OUP user menu

★ fast trackeditor's choice ★

Clinical impact of thrombectomy in acute ST-elevation myocardial infarction: an individual patient-data pooled analysis of 11 trials

Francesco Burzotta, Maria De Vita, Youlan L. Gu, Takaaki Isshiki, Thierry Lefèvre, Anne Kaltoft, Dariusz Dudek, Gennaro Sardella, Pedro Silva Orrego, David Antoniucci, Leonardo De Luca, Giuseppe G.L. Biondi-Zoccai, Filippo Crea, Felix Zijlstra
DOI: http://dx.doi.org/10.1093/eurheartj/ehp348 2193-2203 First published online: 2 September 2009


Aims Thrombectomy in patients with ST-elevation myocardial infarction (STEMI) undergoing percutaneous coronary intervention (PCI) is associated to better myocardial reperfusion. However, no single trial was adequately powered to asses the impact of thrombectomy on long-term clinical outcome and to identify patients at higher benefit. Thus, we sought to assess these issues in a collaborative individual patient-data pooled analysis of randomized studies (study acronym: ATTEMPT, number of registration: NCT00766740).

Methods and results Individual data of 2686 patients enrolled in 11 trials entered the pooled analysis. Primary endpoint of the study was all-cause mortality. Major adverse cardiac events (MACE) were considered as the occurrence of all-cause death and/or target lesion/vessel revascularization and/or myocardial infarction (MI). Subgroups analysis was planned according to type of thrombectomy device (manual or non-manual), diabetic status, IIb/IIIa-inhibitor therapy, ischaemic time, infarct-related artery, pre-PCI TIMI flow. Clinical follow-up was available in 2674 (99.6%) patients at a median of 365 days. Kaplan–Meier analysis showed that allocation to thrombectomy was associated with significantly lower all-cause mortality (P = 0.049). Thrombectomy was also associated with significantly reduced MACE (P = 0.011) and death + MI rate during the follow-up (P = 0.015). Subgroups analysis showed that thrombectomy is associated to improved survival in patients treated with IIb/IIIa-inhibitors (P = 0.045) and that the survival benefit is confined to patients treated in manual thrombectomy trials (P = 0.011).

Conclusion The present large pooled analysis of randomized trials suggests that thrombectomy (in particular manual thrombectomy) significantly improves the clinical outcome in patients with STEMI undergoing mechanical reperfusion and that its effect may be additional to that of IIb/IIIa-inhibitors.

  • ST-elevation myocardial infarction
  • Thrombectomy
  • Primary PCI
  • Long-term clinical outcome
See page 2180 for the commentary on this article (doi:10.1093/eurheartj/ehp349)


Primary percutaneous coronary intervention (PCI) has been shown to provide mortality benefit compared with thrombolysis, mainly because of better and sustained optimal coronary perfusion.1 However, despite epicardial recanalization with TIMI 3 flow, myocardial reperfusion is not achieved in a relevant percentage of ST-elevation myocardial infarction (STEMI) patients, with a significant impact on long-term survival.24 Accordingly, a series of adjunctive devices with different design and mechanism of action have been developed and tested in clinical studies with conflicting results.5 A meta-analysis of prospective randomized trials6 suggested that the use of thrombectomy devices, but not of distal protection devices, is associated with a significant reduction of angiographically evident distal embolization and no-reflow (as assessed by post-procedural myocardial blush grade (MBG) and ST-segment resolution). As angiographic and electrocardiographic markers of myocardial reperfusion are well-known predictors of late clinical events,3,710 the use of thrombectomy may also translate into improved clinical outcome. Unfortunately, no published study was specifically designed and adequately powered to asses long-term clinical outcome. Yet, recent data from a large single-centre trial showed an advantage of thrombus-aspiration use in terms of mortality at 1 year follow-up11 and a recent meta-analysis of nine randomized trials showed an advantage of thrombus-aspiration in terms of early (up to 30 days) mortality.12 As such promising observations deserve further evaluations,13 we have designed and performed a pooled analysis of the individual patient data of prospective randomized trials comparing standard PCI with or without thrombectomy to evaluate the impact of thrombectomy use on clinical outcome.


The study protocol has been registered in the clinicaltrials.gov website (number of registration: NCT00766740), has been drafted and submitted for publication before the first analysis was undertaken and has been published in details in a study design manuscript.14

A systematic MEDLINE database search (www.ncbi.nlm.nih.gov) for studies comparing PCI with thrombectomy with standard PCI was conducted according to a modified Robinson and Dickersin strategy.15 Keywords were ‘STEMI’, ‘randomized’, ‘thrombus aspiration’, and ‘thrombectomy’. Furthermore, the TCT (http://www.tctmd.com), EuroPCR (www.europcr.com), ACC (www.acc.org), AHA (http://www.americaheart.org), and ESC (www.escardio.org) websites were searched for pertinent abstracts and expert slides presentations between October 2003 and February 2008. No language restriction was applied.

Inclusion criteria for selected studies were: (i) comparison of thrombectomy with standard PCI in patients with STEMI; (ii) randomized treatment allocation. Exclusion criteria were: (i) equivocal treatment allocation processes.

Fifteen studies were published as full papers1630 and two additional studies3132 were reported as abstract31 or expert slides presentation.32

The 15 principal investigators of these 17 identified studies were contacted by mail or by fax to participate into the ATTEMPT study (pooled Analysis of Trials on ThrombEctomy in acute Myocardial infarction based on individual PatienT data). The study flowchart is depicted in Figure 1. Each principal investigator who agreed to participate into the study was asked to complete a structured patient-level database including a series of key baseline clinical and angiographic data as well as the longest available clinical outcome data of each patient previously enrolled in the corresponding trial. The modality of follow-up update was not defined in the study protocol.

Figure 1

Flow-chart of the systematic search process (RCT, randomized controlled trials; PI, principal investigator).

The requested dataset included: sex (male or female), age (years), diabetes (yes or no), rescue PCI (yes or no), usage of IIb/IIIa-inhibitors (yes or no), infarct-related artery [left anterior descending artery (LAD), left circumflex artery (LCX), or right coronary artery (RCA)], multivessel disease (yes or no), baseline TIMI flow 0 or 1 (yes or no), time from symptoms to balloon/cath-lab (minutes), thrombectomy device used (name of the device), device efficacy (device able to reach and treat the culprit lesion), and long-term clinical follow-up [death for any cause, time to death, myocardial infarction (MI), time to MI, target lesion/vessel revascularization (TLR/TVR), time to TLR/TVR]. Such individual patient data have been sent to the study coordinator (M.D.V.) who was responsible for data consistency checking (matching against previous publication of the trials as well as coherence controls) and for final pooling in a single database. A statistical expert in the field of meta-analysis (G.B.Z.) was responsible for statistical analyses.

Sample size calculation

A meta-analysis of randomized trials comparing thrombectomy with standard PCI in patients with STEMI6 showed a rate of post-procedural MBG 3 significantly higher in the thrombectomy group with an OR estimate of 2.3. Moreover, van't Hof et al. demonstrated that post-PCI MBG is a strong predictor of long-term mortality in patients with STEMI treated with primary PCI. In particular, they reported a long-term total mortality rate of 3% in patients with post-PCI MBG 3 and of 29% in patients with post-PCI MBG<3.3 Considering these results, we anticipated a sample size of 1350 patients (675 for each randomization arm) to demonstrate, with an alpha risk of 5% and a beta risk of 20%, a survival advantage at 1 year using thrombectomy compared with standard PCI.

Primary endpoint

The primary endpoint of the study was the comparison of all-cause mortality between patients randomized to thrombectomy or standard PCI.

Secondary endpoints

Secondary endpoints of the study were survival free from MI, TLR or TVR, major adverse coronary events (MACE: death + MI + TLR/TVR) and death + MI between patients randomized to thrombectomy or standard PCI.

Pre-defined subgroup analyses

The comparison of all-cause mortality between patients randomized to thrombectomy and standard PCI was performed according to the following pre-defined subgroups:14

  1. treated by manual thrombectomy devices or non-manual thrombectomy devices

  2. with or without diabetes mellitus

  3. undergoing primary PCI or rescue PCI (after failed thrombolysis)

  4. treated or not treated with IIb/IIIa-inhibitors

  5. different ischaemic time (pre-defined time intervals: ≤3, >3 ≤ 6, >6 h)

  6. infarct-related artery (LAD, LCX, or RCA)

  7. pre-PCI TIMI flow (TIMI 0–1 or TIMI 2–3)

Statistical analysis

Continuous variables are reported as mean ± standard deviation or median (first to third quartiles) and categorical variables as n (%), unless otherwise stated. Statistical pooling has been performed with the Peto fixed effect method for patient-level analysis (according to event counts reported at the longest available follow-up) as well as with a random effect method with generic inverse variance weighting (according to risk estimates obtained with Cox proportional hazard analysis). Thus, we have been able to compute pooled odds ratios (OR) with their corresponding 95% confidence intervals. Kaplan–Meier curves have been computed for survival and event-free survival analyses, both crude and stratified by study, with statistical testing based on log-rank test. A two-tailed P-value of 0.05 was chosen as cut-off for statistical significance at hypothesis testing, whereas statistical inconsistency was appraised by means of I2, with values more than 50% identifying subsets with at least moderate heterogeneity. The internal validity and quality of the individual studies entering the pooled analysis have been performed according to the Cochrane collaboration's tool for risk of bias assessment.33 Publication bias has been appraised by means of funnel plot inspection and Egger test.34


Study population

Ten principal investigators who authored 11 of 17 eligible randomized studies agreed to participate into the study and provided the requested data for each patient enrolled in the corresponding trial.

A total of 2686 patients entered the present pooled analysis: 1347 (50.1%) subjects randomized to PCI with thrombectomy device use and 1339 (49.9%) randomized to standard PCI.

The key protocol characteristics of the 11 randomized studies are summarized in Table 1. As shown in Table 1, the updated clinical follow-up available for the present analysis was significantly extended compared with that previously published for the majority of the studies. As a result, from a total of 2686 patients clinical follow-up was available in 2674 (99.6%) patients at a median of 365 days (first to third quartiles 232–365; mean 380 ± 272 days; follow-up time interval: 6–1594 days). In particular, clinical follow-up was available in 2470 patients at 6 months, in 1896 patients at 9 months, in 1685 patients at 12 months, in 374 patients at 18 months, in 296 patients at 24 months, and in 281 patients at more than 24 months.

View this table:
Table 1

Key characteristics of the trials entered the ATTEMPT study and ATTEMPT FU length

StudyDesignThrombectomy deviceTiming of randomizationAngiographic inclusion criteriaMain exclusion criteriaNumber of patientsStenting rateComplications of thrombectomy device useLongest published clinical FU (days)FU length (days) for ATTEMPT study (median, 1st–3rd quartiles)
Antoniucci et al.18Single centreNon-manual T (Angiojet)After coronary angiographyIRA diameter ≥2.5 mmIschaemic time >12 h505098%Not reported30180 (180–180)
Previous MI
Rescue PCI
REMEDIA20Single centreManual T (Diver CE)Before coronary angiographyNoneIschaemic time >12 h4950100%None30371 (212–516)
X-AMINE ST21MulticentreNon-manual T (X-Sizer)After coronary angiographyDe novo lesionIschaemic time >12 h10110099%One case of coronary arterio-venous fistola at angio-FU180180 (180–180)
Single V disease baseline TIMI flow 0–1Previous PCI in IRA
Thrombus containing lesionRescue PCI
IRA diameter ≥2.5 mmKillip class ≥3
Absence of tortuosity or severe calcification in IRA
Noel et al.31Single centreManual T (Export)After coronary angiographyBaseline TIMI flow <3Ischaemic time >12 h2624100%Not reportedIn hospital6 (6–6)
DEAR-MI22Single centreManual T (Pronto)Before coronary angiographyNoneIschaemic time >12 h747498%Not reportedIn hospital657 (429–862)
Cardiogenic shock
Previous MI
Contraindication to Gp IIb/IIIa-inhibitors
VAMPIRE28MulticentreNon-manual T (TVAC)Before coronary angiographyAbsence of LM diseaseIschaemic time >24 h17518094%Not reported240895 (239–1192)
IRA diameter ≥2.5 and ≤5 mmRescue PCI
Cardiogenic shock
Previous CABG
LM disease
Kaltoft et al.23Single centreNon-manual T (Rescue)After coronary angiographyAbsence of LM diseaseIschaemic time >12 h10710896%None30365 (365–365)
IRA suitable for thrombectomy according to treating physicianPrevious CABG
Rescue PCI
LM disease
Cardiogenic shock
De Luca et al.24Single centreManual T (Diver CE)After coronary angiographyIdentifiable thrombus on IRAIschaemic time >12 h3838100%One case of coronary dissection180468 (384–520)
Culprit vessel LADPrevious MI
3 V disease Baseline TIMI flow 0–1Previous CABG
Severe valvular heart disease
PIHRATE32MulticentreManual T (Diver CE)After coronary angiographyBaseline TIMI flow 0–1Ischaemic time >6 h9610098%Not reported180180 (180–180)
ST-elevation >3 mm in one lead
EXPIRA30Single centreManual T (export)After coronary angiographyIRA diameter >2.5 mmIschaemic time >12 h8788100%Not reported270270 (270–270)
TS>3Previous MI
Baseline TIMI flow 0–1Previous CABG
De novo lesionRescue PCI
Absence of LM or 3 V diseaseLM/3 V disease
Cardiogenuic shock
Contraimdication to Gp IIb/IIIa-inhibitors
Severe valvular heart disease
TAPAS trial27Single centreManual T (Export)Before coronary angiographyNoneIschaemic time >12 h53653592%None365365 (365–365)
Rescue PCI
Life expectancy <6 month
ATTEMPT study population13391347365 (232–365)

The clinical and angiographic data collected in the pooled ATTEMPT population were similar between patients randomized to thrombectomy or standard PCI (Table 2).

View this table:
Table 2

Baseline key clinical and angiographic characteristics of the ATTEMPT study population

ATTEMPT thrombectomy (n = 1347)ATTEMPT standard PCI (n = 1339)P-value
Age (mean ± SD)63 ± 1263 ± 121.0
Sex (M), n (%)891 (66%)894 (67%)0.73
Diabetes, n (%)183 (13%)185 (14%)0.86
Failed TL, n (%)16 (1%)14 (1%)0.72
IIb/IIIa-inhibitors, n (%)907 (67%)880 (66%)0.37
Time to reperfusion (min) (mean ± SD)269 ± 184281 ± 2120.09
MVD, n (%)555 (41%)566 (42%)0.57
Baseline TIMI flow 0–1, n (%)944 (70%)968 (72%)0.20
Culprit vessel
 LAD, n (%)578 (43%)600 (45%)0.32
 LCX, n (%)181 (13%)167 (12%)0.45
 RCA, n (%)462 (34%)461 (34%)0.94
Crossover105 (7.7%)29 (2.2%)
FU length (median; 1st–3rd quartiles) (mean ± SD)365 (197–365) 384 ± 275365 (180–365) 376 ± 2690.43
  • PCI, percutaneous coronary intervention; SD, standard deviation; TL, thrombolysis; MVD, multi-vessels disease; LAD, left anterior descending artery; LCX, left circumflex artery; RCA, right coronary artery; FU, follow-up.

The assessment of internal validity and quality for each study included in the analysis is reported in Table 3.

View this table:
Table 3

Appraisal of internal validity and quality of included studies

StudyAdequate sequence generationAllocation concealment usedBlindingIncomplete outcome data addressed in the published FUIncomplete outcome data addressed in the extended FULosses to follow-up <10% in published and extended FUUniform and explicit outcome definitionsFree of selective published outcome reportingFree of selective extended outcome reportingFree of other biasOverall risk of bias
Antoniucci et al.18Yes (computer generated)Yes (sealed envelopes)Yes (angiogram, EKG, SPECT and clinical outcome assessors)NoYes (1 patient lost to FU)YesYesYesUnclearYesLow
REMEDIA20Yes (computer generated)YesYes (angiogram, EKG and clinical outcome assessors)NoYes (two patients lost to FU)YesYesYesYesYesLow
X-AMINE-ST21UnclearUnclearYes (angiogram, EKG and clinical outcome assessors)NoNoYesYesYesUnclearYesModerate
Noel et al.31UnclearUnclearUnclearNoYesYesNoUnclearUnclearUnclearHigh
DEAR-MI22UnclearUnclearYes (angiogram, and EKG assessors)NoNoYesYesYesUnclearYesModerate
VAMPIRE28UnclearUnclearYes (angiogram, EKG and clinical outcome assessors)NoYes (eight patients lost to FU)YesYesYesUnclearYesModerate
Kaltoft et al.23Yes (computer generated)Yes (external personnel)Yes (angiogram, EKG, SPECT and clinical outcome assessors)NoNoYesNoUnclearUnclearYesModerate
De Luca et al.24UnclearUnclearYes (angiogram, EKG, echo and clinical outcome assessors)NoNoYesYesYesUnclearYesLow
PIHRATE32Yes (computer generated)Yes (sealed envelopes)Yes (angiogram, EKG and clinical outcome assessors)NoYes (1 patient lost to FU)YesYesUnclearUnclearYesLow
EXPIRA30UnclearUnclearYes (angiogram, EKG and clinical outcome assessors)NoNoYesYesYesYesYesLow
TAPAS27Yes (computer generated)Yes (computerized voice–response system)Yes (angiogram, EKG and clinical outcome assessors)NoNoYesYesYesYesYesLow

Primary endpoint

Kaplan–Meier analysis at the longest available follow-up, either crude (Figure 2) or stratified by study (which provided similar results for direction and magnitude of statistical significance), showed that allocation to thrombectomy was associated with reduced all-cause mortality (log-rank P = 0.049). A similar result was provided by Peto fixed effect analysis, showing that thrombectomy was associated with significantly fewer deaths (OR = 0.71, 95% CI 0.49–1.00; P = 0.05) when compared with standard PCI. Notably, no evidence of heterogeneity (P for heterogeneity >0.10), statistical inconsistency (I2 < 50%), or small study bias (Egger test >0.05) was evident (Figure 3). The result was similar in the analysis performed excluding the two studies not yet published as full paper.

Figure 2

Kaplan–Meier curves for cumulative survival; log-rank P = 0.049.

Figure 3

Funnel plot graph for mortality.

Secondary endpoints

Kaplan–Meier analyses at the longest follow-up available, either crude or stratified by study (which provided similar results for direction and magnitude of statistical significance), showed that allocation to thrombectomy was associated with significantly fewer MACE (log-rank P = 0.011) (Figure 4) and death + MI (log-rank P = 0.015), but non-significant differences in MI (log-rank P = 0.126) or in TLR/TVR (log-rank P = 0.126). Similar results were provided by Peto fixed effect analysis, showing that thrombectomy was associated with significantly fewer MACE (OR = 0.80, 95% CI 0.65–0.98; P = 0.03) and death + MI (OR = 0.70, 95% CI 0.52–0.93; P = 0.02) when compared with standard PCI, whereas there were non-significant differences in MI (OR = 0.72, 95% CI 0.47–1.10; P = 0.13) or TLR/TVR (OR = 0.87, 95% CI 0.67–1.12; P = 0.27). No evidence of heterogeneity (all P for heterogeneity >0.10), statistical inconsistency (all I2 < 50%), or small study bias (all P at Egger test >0.05) was evident in any of the analyses.

Figure 4

Kaplan–Meier curves for MACE-free survival; log-rank P = 0.011.

Subgroups analysis

Type of thrombectomy device

The ATTEMPT study population was divided into two groups considering the type of thrombectomy device used: manual thrombectomy group (1815 patients enrolled in trials with use of Diver CE, Pronto, and Export catheters) and non-manual thrombectomy group (871 patients enrolled in trials with use of X-Sizer, Angiojet, Rescue, and TVAC devices).

In the manual thrombectomy group, Kaplan–Meier analyses at the longest follow-up available showed that allocation to thrombectomy was associated with significantly fewer deaths (log-rank P = 0.011) (Figure 5A), whereas in the non-manual thrombectomy group, the allocation to thrombectomy was associated to similar mortality compared with standard PCI (log-rank P = 0.481) (Figure 5B).

Figure 5

(A) Kaplan–Meier curves for cumulative survival in the manual thrombectomy group and corresponding control group; log-rank P = 0.011. (B) Kaplan–Meier curves for cumulative survival in the non-manual thrombectomy group and corresponding control group; log-rank P = 0.481.

Clinical and angiographic subgroups

There was no qualitative difference in mortality when splitting the study population according to the presence or absence of diabetes, to shorter, intermediate or longer time-to-reperfusion, to type of culprit artery (left anterior descending or circumflex artery or RCA) and to pre-PCI TIMI flow (0–1 or 2–3).

Conversely, subgroup analysis according to administration of IIb/IIIa-inhibitors showed that randomization to thrombectomy was associated to a survival benefit in the subgroup of patients treated with IIb/IIIa-inhibitors (1787 patients; log-rank P = 0.045; HR 0.61, 95% CI 0.38–0.90) and not in those not receiving this drugs (899 patients; log-rank P = 0.843; HR 0.93, 95% CI 0.48–1.80).


The failure to achieve myocardial tissue reperfusion is known to be the main complication limiting the early and long-term clinical benefit of mechanical reperfusion in patients with STEMI. Among the different strategies to limit the ‘no reflow phenomenon’, reducing distal embolization by thrombectomy devices use is a promising one.13 The present pooled analysis on 2686 individual patients data (from 11 randomized trials) shows that the adjunct of thrombectomy during mechanical reperfusion is associated to a detectable improvement of survival at follow-up with an estimated number needed to treat to prevent one death at 1 year of 62. Moreover, we found that the survival advantage of thrombectomy is confined to patients treated by simple manual thrombectomy catheters with an estimated number needed to treat to prevent one death at 1 year of 34.

Such findings extend the recently published single-centre TAPAS trial results11 which showed a significantly lower cardiac mortality at 12 months in the group of patients treated with manual thrombectomy. However, the TAPAS trial was not specifically designed and powered to assess the difference between thrombectomy and standard PCI at the long-term clinical follow-up.35 In the absence of large, multicentre trials adequately powered to assess long-term outcome in patients with STEMI treated with PCI and thrombectomy, the best way to evaluate the role of this treatment is pooling individual patient data collected by different investigators in different randomized studies of standard PCI vs. PCI with thrombectomy. Indeed, pooled analyses of randomized trials using individual patient data are known to provide more accurate results compared with simple meta-analyses,36 so that they have been used to answer the key questions in the field of cardiovascular medicine like the benefit of prolonged anti-platelet therapy,37 the benefit of beta-blockade after MI,38 the impact of mechanical over pharmacological reperfusion in STEMI39 and the safety of drug-eluting stents in selected patients with coronary artery disease.40 Thus, we designed and carried out this collaborative study, which allowed not only to collect a large number of patients, pooling the majority of available randomized trials, but also to significantly extend the clinical follow-up of some of the studies previously published with a short clinical follow-up. As a consequence, the number of patients with at least 1 year follow-up (i.e. 1685) was larger than the calculated sample size for a reliable assessment of the primary endpoint. Moreover, as the very late outcome of patients treated by thrombectomy was not previously investigated, the present study provides novel information on a subgroup of patients with follow-up extended beyond 1 year. In keeping with the hypothesis of a long-lasting thrombectomy-induced benefit, the survival curves of thrombectomy-treated patients continued to differ from that of control patients throughout the study period.

Large studies are also needed to assess if subgroups of patients have more to gain from the application of this novel treatment. Thus, we planned to collect data allowing a series of pre-defined clinical and angiographic subgroup analyses. Interestingly, the benefit of thrombectomy was more evident in patients who received IIb/IIIa-inhibitors thus suggesting a possible additive benefit of thrombectomy in patients treated with IIb/IIIa-inhibitors. It might be speculated that pharmacological and mechanical thrombus remodelling are synergic to obtain the best myocardial reperfusion and, consequently, the best clinical outcome. Indeed, in the ATTEMPT study, patients treated by both thrombectomy and IIb/IIIa-inhibitors had the lower mortality rate, those who had none of these treatments had the higher mortality rate, patients receiving only one of these therapies exhibiting intermediate outcome (Figure 6). Accordingly, such observations seem to support the last release of the ESC guidelines on the management of patients with STEMI which recommends the use of both IIb/IIIa-inhibitors and thrombus-aspiration to prevent no-reflow.41

Figure 6

Mortality rates observed in the ATTEMPT database according to thrombectomy and to administration of IIb/IIIa-inhibitors. Comparison between four treatment subgroups performed by Fisher test.

Finally, once the use of thrombectomy to reduce distal embolization is accepted, the issue of device selection emerges. More complex (non-manual) devices are probably more effective than manual thrombus-aspiration catheters in extracting athero-thrombotic particles from the coronary arteries. Yet, they are bulky, require longer learning curves and selected coronary anatomies. Conversely, manual thrombus-aspirating catheters are user-friendly, not associated with specific complications, and suitable for most coronary anatomies. In the absence of large studies comparing the simple manual thrombectomy catheters with the more complex non-manual thrombectomy devices, the present study does not support the routine use of non-manual thrombectomy. Possible benefits of complex non-manual thrombectomy devices might be confined to selected subgroups of patients with larger thrombus burden and device-favourable coronary anatomy and need to be tested in ad hoc prospective studies.

Study limitations

The results obtained in the present study may have been influenced by the quality of each original trial. In particular, no further quality control has been performed nor any restriction in trial size, publication status, and updated clinical follow-up length or modality of assessment have been applied. All these limitations were anticipated14 and have been accepted in the effort to achieve the goal to collect in a single database the largest number of patients randomized in thrombectomy trials. Moreover, the validated statistical analyses applied allowed to show no evidence of heterogeneity, statistical inconsistency, or small study bias in any of the analyses.

Another limitation of the present study is the exclusion of six eligible trials (accounting for 1019 patients) due to lack of agreement by the principal investigators to participate the study. As a consequence, the participation was unbalanced in favour of manual thrombectomy trials (indeed 88% of studies testing manual thrombectomy were included when compared with 44% of studies testing non-manual thrombectomy). Thus, the hypothesis that the overall positive result may have been driven by manual thrombectomy studies cannot be discarded.

The subgroup analysis generated small subgroups of patients so that the survival analyses were underpowered. Yet, as no large assessment of clinical benefit of thrombectomy according to the key angiographic and clinical parameters has previously been conducted, the present results should be considered as hypothesis-generating and deserve further evaluations.

Finally, the different endpoint definitions for MI, TVR, or TLR adopted in the trials included may have limited the strength of the secondary endpoints analysis. However, the presence of such possible heterogeneity cannot have influenced the primary endpoint which was the unequivocal all-cause mortality.


The pooled analysis of patients with STEMI enrolled in 11 randomized trials showed that the adjunct of thrombectomy (in particular manual thrombectomy) during mechanical reperfusion improves late clinical outcome and that this benefit is evident in patients receiving IIb/IIIa-inhibitors.

Conflict of interest: none declared.


  • The first two authors contributed equally to the study.


View Abstract