A systematic literature search of Western articles on EVAR and OSR of infrarenal AAA was carried out by two independent researchers (Y. H and J. Z) using the MEDLINE (PubMed), journals@Ovid Full Text, and BIOSIS previews databases (OVID). The following Medical Subject Headings (MeSH) search terms were used: aortic aneurysm, abdominal, vascular surgical procedures, stents, and randomized controlled trial. In addition, exploding keywords included endovascular, comparative, and mortality. The corresponding search strategy is provided (Supplementary files 1). We limited our search to reports on human subjects in the English language. We searched databases for articles published between November 1991 (because (Parodi et al., 1991) published their results on the first clinical application of endovascular repair for abdominal aortic aneurysm in 1991) and December 2018. Every relevant article retrieved had its references thoroughly checked. Following this the articles were entered into the ISI WEB OF KNOWLEDGE: WEB OF SCIENCE database (1991 to 2018) to find the references cited in order to find any related literature. Relevant medical journals were also thoroughly checked.

We performed a comprehensive search of the Chinese literature. The key words abdominal aortic aneurysm and endovascular repair were used in a comprehensive search of the CBMdisc (Chinese Biomedical Database) and CNKI (Chinese National Knowledge Infrastructure) databases. All Chinese studies on abdominal aortic aneurysm treatment between January 1998 (because in 1998, Jing and colleagues (Jing et al., 1998) published their initial article of endovascular repair AAA in China) and 24th December 2018 were identified. Reference lists were also examined and relevant articles added to the list. Furthermore, several Chinese journals were hand searched from 1998 to 2015, specifically the Chinese Journal of Practical Surgery, Chinese Journal of General Surgery, Chinese Medical Journal, and Chinese Journal of Vascular Surgery.

Inclusion criteria: Randomized controlled, prospective, and retrospective, studies were included if they met the following criteria:

(a): patients diagnosed with an infrarenal and non-ruptured abdominal aortic aneurysm undergoing elective endovascular repair were compared with patients undergoing elective open surgical repair;

(b): at least 20 patients;

(c): sufficient data (> 25 percent of predefined variables) such as patient baseline characteristics and short- and long-term outcomes;

(d): in the case of multiple reports from the same institution, only the most rigorous published report was used to avoid duplication of data;

(e): however, when two or more studies from the same institution reported on medium- or long-term outcomes on the same patients, the results from all linked studies were included synergistically.

Excluded criteria: articles were excluded if they:

(a): did not report outcomes in a comparable fashion;

(b): included patients with other aortic pathologies such as ruptured abdominal aortic aneurysms, thoracic aortic aneurysms and aortic dissections;

(c): contained insufficient data (< 25 percent of predefined variables);

(d): did not meet the criteria for abstracts and unpublished data;

(e): were confined to select subgroups of patients (e.g. octogenarians).

Complications were classified and graded according to the reporting standards of the Ad Hoc Committee for Standardized Reporting Practices in Vascular Surgery of the Society for Vascular Surgery International Society for Cardiovascular Surgery. Three classes of complications (systemic, local-nonvascular, and local-vascular or implant-related) and three grades of severity (mild, moderate, and severe) were used. A reintervention was defined as any surgical or endovascular procedure that was related to the primary aneurysm-repair procedure.

We aimed to reduce the possibility of publication bias through the following methods: searches of meeting abstracts, theses, dissertations and contacting authors for any additional unreported data. When necessary, we contacted the authors of the original papers to receive further information. The risk of bias was assessed using the methods of (Lundh and Gøtzsche, 2008).

Randomized controlled trials (RCTs): the quality of the methodology was assessed independently by two reviewers (F.S. and Y.H.) using the Cochrane risk of bias tool (Higgins et al., 2011). The risk of bias was divided into low (all criteria met), moderate (one or more criteria met), and high (no criteria met). In addition, the quality of the methodology used in each study was evaluated by the same authors, using the system described by (Jadad et al., 1996). Non-randomized studies (NORS): The quality of non-randomized studies was assessed by two reviewers (F. S. and Y. H.) using the MINORS scoring system describe by (Slim et al., 2003). This method uses (Moher et al., 1999) items for comparative studies with each item scored 0, 1 or 2 corresponding to not reported, inadequately reported, and adequately reported, respectively. Whether RCTs or NORS, discrepancies in ratings were resolved by discussion between the two reviewers (C. J. and Y. H.). Additionally, when discrepancies arose, a third party (J. Z.) was consulted.

After the initial assessment for eligibility, two authors (H. Y. and S. W.) independently abstracted the data from the primary sources. They independently reviewed all the articles. Different results between the two authors were resolved by consensus. For each study we extracted rudimentary information including author(s), study design, journals published, geographical location, date published, etc. The following predefined variables were recorded using an electronic extraction form: preoperative characteristics and postoperative outcomes (we conducted a separate meta-analysis for two different postoperative time periods: short-term and mid-long term). Follow-up visits were scheduled 30 days and 12, 24, 36 months and 5 years after the procedure. Any death or complications occurring within 30 days after the original procedure were defined as short-term, and any death, complication and reintervention occurring more than 30 days after the original procedure were defined as mid-long term. For the results during the follow-up period in the long-term analysis we used data on all-cause mortality at 1, 2-3 and more than 5 years after discharge, the rate of re-intervention at 1 year and more than 4 years and aneurysm-related mortality at 2-3 years and more than 5 years.

This study was performed in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement (available at http://www.prisma-statement.org/) This study was also conducted in accordance with the Quality Reporting of Meta-analyses (QUROM) guidelines (Moher et al., 1999) and reporting Meta-analyses of observational studies (MOOSE) guidelines (Stroup et al., 2000). We measured the odds ratio (OR) for dichotomous data, reported with 95% confidence intervals (CI). In addition, we calculated the weighted mean difference (WMD) with 95% CI for continuous data. The analysis was performed by comparing EVAR with OSR; an OR greater than 1 denoted that an event was more common following EVAR. We used the fixed effect model if evidence of nonsignificant heterogeneity was found. If there was evidence of significant heterogeneity, a random effects model as described by (DerSimonian and Laird, 1986) was used, and the degree of heterogeneity was assessed using the I² (Lovegrove et al., 2008) test. I² (Lovegrove et al., 2008) values of 25%, 25-50% and >50% were considered to represent small, medium and large amounts of inconsistency.

A comparison between Western and Chinese cohorts using the method of subgroups was performed to analyze the diversity between the two groups. However, for the mid-long term survival outcomes the hazard ratios (HRs), Kaplan-Meier survival curves and log-rank test P value (Tierney et al., 2007) were extracted from individual studies. Furthermore, pooling of individual log HRs was performed using the method described by (Parmar et al., 1998) Subgroup analysis and HRs which were calculated by a statistical method were used to contrast the two cohorts. An HR less than 1 indicated a better survival rate for EVAR.

Publication bias was assessed using a funnel plot. Furthermore, to assess heterogeneity, Egger’s test (Egger et al., 1997) and graphical exploration with Egger’s regression asymmetry plot were used to evaluate publication bias. We used the results of 30-day mortality and all-cause mortality for mid-long term as the effect index, because this was the most standardized outcome presented in most studies.

A difference for P < 0.05 was considered statistically significant. All statistical analyses were conducted using spreadsheet software (Microsoft Excel 2007; Microsoft, Redmond, Washington, USA) and performed using the meta. ado module of STATA (version 11.0, College Station, Texas, USA)

The Western literature search identified 172 studies for potential inclusion in the meta-analysis; 127 of these papers were excluded: 96 did not compare endovascular repair (EVAR) with open surgical repair (OSR), four papers described only a selected patient subgroup such as obese or octogenarian patients, another 18 papers contained insufficient data, and nine overlapped with included studies from the same institution. Thus, 45 papers met our inclusion criteria. Trial flow charts are provided in Fig 1. The trial flow chart for Chinese language articles is provided in Fig 2.

Figure 1.

Flow diagram of this meta-analysis in accordance with the QUOROM statement from Western studies.

Figure 2.

Flow diagram of this meta-analysis in accordance with the QUOROM statement from Chinese studies.

Details of the selected articles for this study are given in Table 1 for the English language articles. The included 45 papers assessed outcomes following elective AAA repair: three reports from the Dutch Randomized Endovascular Aneurysm Management (DREAM) Trial study (De Bruin et al., 2010; Blankensteijn et al., 2005; Prinssen et al., 2004) and 3 articles from the EVAR trial 1 study (United Kingdom EVAR Trial Investigators et al., 2010; EVAR trial participants, 2005; Greenhalgh et al., 2004) were treated as one study each for the purpose of analysis because in each case one of the reports represented a mid-long term follow-up study subsequent to an initial report. In addition, because (Johnson et al., 2006) and (Bush et al., 2006), Carpenter (Carpenter and Endologix, 2004) and (Wang et al., 2008; Matsumura et al., 2003; Peterson et al., 2007) reported from the same institution; they were treated as one for the purpose of analysis. The former dates were replenished respectively by Bush, Wang and Peterson, consequently these both fitted into our study. In total, 38 studies published up to January 2011 ultimately met our predefined inclusion criteria. These articles were published from June 1998 to June 2010 and used patients enrolled from 1994 to 2007. These articles consisted of four randomized controlled trials, 12 prospective non-randomized trials, and 22 retrospective conducted studies; 13 were from Europe and 25 were from North America.

The twenty-one Chinese language papers contained 1,405 AAA procedures. The Chinese language articles were published between 1993 and 2010 (Table 2).

Within the Western papers, a total of 31,074 elective AAA procedures were reported, of which 12,774 were EVAR and 18,300 were OSR. By comparison, a total of 1,405 operations were performed for unruptured AAA in the Chinese papers, of which 625 were EVAR and 780 were OSR. There are significant difference across racial groups, with Chinese patients more likely to receive EVAR compared to Western patients (P = 0.012) (Table 3).

Table 4 shows that four RCT studies adequately performed all methodological requirements as assessed by the Cochrane risk of bias tool and Jadad’s scoring system, and three were considered to be at low risk of bias, with the remaining trial was graded at moderate risk. Furthermore, for Western non-randomized studies, 19 (21 papers) of 34 studies (37 papers) achieved an ideal score of 24 points. However, 15 studies did not report follow-up data and therefore did not meet the criterion for the item “loss of follow-up less than 5%” and were thus allotted 22 points.

Patient characteristics:

In Table 5 we provide the demographic and clinical characteristics of all the patients included in this review, broken down by treatment group. A significant difference was present for sex in Western cohorts, with more male patients in the EVAR group (OR = 1.692, P < 0.001) whereas this difference was not present in Chinese cohorts (OR = 1.007, P = 0.972). Whether Western or Chinese, patients in the EVAR group were significantly older than those in the OSR group (P < 0.001 and P = 0.008), nevertheless, the age range was larger in Chinese cohorts (WMD_Western 2.413 years versus WMD_Chinese 3.63 years). The maximum diameter of the aneurysms in the EVAR group was significantly smaller than in the OSR group for both cohorts (P < 0.001 and P < 0.001). Also, the range was larger for the Chinese cohorts than for the Western cohorts (WMD_Western -2.491 mm versus WMD_Chinese -5.793 mm).

We found that the EVAR group had a higher rate of co-morbid conditions in Western studies, with a statistically significant difference in three of the baseline medical conditions; these were diabetes mellitus (OR = 1.248, P = 0.025), respiratory disease (OR = 1.515, P < 0.001) and cerebrovascular disease (OR = 1.287, P = 0.003). However, these differences were not present in the Chinese cohort. In addition, Western patients with the American Society of Anesthesiologists classification III or IV were more likely to receive elective EVAR than OSR (OR = 2.637, P = 0.007). Of note, thirteen Western studies showed a statistically significant difference in the odds ratio (OR = 2.901, P < 0.001) for congestive heart failure between the EVAR group and OSR group. No significant difference, however, was found in other aspects of cardiac disease in the Western and Chinese cohorts.

Characteristics for surgical and post-operative data are given in Table 6. Information on the type of anesthesia was available for seven Western studies: 36.1 percent of 1,612 patients receiving endovascular repair received a general anesthetic compared with 94.3 percent of 1,393 patients receiving open surgical repair (OR = 0.055, P < 0.001). Most of the surgical characteristics showed significant differences between the EVAR and OSR groups both for Chinese and Western cohorts. The duration of a typical EVAR procedure (mean 183.8 minutes versus 149.9 minutes) was significantly different than that of a typical OSR procedure (mean 261.9 versus 184.1 minutes) (WMD_Chinese -82.2 minutes versus WMD_Western -33.6 minutes, P < 0.001 versus P < 0.001, respectively).

Considering either the volume of blood loss or blood transfusion, the superiority of endovascular repair over open surgical repair was clear. Specifically, in the Western cohort the range of blood loss between EVAR and OSR was even larger (WMD_Western -989.1 ml versus WMD_Chinese -640.5 ml). Nevertheless, the range of blood transfusion between the two procedures was even smaller (WMD_Western -312.7 ml versus WMD_Chinese -517.3 ml), namely, contract by OSR, Chinese patients undergoing EVAR were transfused less blood (approximately a unit) than Western patients.

In addition, patients who underwent EVAR experienced significantly shorter stays in the intensive care unit (ICU) and the hospital for both Western and Chinese cohorts. The Western patients who underwent EVAR had significantly shorter stays in the ICU (WMD_Western -52.785 hours, P < 0.001) and significantly shorter duration in the hospital (WMD_Western -5.608 days, P < 0.001) than Chinese patients.

Table 7 presents a more detailed overview of the reported complications; the Western pooled estimate of the cardiac complication rate for EVAR was significantly lower than that for OSR (ORWestern = 0.411, P < 0.001). This superiority for EVAR was not replicated in the Chinese cohorts, however. Whether in Chinese or Western cohorts, endovascular repair was associated with a significant reduction in postoperative respiratory complications, however, the Chinese complication rate for the respiratory system (OR = 0.270, P < 0.001) in the EVAR group was lower than in Western patients (OR=0.552, P = 0.029) receiving EVAR. There was no significant difference in the incidence of renal, cerebrovascular, and wound complications. Furthermore, Chinese limb ischaemia or embolization were more common (ORChinese = 1.539, P = 0.049) in patients receiving EVAR, although Western studies did not contain this difference. In addition, the pooled estimate of 30-day mortality in Western studies for the EVAR group was significantly lower than that for the OSR group (OR = 0.481, P < 0.001), however this difference was not present in the Chinese cohort (OR = 0.733, P = 0.425) (Fig. 3).

Figure 3.

Forest plot of 30-day mortality following endovascular aortic aneurysm repair (EVAR) or open surgery repair (OSR) to the treatment of elective abdominal aortic aneurysms (AAAs) between China and Western countries.

3.8.1 Pooled ORs for Western studies

In Western countries the amount of medium-long term research in this field is far more than in China and it is therefore far easier to reach a satisfactory outcome using Western studies. Consequently, we have primarily based the analysis of the mid-long term comparison of the two AAA treatment methods on Western studies.

Fifteen, four, and nine articles reported data for all-cause mortality at one year, two-three years and more than five years, respectively. In an analysis of the pooled data we found no significant difference in all-cause mortality at one year, 2-3 years and more than five years (OR = 0.932, 95% CI: 0.723-1.203; P = 0.590 for one year) (OR = 0.850, 95% CI: 0.407-1.774; P = 0.665 for 2-3 years) (OR = 1.301, 95% CI: 0.982-1.724; P = 0.067 for more than 5 years) respectively (Fig. 4).

Figure 4.

Forest plot of all cause mortality following endovascular aortic aneurysm repair (EVAR) or open surgery repair (OSR) during follow-up time at one year, 2-3 years and more than five years after discharge.

Eight and three articles, respectively, compared aneurysm-related mortality for medium term aneurysm-related mortality and more than 5 years between endovascular repair group and open surgery groups, which showed that medium-term (mean follow-up time was 2-3 years) aneurysm-related mortality (OR = 0.566, 0.402-0.796; P = 0.001) was significantly lower after EVAR compared to OSR. Endovascular repair did not, however, influence long-term (mean follow-up time was more than five years) aneurysm-related mortality (OR = 0.639, 95% CI: 0.360-1.133; P = 0.125) (Fig. 5).

Figure 5.

Forest plot of aneurysm-related mortality following endovascular aortic aneurysm repair (EVAR) or open surgery repair (OSR) during follow-up time at 2-3 years and more than five years after discharge.

An analysis of eight studies with 1-year follow-up data showed that patients receiving EVAR were more likely to experience re-intervention, (OR = 2.042, 95 % CI: 1.467-2.843; P < 0.001) compared to OSR patients. Additionally, an analysis of four studies involving 3,214 participants with 3-year follow-up data showed that EVAR patients were more likely to experience re-intervention (OR = 4.107, 95% CI: 2.304-7.318; P < 0.001) than OSR patients (Fig. 6).

Figure 6.

Forest plot of rate of re-intervention following endovascular aortic aneurysm repair (EVAR) or open surgery repair (OSR) during follow-up time at one years and more than three years after discharge.

3.8.2 Pooled HRs for Western and Chinese studies

Although there were few studies reporting aneurysm-related mortality and the rate of re-intervention, we included the three Chinese articles which reported all-cause mortality (Fig. 7). These studies had a mean follow-up time of 25.4 to 37.6 months and involved 265 participants (95 in EVAR, 170 in OSR). The medium-long term all-cause mortality to a maximum follow-up period of 4 years was not significantly higher (HR = 1.173, 95% CI: 0.895-1.538; P = 0.247) for the EVAR group than the OSR group.

Figure 7.

Forest plot of all-caused mortality pooled HR (hazard ratio) following endovascular aortic aneurysm repair (EVAR) or open surgery repair (OSR) during follow-up time.

Data on all-cause mortality were reported in seventeen trials from Western studies. These studies included 9,994 participants, with a mean follow-up time ranging from 19.2 to 76.8 months. We found no statistically significant difference in the risk of all-cause mortality between the EVAR and OSR patients (HR = 1.026, 95% CI: 0.956-1.100; P = 0.483).

Publication bias was evaluated with a funnel plot of all analyses reporting on 30-day mortality following elective abdominal aortic aneurysm repair. all the points were within the trilateral region (Fig. 8A). Furthermore, publication bias was evaluated using the Begg test (result: P = 0.122) and Egger test (result: P > |t| was 0.920), and the Egger linear regression asymmetry plot frequently tended to show the presence of publication bias (Fig. 8B). The plot displays that the regression line passes through the point of origin, which may indicate the absence of bias.

Figure 8.

a. The funnel plot to assess publication bias from 30-day mortality data from Chinese and Western studies. b. Egger’s publication bias plot to assess publication bias from 30-day mortality data from Chinese and Western studies.