Comparable Perioperative Mortality Outcomes in Younger Patients Undergoing Elective Open and Endovascular Abdominal Aortic Aneurysm Repair

Abstract

Background

Evidence for benefit of endovascular (EVAR) over open surgical repair for de novo infrarenal abdominal aortic aneurysms (AAA) in younger patients remains conflicting due to heterogeneous study populations and small sample sizes. The objective of this study is to compare perioperative and short-term outcomes for EVAR and open surgery in younger patients using a large national disease and procedure specific dataset.

Methods

We identified patients 65 years of age or younger undergoing first-time elective EVAR or open AAA repair from the Vascular Quality Initiative (2003–2014). We excluded patients with pararenal or thoracoabdominal aneurysms, those medically unfit for open repair, and EVAR for isolated iliac aneurysms. Clinical and procedural characteristics were balanced using inverse propensity of treatment weighting. A supplemental analysis extended the study to those under 70 years old.

Results

We identified 2641 patients; 73% (n=1928) EVAR and 27% (n=713) open. The median age was 62 (IQR 59–64) and 13% were female. The median follow-up time was 401d (IQR 357–459). Unadjusted perioperative survival was 99.6% overall (open: 99.1%; EVAR: 99.8%; P<0.001) with 97.4% 1-year survival overall (open: 97.3%; EVAR: 97.4%; P=0.9). Unadjusted reintervention rates were 5 (open) and 7 (EVAR) reinterventions per 100 person-years (P=0.8). After propensity weighting, the absolute incidence of perioperative mortality was less than one percent in both groups (Open: 0.9%, EVAR: 0.2%; P<.001), and complication rates were low. Propensity-weighted survival (HR 0.88, 95% CI [0.56–1.38]; P=0.6) and reintervention rates (open: 6; EVAR: 8; reinterventions per 100 person-years; P=0.8) did not differ between the two interventions. The under-70 analysis showed similar results.

Conclusions

In this study of younger patients undergoing repair of infrarenal AAA, 30-day morbidity and mortality for both open surgery and EVAR are low and the absolute mortality difference is small. The prior published perioperative mortality and 1 year survival benefit of EVAR over open AAA repair is not observed in younger patients. Further studies of long-term durability are needed to guide decision making for open versus EVAR in this population.

INTRODUCTION

Endovascular aneurysm repair of abdominal aortic aneurysms (AAA, EVAR) has increasingly replaced open surgical repair as the standard of care for anatomically suitable candidates^1,2. By 2006, EVAR utilization had risen to over 70%³ and continued at this rate through 2010⁴. EVAR has been shown to have a significant perioperative mortality and morbidity advantage that decreases over time^5,6, especially amongst the elderly or those with significant comorbid conditions but even in acceptable-risk patients. However, this perioperative advantage may be less in lower risk populations such as patients with fewer medical comorbidities or those of a younger age with longer life expectancies⁷.

EVAR has been shown to be potentially less durable compared to open repair over time requiring more reinterventions and with a small but persistent risk of aneurysm rupture^6,8. The declining effectiveness of EVAR over time may make open repair an acceptable alternative for younger patients if the perioperative morbidity and mortality risks in this population are comparable. Therefore, the objective of our study is to examine the perioperative and one-year outcomes of younger patients receiving open repair or EVAR.

METHODS

This study of deidentified national registry data was approved by the Society for Vascular Surgery Patient Safety Organization Research Advisory Committee and our institutional review board prior to data acquisition and analysis. This study was exempted from informed consent due to the deidentified nature of the data.

We identified and included patients 65 years of age or younger undergoing AAA repair using the Vascular Quality Initiative (VQI) EVAR and open repair registries from 2003–2014. We sought to compare only patients with AAAs that were amenable to either open repair or EVAR. In an attempt to approximate these parameters, exclusion criteria included any pararenal EVAR chimney or fenestrated operations, as well as any open repairs involving suprarenal clamping/pararenal or thoracoabdominal aneurysms. Repairs performed for isolated iliac aneurysm were also excluded. To preserve the comparability of the two groups, EVAR patients who were deemed medically unfit for open repair as designated on the intake form by the operative surgeon were also excluded. For patients with multiple recorded abdominal aortic aneurysm repairs in the database, the earliest repair was used as the index procedure. The primary comparison groups were EVAR and open repair. Although these two procedures were recorded in two distinct registries, the outcome measure variables were consistent across both datasets. The primary outcome measures were in-hospital mortality and morbidity rates. Primary morbidity measures were myocardial infarction, respiratory failure (defined in the VQI registry as prolonged intubation), need for new-onset hemodialysis, or acute stroke. Biomarker-only elevations such as transient elevated troponin or creatinine elevation without clinical diagnosis of MI or need for hemodialysis were not included as these may be subject to significant selection bias from differing baseline hospitalization courses of the two procedures. Secondary outcomes were reintervention rates over time and one-year survival based on SSDI-linked death records present in the VQI data. Follow up rates were limited to one year as institutional follow-up past one year is not required for participation in VQI; those without follow-up were excluded from longitudinal analysis.

Propensity weighting was used to further balance and adjust for clinical and comorbid characteristics between the EVAR and open repair groups. The propensity for receiving EVAR vs open repair was fit using a logistic model and clinically relevant covariates. Covariates were included following the stepwise inclusion method described by Imbens and Rubin⁹ or forced into the model if deemed clinically relevant (Supplemental Table 1). The comparability of the two initial cohorts was confirmed by qualitatively examining the overlapping distributions of propensity scores (common support, Supplemental Figure 1). An inverse probability of treatment weight based on the propensity score was then calculated for each subject and applied to both cohorts; stabilized weights were used to correct for outliers¹⁰. Adequate balance between the weighted EVAR and open groups was confirmed using standardized differences⁹. The final adjusted analyses were conducted using these weighted cohorts. A supplemental analysis extended this study to those up to the age of 70 years.

Statistical Analysis

Two-sample t, chi-square, Fisher exact, Wilcoxon rank-sum, and Kruskal-Wallis tests were used for unadjusted comparisons between the two cohorts where appropriate. Normality was assessed qualitatively using histogram plots for continuous variables. Analysis of in-hospital and one-year binary outcomes was performed using logistic regression, while analysis of reintervention rates over time was done using negative binomial regression in order to account for multiple reinterventions per patient. All analyses were performed using Stata 13.1 (Statacorp, College Station, TX) and R 3.1.0 (R Foundation for Statistical Computing, Vienna, Austria).

RESULTS

Out of 22,268 patients undergoing AAA repair from 2003–2014, we identified 2641 (11.9%) patients meeting the inclusion criteria. We excluded 17,538 (78.8%) due to age criteria; subsequently, 416 were excluded for unfit-for-open-repair status, 576 for suprarenal clamping, 958 for urgent or ruptured status, and 139 who underwent concomitant renal artery bypass. EVAR was performed in the majority of patients (73%; n=1928) compared to open repair (27%; n=713). The median age was 62 years (IQR 59–64) and 13% were female (n=337). The median follow-up time was 401 days (IQR 357–459). EVAR patients were significantly older and had higher rates of comorbidities, including diabetes, coronary heart disease, congestive heart failure, emphysema, and previous coronary bypass or intervention (Table 1). The maximum AAA diameter was significantly larger in the open group (EVAR: 54mm, IQR 50–59; open 55mm, IQR 50–62; P<0.001), although this difference is likely clinically indistinct. Patients in the open group were also more likely to have concomitant iliac aneurysms (EVAR: 25.1%, n=481; open: 35.2%, n=250; P<0.001).

Table 1.

Demographics and Intervention Characteristics

	EVAR	Open	P
	N=1928	N=713
Age (yr)	62 (59, 64)	61 (58, 64)	<.001
Male Gender	1696 (88.0%)	608 (85.3%)	.076
Hypertension	1556 (80.7%)	566 (79.5%)	.47
Diabetes	406 (21.1%)	96 (13.5%)	<.001
Coronary Artery Disease	557 (28.9%)	171 (24.1%)	.014
Heart Failure	139 (7.2%)	33 (4.6%)	.016
Emphysema	525 (27.3%)	204 (28.7%)	.46
History of CABG/PCI	604 (31.4%)	194 (27.3%)	.045
History of Peripheral Bypass	58 (3.0%)	28 (3.9%)	.27
History of CEA/CAS	65 (3.4%)	30 (4.2%)	.35
History of PVI	93 (4.8%)	25 (3.5%)	.17

Caption. All values are N(%) or median (IQR). Abbreviations: CABG: coronary artery bypass graft; PCI: percutaneous coronary intervention; CEA: carotid endarterectomy; CAS: carotid stenting; PVI: peripheral vascular intervention

The unadjusted in-hospital mortality rate was very low for both methods of repair but was higher for open repair (EVAR: 0.2%, n=3; open: 1.0%, n=7; P<0.002). Open repair also carried a significantly higher rate of myocardial infarction (EVAR: 0.1%, n=2; open: 1.3%, n=9; P<0.001), respiratory failure requiring prolonged intubation (EVAR: 0.5%, n=10; open: 5.6%, n=40), acute kidney injury requiring hemodialysis (EVAR: 0.1%, n=2; open 1.4%, n=10; P<0.001), and limb ischemia (EVAR: 0.5%, n=9, open: 1.3%, n=9; P=0.04). The postoperative risk of stroke was not significantly different between the two cohorts (Table 2). Unadjusted reintervention rates were 5 (open) and 7 (EVAR) reinterventions per 100 person-years (P=0.8). Unadjusted one-year survival rates were not significantly different between the two interventions (both open and EVAR: 3.0%, P=0.98).

Table 2.

Unadjusted Perioperative Outcomes

	EVAR	Open	P
	N=1928	N=713
30-Day Mortality	3 (0.2%)	7 (1.0%)	.002
Transfusion	19 (1.0%)	52 (7.6%)	<.001
Myocardial Infarction	2 (0.1%)	9 (1.3%)	<.001
Dysrhythmia	23 (1.2%)	48 (6.7%)	<.001
Respiratory Failure	10 (0.5%)	40 (5.6%)	<.001
Renal Failure	2 (0.1%)	10 (1.4%)	<.001
Stroke	5 (0.3%)	3 (0.6%)	.26
Endoleak at Completion	413 (22%)	-
Type 1	30 (1.7%)	-
Type 2	346 (18.1%)	-
Type 4	31 (1.7%)	-
Open Conversion	5 (0.3%)	-
Transperitoneal Exposure	-	590 (83%)
Length of Stay (d)	1 (1, 2)	6 (5, 8)	<.001
Length of ICU Stay (d)	0 (0, 1)	2 (1, 3)	<.001

Caption. All values are N(%) or median (interquartile range)

After propensity weighting, the two weighted cohorts were found to be adequately balanced over the relevant covariates (Supplemental Figure 2). Analysis of the propensity-weighted cohorts continued to demonstrate a statistical difference in the rates of in-hospital mortality and morbidity, although the rates of mortality (Open: 0.9%, EVAR: 0.2%) and myocardial infarction were extremely low at less than 1%. No significant difference was demonstrated in the risk of postoperative stroke between the two groups (Figure 1). Propensity-weighted survival (HR 0.88, 95% CI 0.56–1.38; P=0.6) and reintervention rates (IRR 1.35, 95% CI 0.57–3.21; P=0.5) did not differ between the two methods of repair. An extension analysis of perioperative outcomes in those 70 years of age or younger showed similar results (Figure 2).

Figure 1. Propensity-Weighted Perioperative Outcomes.

Propensity-weighted perioperative outcomes for the 65 and under primary cohort.

Figure 2. Propensity weighted perioperative outcomes, 70-and-under subcohort.

Propensity weighted perioperative outcomes in the subgroup analysis of those 70 years of age and under.

DISCUSSION

Large nonrandomized contemporary studies of AAA repair have suffered from ascertainment bias due to the inherent incomparability of real-world open surgical and endovascular cohorts. This is often unavoidable due to the necessary tradeoff between granularity and sample size in large administrative databases^6,8. In these situations, open surgical and endovascular repair groups may not be comparable as open surgical candidates are likely patients who have unfavorable aortic anatomy for EVAR. Therefore, perioperative morbidity and mortality figures in these comparison studies will unfairly demonstrate higher rates for open surgery if greater anatomic or procedural complexity is undocumented or otherwise unaccounted for. Regardless of the comparability of the open repair and EVAR cohorts, it is unlikely that open surgery would result in lower morbidity or mortality rates than that achieved with EVAR simply due to the nature of the two procedures. Pivotal randomized trials comparing open to EVAR in fit individuals have consistently demonstrated reductions in perioperative mortality for EVAR compared to open (OR 0.33 [95% CI 0.2 – 0.55])¹¹. These trials have studied these interventions over a variety of age ranges. However, recent studies have suggested that highly selected populations such as younger patients or those with few comorbidities may do initially as well with open surgery as with EVAR¹². While EVAR has been employed for over two decades, early endografts have shown a propensity toward device related complications such as material fatigue and migration in extended follow up¹³. The current generation of endografts have only been in use for the past decade; long-term durability of these devices remains unknown, but have a clear dependence on adherence to the device indications for use (IFU)¹⁴. As such, young or fit patients who have equivalent morbidity and mortality outcomes with open or endovascular repair may warrant further long-term comparative study during which endograft failure and persistent risk of rupture may prove to be significant.

Our study sought to examine perioperative results in younger patients. Although various cutpoints have been suggested for defining “younger” patients, the majority of previous studies have identified those 65 years of age or less as “younger”¹²; in addition to this cutpoint, we extended this study in a subgroup analysis of those under the age of 70 years. In order to create two cohorts of comparable patients, we excluded those with open pararenal or thoracoabdominal repairs, EVAR patients deemed medically unfit for open repair as designated by the surgeon, and those undergoing EVAR for isolated iliac aneurysms.

Both our unadjusted and propensity-weighted results indicate that the procedural mortality of EVAR and open repair are fairly comparable. Although EVAR was statistically superior compared to open surgery, the actual percentage of perioperative mortality in both cohorts was very low at under one percent, under the practical threshold for determining high-risk operations¹⁵. Classical randomized studies of EVAR in the general population have shown significant perioperative advantages for EVAR compared to open surgery in all comers¹⁶. Initial results from the OVER trial¹⁷ demonstrated an absolute risk reduction in perioperative mortality of 2.5% (3% open vs. 0.5% EVAR), an advantage which persisted up to two years. Likewise, the United Kingdom-based EVAR-1 trial¹⁸ showed a 3.5% absolute risk reduction in perioperative mortality (4.3% open vs 1.8% EVAR). By focusing our study only on a younger patient cohort, we demonstrate both a lower perioperative mortality overall for both EVAR and open surgery as well as a lower absolute risk reduction (0.7% in adjusted comparisons) compared to these trial benchmarks.

Large-scale population subgroup analyses focusing on younger patients have identified similar advantages for EVAR with perioperative mortality rates well under one percent^3,6 and perioperative mortality for open repair ranging from 1.2%³ to 2.4%⁶. These mortality figures for open repair are higher than in our study but are likely due to the inability to exclude complex aneurysms such as those requiring suprarenal clamping. Rates of perioperative myocardial infarction were likewise statistically significant in favor of EVAR but less than one percent in both cohorts, and the rate of perioperative stroke was nonsignificant.

Similar but much smaller studies in select patient populations have also been carried out utilizing registry data from the Vascular Study Group of New England (VSGNE)⁷ and the National Surgical Quality Improvement Program (NSQIP)¹⁹, primarily in patients younger than 60 years of age. Neither of these studies could demonstrate statistically significant mortality differences between the two cohorts, likely due to small sample size (169 in the VSGNE and 651 in the NSQIP study). In contrast, the large sample size of our study allows for more certainty in determining incidence of perioperative outcomes and true differences between the two groups. Open AAA repair did, however, have a significantly higher rate of respiratory failure and renal failure requiring hemodialysis that reflect clinically significantly differences. This is not surprising given that open repair requires a laparotomy and manipulation of the pararenal aorta. However, this should figure prominently into any surgeon’s risk-benefit analysis given the morbidity carried by respiratory failure and prolonged intubation²⁰.

The increased need for aortic-related reintervention after EVAR as opposed to open repair has been suggested based on both randomized^17,21 and large cohort data^6,8 for both minor and major aortic-related reinterventions. In addition, the mortality benefit of EVAR over open repair has been shown to decrease over time at least in the general population. At one year in our study, survival rates were almost identical between the cohorts; however, the difficulty of tracking true mortality and large percentage of loss to follow-up in VQI data makes assessment of outcomes outside of the immediate perioperative period difficult.

This suggests that further study is needed to define concrete benefits related to the use of EVAR in younger patients with longer lifespans. Recent studies have suggested that patients with less anatomically favorable criteria for EVAR may be more likely to experience more long-term complications compared to those with favorable aneurysm anatomy^14,22, although solid evidence to support this conclusion remains limited and conflicting²³. Studying younger patients within this “off-IFU” cohort with unfavorable aneurysm anatomy for EVAR may identify a population potentially benefitting from open repair.

It is worth noting that perioperative outcomes such as mortality may not be the only driver for decisions for the type of aneurysm repair in younger patients. Other factors such as short-term quality of life and return to work may figure significantly in these decisions. Further research is needed in this area and in this specific population, as this type of data is not available in many clinically-focused studies such as ours.

Our study does have some limitations which must be considered when interpreting these results. The VQI is a voluntary registry, with the potential for bias arising from both the self-reported nature of the data as well as selection bias from the centers participating in the registry, as centers participating in the registry may have systematically different treatment algorithms, patient populations, or disease characteristics than those not participating. Despite our best attempts to adjust for differences between the two cohorts using all of the available information, first by excluding complex aneurysms and then using propensity methods, the absence of anatomical data or other unmeasured confounders may impact the comparability of the two groups. Analysis of the data is also complicated by significant amounts of missing data for one-year follow-up due to the voluntary submission nature of the database. The nature of the improvement of endovascular technology and techniques over the ten-year collection period must also be accounted for when interpreting the results of these studies. Finally, the one-year follow-up period is a very limited window into the durability of either open or EVAR and cannot reliably determine long-term aneurysm behavior – further long-term study is necessary to determine the durability of these interventions in this younger patient population.

CONCLUSIONS

Though EVAR has largely supplanted open AAA repair as the standard treatment modality of choice due to favorable morbidity and mortality profiles, the long-term durability of these devices remains unclear. Our study demonstrates that open repair carries very similar perioperative mortality risks compared to EVAR in highly select populations such as those of younger age, although with a concomitantly higher rate of respiratory complications. Further study is needed regarding the long-term mortality and aortic-related reintervention rates for both treatment paradigms to identify populations in which open AAA repair may provide a durable long-term alternative to EVAR.