Median follow-up time was 2.8 years (interquartile range, 2.7 years) for open repair and 2.4 years (interquartile range, 1.3 years) for endovascular repair.
Jackson RS, Chang DC, Freischlag JA. Comparison of long-term survival after open vs endovascular repair of intact abdominal aortic aneurysm among Medicare beneficiaries. JAMA. 2012;307(15):1621-1628.
eTable. Clinical Classifications Software (CCS) Diagnostic Categories Used to Generate Propensity Scores
This supplementary material has been provided by the authors to give readers additional information about their work.
Jackson RS, Chang DC, Freischlag JA. Comparison of Long-term Survival After Open vs Endovascular Repair of Intact Abdominal Aortic Aneurysm Among Medicare Beneficiaries. JAMA. 2012;307(15):1621-1628. doi:10.1001/jama.2012.453
Author Affiliations: Department of Surgery, Georgetown University Hospital, Washington, DC (Dr Jackson); Department of Surgery, University of California, San Diego (Dr Chang); and Department of Surgery, the Johns Hopkins Medical Institutions, Baltimore, Maryland (Dr Freischlag).
Context Endovascular repair of abdominal aortic aneurysm (AAA) compared with open repair increases perioperative survival, but it is not known if it increases long-term survival.
Objective To compare long-term outcomes after open vs endovascular repair of AAA.
Design, Setting, and Patients Retrospective analysis of patients 65 years or older in the Medicare Standard Analytic File, 2003-2007, who underwent isolated repair of intact AAA. Cause of death was determined from the National Death Index.
Main Outcome Measures The primary outcome was all-cause mortality. Secondary outcomes were AAA-related mortality, hospital length of stay, 1-year readmission, repeat AAA repair, incisional hernia repair, and lower extremity amputation.
Results Of 4529 included patients, 703 were classified as having undergone open repair and 3826 as having undergone endovascular repair. Mean and median follow-up times were 2.6 (SD, 1.5) and 2.5 (interquartile range, 2.4) years, respectively. In unadjusted analysis, both all-cause mortality (173 vs 752; 89 vs 76/1000 person-years, P = .04) and AAA-specific mortality (22 vs 28; 11.3 vs 2.8/1000 person-years, P < .001) were higher after open vs endovascular repair. After adjusting for emergency admission, age, calendar year, sex, race, and comorbidities, there was a higher risk of both all-cause mortality (hazard ratio [HR], 1.24 [95% CI, 1.05-1.47]; P = .01) and AAA-related mortality (HR, 4.37 [95% CI, 2.51-7.66]; P < .001) after open vs endovascular repair. The adjusted hospital length of stay was, on average, 6.5 days (95% CI, 6.0-7.0 days, P < .001) longer after open repair (mean, 10.4 days), compared with endovascular repair (mean, 3.6 days). Incidence of incisional hernia repair was higher after open AAA repair (19 vs 23; 12 vs 3 per 1000 person-years; adjusted HR, 4.45 [95% CI, 2.37-8.34, P < .001]), whereas the incidence of 1-year readmission (188 vs 1070; 274 vs 376/1000 person-years; adjusted HR, 0.96 [95% CI, 0.85-1.09, P = .52]), repeat AAA repair (15 vs 93; 9.7 vs 12.3/1000 person-years; adjusted HR, 0.80 [95% CI, 0.46-1.38, P = .42]), and lower extremity amputation (3 vs 25; 1.9 vs 3.3/1000 person-years; adjusted HR, 0.55 [95% CI, 0.16-1.86, P = .34]) did not differ by repair type.
Conclusion Among older patients with isolated intact AAA, use of open repair compared with endovascular repair was associated with increased risk of all-cause mortality and AAA-related mortality.
Endovascular repair of abdominal aortic aneurysm (AAA), initially introduced as an option for high-risk patients,1 has surpassed open surgery as the most common technique for elective management of AAA among Medicare beneficiaries in the United States.2 In randomized clinical trials (RCTs), endovascular AAA repair has been shown to decrease 30-day and in-hospital mortality,3,4 blood transfusion requirements, duration of mechanical ventilation, and intensive care unit and hospital length of stay after repair.4 However, RCTs have failed to demonstrate a long-term survival advantage of endovascular compared with open repair.3- 5 Furthermore, compared with open repair, endovascular repair incurs higher costs3 and a need for long-term surveillance because of a 25% to 40% late complication rate,6,7 leading to ongoing controversy over the elective use of endovascular repair, especially in healthy patients with anticipated long-term survival.
Although clinical trials have failed to demonstrate a long-term mortality difference between open and endovascular repair, certain characteristics of RCTs limit the applicability of their results in clinical practice. In RCTs, repair of AAA is usually performed at high-volume centers by vascular surgeons experienced in endovascular technique. Participants in RCTs comparing endovascular with open AAA repair have, on average, fewer and less severe comorbidities and are more likely to be male than patients encountered in clinical practice.8 Most importantly, with rapidly evolving technology such as endovascular surgery, outcomes may change over time, rendering results of some previous RCTs obsolete.8 Well-designed retrospective cohort studies of open repair yield results that are both valid9 and more broadly generalizable than the results of RCTs. The objective of this study was to compare overall and AAA-specific mortality, readmission, and reintervention after endovascular vs open repair of intact AAA in a cohort of Medicare beneficiaries.
The Johns Hopkins Medicine institutional review board, Centers for Medicare & Medicaid Services, and National Death Index approved this study. The Johns Hopkins Medicine institutional review board waived the requirement for patient consent.
The study cohort was composed of eligible patients from the Medicare Standard Analytic Files (SAF), from January 1, 2003, through December 31, 2007. The Medicare SAF contain data from a 5% sample of Medicare inpatient discharges. The 5% sample is created based on selecting records that report a 05, 20, 45, 70, or 95 in positions 8 and 9 of the Health Insurance Claim number.10 Patient records contain longitudinal data from the index admission and subsequent hospital admissions, vital status, and date of death for deceased beneficiaries. Date of death in the Medicare SAF database is a Social Security Administration–validated date of death and is considered highly reliable. Therefore, follow-up for the primary study outcome, mortality, was considered complete. Longitudinal inpatient records were available through December 31, 2007, and vital status was ascertained through September 11, 2008.
Patients 65 years or older with an International Classification of Diseases, Ninth Revision (ICD-9) diagnosis code corresponding to intact AAA (441.4) and ICD-9 procedure codes corresponding to open (38.34, 38.44, 38.64, 39.52) or endovascular (39.71) AAA repair were included. Exclusion criteria were patient younger than 65 years; ICD-9 diagnosis codes for syphilitic (093.0), traumatic (901.0, 902.0), thoracoabdominal (441.6, 441.7), ruptured (441.3), or unspecified site aortic aneurysms (441.9); or ICD-9 procedure code for endovascular repair of thoracic aortic aneurysm (39.73). Patients with procedure codes for both open and endovascular repair recorded on the same date were also excluded, because it was not possible to determine procedure sequence in these patients.
Medicare SAF data were provided in deidentified form. For known decedents, patient identifiers (social security number, date of birth) were obtained to allow for linkage with the National Death Index (NDI). The NDI is a central computerized index of states' death record information that exists to facilitate epidemiologic and health services research.11 For patients who died during the study period, the cause of death was determined from the NDI.
Demographic variables (age, sex, and race) were obtained from the Medicare SAF. Age was categorized in 5-year increments. Race in the Medicare SAF is obtained from the Social Security Administration's master beneficiary record. Race was classified as black or not black, because the cohort included very few nonblack minorities, compared with other racial groups. Calendar date of repair was treated as a continuous variable.
In the data set provided, comorbidities were assigned using the Clinical Classifications Software (CCS) developed by the Healthcare Cost and Utilization Project of the Agency for Healthcare Research and Quality.12 The CCS is a categorization scheme that consolidates more than 14 000 ICD-9 diagnosis codes into a smaller number of CCS categories. Each patient had up to 25 CCS diagnostic categories assigned based on corresponding ICD-9 diagnosis codes from his or her inpatient, outpatient, skilled nursing facility, and home health agency Medicare files from the year preceding the index AAA repair. For patients in the 2003 Medicare SAF file, CCS diagnostic categories were assigned based on ICD-9 diagnosis codes from the year of the procedure, because files from the preceding year were not available.
The number of CCS diagnostic categories assigned to each patient was also recorded. Sixty-one diagnostic categories (corresponding to cardiovascular disease and risk factors, pulmonary disease, malignancy, and renal disease) known to be risk factors for poor outcomes after AAA repair13- 18 were selected as adjustment variables for the propensity scores (eTable).The selected diagnostic categories were treated as binary variables, with each patient classified as having or not having each of the 61 diagnoses.
The primary outcome was overall mortality following AAA repair, comparing open vs endovascular repair. Patients without recorded death dates were censored on September 11, 2008, because this was the last date at which death was ascertained in the Medicare SAF file used for analysis. Thirty-day mortality was used to compare perioperative mortality between repair types, because this has been shown to be more accurate than comparison of in-hospital mortality.19 Mortality within 30 days of repair was examined by censoring living patients at 30 days; mortality among patients who survived at least 30 days after repair was determined by excluding patients who died within 30 days after repair.
Secondary outcomes included AAA-related mortality, hospital length of stay, readmission (for any cause) within 1 year, repeat AAA repair, lower extremity amputation, and incisional hernia repair.
AAA-related mortality was defined by an ICD-9 diagnosis of intact or ruptured AAA denoting the underlying cause of death in the NDI record. AAA-related mortality occurring within 30 days of AAA repair and at more than 30 days after repair was examined using the method described for all-cause mortality. Repeat AAA repair was defined as open or endovascular repair after the index repair. Patients with both a diagnosis of iliac artery aneurysm (ICD-9 code 44.22) and an ICD-9 procedure code for aneurysm repair not specific to the aorta (39.52) were excluded from the analysis of repeat AAA repair, because these could represent primary repair of iliac artery aneurysm rather than repeat AAA repair. Indications for repeat AAA repair were categorized based on ICD-9 diagnosis codes as mechanical complication of vascular graft (996.1, 996.59), graft infection (996.60, 996.62), AAA rupture (441.3, 441.5), aortic (440.0) or graft (996.74) atherosclerosis, or embolization from the graft (444.2, 444.22, 444.81).
Lower extremity amputation was defined using ICD-9 procedure codes for lower extremity amputation at any level (84.10-84.18). Incisional hernia repair was defined using ICD-9 procedure codes specific to incisional hernia repair (53.51, 53.61, 53.62) or a combination of a procedure code for abdominal wall hernia repair (53.59, 53.63, 53.69) with a diagnosis code for incisional hernia (551.21, 552.21, 553.21, 998.31).
For readmission, event-free patients were censored 1 year after their index AAA repair or on December 31, 2007 (because this was the last date for which data on inpatient admissions were available), whichever was first. For the other secondary outcomes, event-free patients were censored on December 31, 2007.
Baseline Characteristics and Event Rates. Baseline patient characteristics were compared between the open and endovascular AAA repair groups using the t test for continuous variables and the χ2 statistic for binary and categorical variables. Incidence rates per person-time were calculated using Kaplan-Meier methods, and log-rank tests were used to compare survival functions between the open and endovascular repair groups. The t test was used to compare mean hospital length of stay between repair groups. The χ2 statistic was used to examine the diagnosis codes related to repeat repair, comparing the initial open repair and initial endovascular repair groups.
Propensity Scores. Logistic regression was used to calculate a propensity score for each patient. Propensity scores were equal to the adjusted odds of a patient undergoing open AAA repair (vs endovascular repair), based on emergency presentation, age, calendar year of repair, sex, race, number of CCS diagnostic categories, and the 61 selected CCS diagnostic categories. To assess for treatment group balance, box plots of propensity scores by AAA repair type were examined, and median propensity score and interquartile range were compared between the open and endovascular repair groups.
Because the relationship of propensity score to study outcomes was nonlinear, it was preferable to use the propensity score as a categorical variable. Categorization of the propensity score also allowed the relationship between propensity score and outcome to vary between the different study outcomes. Propensity scores were divided into quintiles by repair type, and propensity score quintile was treated as a categorical variable in the multivariable models.
Univariable and Multivariable Analysis. In all multivariable analyses, models were adjusted for propensity score quintile. For 30-day mortality, univariable and multivariable logistic regression analysis were used. For hospital length of stay, univariable and multivariable linear regression were used.
For all other outcomes, univariable and multivariable analyses were performed using Cox proportional hazard models. Log-log plots and Schoenfeld residuals were checked for violations of Cox proportional hazard model assumptions.
Sensitivity Analyses. Two sensitivity analyses were performed to examine the effect of model assumptions on study results. The first sensitivity analysis examined the influence of possible misclassification of surgical complications as comorbidities in the portion of the cohort whose CCS categories were drawn from the same calendar years as the index AAA repair. The primary analysis adjusted for all CCS categories as comorbidities. A sensitivity analysis was performed in which for any patient in whom the CCS categories were assigned from the same calendar year as the index AAA repair, any CCS category that could have represented a postoperative complication was not adjusted for as a comorbid condition (eTable). Propensity scores were recalculated based on these altered assumptions, and all multivariable analyses were performed as described above, adjusting for the recalculated propensity scores. These results were compared with the results of the primary analysis to assess whether uncertainty about comorbidities in this subset of patients confounded study results.
The second sensitivity analysis examined the influence of emergency presentation on study results. In the primary analysis, emergency presentation was adjusted for as a component of the propensity score. In the sensitivity analysis, patients with emergency presentation were excluded from the analysis and multivariable analysis was performed as described above, adjusting for propensity score quintile.
Two-sided tests for statistical significance were used, and statistical significance was defined as P ≤ .05 for all outcomes. All analyses were performed using Stata 10.20
During the study period, 4561 patients 65 years or older underwent repair of an intact AAA. Of these, 702 underwent open repair alone, 3824 underwent endovascular repair alone, and 35 underwent both repair types at the index admission. Of the 35 patients who underwent both open and endovascular repair, 32 underwent both procedures on the same date and were excluded from the analysis. The remaining 3 patients were reclassified according to the type of repair performed first, yielding 703 patients in the open repair group and 3826 patients in the endovascular repair group, for a total of 4529 patients. Race information was incomplete in 13 patients (0.03%). Data on other demographic variables were complete. Casewise deletion was used to exclude patients with missing data.
Patients who underwent open AAA repair were more likely to be women (29.2% vs 20.1%, P < .001) and were 1 year younger in mean age (75.2 years vs 76.4 years, P < .001) than patients who underwent endovascular repair (Table 1). A greater proportion of open repairs were performed in the earlier years of the study, whereas a greater proportion of endovascular repairs were performed in the later years (P < .001) (Figure 1). A higher proportion of open repairs were performed during an emergency hospital admission, compared with endovascular repairs (23.3% vs 14.7%, P < .001).
The mean number of recorded CCS comorbidity categories was slightly lower in the open repair group than in the endovascular repair group (4.6 vs 5.0, P = .004) (Table 1). The proportion of patients with a given comorbidity diagnosis differed significantly between open and endovascular repair for only 5 of the 61 CCS diagnostic categories used to generate the propensity scores, and these differences were all relatively small. Patients who underwent open repair were less likely to have a diagnosis of diabetes mellitus without complication (14.9% vs 18.8%, P = .02), essential hypertension (53.9% vs 59.1%, P = .01), or cardiac conduction disorders (5.3% vs 8.2%, P = .008), compared with patients who underwent endovascular repair.
Comparison of propensity scores between the open and endovascular repair groups showed similar distributions between the 2 repair types. For open repair, the median propensity score was 0.18 (interquartile range, 0.12), whereas for endovascular repair the median propensity score was 0.13 (interquartile range, 0.10).
In the analysis of survival, the mean follow-up time was 2.6 years (SD, 1.5 years) and the median follow-up time was 2.5 years (interquartile range, 2.4 years), with a maximum follow-up time of 5.7 years.
The overall mortality rate was higher after open AAA repair, compared with endovascular repair (173 vs 752; 89 vs 76 deaths/1000 person-years, P = .04) (Figure 2). The number needed to treat to save 1 person-year of life was 77 patients.
Of the 925 patients who died during follow-up, cause of death was determined for 913 (98.7%). Cause-specific mortality rates are shown in Table 2. The AAA-specific mortality rate was substantially higher after open AAA repair than after endovascular repair (22 vs 28; 11.3 vs 2.8 deaths/1000 person-years, P < .001). The majority of deaths within 30 days after surgery were recorded as related to aortic aneurysm, and when patients who died within 30 days after surgery were excluded, there was no difference in the unadjusted risk of AAA-specific mortality after open vs endovascular repair (6 vs 22; 3.1 vs 2.2/1000 person-years, P = .36).
Hospital length of stay was significantly longer after open AAA repair (mean, 10.43 days [95% CI, 9.60-11.26 days]) vs endovascular repair (mean, 3.58 days [95% CI, 3.42-3.75 days]) (P < .001). The rate of incisional hernia repair was higher after open vs endovascular AAA repair (19 vs 23; 12 vs 3/1000 person-years, P < .001). However, the incidence of 1-year readmission (188 vs 1070; 274 vs 376/1000 person-years, P = .50), repeat AAA repair (15 vs 93; 9.7 vs 12.3/1000 person years, P = .20), and lower extremity amputation (3 vs 25; 1.9 vs 3.3/1000 person-years, P = .20) did not differ by repair type.
Primary Outcomes. The adjusted hazard of all-cause mortality over the entire follow-up period was higher after open AAA repair, compared with endovascular repair (hazard ratio [HR], 1.24 [95% CI, 1.05-1.47]; P = .01) (Table 3). Within the first month after repair, patients in the open repair group had a higher risk of all-cause mortality (HR, 5.54 [95% CI, 3.47-8.82]; P < .001), compared with patients who underwent endovascular repair. Among patients who survived longer than 1 month, however, the risk of all-cause mortality did not differ by repair type (HR, 1.01 [95% CI, 0.84-1.22]; P = .91).
Secondary Outcomes. The adjusted risk of AAA-related mortality was higher among patients who underwent open AAA repair, compared with those who underwent endovascular repair (HR, 4.37 [95% CI, 2.51-7.66]; P < .001) (Table 3). Within the first month after repair, patients in the open repair group had a higher risk of AAA-related mortality (HR, 16.99 [95% CI, 4.62-62.54]; P < .001), compared with patients who underwent endovascular repair. Among patients who survived longer than 1 month, however, the risk of AAA-related mortality did not differ by repair type (HR, 1.35 [95% CI, 0.44-4.11]; P = .60).
The adjusted length of hospital stay was 6.5 days (95% CI, 6.0-7.0 days) longer after open AAA repair, compared with endovascular repair (P < .001). The adjusted hazard of incisional hernia repair was higher after open AAA repair, compared with endovascular repair (HR, 4.45 [95% CI, 2.37-8.34]; P < .001) (Table 3). There was no difference between the AAA repair types in the adjusted hazard of 1-year readmission (HR, 0.96 [95% CI, 0.85-1.09]; P = .52), repeat AAA repair (HR, 0.80 [95% CI, 0.46-1.38]; P = .42), or lower extremity amputation (HR, 0.55 [95% CI, 0.16-1.86]; P = .34). Statistical power to show a 20% reduction in the risk of 1-year readmission between the open and endovascular repair groups was 98%, but power was very low to detect a 20% reduction in risk of either repeat repair (13%) or lower extremity amputation (4%).
Among patients who underwent repeat AAA repair, patients who initially underwent endovascular repair were more likely to have a diagnosis of mechanical graft complication than were patients who initially underwent open repair (38.7% vs 0%, P = .003) (Table 4). Other diagnoses associated with repeat repair did not differ between the open and endovascular repair groups.
For the Cox proportional hazard models, log-log plots and Schoenfeld residuals did not identify any violations of proportional hazards or other model assumptions.
Sensitivity Analyses. In 27.2% of patients (36.2% undergoing open AAA repair vs 25.5% undergoing endovascular repair, P = .001) it was not possible to differentiate comorbidities from surgical complications, because their comorbidities were determined from the same calendar year as the index AAA repair. After removing from these patients' comorbidities any CCS category that could have represented a postoperative complication (eTable), all multivariable outcomes remained similar to those reported in the primary analysis. The adjusted risk of overall mortality over the entire study period was higher for open repair, compared with endovascular repair (HR, 1.24 [95% CI, 1.05-1.48]; P = .01). This was related to a higher adjusted risk of all-cause 30-day mortality (HR, 4.61 [95% CI, 2.98-7.15]; P < .001), because the subsequent adjusted risk of all-cause mortality among patients who survived at least 30 days after the index repair did not differ by repair type (HR, 1.02 [95% CI, 0.84-1.23]; P = .86).
Similarly, both the overall (HR, 3.98 [95% CI, 1.99-7.96]; P < .001) and 30-day (HR, 16.99 [95% CI, 4.62-62.54]; P < .001) risk of AAA-related mortality were increased in patients who underwent open AAA repair, compared with endovascular repair. In patients who survived beyond 30 days after repair, there was no subsequent difference in the risk of AAA-related mortality between the open and endovascular repair groups (HR, 1.35 [95% CI, 0.44-4.11]; P = .60). As in the primary analysis, the sensitivity analysis found that the length of hospital stay was, on average, 6.5 days (95% CI, 6.0-7.0 days) longer for patients who underwent open repair, compared with endovascular repair (P < .001). The adjusted risk of incisional hernia repair was increased after open AAA repair (HR, 4.41 [95% CI, 2.36-8.25]; P < .001), and the adjusted risk of 1-year readmission (HR, 0.96 [95% CI, 0.84-1.09]; P = .49), repeat AAA repair (HR, 0.77 [95% CI, 0.44-1.34]; P = .36), and lower extremity amputation (HR, 0.52 [95% CI, 0.16-1.78]; P = .30) were similar between repair groups.
In a second sensitivity analysis, 23.3% of patients who underwent open AAA repair (n = 64) and 14.7% of patients who underwent endovascular repair (n = 562) during emergency admission were excluded (Table 1). Results of this analysis were similar to those of the primary analysis. Open repair was associated with increased risk of overall (HR, 1.47 [95% CI, 1.06-2.03]; P = .02) and 30-day (HR, 4.70 [95% CI, 2.23-9.94]; P < .001) all-cause mortality, but there was no difference between repair types in the risk of all-cause mortality among patients who survived beyond 30 days (HR, 1.09 [95% CI, 0.74-1.60]; P = .66). Open AAA repair was also associated with increased risk of AAA-related mortality (HR, 4.85 [95% CI, 2.04-11.52]; P < .001) and repair of incisional hernia (HR, 20.75 [95% CI, 2.41-178.41]; P = .006), and hospital length of stay was longer after open AAA repair (6.1 days [95% CI, 4.5-7.6 days]; P < .001), compared with endovascular repair. There was no difference in the risk of 1-year readmission (HR, 1.18 [95% CI, 0.90-1.54]; P = .23), repeat AAA repair (HR, 1.39 [95% CI, 0.55-3.49]; P = .49), or lower extremity amputation (HR, 0.67 [95% CI, 0.14-3.21]; P = .62) between the 2 repair types.
This study of Medicare beneficiaries demonstrates a long-term overall and AAA-specific survival advantage associated with endovascular repair of intact AAA, compared with open repair, in patients 65 years or older. The long-term survival advantage associated with endovascular repair persisted in multivariable analysis, suggesting that this relationship was independent of patient demographics, emergency presentation, and comorbidities. Improved long-term survival was entirely associated with reduction in perioperative deaths after endovascular repair, compared with open repair. However, the survival advantage generated in the perioperative period persisted throughout the entire follow-up period, suggesting that, compared with open repair, endovascular repair was associated with similar protection against late AAA rupture.
Similar to our study, the EVAR 1 (Endovascular Aneurysm Repair 1) and OVER (Open Versus Endovascular Repair) trials both reported a perioperative survival advantage of endovascular AAA repair over open repair.3,4 In these randomized trials, however, the survival advantage was not statistically significant at 2 years of follow-up.3,4 An earlier study of Medicare beneficiaries undergoing AAA repair between 2001 and 2004 showed an initial survival benefit for endovascular repair but equivalent survival beyond 3 years.21 The present study demonstrates an even longer survival advantage of endovascular over open repair, which was maintained throughout the entire 5.7 years of follow-up. Improved durability of endovascular repair over time, as well as the large sample size of the present study, may be important in explaining differences between the findings presented here and those of earlier3,21 and smaller4 studies.
Some previous studies have raised concerns about the possible need for more reinterventions after endovascular AAA repair, compared with open repair. This analysis demonstrates substantially increased risk for incisional hernia repair among patients who underwent open AAA repair, compared with endovascular repair. Although this result makes intuitive sense, it deserves mention because comparisons of reintervention after open vs endovascular AAA repair have sometimes overlooked incisional hernia repair.3
In our study, there was no evidence of a difference in hazard of 1-year readmission, repeat AAA repair, or lower extremity amputation between the 2 repair types. These results are comparable with those of the OVER trial, which demonstrated similar rates of procedure failure and secondary therapeutic procedures between the 2 repair types but higher rates of incisional hernia repair after open AAA repair.4 In contrast, the EVAR 1 trial group reported higher rates of reintervention for graft-related complications after endovascular AAA repair.3 The Medicare SAF uses ICD-9 (rather than Current Procedural Terminology) procedure codes, and it was therefore not possible to identify minor endovascular reinterventions, such as placement of an extension cuff, which may explain the difference between our findings and those of the EVAR 1 trial. However, the present study demonstrates that there was no increased risk of major AAA-related reintervention (repeat open or endovascular repair) between the repair types. Among patients undergoing repeat AAA repair, the present study did find a more frequent diagnosis of graft-related mechanical complications in patients who initially underwent endovascular repair, compared with those who initially underwent open repair. However, most repeat repairs after initial open repair did not have an associated explanatory diagnosis, suggesting that this difference may be attributable to reporting bias.
Although endovascular AAA repair was once considered an option for patients with significant comorbidities who were not good candidates for open repair, it is now accepted that patients who are unfit for open repair also have poor outcomes after endovascular repair.22 The criteria for endovascular repair are anatomical: candidates must have a length of at least 1.5 cm at both the infrarenal neck and the common iliac arteries for proximal and distal graft fixation as well as appropriate iliofemoral access without excessive tortuosity.23 Patients undergoing open and endovascular AAA repair had similar baseline characteristics and propensity score distributions, suggesting that surgeons selected patients for endovascular repair based on anatomy rather than age and comorbidity. Similarly, previous research among Medicare beneficiaries has shown no increase in overall AAA repair procedure volume, despite a more than 2-fold increase in endovascular repair procedure volume between 2001 and 2006.2 The finding of similar age and comorbidity distributions in patients undergoing open vs endovascular repair suggests an appropriate shift away from selection of repair type based on patient fitness toward selection based on patient anatomy.
Our study has some limitations related to the use of administrative data. First, our study was performed using Medicare data and only included patients 65 years or older. This is a minor limitation, because the prevalence of AAA is low in younger patients.24
Second, our study was not a randomized trial, and patient comorbidities and age may have influenced the repair type chosen. However, if elderly patients or those with serious comorbidities were selectively chosen for endovascular repair, this would be expected to bias study results to show a survival disadvantage for endovascular repair. This limitation, therefore, cannot account for our finding of a mortality advantage for patients who underwent endovascular repair.
Third, the Medicare database does not contain information about aneurysm size or anatomical features, which influence mortality after AAA repair.25 Medicare data also do not allow adjustment for hospital or surgeon repair volume, which correlate positively with survival, particularly after open repair.26,27 However, previous research has shown that registry data and RCTs comparing endovascular and open repair yield similar results after controlling for comorbidity,9 suggesting that these limitations should not invalidate the findings of our study.
Fourth, the Medicare SAF database provides information only on inpatient procedures, and secondary interventions performed on an outpatient basis after the index AAA repair would not be captured by this study. However, although this strategy could overlook some repairs of incisional hernia, it is unlikely to miss any repeat AAA repairs, because these procedures almost always necessitate inpatient admission.
We have demonstrated that, after adjusting for demographics and comorbidities, endovascular AAA repair was associated with a long-term survival advantage when compared with open repair in patients 65 years or older. This survival difference was related to higher mortality within the first month after open repair but persisted for the entire 5-year follow-up period. We demonstrated a longer average hospital stay after open AAA repair and a higher risk for repair of incisional hernia after open AAA repair but did not find evidence of differences in the hazard of rehospitalization within 1 year after AAA repair, repeat repair, or lower extremity amputation, comparing open vs endovascular repair.
Corresponding Author: Rubie Sue Jackson, MD, MPH, Department of Surgery, Georgetown University Hospital, 3800 Reservoir Rd, Washington, DC 20007 (email@example.com).
Author Contributions: Dr Jackson had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study conception and design: Jackson, Chang, Freischlag.
Acquisition of data: Jackson, Chang.
Analysis and interpretation of data: Jackson, Chang, Freischlag.
Drafting of manuscript: Jackson.
Critical revision of manuscript: Chang, Freischlag.
Statistical analysis: Jackson, Chang.
Study supervision: Chang, Freischlag.
Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest and none were reported.
Funding/Support:Dr Chang is supported in part by SCANNER grant R01 HS19913-01 awarded by the Agency for Healthcare Research and Quality.
Role of the Sponsor: The Agency for Healthcare Research and Quality had no role in the design and conduct of the study; the collection, analysis, and interpretation of the data; or the preparation, review, or approval of the manuscript.
Additional Contributions: We thank Gerard Anderson, PhD (the Johns Hopkins Bloomberg School of Public Health), for his generosity in preparing the Medicare database used for this study. Dr Anderson received no compensation for his contributions.