Abstract
Background The optimal pulmonary revascularisation strategy in high-risk pulmonary embolism (PE) requiring implantation of extracorporeal membrane oxygenation (ECMO) remains controversial.
Methods We conducted a systematic review and meta-analysis of evidence comparing mechanical embolectomy and other strategies, including systemic thrombolysis, catheter-directed thrombolysis or ECMO as stand-alone therapy, with regard to mortality and bleeding outcomes.
Results We identified 835 studies, 17 of which were included, comprising 327 PE patients. Overall, 32.4% were treated with mechanical pulmonary reperfusion (of whom 85.9% had surgical embolectomy), while 67.6% received other strategies. The mortality rate was 22.6% in the mechanical reperfusion group and 42.8% in the “other strategies” group. The pooled odds ratio for mortality with mechanical reperfusion was 0.439 (95% CI 0.237–0.816) (p=0.009; I2=35.2%) versus other reperfusion strategies and 0.368 (95% CI 0.185–0.733) (p=0.004; I2=32.9%) for surgical embolectomy versus thrombolysis. The rate of bleeding in patients under ECMO was 22.2% in the mechanical reperfusion group and 19.1% in the “other strategies” group (OR 1.27, 95% CI 0.54–2.96; I2=7.7%). The meta-regression model did not identify any relationship between the covariates “more than one pulmonary reperfusion therapy”, “ECMO implantation before pulmonary reperfusion therapy”, “clinical presentation of PE” or “cancer-associated PE” and the associated outcomes.
Conclusions The results of the present meta-analysis and meta-regression suggest that mechanical reperfusion, notably by surgical embolectomy, may yield favourable results regardless of the timing of ECMO implantation in the reperfusion timeline, independent of thrombolysis administration or cardiac arrest presentation.
Abstract
Mechanical reperfusion, notably by surgical embolectomy, yields favourable results regardless of the timing of ECMO implantation in the reperfusion timeline, independent of thrombolysis administration or cardiac arrest presentation. https://bit.ly/37uQvWq
Introduction
High-risk pulmonary embolism (PE) refers to a large embolic burden causing right ventricular failure and haemodynamic instability. It accounts for ∼5% of all PE [1], but contributes significantly to the overall mortality associated with PE, due to circulatory collapse, with in-hospital mortality rates ranging from 25% in patients with cardiogenic shock to 65% in those requiring cardiopulmonary resuscitation [2, 3]. Systemic thrombolysis is the first-line revascularisation therapy in high-risk PE. Surgical embolectomy or catheter-directed therapy is recommended in patients with an absolute contraindication to systemic thrombolysis [4, 5].
Extracorporeal membrane oxygenation (ECMO) provides respiratory and haemodynamic support as a bridge to recovery for the most critically ill PE patients with refractory cardiogenic shock or cardiac arrest [6, 7]. However, the optimal pulmonary reperfusion strategy from among mechanical reperfusion, systemic thrombolysis or catheter-directed thrombolysis (CDT), or ECMO plus heparin as a stand-alone therapy remains vigorously debated. Indeed, available data derive from case series or cohort studies with small sample sizes. The updated European Society of Cardiology (ESC) guidelines for the management of acute PE suggested mechanical pulmonary artery clot removal, either by surgery or using a catheter-based approach, albeit with a class IIb recommendation [5].
We performed a systematic review and meta-analysis of available evidence comparing mechanical reperfusion and other strategies, including systemic thrombolysis or CDT, or ECMO as stand-alone therapy, and the associated mortality and bleeding outcomes in patients with acute high-risk PE requiring mechanical haemodynamic support.
Methods
We followed the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analysis) and MOOSE (Meta-Analyses of Observational Studies in Epidemiology) guidelines for systematic reviews and meta-analyses [8, 9]. The protocol for this systematic review and meta-analysis is registered at PROSPERO with identifier number CRD42020165742.
Eligibility criteria
For the purpose of this analysis, eligible studies were those reporting reperfusion strategies and outcomes of adult patients who underwent ECMO for acute high-risk PE. Eligible studies for this analysis were required to meet all of the following inclusion criteria: 1) provide data on revascularisation strategies in patients who required ECMO for the management of acute high-risk PE; 2) include patients aged ≥18 years; 3) report data on at least in-hospital mortality; 4) be published in English as a full article; 5) include at least four patients to avoid inaccuracy in the assessment of effect estimates; and 6) be published since 2010 (to account for contemporary practices). Articles were ineligible for inclusion in this study if they: 1) did not provide data on patients’ revascularisation status or if the results on survival remained unknown despite careful analysis of the full-length manuscript, the supplementary material and no response from the authors despite repeated e-mail contact requesting additional insights; 2) reported ambiguous or inaccurate data (discrepancies between data reported in the text and tables); or 3) were case reports, which were excluded due to the likelihood of positive outcome bias.
Search strategy
A systematic literature review of PubMed and Ovid MEDLINE from January 2010 through December 2020 was conducted by two authors (R.C. and H.P.), who jointly performed a systematic literature search using the search terms “extracorporeal membrane oxygenation” OR “ECMO” OR “extracorporeal life support” OR “ECLS” AND “pulmonary embolism” OR “PE”. Further details regarding the search strategy used for each database are provided in the supplementary material. The reference lists of included studies were manually searched to identify any further related articles. The bibliographic records retrieved were downloaded, imported and de-duplicated using EndNote version X7 (Clarivate Analytics, Philadelphia, PA, USA).
Outcomes and definition
The primary outcome of this analysis was study-defined mortality. The secondary outcome was any study-defined bleeding event. The mechanical reperfusion group included surgical embolectomy and catheter-based thrombus removal. The “other strategies” group included systemic thrombolysis, CDT and ECMO as stand-alone therapy with physiological fibrinolysis and heparin. Patients who were treated with systemic thrombolysis or CDT before mechanical reperfusion (i.e. surgical embolectomy or catheter-based thrombus removal) were included in the mechanical reperfusion arm. Patients with systemic thrombolysis or systemic thrombolysis plus CDT were classed as having received thrombolysis.
Data collection
Two reviewers working independently (R.C. and H.P.) used a standardised data extraction form to extract data from eligible studies. We extracted descriptive data for the patients included in each study, study characteristics and outcome data. The quality of the included studies was assessed by the two investigators to evaluate risk of bias using the National Heart, Blood, and Lung Institute (NHBLI) criteria for study quality assessment of case–control series (www.nhlbi.nih.gov/health-topics/study-quality-assessment-tools). The entire tabulated dataset was reviewed and disagreements were resolved via consensus or by a third author (N.M.).
Statistical analysis
Meta-analysis
Meta-analyses were performed assuming a random effects model, which accounts for heterogeneity across included studies. Computations of odds ratios were done on a log scale. We computed the log odds ratio and the standard error of the log odds ratio, and used these numbers to perform all steps in the meta-analysis. Then, we converted the results back to the original metric using Comprehensive Meta-Analysis version 3.3 (Biostat, Englewood, NJ, USA). Statistical heterogeneity between studies is presented using the I2-statistic, where an I2-value ≤30%, 30–50%, 50–75% and ≥75% is considered to indicate low, moderate, substantial and considerable heterogeneity, respectively [10]. p-values for I2-statistics were derived from the χ2 distribution of the Cochran Q-test. We assessed within-study variance using the Cochran Q-statistic (i.e. τ2) [11]. We performed a one-study-omitted sensitivity analysis to investigate the robustness of our results. Studies with an event rate of 0% or 100% in both groups were not displayed in forest plots summarising the study results for mortality [12] and bleeding [13].
Meta-regression and sensitivity analyses
Meta-regression was conducted on the transformed proportion using mixed effects to explore potential heterogeneity and to investigate whether four prognostically relevant covariates (i.e. reperfusion strategy requiring “more than one pulmonary reperfusion therapy” (%), “ECMO implantation before pulmonary reperfusion therapy” (%), “clinical presentation of PE” (cardiac arrest versus refractory cardiogenic shock) (%) and “cancer-associated PE” (%)) affected the results of the main analyses. To compute the significance of covariates, we used the Knapp–Hartung adjustment to obtain robust estimates [14].
Results
Literature search and study selection
The initial literature search identified 598 records in Ovid MEDLINE and 249 in PubMed (835 studies after removal of duplicates). After first selection based on titles and abstracts, we excluded 695 studies and assessed 140 full-text articles. We identified two additional relevant studies by reviewing the references of the included studies [15, 16]. After full-text review, we included 17 studies in the systematic review and in the meta-analysis (supplementary figure S1) [12, 13, 15–29].
Table 1 provides the main characteristics of the 17 studies, comprising 327 acute PE patients who required ECMO implantation. All studies but one [23] involved data from a single centre. The mean±sd age of patients across the studies was 52.0±4.6 years and 89.3.2% were male. 24 patients (7.3%) had cancer-associated PE. The time between PE diagnosis and pulmonary reperfusion therapy was specified in only one study including six patients, where the mean±sd duration was 30.8±15.6 min [16]. A pulmonary embolism response team (PERT) was activated for management in three of the 17 studies, corresponding to 85 out of 327 patients (25.6%). ECMO was implanted for refractory shock in 140 patients (42.8%) and for cardiac arrest in 187 patients (57.2%). 124 patients (37.9%) received ECMO before reperfusion therapy.
Baseline characteristics of the populations from eligible studies (n=17) evaluating reperfusion strategies in high-risk pulmonary embolism (PE) patients with extracorporeal membrane oxygenation (ECMO)
44 patients (13.4%) received more than one form of reperfusion therapy (table 2). Additional information on the individual studies, outcomes and ECMO configuration is provided in supplementary tables S1–S3. All 17 studies were assessed with the NHBLI quality criteria; eight out of the 17 studies (47.0%) were rated as good quality (supplementary table S4).
Reperfusion strategies used across eligible studies (n=17)
Association between reperfusion strategies and clinical outcomes
In total, 106 patients (32.4%) were treated with mechanical pulmonary reperfusion, including 91 patients (85.8%) with surgical embolectomy (comprising surgical embolectomy alone (n=64) or surgical embolectomy with prior thrombolysis (n=27)) and 15 patients (14.1%) with catheter-based embolectomy alone, while 221 patients (67.6%) received other strategies (comprising systemic thrombolysis alone (n=92), systemic thrombolysis with CDT (n=9), CDT alone (n=20) or stand-alone ECMO (n=100)). Table 2 describes the reperfusion strategies used across included studies. The mortality rate was 22.6% in the mechanical reperfusion group and 42.8% in the “other strategies” group, yielding an absolute difference of 20.2% (95% CI 5.2% to 27.6%) (figure 1). The pooled OR for mortality with mechanical reperfusion was 0.439 (95% CI 0.237–0.816) (p=0.009; I2=35.2%; τ2=0.02 (se=0.62)) versus other reperfusion strategies. The rate of bleeding in patients under ECMO was 22.2% in the mechanical reperfusion group and 19.1% in the other reperfusion group (absolute difference 3.1% (95% CI −5.8% to 12.0%)) (OR 1.27 (95% CI 0.54–2.96); I2=7.7%; τ2=1.38 (se=1.81)) among 10 eligible studies with available bleeding data. One-study-omitted sensitivity analyses showed that individual study data did not influence the overall results (OR 0.46 (95% CI 0.21–0.99); p=0.05 for mortality events and OR 1.09 (95% CI 0.46–2.54); p=0.84 for bleeding events).
Pooled odds ratio for a) mortality and b) bleeding between the mechanical pulmonary reperfusion group and the “other strategies” group in patients with acute high-risk pulmonary embolism requiring extracorporeal membrane oxygenation. Weights are from random effects analysis.
Subanalyses according to specific reperfusion strategies
We performed multiple subanalyses according to specific reperfusion strategies. In particular, the mortality rate was lower in patients treated with surgical embolectomy (22.4% (95% CI 14.2–32.5%)) than in the thrombolysis group (43.0% (95% CI 36.4–49.8%)) (absolute difference 20.6% (95% CI 8.7% to 32.4%)), with a pooled OR of 0.368 (95% CI 0.185–0.733) (p=0.004; I2=32.9%; τ2=0.01 (se=0.42)) (figure 2). The OR was 0.28 (95% CI 0.10–0.78) between mechanical embolectomy and thrombolysis (absolute difference 26.8% (95% CI 11.7% to 41.8%)) and 0.48 (95% CI 0.19–1.24) between mechanical embolectomy and ECMO as stand-alone therapy (absolute difference 23.1% (95% CI 8.4% to 37.7%)) (figure 3). Bleeding rates were lower in the embolectomy group, independent of the subanalyses (figures 2 and 3). Comparison between mechanical reperfusion strategy and CDT included 31 patients with mortality events and none with bleeding events.
Pooled odds ratio for a) mortality and b) bleeding between the surgical embolectomy group and the thrombolysis group in patients with acute high-risk pulmonary embolism requiring extracorporeal membrane oxygenation. Weights are from random effects analysis.
Subanalyses according to specific reperfusion strategies.
Meta-regression
The heterogeneity of the main analyses was moderate for mortality events (I2=35.2%) and low for bleeding events (I2=0.0%). The meta-regression model did not identify an association between the covariates “more than one pulmonary reperfusion therapy”, “ECMO implantation before pulmonary reperfusion therapy”, “clinical presentation of PE” (cardiac arrest versus refractory cardiogenic shock) or “cancer-associated PE” and mortality events (p=0.95, p=0.52, p=0.81 and p=0.68, respectively) (figure 4). Regarding bleeding, there was no association with the covariates “more than one pulmonary reperfusion therapy”, “ECMO implantation before reperfusion” or “cancer-associated PE” (p=0.19, p=0.58 and p=0.84, respectively), but we observed p=0.03 for the relation with clinical presentation of high-risk PE requiring ECMO (supplementary figure S2).
Meta-regression analysis on the transformed proportion using mixed effects to describe the relationship between prognostically relevant covariates: a) “more than one pulmonary reperfusion therapy” (%), b) “ECMO implantation before pulmonary reperfusion therapy” (%), c) “clinical presentation of PE” (cardiac arrest versus refractory cardiogenic shock (%) and e) “cancer-associated PE” (%) and the logit odds ratio of mortality. Goodness of fit: τ2=2.01; I2=54.9%; Q=19.9. PE: pulmonary embolism.
Publication bias
Visual inspection of funnel plots of effect size (natural log of odds ratio) against inverse standard error showed a symmetric distribution around the effect size. Similarly, the Egger's regression asymmetry test suggests absence of significant publication bias (figure 5).
Funnel plot and corresponding Egger's regression asymmetry test to assess publication bias in a) mortality and b) bleeding outcomes for comparison between mechanical pulmonary reperfusion therapy and other strategies in patients with acute high-risk pulmonary embolism requiring extracorporeal membrane oxygenation.
Discussion
The results of this meta-analysis including 327 patients from 17 studies with moderate heterogeneity suggest that mechanical pulmonary reperfusion, predominantly surgical embolectomy, seems to be more effective than other strategies, especially thrombolysis, for mitigating the mortality rate and with an apparently similar risk of bleeding. The timing of ECMO implantation (before or after pulmonary reperfusion), the use of more than one reperfusion strategy, the clinical presentation of PE (i.e. cardiac arrest or refractory cardiogenic shock) and cancer-associated PE did not affect the observed benefit of mechanical therapies in the meta-regression model. However, these findings should be interpreted in light of the potential bias in the individual studies included and should not be construed to imply causation, but rather should be seen as hypothesis generating.
The favourable association of surgical embolectomy on mortality compared with other strategies was mainly driven by two studies published within the last 4 years. First, in 2018, we published the largest (n=52 patients) and the only multicentre (n=11 centres) cohort study to date on this topic [23]. Our results showed an overall mortality rate of 41.2% with the surgical embolectomy approach (23.5% when surgical embolectomy was the only reperfusion strategy used) versus 76.5% with systemic thrombolysis and 77.8% with ECMO as stand-alone therapy [23]. Second, in 2019, Ius et al. [22] reported a mortality rate of 5.0% in patients who underwent surgery (including 50% who received prior thrombolysis) and 69.0% in those who did not undergo surgery, among 36 patients with high-risk PE managed with ECMO.
The type of catheter-based embolectomy used in eligible studies (n=5) for the present meta-analysis is no longer approved by the US Food and Drug Administration for reperfusion in the PE setting [30]. Two devices, i.e. the FlowTriever System (Inari Medical, Irvine, CA, USA) and the Indigo Aspiration Catheter (Penumbra, Alameda, CA, USA), recently demonstrated an improvement in right ventricular function with a low rate of bleeding events in PE patients with systolic blood pressure ≥90 mmHg and right/left ventricular ratio ≥0.9 included in two prospective, single-arm studies [31, 32]. Nonetheless, their performances were not tested in ECMO patients. Theoretically, these devices could be an attractive alternative to thrombolysis and cardiopulmonary surgery (which requires a high dose of heparin during the cardiopulmonary bypass) in patients at very high bleeding risk. However, well-designed studies are warranted to investigate this treatment option.
Subanalyses by reperfusion strategy showed a lower mortality rate with mechanical pulmonary reperfusion compared with thrombolysis. The pooled odds ratio for mortality was also in favour of the mechanical strategy when compared with ECMO as stand-alone therapy, albeit without reaching statistical significance (OR 0.48 (95% CI 0.19–1.24)). Finally, data regarding the use of CDT in PE patients requiring ECMO are too sparse to draw any conclusions, with only three published studies [16, 17, 20]. George et al. [20] reported encouraging results in PE patients under ECMO, treated with CDT. The mortality rate was 25% among 16 patients treated with CDT versus 100.0% among five patients who received systemic thrombolysis, but further data are needed. Recent data reported that accelerated lower dose thrombolysis regimens for ultrasound facilitated using the EkoSonic Endovascular System (Boston Scientific, Marlborough, MA, USA) improved right ventricular function at 2 days [33–35], and resulted in sustained right ventricular recovery and improvements in functional status and quality of life over 1 year [36].
The 2019 ESC guidelines for the management of PE propose an algorithm for reperfusion management of PE patients with refractory shock or cardiac arrest who require ECMO life support [5]. The guidelines propose referring patients to surgical or catheter-based embolectomy if ECMO is already initiated, while systemic thrombolysis should be used if ECMO is not initiated. It is not clear from evidence-based clinical practice guidelines whether ECMO is recommended in patients who remain unstable after thrombolysis. In this context, the present results confirm and strengthen the ESC recommendations by providing additional evidence in favour of surgical embolectomy for PE patients under ECMO. Moreover, our meta-regression suggested that surgery remains a preferable option, regardless of whether thrombolysis has been administered or not, regardless of the timing of ECMO implantation in the reperfusion timeline and regardless of the clinical presentation at the time of ECMO implantation. Data from cancer-associated PE in PE patients requiring ECMO should be interpreted with caution. Indeed, the Extracorporeal Life Support Organization (ELSO) guidelines do not support the use of ECMO in patients with extended malignancy [6].
Based on these findings, we propose a hypothetical algorithm for the management of acute high-risk PE, which takes account of the activation of the PERT [37], international guidelines for the management of acute PE [4, 5], ELSO guidelines for the appropriate use of ECMO [6], and the Journal of the American College of Cardiology Scientific Expert Panel [7], the European Resuscitation Council and European Society of Intensive Care Medicine guidelines 2021 [38], as well as the results of our meta-analysis and meta-regressions (figure 6). Although this algorithm has not been validated in clinical trials, it represents a synthesis of evidence-based approaches to the management of high-risk PE, which may help guide clinicians until further evidence becomes available.
Proposed algorithm for the management of acute high-risk pulmonary embolism (PE). #: hypotension with systolic blood pressure (SBP) <90 mmHg for at least 15 min or requiring inotropic support, or shock with signs of tissue hypoxia (e.g. altered mental status, cold clammy skin, oliguria or elevated blood lactates) [4, 5]; ¶: refractory shock defined as 1) sustained SBP <90 mmHg, 2) evidence of end-organ hypoperfusion (e.g. altered mental status, cold, clammy skin or serum lactate ≥2.3 mmol·L−1) and 3) high-dose vasoactive drug infusion of at least two inotropes or vasopressors [7]; +: extracorporeal membrane oxygenation (ECMO) if aged <75 years, no end-stage renal or liver disease, no extend malignancy, or if cardiac arrest within the first 60 min, according to the 2021 Extracorporeal Life Support Organization guidelines [6]; §: ECMO should be used as a stand-alone therapy if absolute contraindication to surgical embolectomy, including recent neurosurgery, recent intracranial haemorrhage and other high bleeding risk conditions; ƒ: according to the 2021 European Resuscitation Council/European Society of Intensive Care Medicine guidelines [38]; ##: preferred therapeutic option if thrombus is crossing the interatrial septum through a patent foramen ovale or is in transit passing through the right atrium and ventricle [4, 5]. This algorithm has not been validated in clinical trials, but represents a synthesis of evidence-based approaches to the management of high-risk PE. PERT: pulmonary embolism response team; i.v.: intravenous; ACS: acute coronary syndrome; RV: right ventricular; CTPA: computed tomography pulmonary angiography.
This study has some limitations. First, the low number of patients included, the magnitude of odds ratio dispersion across studies, the width of the pooled confidence intervals and the potential for selection bias in the studies included call for caution in the interpretation of our results. Further larger cohort studies or prospective trials are warranted to corroborate our findings. Nevertheless, randomised trials in such a severe patient category have previously been halted as a result of low inclusion rates [39]. Second, we were able to obtain accurate data regarding reperfusion strategy allocation and mortality for all eligible studies except one, which included ≤10 ECMO patients [40]. In this claims-based dataset, the authors were unable to share details of patients, due to privacy issues related to the database used on a small number of subjects [40]. Third, differences between institutions in terms of patient selection, volume and expertise, treatment strategy, the recent introduction of PERTs [41], and availability of ECMO support in remote locations and in the pre-hospital setting to enable transfer to an experienced ECMO centre might have had a significant impact on outcomes [7, 42]. Fourth, all studies but one [17] reported early outcomes, precluding any evaluation of the durability of this salvage therapy in the medium to long term. Fifth, we chose to include patients who had previous lysis in the surgery group, which may be associated with potential reporting bias. However, it should be noted that the meta-regression investigating the effect of “more than one reperfusion therapy” did not find any significant impact on the overall result, suggesting that our choice to include previously thrombolysed patients in the surgery group did not have a major impact on the analysis. Finally, the localisation of PE (i.e. central versus distal) may have led physicians to prefer one option to the other, at least in some cases (e.g. surgical embolectomy for central PE). Nevertheless, computed tomography scan data are lacking in the present analysis to draw any conclusion or hypothesis.
Perspectives for future research
We propose that an adequately powered, multicentre, prospective registry, with independent adjudication of events, is warranted to identify potential differences in mortality between groups in patients with high-risk PE requiring ECMO. To calculate the optimal sample size, on the basis of four different arms of pulmonary reperfusion with ECMO (i.e. surgical embolectomy, systemic thrombolysis, CDT and ECMO alone), we performed a post hoc analysis of the primary outcome (death) from the present study and calculated the effect sizes. Using these effect sizes, we estimated that a minimum of 198 PE patients with ECMO would be needed to demonstrate a statistically significant difference in mortality between pulmonary reperfusion groups, after multivariable adjustment for potential confounders.
Conclusions
The results of this meta-analysis suggest that mechanical pulmonary reperfusion, especially surgical embolectomy, may be an effective strategy for mitigating mortality rates, at a similar risk of bleeding compared with other reperfusion strategies. The timing of ECMO implantation (before or after pulmonary reperfusion), the use of more than one reperfusion strategy and the clinical presentation of high-risk PE did not appear to impair the benefit of mechanical reperfusion therapies. These findings should be interpreted in light of the potential bias in the individual studies included and should not be construed to imply causation, but rather should be seen as hypothesis generating. Accordingly, additional studies are warranted to confirm our results and to better define management of severe PE patients at very high bleeding risk.
Supplementary material
Supplementary Material
Please note: supplementary material is not edited by the Editorial Office, and is uploaded as it has been supplied by the author.
Supplementary material ERJ-02977-2021.Supplement
Shareable PDF
Supplementary Material
This one-page PDF can be shared freely online.
Shareable PDF ERJ-02977-2021.Shareable
Footnotes
This study is registered at PROSPERO with identifier number CRD42020165742.
Conflict of interest: G. Piazza has received research grant support from Bristol Myers Squibb/Pfizer Alliance, Janssen, Boston Scientific Corporation, Amgen and Bayer, and consulting fees from Amgen, Agile, Pfizer, and the Prairie Education and Research Cooperative. No other author has any conflict of interest to declare.
- Received November 19, 2021.
- Accepted April 11, 2022.
- Copyright ©The authors 2022. For reproduction rights and permissions contact permissions{at}ersnet.org
References
