Occurrence and outcomes of type 3 endoleaks in endovascular aortic repair within the Vascular Quality Initiative database

Vascular Quality Initiative (VQI) database

The Society for Vascular Surgery (SVS) VQI is a patient safety organization initially formed in 2001 as the Vascular Study Group of New England which served as a voluntary cooperative regional database for local quality improvement and expanded yearly thereafter. It joined with the SVS and became national in 2011 and is now composed of 18 regional groups with registry data representing more than 650 community and academic hospitals in the USA and Canada.9 Entry of data and procedures, now totaling well over 800 000 procedures crossing nine major vascular catagories, is maintained by individual centers and physicians and ensured by annual audits against hospital claims data.10 Follow-up data are required for 1 year following procedure but, at the discretion of individual centers, can be entered indefinitely. Data entry and completeness, and therefore integrity, are assisted at the time of data entry by detailed definitions, prompts and incomplete data warnings.11 The VQI is the largest database dedicated to quantifying vascular procedural outcomes and is unique from other large databases in its longitudinal follow-up.12 This project was screened and approved by the VQI Research Advisory Council and was reviewed and approved by the institutional review board at the University of Alabama at Birmingham.

Cohort selection

A retrospective cohort of deidentified patients undergoing standard elective endovascular abdominal aortic aneurysm repair for non-ruptured aneurysms within the VQI from January 2003 to September 2018 was identified. Standard EVAR was defined as infrarenal device deployment without use of complex techniques such a custom/physician modified devices, fenestrated/branched EVAR or parallel stenting techniques. Analysis of case variables allowed duplicate patient entries to be consolidated and duplicate data eliminated. Manufacturer data for each stentgraft was not directly characterized, although the restricted data were collected in the VQI.

Standard endoleak definitions were used (full definitions are provided in online supplemental table 1). T3ELs are defined as persistent flow into the aneurysm sac through graft components (modular disconnections or graft fabric defects).13 Cases were included in the final analysis if they had non-missing responses to initial endoleak evaluation. If no evaluation for endoleak was ever documented, the cases were removed. Two time frames of endoleaks were defined; index and incident endoleaks. Index endoleaks were defined as any endoleak discovered during the index hospitalization, which was defined as within the procedure to time of hospital discharge. Incident endoleaks were defined as any new endoleak discovered at follow-up that was not present during initial hospitalization. This categorization prevented endoleaks from being included as duplicate entries and ensured that a patient could not have both an index hospital T3EL and incident T3EL. Patients without documented follow-up were included in procedural and index hospitalization analysis but were not included in the long-term follow-up (incident) cohort. Long-term follow-up data were defined as the period between 9 and 21 months after initial procedure and index hospitalization. All patients included in this study had endoleak documentation at index procedure and at long-term follow-up within the 9–21 months of follow-up reporting interval. The timing of post-operative CT scans for endoleak surveillance in this procedural registry is up to individual surgeons based on clinical discretion, though time to imaging and time to reinterventions from index procedure are captured.

Covariates and outcomes examined

Preoperative demographics analyzed included age, race, gender, primary insurer and weight (kg). Comorbidities assessed included coronary artery disease, chronic obstructive pulmonary disease, cerebrovascular disease, congestive heart failure, hypertension, dialysis-dependent end-stage renal disease (ESRD), preoperative creatinine, preoperative ejection fraction, stress test results, diabetes mellitus (insulin or non-insulin requiring), smoking (current/former/never), prior percutaneous coronary intervention and/or coronary artery bypass grafting, and preoperative ambulatory status), where available.

Intraoperative variables analyzed included procedure time (in hours), American Society of Anesthesiologists class, intraoperative heparin administration, proximal extent of aortic disease by zone, intraoperative crystalloid volume infused (in litres), intraoperative transfusions, intraoperative estimated blood loss (in litres), return to operating room, concomitant procedures and intraoperative complications.

Postoperatively, variables were assessed at two time-frames: index hospitalization, defined at the period between procedure and first discharge, and follow-up, defined as 9–21 months post-operatively. Variables examined included reinterventions which was defined per VQI standards as any repeat surgical procedure to address a complication of the primary procedure. Aortic reinterventions were defined per VQI standards as involving any of the following: aneurysm sac growth, aneurysm rupture, device migration or misalignment, device occlusion or stenosis, device infection or endoleak development (with subcatagories of endoleaks represented). Access reinterventions were defined as bleeding, stenosis, thrombosis, pseudoaneurysm or distal embolization of any vessel accessed for device delivery.

Statistical analysis

Collected data were analyzed in SPSS statistical software V.23.0. Continuous variables were presented as mean±SD and were compared using an independent Student t-test. Categorical variables were presented as number and percentage and were compared using a χ² test. Kaplan-Meier curve analysis and log-rank analysis were used to plot outcomes related to survival and freedom from intervention. A p value of <0.05 defined statistical significance. Missing values were automatically recognized by SPSS and treated as system-missing and corresponding cases were omitted in a pairwise manner from descriptive statistics, frequencies, correlations, and regressions.

Study endpoints

Primary endpoints of this study were to determine the occurrence rate of T3ELs at both placement and follow-up, and to determine the association of T3ELs with reinterventions and all-cause mortality.

Demographics

Between 2003 and 2018, a total of 42 246 individual EVAR cases with recorded endoleak evaluation were entered into the VQI database. Of those, 41 604 had appropriate procedural data available and were included in final analysis. Out of the initial EVAR cohort, 63.5% (n=26 422) had any follow-up data recorded. All patients included in follow-up had appropriate endoleak documentation during index hospitalization and up to at least 9 months postoperatively, per VQI follow-up standards. Cohort demographics are presented in table 1 and are notable for predominately male populations of current and former smokers with expected aneurysm comorbidities. Within the VQI, EVAR procedure case numbers demonstrated a linear year-over-year increase in case entry starting in 2003. Time until first imaging was highly variable, and this variable had significant incompleteness (84%) and was therefore not analyzed. Figure 1 demonstrates cohort paths based on occurrence of T3EL, mortality, and follow-up.

Request reuse permission

Index hospitalization T3EL rates and reinterventions

At index hospitalization, the rate of any endoleak around initial EVAR placement was 23% (n=9548). Overall endoleak rates and distribution of endoleak types for standard EVAR at index hospitalization are presented in table 2. The rate of T3EL in EVAR at index hospitalization was 0.4% (n=157), which accounted for 1.6% of all the initial endoleaks (table 2).

Within these 157 endoleaks, 85% (n=133) were attributed to mid-graft separation and 15% (n=24) were attributed to mid-graft fabric holes. The raw number of T3EL peaked in 2014 and has since been downtrending, likely due to changes in device availability and overall device design. Similarly, the rate of T3EL has been downtrending since 2014 (online supplemental figure 1). Within standard EVAR, Type 2 endoleaks were the most common, comprising 71% of all initial endoleaks. The rate of indeterminate endoleaks within standard EVAR at placement was 2.3% (n=942), which accounts for 9.8% of all of the endoleaks.

In standard EVAR at index hospitalization, the overall rate of reintervention for any indication was 1.8% (n=766). The rate of reinterventions for any endoleak type was 0.12% (n=50/41,604), which accounts for 6.5% of all reinterventions (n=50/766) in standard EVAR during index hospitalization (table 3).

The rate of reintervention for T3EL during index hospitalization was 0.02% (n=7). Out of the 157 T3EL discovered during index hospitalization, 4.5% received procedural intervention during that hospitalization. Reinterventions for T3EL account for 14% of endoleak reinterventions and 1% of all reinterventions (table 4).

Type 3 endoleak rates and reinterventions at follow-up

The overall rate of any endoleak at follow-up was 16%. Endoleak rates overall and by endoleak type for standard EVAR at follow-up are presented in table 2. The rate of incident T3EL in standard EVAR at follow-up was 0.7% (n=205), which accounts for 4.7% of all the endoleaks seen at follow-up. Type 2 endoleaks were the most common endoleak, comprising 58% (n=196) of all endoleaks in standard EVAR. The rate of indeterminate endoleaks within standard EVAR at follow-up was 2% (n=509), which accounts for roughly 11.7% of all the endoleaks.

The overall follow-up reintervention rate for standard EVAR for any indication was 3.4% (n=899). The overall rates of reintervention for endoleaks at follow-up in standard EVAR are presented in table 3. The overall rate of reintervention for any endoleak was 1.3% (n=335), which accounts for 37% of all reinterventions. Of the 205 incident T3EL discovered at long-term follow-up, only 30 (15%) were recorded as treated with reintervention. Reintervention for T3EL occurred in 0.1% of the population, accounting for 8.9% (n=30/335) of the endoleak reinterventions and 3.3% of all reinterventions (table 4).

Survival and type 3 endoleaks

By the end of the observational study time frame concluding in 2018, 8.7% (n=2455) of the cohort was deceased. The mean follow-up time in surviving patients was 4.4 years.The presence of an index hospitalization T3EL (present in 0.4% of standard EVAR) did not affect overall mortality. Analysis of long-term follow-up data in standard EVAR did demonstrate a significant relationship between the presence of an incident T3EL and overall survival (figure 2). The overall mortality rate for standard EVAR patients with incident T3EL was 25%. We did not identify any mortalities within 30 days of T3EL reintervention. The presence of an incident T3EL was associated with decreased survival at 5 years compared with cases without incident T3EL (74% vs 80%, p=0.041).

Request reuse permission

Limitations

Although every effort was made to ensure accuracy and reproducibility of this study, several limitations must be addressed. While the 2019 Annual VQI report indicates strong sustained enrollment of centers across the USA with the addition of 110 new reporting centers, the VQI is not yet entirely comprehensive in its representation of EVAR implantations in the country.24 Additionally, early procedural volumes in the VQI are less robust, given given the national expansion that occurred after 2011 and subsequent extension of the registry to Canadian centers; however, annual EVAR procedural volumes in VQI remain significantly higher than those available in other US national surgical quality databases.12 25 An additional limitation is inherent to analysis of the VQI, in that the data are derived on a voluntary basis from individual surgeons, and assumptions must be made regarding the veracity of the data. Thus, although the VQI compares entries against hospital claims data, it is possible that discrepancies, intentional or unintentional, exist within the data due to its self-reported nature. Despite these factors, specificity in endoleak reporting for this EVAR-specific complications is unavailable in other surgical quality databases or claims-based administrative data, currently making VQI our most representative sample of clinical practice in North America.

With regard to diagnosis of T3EL, there is no absolute certainly of when a T3EL occurred, unless it occurred during the index procedure where completion angiographic runs are standard of care. Postprocedural imaging studies prior to discharge are not mandatory or advised in the absence of patient clinical indications. Therefore, a T3EL could develop during index hospitalization but not be diagnosed until follow-up imaging; modalities and duration of long-term follow-up imaging in the window of 9–21 months are recorded but pose potential for lack of documentation in the registry of findings from commonly employed 1-month procedural scans. Ultimately, the nature of the endoleak type (fabric hole or device modular disconnection) makes it unlikely that a T3EL would spontaneously resolve or be missed at the long-term follow-up interval if a reintervention had not been recorded. In the future, if needed to better understand the exact timing, a more explicit time to diagnosis variable in VQI could reduce uncertainty or creation of type-specific endoleak billing codes could facilitate linkage to claims data. A final significant limitation seen in this study is the lack of continuity within patients’ follow-up. Of the initial patients who had an EVAR device implanted, 37% were lost to follow-up. This presents several analytical difficulties, and whether this represents poor documentation of the behalf of physicians or if patients are truly not returning for surveillance after surgical intervention is unknown. Further, the lack of follow-up makes drawing conclusions regarding survival, both positive and negative, difficult. Initial attempts to address the limitations presented earlier include the MDEpiNet Vascular Implant Surveillance & Interventional Outcomes Network coordinated research network initiative expanding administrative claims and other data source linkages to VQI data which can aid in extension of long-term capture of patient care.26

Despite the aforementioned shortcomings, the VQI database represents the single largest well-regulated data available for analysis with the granular detail to study T3ELs. While these limitations should be taken into consideration when evaluating the conclusions presented in this paper, all studies that interpret the VQI data face the aforementioned difficulties and the limitations do not invalidate the findings or conclusions.

Finally, the presence of a T3EL at 9–21 months was associated with trends toward decreased survival; we suspect this trend would continue if data were available for follow-up in ensuing years; however, consistent reporting of 5 and 10 years is unavailable in this observational database at present. Future analyses using linkage of EVAR patients to Medicare claims data may allow for further longitudinal assessment of the mortality impact.