients in all the studies was 3.2% for adjunctive devices vs. 3.7% for PCI alone (RR = 0.87, 95% CI = 0.67–1.13, P = 0.29, I2 = 0%, P for bias = 0.62) (Figure 2). Among the thrombus aspiration studies at a mean of 6.2 months, mortality was 2.7 vs. 4.4% (P = 0.018), among the mechanical thrombectomy studies at a mean of 4.6 months, mortality was 5.3 vs. 2.8% (P = 0.050), and among the embolic protection studies at a mean of 3.7 months, mortality was 3.1 vs. 3.4% (P = 0.69), respectively, for adjunctive device vs. PCI alone (Figure 3). The results for early mortality (i.e. hospital discharge to 30 days) are shown in Table 3.

Figure 2

Summary plot of mortality. Duration of follow-up is documented after each trial name. Trials with no events in either arm are shown as ‘excluded’, whereas trials that did not adjudicate this outcome are not listed (the same applies for Figures 47).

Figure 3

Incidence of mortality with similar type adjunctive thrombectomy devices grouped together.

View this table:
Table 3

Measures of effect for clinical outcomes and surrogate markers of reperfusion for similar type adjunctive devices

Among all the studies, the incidence of myocardial infarction was 1.3 vs. 1.9% (Figure 4), the incidence of target vessel revascularization was 5.9 vs. 6.4% (Figure 5), the incidence of stroke was 1.3 vs. 0.59% (Figure 6), and the incidence of MACEs was 11.2 vs. 12.7% (Figure 7). The clinical outcomes for similar type adjunctive devices are presented in Table 3, and sensitivity analysis is presented in Table 4.

Figure 4

Summary plot of myocardial infarction.

Figure 5

Summary plot of target vessel revascularization.

Figure 6

Summary plot of stroke.

Figure 7

Summary plot of major adverse cardiac events.

View this table:
Table 4

Sensitivity analyses on clinical outcomes and surrogate markers of reperfusion

Myocardial perfusion

Among studies that examined TIMI blush, a grade of 3 was reported in all studies, except for three studies that reported a grade of 2 or 3 together.23,28,42 Among studies that examined complete ST-segment resolution, four studies reported ST-segment resolution of >50% (instead of 70%),3235 eight studies reported ST-segment resolution immediately after the procedure20,23,25,26,3235, and four studies reported ST-segment resolution 90 min after the procedure.13,21,22,28

Among all the studies, the incidence of TIMI blush grade of 3 after revascularization was 53% for adjunctive devices vs. 40% for PCI alone (RR = 1.38, 95% CI = 1.20–1.58, P < 0.0001, I2 = 84%, P for bias = 0.06). The incidence of complete ST-segment resolution was 63% for adjunctive devices vs. 53% for PCI alone (RR = 1.27, 95% CI = 1.15–1.41, P < 0.0001, I2 = 69%, P for bias = 0.032). The outcomes of surrogate markers for similar type adjunctive device studies are presented in Table 3.


Our analysis of 30 randomized trials in 6415 patients with acute myocardial infarction shows that thrombectomy devices produce heterogeneous clinical outcomes depending on the type of the device used. Cather thrombus aspiration (for example, Export and Pronto) prior to PCI significantly reduces mortality, compared with PCI alone. This translates into ∼59 patients who need to be treated with a thrombus aspiration device to save a life. In contrast, mechanical thrombectomy (for example, AngioJet and X-sizer) appears to significantly increase mortality, with 38 patients who need to be treated to result in an excess death. With the available data, embolic protection devices (for example, GuardWire and FilterWire) appear to neither significantly increase nor decrease the mortality in this setting.

Our findings apply to the studied patient population which is largely individuals with native vessel ST-elevation myocardial infarction who present within a relatively early period (ischaemic range, 2.4–7.9 h). Our findings may not be as applicable to non-ST-elevation acute coronary syndromes, patients who present after 12–24 h of symptom onset, or patients with acute myocardial infarction due to a saphenous vein graft lesion.

The incidence of mortality with PCI alone varied among the trials, although it was highest among the thrombus aspiration studies. This was probably an effect of the TAPAS trial which had the longest follow-up among all the studies (i.e. 12 months). This extended follow-up allowed additional deaths to accrue. The relative increase in mortality with mechanical thrombectomy compared with PCI alone was driven by excess deaths in the AiMI trial; however, the trial investigators did not find a specific device-related explanation for the findings.13 Owing to these concerns, the data safety monitoring board extended the follow-up to 6 months, which documented three excess deaths in each group.

There were eight strokes in the thrombus aspiration studies (seven with aspiration vs. one with PCI alone) and nine strokes in the mechanical thrombectomy studies (seven with thrombectomy vs. two with PCI alone). Although there is insufficient power to fully make a conclusion on this outcome, the trend is worrisome. When the data are combined from these two study groups, there is a significant increase in the risk for stroke with adjunctive thrombectomy (RR = 3.01, P = 0.024). If this association is proved to be true, there are several explanations why it may exist. In the AiMi trial, a significant proportion of the patients had received rescue PCI, which is associated with an increased stroke risk.48 Catheter thrombus aspiration devices can entrain air, and it is also possible that disruption of thrombus may result in proximal as well as distal embolization.

We also explored the association between adjunctive devices and markers of myocardial perfusion: TIMI blush of 3 and complete ST-segment resolution. Among all the studies, there was evidence for significant heterogeneity and publication bias, therefore, we caution against placing too much weight on these outcomes. These surrogate markers were variably defined and reported in the individual trials. The adjudication of these markers is also subjective, such as TIMI blush, which is a visual assessment of myocardial reperfusion. These characteristics probably affect the precision, reproducibility, and validity of these outcomes.

A strength of this analysis is that it studied homogeneous patient populations by grouping together similar type adjunctive devices. Within similar type thrombectomy or embolic protection devices, there was no evidence for heterogeneity (except for moderate heterogeneity with the assessment of MACEs among the mechanical thrombectomy studies). This was important since the analysis of all available studies failed to reveal statistical heterogeneity, even though we felt that there was clinical heterogeneity. For example, the important technical differences between the devices include the fact that embolic protection requires optimal coronary anatomy, especially distally, to deploy a filter or occlusive balloon, while the larger calibre mechanical thrombectomy devices require large and non-tortuous coronary anatomy to be used successfully. If anything, this could have produced a bias against catheter aspiration devices since they were probably used in lesions that would have been prohibitive to mechanical thrombectomy or embolic protection devices.

To maximize the utilization of all available data, we included abstract presentations that have not been subjected to peer-review and scrutiny and may not be as high of quality; however, we felt that this was necessary for two reasons. Including abstracts served as an additional mechanism to evaluate for any potential publication bias (although none was formally detected by statistical testing among the clinical outcomes). The second reason is that the magnitude of treatment effect can be overestimated by analysing only the published data.49

Mechanical aspiration devices represented only a small fraction of the total study weight. Moreover, the mechanical aspiration results were strongly influenced by a single study, the AiMi trial. Therefore, it remains possible that such devices may have a beneficial role in the future. We look forward to update this analysis with the inclusion of future studies examining these types of devices.50 Also needed are studies that specifically address the role of adjunctive devices in saphenous vein grafts during acute myocardial infarction. Our findings do not give any insight into how one adjunctive thrombectomy device performs against another and it should be noted that this study did not analyse patient level data. Lastly, not all study outcomes could be fully analysed due to sparse reporting and inconsistent measurement in variables such as left ventricular function, infarct size, and heart failure. There were only four studies that reported follow up data on left ventricular ejection fraction, which was 55% in the adjunctive device group vs. 54% in the PCI alone group. For rehospitalization due to heart failure, there were data from five studies. This outcome occurred in 8.1 vs. 7.3%, respectively, for the adjunctive device group vs. PCI alone (P = 0.39).

In summary, not all adjunctive thrombectomy devices are similar. Specifically, catheter thrombus aspiration devices appear to be the most attractive group by significantly reducing mortality, compared with PCI alone. This may speak to the relatively unobtrusive, yet effective means of removing thrombus by catheter aspiration. Pending further data, the routine use of mechanical thrombectomy devices should be avoided in the emergent management of ST-elevation myocardial infarction since they appear to increase the mortality. The potential increase in strokes with adjunctive devices also needs to be carefully monitored, and minimized with careful technique where possible. There is no obvious benefit or harm with embolic protection devices; therefore, the coronary utility of these devices remains in the revascularization of saphenous vein grafts in stable patients. TIMI blush grade and ST-segment resolution are limited in their ability to act as surrogate markers and therefore they should not be used solely in place of clinical outcomes in designing future acute myocardial infarction studies. Catheter aspiration of thrombus represents a relatively simple mechanism to improve cardiovascular outcomes including survival in ST-elevation myocardial infarction.

Conflict of Interest: A.A.B. discloses the following relationships: Honoraria from Boston Scientific and Access Closure. D.L.B. discloses the following relationships: Research Grants (directly to the institution)—Bristol Myers Squibb, Eisai, Ethicon, Heartscape, Sanofi Aventis, and The Medicines Company; Honoraria (donated to non-profits for >2 years)—Astra Zeneca, Bristol Myers Squibb, Centocor, Daiichi-Sankyo, Eisai, Eli Lilly, Glaxo Smith Kline, Millennium, Paringenix, PDL, Sanofi Aventis, Schering Plough, The Medicines Company, and tns Healthcare; Speaker’s bureau (>2 years ago)—Bristol Myers Squibb, Sanofi Aventis, and The Medicines Company; Consultant/Advisory Board (donated to non-profits for >2 years)—Astellas, Astra Zeneca, Bristol Myers Squibb, Cardax, Centocor, Cogentus, Daiichi-Sankyo, Eisai, Eli Lilly, Glaxo Smith Kline, Johnson & Johnson, McNeil, Medtronic, Millennium, Molecular Insights, Otsuka, Paringenix, PDL, Philips, Portola, Sanofi Aventis, Schering Plough, Scios, Takeda, The Medicines Company, tns Healthcare, and Vertex; Expert testimony regarding clopidogrel (>2 years ago; the compensation was donated to a non-profit organization); Cleveland Clinic Coordinating Center currently receives or has received research funding from: Abraxis, Alexion Pharma, AstraZeneca, Atherogenics, Aventis, Biosense Webster, Biosite, Boehringer Ingelheim, Boston Scientific, Bristol-Myers Squibb, Cardionet, Centocor, Converge Medical Inc., Cordis, Dr Reddy’s, Edwards Lifesciences, Esperion, GE Medical, Genentech, Gilford, GSK, Guidant, J&J, Kensey-Nash, Lilly, Medtronic, Merck, Mytogen, Novartis, Novo Nordisk, Orphan Therapeutics, P&G Pharma, Pfizer, Roche, Sankyo, Sanofi-Aventis, Schering-Plough, Scios, St Jude Medical, Takeda, TMC, VasoGenix, and Viacor.


View Abstract