Correspondence to: Mr P. W. Stather, Vascular Surgery Group, Department of Cardiovascular Sciences, University of Leicester, Leicester LE2 7LX, UK (e-mail: email@example.com)
Any possible long-term benefit from endovascular (EVAR) versus open surgical repair for abdominal aortic aneurysm (AAA) remains unproven. Long-term data from the Open Versus Endovascular Repair (OVER) trial add to the debate regarding long-term all-cause and aneurysm-related mortality. The aim of this study was to investigate 30-day and long-term mortality, reintervention, rupture and morbidity after EVAR and open repair for AAA in a systematic review.
Standard PRISMA guidelines were followed. Random-effects Mantel–Haenszel meta-analysis was performed to evaluate mortality and morbidity outcomes.
The existing published randomized trials, together with information from Medicare and SwedVasc databases, were included in a meta-analysis. This included 25 078 patients undergoing EVAR and 27 142 undergoing open repair for AAA. Patients who had EVAR had a significantly lower 30-day or in-hospital mortality rate (1·3 per cent versus 4·7 per cent for open repair; odds ratio (OR) 0·36, 95 per cent confidence interval 0·21 to 0·61; P < 0·001). By 2-year follow-up there was no difference in all-cause mortality (14·3 versus 15·2 per cent; OR 0·87, 0·72 to 1·06; P = 0·17), which was maintained after at least 4 years of follow-up (34·7 versus 33·8 per cent; OR 1·11, 0·91 to 1·35; P = 0·30). There was no significant difference in aneurysm-related mortality by 2 years or longer follow-up. A significantly higher proportion of patients undergoing EVAR required reintervention (P = 0·003) and suffered aneurysm rupture (P < 0·001).
There is no long-term survival benefit for patients who have EVAR compared with open repair for AAA. There are also significantly higher risks of reintervention and aneurysm rupture after EVAR.
Endovascular aneurysm repair (EVAR) has become established as the treatment of choice for abdominal aortic aneurysm (AAA) in many vascular centres. Several similar randomized controlled trials (RCTs) have been undertaken to determine both the short- and long-term outcomes after EVAR compared with open surgical repair. Although these trials are not yet complete, with true long-term 10-year outcomes awaited, they have highlighted a significant reduction in 30-day operative mortality and length of hospital stay in favour of EVAR[1, 2].
Medium- and long-term follow-up from these trials varies, with some showing that the early survival benefit of EVAR may be lost over time. In addition, data from registries including EUROSTAR (EUROpean collaborators on Stent-graft Techniques for abdominal aortic Aneurysm Repair) have indicated the need for close surveillance of endografts, as complications arise in 25–40 per cent of patients. These often require additional intervention including conversion to open surgery. A number of concerns have arisen regarding the long-term durability of EVAR and this, in addition to added cost for lifelong surveillance after EVAR, may negate any early survival advantage. Currently, limited data are available on long-term reintervention and mortality rates after both EVAR and open repair.
The aim of this systematic review and meta-analysis was to investigate short-, medium- and long-term outcomes, including morbidity, reintervention and mortality, associated with EVAR and open aneurysm repair.
Standard reporting guidelines set by the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) Group were followed to identify RCTs and other high-level evidence reporting comparative outcomes of endovascular and open repair of AAA. Study titles and abstracts were searched using MEDLINE, Embase, and Health and Psychosocial Instruments databases, using Ovid Online (version: OvidSP_UI03.04.02.112) in July 2012, with the most recent search carried out on 31 December 2012, separately by two researchers. No language restrictions or filters used to restrict study designs were applied. Reference lists were searched for further studies to be included.
Eligibility criteria and study selection
A comprehensive literature search was performed using the search terms ‘endovascular AND open AND repair AND abdominal AND aortic AND aneurysm’ AND ‘randomised OR randomized’ AND ‘trial’ NOT ‘thoracic’. Two reviewers individually reviewed potential studies according to a set of eligibility criteria, with discussion of discrepancies. Inclusion criteria were that the study must have compared EVAR with open surgery; be an original publication; be an RCT or validated age–sex-matched non-randomized cohort study; contain more than 200 patients in the case of RCTs or greater than 2000 patients in the case of cohort studies; and report on 30-day and longer-term mortality. The following articles were excluded: review articles, studies in which duplicated data were published and those pertaining to thoracic aortic aneurysms.
Data were extracted independently by two reviewers, with discussion of any discrepancies. The outcome measures were in accordance with the reporting standards for EVAR. The following outcomes were recorded: death within 30 days of surgery; postoperative complications including myocardial infarction, stroke and renal failure; 2-year all-cause mortality, 4-year or greater all-cause mortality; 2-year or greater aneurysm-related mortality; aortic rupture; and secondary intervention, Reported percentages were converted into raw numbers for the purpose of meta-analysis. One study published only graphical representation of an outcome (with no raw data). The lead author of this study was contacted, but the raw data were not released; therefore, the graph was extrapolated to estimate the numbers.
The data were analysed using Review Manager 5·1. Separate analyses were performed for each individual outcome, with inclusion of all papers that had published results on the outcome under analysis. Sensitivity analysis was carried out by excluding studies on the basis of design. Meta-analysis was undertaken using the Mantel–Haenszel method, with a standard continuity correction of 0·5. A random-effects model was used, owing to the variability in baseline characteristics in each paper. An α level of ≤ 0·05 was used to determine statistical significance. The heterogeneity of treatment effect among trials was assessed using the I2 statistic. This describes the percentage of total variation across studies that is due to heterogeneity rather than chance or random error. A value greater than 50 per cent reflects significant heterogeneity owing to real differences in study populations, protocols, interventions and outcomes. Outcomes are reported as odds ratios (ORs) with 95 per cent confidence intervals (c.i.).
The quality of randomized studies was assessed using the Jadad score. The Newcastle–Ottawa Scale was used for non-randomized studies, with study quality assessed by examining patient selection methods, comparability of groups and assessment of outcome. Studies achieving at least seven of nine stars were considered to be of high quality. The potential for publication and reporting bias was assessed by means of funnel plots.
A total of 906 abstracts were identified through the literature search, with no further studies identified by manual searching of reference lists from these articles. Following removal of duplicates, the titles and abstracts of 709 papers were reviewed. A total of 29 papers were read in full. Eighteen studies were excluded for the following reasons: two were meta-analyses[15, 16], two were RCTs with fewer than 200 patients[17, 18], six non-randomized studies had fewer than 2000 patients[19-24], one study investigated challenging aortic necks, one investigated aortoiliac aneurysms, two investigated high-risk patients[27, 28], one was a prospective non-randomized trial with fewer than 2000 patients, one was a systematic review, one included only patients aged more than 80 years, and one excluded patients aged over 60 years (Fig. 1). Eleven studies met the eligibility criteria and were included in further analyses: nine papers that reported on the outcomes of four independent trials at different time points[1, 2, 33-39], one reporting an analysis of the US Medicare database and one representing the Swedish National Registry for Vascular Surgery (SwedVasc) database (Table1).
Table 1. Studies included in meta-analysis of endovascular and open repair of abdominal aortic aneurysm
Year of publication
No. of subjects
Maximum length of follow-up (years)
Mean length of follow-up (years)
Study quality score
The Jadad score (maximum 5) was used to assess the quality of randomized controlled trials, and the Newcastle–Ottawa Scale (maximum 9) for non-randomized studies. EVAR, endovascular aneurysm repair; ACE, Anévrysme de l'aorte abdominale, Chirurgie versus Endoprothèse; DREAM, Dutch Randomized Endovascular Aneurysm Management; OVER, Open Versus Endovascular Repair; SwedVasc, Swedish National Registry for Vascular Surgery.
The French Anévrysme de l'aorte abdominale, Chirurgie versus Endoprothèse (ACE) trial reported at a single time point; the Open Versus Endovascular Repair (OVER) Veterans Affairs Cooperative Study produced two papers looking at short- and long-term outcomes[38, 39]; and both the UK EndoVascular Aneurysm Repair (EVAR1) trial[1, 36, 37] and the Dutch Randomized Endovascular Aneurysm Management (DREAM) trial[2, 34, 35] each produced three papers, reporting outcomes at 30 days, a median follow-up of 2 years and then 6 years. A total of 1393 patients undergoing EVAR and 1390 undergoing open surgical repair were reported in the RCTs. The non-randomized studies included age- and sex-matched cohorts of 23 685 patients who had EVAR and 25 752 who had open repair.
Study quality and publication bias
Study quality, assessed using the Jadad score for RCTs and the Newcastle–Ottawa Scale for non-randomized studies, is shown in Table1. Funnel plots were performed for all outcomes, and suggested minimal publication bias (Fig. S1, supporting information).
The background analysis was similar in all studies (Table S1, supporting information). Each study included a well matched cohort of patients. However, DREAM 2010 had a higher prevalence of pulmonary disease in its EVAR group, and the OVER trial had greater use of aspirin before operation in its open surgery group. The stent-graft devices used are summarized in Table S2 (supporting information).
Patients undergoing EVAR had a significantly lower 30-day or in-hospital mortality rate than those undergoing open AAA repair: 1·3 versus 4·7 per cent respectively (OR 0·36, 95 per cent c.i. 0·21 to 0·61; P < 0·001) (Fig. 2). At 2-year follow-up there was no difference in overall mortality between EVAR or open repair: 14·3 versus 15·2 per cent (OR 0·87, 0·72 to 1·06; P = 0·17). At 4 years or more there was no difference in mortality between patients who had EVAR or open repair: 34·7 versus 33·8 per cent (OR 1·11, 0·91 to 1·35; P = 0·30). After 2 years or more of follow-up, there was no significant difference in AAA-related mortality between EVAR and open repair: 3·9 versus 4·5 per cent (OR 0·81, 0·42 to 1·58; P = 0·54) (Fig. 3).
Complication data were extracted from the latest published report from each study. A significantly higher proportion of patients who had EVAR required reintervention for graft-related complications (P = 0·003) or had suffered aneurysm rupture (P < 0·001) compared with patients who had open AAA repair (Fig. 4, Table2). A higher proportion of patients who had open repair suffered perioperative myocardial infarction (P < 0·001). There was no significant difference in the rate of renal failure or stroke.
Table 2. Comparison of complication rates following endovascular or open abdominal aortic aneurysm repair, including and excluding non-randomized trials
Open repair (%)
Open repair (%)
Values in parentheses are 95 per cent confidence intervals. RCT, randomized controlled trial; EVAR, endovascular aneurysm repair.
0·73 (0·68, 0·78)
0·82 (0·41, 1·61)
1·05 (0·29, 3·76)
1·80 (0·45, 7·11)
1·28 (0·76, 2·16)
1·28 (0·76, 2·16)
2·08 (1·27, 3·39)
2·46 (1·51, 4·01)
5·94 (2·33, 15·14)
7·20 (1·00, 52·07)
Sensitivity analysis was performed by analysing the RCTs and the non-randomized data separately. In the RCTs, patients undergoing EVAR had a significantly lower 30-day mortality rate than patients having open repair: 2·0 versus 5·9 per cent respectively (OR 0·33, 0·18 to 0·59; P < 0·001); this tended towards significance in the non-randomized data: 1·2 versus 4·6 per cent (OR 0·39, 0·14 to 1·09; P = 0·07) (Fig. S2, supporting information). In the RCTs, at 2-year follow-up there was no difference in mortality between patients undergoing EVAR or open repair: 6·1 versus 8·4 per cent (OR 0·74, 0·50 to 1·10; P = 0·13); the non-randomized data yielded a similar result: 14·8 versus 15·5 per cent (OR 0·94, 0·79 to 1·11; P = 0·46). In the RCTs, there was no significant difference after at least 4 years between EVAR and open repair: 34·5 versus 34·7 per cent (OR 0·99, 0·85 to 1·17; P = 0·94); again, the non-randomized data reflected this finding: 34·7 versus 33·7 per cent (OR 1·23, 0·81 to 1·86; P = 0·33).
Sensitivity analysis for aneurysm rupture revealed that the significant increase in late aortic rupture remained when the RCTs were analysed independently (P = 0·05) (Table2; Fig. S3, supporting information). However, the rate of myocardial infarction was no longer significantly different. In the RCTs, there was no significant difference in the rate of renal failure between EVAR and open repair.
This meta-analysis compares outcomes following EVAR and open repair, and includes the long-term follow-up from the OVER trial, and updated 30-day and 2-year mortality data. It demonstrates that EVAR confers a clear survival advantage over open AAA repair with respect to 30-day mortality (OR 0·33; P < 0·001); however, this survival benefit does not persist beyond the early postoperative period.
The findings from this study are in keeping with previous meta-analyses comparing outcomes among patients undergoing EVAR and open AAA repair. The largest previous meta-analysis, by Lovegrove and colleagues, reported improved 30-day mortality outcomes after EVAR; more recently, Dangas and co-workers published a meta-analysis of all six RCTs comparing EVAR and open repair, drawing similar conclusions. The present study also includes data from non-randomized studies, increasing the generalizability of these results.
Thirty-day mortality had different definitions between studies, with DREAM and OVER publishing a combination of 30-day and in-hospital mortality[34, 38], ACE publishing 30-day mortality only, and EVAR1 publishing both separately. Dangas et al. used 30-day mortality data from the initial publications, rather than updated data from the latest EVAR and OVER trials, and did not include in-hospital mortality from the EVAR1 trial. Inclusion of the latest data and in-hospital mortality refined the OR to 0·36 (0·33 for RCTs only, 0·39 for non-randomized data only). Although the present study found no difference in the rate of postoperative stroke or renal failure between EVAR or open AAA repair, there was a significantly lower rate of perioperative myocardial infarction among those who had EVAR (OR 0·73; P < 0·001), which may explain the improved 30-day mortality in this group.
EVAR and open repair were equivalent with respect to all-cause mortality at both 2-year and 4-year or greater follow-up, and late AAA-related mortality. Lovegrove and colleagues found no difference in long-term all-cause mortality, but they did report that aneurysm-related mortality was significantly lower in the EVAR group at 6-year follow-up. Dangas and co-workers found no significant difference in all-cause mortality after 2 years, and a loss of the initial improvement in aneurysm-related mortality in the EVAR group by 4 years of follow-up.
The findings of the present study are similar, whether or not randomized patients only were included in the analysis. Exclusion of non-randomized data left a meta-analysis of only large RCTs (more than 200 patients/study). Two small RCTs were excluded[17, 18]. The inclusion of data from the Medicare and SwedVasc databases, two of the largest validated healthcare outcomes databases in the world, significantly increases the number of patients included in the analysis, but also examines whether the outcomes reported in the major RCTs reflect those seen in the registries.
The results of this study are another challenge to the durability of EVAR. The study reports a significantly higher rate of reintervention (OR 2·08; P = 0·003), as reported previously. The most recent OVER report, however, identified equivalent reintervention rates after EVAR and open repair, which may reflect a lack of reported complications after open repair in previous studies. In the EVAR1 trial, for example, the rate of incisional hernia repairs was not reported after open AAA surgery. Of more concern, the present meta-analysis reports a significantly higher rate of aortic rupture after EVAR (OR 5·94; P < 0·001), which remains significant even when the RCTs are analysed alone. Both reinterventions and rupture may be contributing to the erosion of survival benefit beyond 30 days, as graft rupture is associated with a high death rate. It is only by the inclusion of the long-term OVER data and the two possible (but not confirmed) EVAR ruptures reported in DREAM that the rate of aortic rupture remains significantly increased after EVAR. It should, however, be noted that the OVER trial used a significant proportion of Medtronic AneurX® (Medtronic, Minneapolis, Minnesota, USA) devices (19·8 per cent), which were shown to be associated with a worse survival rate; these had been used in six of ten patients who an aneurysm-related death, and two of three who experienced a non-fatal rupture.
The rate of rupture after EVAR was higher in the EVAR1 trial (4·0 per cent) than in the other RCTs and the non-randomized studies (1·4 and 1·9 per cent respectively). This raises questions about the level of surgeon experience as well as the types of endograft device used in this study. EVAR1 required all contributing centres to have performed 20 EVARs before entering the trial; only five previous EVARs were needed for DREAM, although experienced proctor assistance was required if fewer than 20 had been performed. OVER required experience of 12 EVARs before inclusion, and in the ACE trial each centre had previously carried out 30 EVARs and continued to perform at least eight each year. Therefore, although eligibility criteria were different in each study, EVAR1 required a respectable experience, suggesting that surgical experience does not account for the increased rate of aortic rupture.
Regarding the endograft devices, all studies used a high percentage of Cook Zenith® (Cook Medical, Bloomington, Indiana, USA) and Gore Excluder® (W. L. Gore, Flagstaff, Arizona, USA) devices; DREAM, EVAR1 and ACE also used the Medtronic Talent™ device, and OVER used a significantly higher proportion of Medtronic AneurX® devices (Table S2, supporting information). The only significant variation in devices used was in the OVER trial, which reported a lower incidence of aortic rupture than EVAR1 despite use of the AneurX® device. In spite of the higher rate of rupture in EVAR1, exclusion of this paper from the overall analysis results in minimal change in the odds of AAA rupture (OR 5·94 with EVAR1, 5·36 without EVAR1). When the RCTs are analysed alone, however, exclusion of the EVAR1 trial significantly decreases the OR from 7·20 to 3·37 and results in a non-significant difference in rupture risk between EVAR and open AAA repair (P = 0·16). The reason for the higher rate of aortic rupture in the EVAR1 trial therefore remains unknown.
The main limitation of the present study was that the information had to be extracted from the published reports of each trial, without access to the raw data. As a consequence, the analysis was confined to the time points specified in each study, which were not always directly comparable. Another limitation was that recruitment into these trials was completed several years ago. Significant improvements have occurred in both operator experience and stent-graft design, allowing EVAR to be achieved in more complex aortic anatomy and in patients previously deemed inoperable. There have also been improvements in both monitoring and patient optimization before surgical intervention, thus decreasing length of hospital stay, and improving outcomes for patients undergoing both open and endovascular intervention. Long-term outcomes for these cohorts of patients will not be available for several years, and a 10-year comparison will be essential.
There remains debate about which method of AAA repair is optimal in the long term. With cost-effectiveness a major driving factor in current healthcare, the additional cost of EVAR in the short term, and the cost of monitoring and reintervention, may see a nudge towards open repair in suitable patients.
P.W.S. is funded by a Royal College of Surgeons/Dunhill Medical Trust Research Fellowship, N.D. by a fellowship from the Circulation Foundation, and M.J.B. by a Higher Education Funding Council for England Clinical Senior Lecturer Fellowship. E.C. is a Cook Endovascular Fellow (British Society of Endovascular Therapy).
Disclosure: The authors declare no conflict of interest.