Defining high-risk patients for endovascular aneurysm repair

Article Outline

• Abstract
• Materials and methods
  • Data sources and study population
  • Statistical analysis
• Results
  • Risk factors: univariate analysis
  • Multivariable model
  • Risk score
• Discussion
• Author contributions
• Appendix (online only)
• References
• Copyright

Background

Endovascular aneurysm repair (EVAR) is commonly used as a minimally invasive technique for repairing infrarenal aortic aneurysms. There have been recent concerns that a subset of high-risk patients experience unfavorable outcomes with this intervention. To determine whether such a high-risk cohort exists and to identify the characteristics of these patients, we analyzed the outcomes of Medicare patients treated with EVAR from 2000-2006.

Methods

We identified 66,943 patients who underwent EVAR from Inpatient Medicare database. The overall 30-day mortality was 1.6%. A risk model for perioperative mortality was developed by randomly selecting 44,630 patients; the other one third of the dataset was used to validate the model. The model was deemed reliable (Hosmer-Lemeshow statistics were P = .25 for the development, P = .24 for the validation model) and accurate (c = 0.735 and c = 0.731 for the development and the validation model, respectively).

Results

In our scoring system, where scores ranged between 1 and 7, the following were identified as significant baseline factors that predict mortality: renal failure with dialysis (score = 7); renal failure without dialysis (score = 3); clinically significant lower extremity ischemia (score = 5); patient age ≥85 years (score = 3), 75-84 years (score = 2), 70-74 years (score = 1); heart failure (score = 3); chronic liver disease (score = 3); female gender (score = 2); neurological disorders (score = 2); chronic pulmonary disease (score = 2); surgeon experience in EVAR <3 procedures (score = 1); and hospital annual volume in EVAR <7 procedures (score = 1). The majority of Medicare patients who were treated (96.6%, n = 64,651) had a score of 9 or less, which correlated with a mortality <5%. Only 3.4% of patients had a mortality ≥5% and 0.8% of patients (n = 509) had a score of 13 or higher, which correlated with a mortality >10%.

Conclusion

We conclude that there is a high-risk cohort of patients that should not be treated with EVAR because of prohibitively high mortality; however, this cohort is small. Our scoring system, which is based on patient and institutional factors, provides criteria that can be easily used by clinicians to quantify perioperative risk for EVAR candidates.

With a greater awareness through the liberal use of cross-sectional imaging and enhanced screening efforts, abdominal aortic aneurysms (AAA) are being identified with increasing frequency.¹, ² With multiple comorbidities that are associated with an increased risk of intervention, this population presents a unique challenge to vascular interventionalists. Endovascular repair of AAA (EVAR) was first introduced in 1991 by Parodi et al.³ Fifteen years later, it appears that EVAR will soon become the predominant method of AAA repair.⁴ Because it is minimally invasive, EVAR potentially holds great advantage for high-risk patients with multiple comorbidities. The procedure does not require general anesthesia or intensive care unit (ICU) admission postoperatively. Additionally, EVAR requires only femoral artery exposure, eliminating the need for a laparotomy and its complications. There is decreased blood loss compared with open repair, and the major perioperative intravenous fluid shifts observed with open repair are avoided. It has been previously demonstrated that the perioperative morbidity and mortality associated with endovascular repair approaches one fourth that of traditional open surgery.⁴

These advantages suggest that a broad spectrum of AAA patients should be appropriate candidates for the endovascular approach. However, the concept that EVAR is applicable to all patients regardless of the severity of their comorbidities has recently been challenged. In the EVAR 2 trial,⁵ Greenlaugh and coauthors identified a cohort of patients “unfit for open repair” and randomized these patients to EVAR versus medical treatment. These investigators found that there was no significant difference in all-cause mortality between medical treatment and EVAR, with EVAR patients having a perioperative mortality of 9%. These authors concluded that for high-risk AAA patients, no surgical intervention is warranted. Despite valuable insights from EVAR 2, there are unanswered questions: how large is the subset of patients who are high risk for EVAR and what are the preoperative characteristics that can identify them?

Several studies have reported risk stratification paradigms for open AAA repair.⁶, ⁷, ⁸, ⁹, ¹⁰ Subgroups of patients, who are at high or even prohibitively high risk for conventional AAA repair, have been identified. Variables that have been commonly defined as pre-operative risk factors for mortality include: increased age, congestive heart failure (CHF), myocardial ischemia, and renal and pulmonary dysfunction.⁶, ⁷, ⁹, ¹¹ By comparison, such risk factors have not been identified for patients undergoing EVAR. Specification of these risk factors is essential to delineate those patients at prohibitive risk for EVAR who would benefit from medical therapy alone. Thus, to better understand the mortality associated with EVAR and, more importantly, to define patients at excessively high risk for this “minimally invasive” procedure, we analyzed the outcomes of Medicare patients treated with EVAR between 2000 and 2006.

Back to Article Outline

Materials and methods 

Data sources and study population 

We used the Medicare Inpatient Standard Analytical file (Medicare part A) to identify hospitalized patients who underwent EVAR between 2000 and 2006. These files contain hospital-discharge abstracts on 100% of Medicare-reimbursed hospitalizations, except for those beneficiaries enrolled in Medicare HMOs (approximately 10% of patients). The data were supplemented with the Medicare Denominator file, which contains demographic, geographic, and vital status data. The data were obtained from the Centers for Medicare and Medicaid Services (CMS).

Patients who underwent AAA repair were identified through a combination of the International Classification of Diseases, Clinical Modification (ICD-9-CM) diagnosis code, 441.4 (aortic abdominal aneurysm without mention of rupture) in the primary or any secondary position, plus the primary or any secondary ICD-9-CM procedure code, 39.71 (endovascular implantation of graft in abdominal aorta). If a patient had multiple AAA repairs, the first AAA procedure was included in the analysis. Only patients with elective admissions were included in this study. To ensure that all comorbidities identified at prior hospitalizations were not missed, we used a longer time period (1995-2006) to define the comorbidities of patients who were ultimately treated with EVAR.

It is often difficult to differentiate between pre-existing comorbidities and postoperative complications (eg, stroke) in large datasets. To address this weakness, we included a diagnosis as a comorbidity if 1) it was present on a previous hospital admission or 2) if it appeared during the index hospitalization and was coded as a chronic or “acute on chronic” disorder. The following comorbidities were assessed (primary and all secondary diagnoses): cardiac disease (coronary artery disease, congestive heart failure [CHF], valvular heart disease, cardiac arrhythmias), diabetes, chronic pulmonary disease, peripheral arterial disease (clinically significant lower extremity ischemia, vascular insufficiency of the intestine, and renal atherosclerosis), renal disease, neurological disorders (cerebrovascular, paralysis, and other neurological diseases), cancer, rheumatoid arthritis, and liver disease. The list of ICD-9 diagnosis codes for comorbidities is provided in the Appendix (online only). The annual hospital volume (number of EVAR/year) and cumulative physician experience with EVAR at the time of the procedure were used to develop a relationship between EVAR volume/experience and outcome. All endovascular repairs (elective and ruptured) were included in calculations of hospital volume and surgeon experience.

Statistical analysis 

To construct a risk model for perioperative mortality after EVAR, all patients were randomly allocated to a dataset for model development (the training set; n = 44,630, 2/3 of cohort) and a dataset for model validation (the test set; n = 22,313, 1/3 of the cohort). In deriving the model, we first analyzed the univariate associations between the independent variables (patient demographics, baseline comorbidities, hospital volume, and surgeon EVAR experience) and 30-day mortality. Continuous variables (age, hospital volume, and surgeon experience) were transformed into categorical variables. Hospitals annual volume and surgeon experience were categorized into 10 groups (deciles) with approximately equal distribution of patients between groups. Patients less than 65 years of age were excluded from the analysis to avoid some confounding issues due to their disability as a criterion for Medicare eligibility. Five-year increments were used for age groupings. A Chi-square test was used to assess the association between potential risk factors and mortality. Variables with a level of significance (P value) < .25 were included in a logistic regression analysis. This multivariable regression model examines dichotomous outcomes (dead/alive), and their associated risk factors. Only variables with P value ≤ .05 were included in the final model. The interpretation of a risk factor included in the final model is that it is independently associated with the event, controlling for other significant covariates, and all risk factors jointly predict the event. Interactions between significant predictors and age, gender, and race/ethnicity were also tested. The diagnostic properties of the training model were then tested using the validation dataset. The area under the receiver operator curve (c statistic) was calculated as a measure of discrimination or predictive ability. A value of 1 indicates perfect discrimination. Calibration of the model (statistical precision) was assessed by the Hosmer-Lemeshow goodness-of-fit statistic. This statistic compares observed number of patients with expected, derived by logistic model. A P value for goodness of fit greater than 0.05 indicates that there is no statistical difference between observed and expected numbers and that the model has a high predictive ability. Risk factors were derived from the training model and verified on validation dataset.

The regression coefficients of the risk factors were used to develop a scoring system to predict 30-day mortality after EVAR. Regression coefficients were multiplied by a scaling factor and then rounded to the nearest integer.¹² The total risk score of a patient was the sum of the scores for each individual risk factor. Additional logistic regression model was constructed to evaluate the relationship between total risk score and mortality. The model was created using training dataset and was validated on test dataset. The 30-day mortality associated with total risk score was the average risk among all patients having the same total score. The accuracy of our scoring system was tested by comparing the predicted mortality associated with each risk score with the observed mortality on the validation dataset. All statistical analyses were performed using the SAS system software version 9.1 (SAS Institute Inc., Cary, NC).

Back to Article Outline

Results 

Risk factors: univariate analysis 

We identified 66,943 patients age 65 or older who underwent EVAR between 2000 and 2006. Since Medicare dataset provides date of surgery, we were able to identify conversion cases. We excluded 48 patients who had an open AAA repair prior to an EVAR procedure during the same hospitalization. However, we retained all conversions from EVAR to open repair in the dataset. In terms of demographics, 8.9% of the study population was 85 years of age or older, 82.9% were males, and 94.8% were Caucasians (Table I). The overall 30-day mortality was 1.6%. The 30-day mortality after EVAR among females was higher than among males (2.5% vs. 1.4%; P < .0001). Perioperative mortality also increased with patient age; this became statistically significant for patients 70 years of age or older (Table I).

Table I. Patient demographics and 30-day mortality by demographic groups (n = 66,943 patients)

Variable	No. of patients	% of cohort	Mortality (%)	P value
Age, years
65-69	12,046	17.99	0.9	(reference group)
70-74	16,994	25.39	1.2	0.04
75-79	18,624	27.82	1.6	<0.0001
80-84	13,311	19.88	2.0	<0.0001
≥85	5,968	8.92	3.2	<0.0001
Male	55,485	82.88	1.4	(reference group)
Female	11,458	17.12	2.5	<0.0001
Whites	63,492	94.84	1.6	(reference group)
Blacks	1,847	2.76	1.9	0.29
Hispanics	354	0.53	2.0	0.55
Native Americans	118	0.17	1.7	0.92
Other races	830	1.24	0.7	0.05

A number of baseline comorbidities were associated with 30-day mortality as shown in Table II. Common risk factors included chronic pulmonary diseases (37.1% of cohort, mortality 2.0%, P < .0001), cardiac arrhythmia (25.2%, mortality 2.3%, P < .0001), and heart failure (14.4%, mortality 3.5%, P < .0001). Patients with renal failure with dialysis represented only 1.1% of the cohort; however, their risk of dying after EVAR was highest (11.8%, P < .0001). Another less common risk factor strongly associated with mortality, was clinically significant lower extremity ischemia (2.1% of cohort, mortality 6.2%, P < .0001).

Table II. Comorbidities and their association with 30-day mortality (N = 66,943 patients)

Comorbidity	No. of patients	% of cohort	Mortality (%)	P value	Odds ratio
Renal failure w/dialysis	718	1.07	11.8	<.0001	9.01[7.12-11.39]
Renal failure w/o dialysis	2,554	3.82	3.8	<.0001	2.64[2.13-3.26]
PAD	4,855	7.25	3.1	<.0001	2.17[1.82-2.58]
LE ischemia	1,414	2.11	6.2	<.0001	4.42[3.52-5.53]
Vascular intestine	141	0.21	2.8	0.23	1.82[0.67-4.93]
Renal atherosclerosis	3,409	5.09	1.9	0.09	1.24[0.97-1.60]
Heart failure	9,644	14.41	3.5	<.0001	2.88[2.53-3.28]
Neurological disorders	7,494	11.19	2.4	<.0001	1.63[1.39-1.92]
Cerebrovascular and/or paralysis	5,282	7.89	2.1	0.0016	1.38[1.13-1.68]
Other neurological	2,596	3.88	3.3	<.0001	2.23[1.79-2.80]
Liver disease	728	1.09	3.2	0.0006	2.05[1.35-3.13]
Cardiac arrhythmia	16,840	25.16	2.3	<.0001	1.78[1.57-2.01]
Rheumatoid arthritis	1,311	1.96	2.3	0.04	1.47[1.02-2.12]
Valvular disease	6,364	9.51	2.2	<.0001	1.44[1.20-1.72]
Chronic pulmonary	24,854	37.13	2.0	<.0001	1.49[1.32-1.68]
Atherosclerosis	5,378	8.03	1.7	0.49	1.08[0.87-1.34]
Cancer	5,135	7.67	1.7	0.57	1.07[0.85-1.33]
Diabetes	11,013	16.45	1.6	0.80	0.98[0.83-1.15]
Coronary disease	36,664	54.77	1.5	0.03	0.88[0.78-0.99]

LE, Lower extremity; PAD, Peripheral arterial disease.

Finally, we evaluated the relationship between hospital annual volume and surgeon cumulative experience with EVAR and perioperative mortality (Table III). Mortality declined from 2.3% to 1.4%, with growing hospital annual volume from less than seven procedures versus volume greater than 73 EVARs (Table III). Thirty-day mortality, when EVAR was performed by surgeons with total experience of ≤ 2 procedures, was 2.4%, whereas the mortality was in the range of 1.3% to 1.6% for surgeons with a cumulative EVAR experience ≥ 3 procedures.

Table III. Annual hospital volume and cumulative surgeon experience over the study period and their association with 30-day mortality (n = 66,943 patients)

Variable	No. of patients	% of cohort	Mortality (%)	P value
Hospital EVAR Volume (Annual, deciles)
1-6	7,924	11.84	2.3	.0001
7-10	7,011	10.47	1.6	.33
11-14	6,649	9.93	1.8	.06
15-18	5,940	8.87	1.5	.85
19-23	6,571	9.82	1.4	.91
24-30	6,790	10.14	1.6	.39
31-37	6,126	9.15	1.4	.86
38-49	6,789	10.14	1.2	.34
50-73	6,469	9.66	1.5	.83
74+	6,674	9.97	1.4	reference
Surgeon's EVAR experience (cumulative, deciles)
1-2	7,895	11.79	2.4	<.0001
3-5	7,539	11.26	1.6	.16
6-8	5,844	8.73	1.8	.02
9-12	6,168	9.21	1.5	.36
13-18	6,957	10.39	1.5	.37
19-25	6,104	9.11	1.5	.36
26-36	6,597	9.85	1.5	.26
37-54	6,728	10.05	1.4	.63
55-93	6,504	9.72	1.4	.48
94+	6,607	9.87	1.3	reference

EVAR, Endovascular aneurysm repair.

Multivariable model 

In a multivariable regression model, the following baseline comorbidities predicted 30-day mortality after EVAR: renal failure with dialysis (odds ratio [OR] = 7.06, P < .0001) and without dialysis (OR = 1.91, P < .0001), clinically significant lower extremity ischemia (OR = 3.55, P < .0001), liver disease (OR = 2.52, P < .0001), CHF (OR = 2.23, P < .0001), neurological disorders (OR = 1.59, P < .0001), and chronic pulmonary diseases (OR = 1.57, P < .0001) (Table IV). The risk of death after EVAR was 68% higher for females versus males (OR = 1.68, P < .0001) and increases with patient age: OR = 1.40 for patients 75-79 years of age to OR = 3.10 for patients ≥85 years of age, controlling for comorbidities, gender, hospital volume, and surgeons experience. Hospital volume (< 7 EVARs per year) remained in the model as a predictor of death after surgery, as did surgeon experience of < 3 EVAR procedures at the time of the index operation.

Table IV. Statistically significant predictors of 30-day mortality after EVAR AAA (based on the results of multivariable logistic regression model, concordance index = 0.735, Hosmer-Lemeshow goodness of fit test P = .25)

Risk factor	Parameter	Odds ratio and 95% CL	P value
Renal failure w/ dialysis	1.95	7.06[5.23-9.53]	<.0001
LE ischemia	1.27	3.55[2.65-4.75]	<.0001
Age≥85 years	1.13	3.10[1.57-2.37]	<.0001
Liver disease	0.93	2.52[1.54-4.12]	.0002
CHF	0.80	2.23[1.89-2.64]	<.0001
Renal failure w/o dialysis	0.65	1.91[1.45-2.51]	<.0001
Age 80-84 years	0.65	1.92[1.56-2.36]	<.0001
Female	0.52	1.68[1.42-1.99]	<.0001
Neurological	0.45	1.59[1.29-1.94]	.0001
Chronic pulmonary	0.45	1.57[1.35-1.83]	<.0001
Hospital annual vol <7	0.37	1.45[1.18-1.80]	.0005
Age 75-79 years	0.34	1.40[1.14-1.71]	0.001
Surgeon EVAR vol <3	0.26	1.30[1.04-1.62]	.002

CHF, Congestive heart failure; EVAR, endovascular aneurysm repair; LE, lower extremity.

Using the receiver operating curve characteristics, we found that the c-indices were 0.735 for the training set and 0.731 for the test set, indicating the robust predictive ability of these models. The Hosmer-Lemeshow goodness of fit statistics (comparison of observed and expected deaths) were 0.25 and 0.24 for training and test datasets respectively, indicating good statistical precision of the models.

Risk score 

Table V depicts risk scores for every statistically significant risk factor. Risk scores ranged from a minimum of one point for chronic pulmonary disorders to a maximum of seven points for renal failure with dialysis. The total risk score was obtained by summing individual risk points. The regression model that evaluated the relationship between total risk score and 30-day mortality was deemed reliable (Hosmer-Lemeshow statistics was P = .06 for the development, P = .83 for the validation model) and accurate (c = 0.73 and c = 0.70 for the development and the validation model, respectively).

Table V. Risk scores for 30-day mortality for EVAR patients

Risk factor	Score
Renal failure w/dialysis	7
LE ischemia	5
Age≥85 years	4
Liver disease	3
CHF	3
Renal failure w/o dialysis	3
Age 80-84 years	2
Female	2
Neurological	2
Chronic pulmonary	1
Surgeon EVAR experience <3	1
Hospital annual volume <7	1
Age 75-79 years	1

CHF, Congestive heart failure; EVAR, endovascular aneurysm repair; LE, lower extremity.

The relationship between predicted 30-day mortality after EVAR and patients' total risk score is presented in Table VI. The estimated mortality ranged from 0.5% to 38.4% for risk scores that ranged from 0 to 20. We then evaluated the agreement between predicted and observed mortality by risk score (Fig 1). The correlation between observed mortality (test dataset) and expected mortality (training dataset) using this model was very strong; r² = 0.83 (P < .0001).

Table VI. Predicted mortality based on scoring system

• 
  • View Large Image
  • Download to PowerPoint

• Fig 1. 

  Relationship between observed and predicted mortality by total score. Predicted mortality was estimated based on logistic regression model of two thirds of the cohort (development sample). Observed mortality was depicted from the remaining one third of the cohort (test sample). Coefficient of correlation between observed and predicted mortality r² = 0.8294. Number of observations in the test sample by score: 1 – 3229, 2 – 2497, 3 – 4051, 4 – 2573, 5 – 2135, 6 – 1555, 7 – 1104, 8 – 800, 9 – 475, 10 – 312, 11 – 186, 12 – 130, 13 – 69, 14 – 24, 15 – 31, 16 – 15, 17 – 11, 18 – 2, 19 – 0, 20 – 3.

The distribution of patients by risk score is shown in Fig 2: 96.6% (n = 64,651) of patients had a score of nine or less, which correlated with a mortality of less than 5%; 3.4% of patients (n = 2,292) had a score >9 and a mortality greater than 5%. Only 0.8% of patients (n = 509) had a score of 13 or higher, which correlated with a mortality of greater than 10%.

• 
  • View Large Image
  • Download to PowerPoint

• Fig 2. 

  Distribution of patients by risk scores.

Back to Article Outline

Discussion 

Numerous studies have compared the outcomes of open repair with EVAR, and the benefits of EVAR in terms of more intermediate outcomes have been well documented.⁶, ¹³, ¹⁴, ¹⁵, ¹⁶, ¹⁷, ¹⁸, ¹⁹, ²⁰ However, concerns have arisen as to whether EVAR is a sufficiently low-risk procedure that it can be used safely in all patients with AAA >5.5 cm. Greenlaugh and colleagues addressed this question with the EVAR trial 2. They identified a perioperative mortality of 9% in a population of patients with large aneurysms “unfit” for open surgery repaired with endovascular techniques.⁵ These authors were the first to recognize and report the limitations of EVAR and raise the notion that this technique should not be used in all high-risk patients with large aneurysms. However, the definitions used for high risk in EVAR trial 2 remain somewhat elusive. Guidelines for determining patient enrollment in EVAR trial 2 have been published;²¹ however, “physician discretion” was also used in determining which patients were ultimately eligible for this trial. The insights provided by EVAR 2 are important. There is most certainly a population of patients with large aneurysms that are better treated medically than with surgical intervention. However, questions still remain about the risk factors that predict mortality in patients undergoing EVAR and the size of the population of patients that are truly high risk.

Sicard et al retrospectively analyzed data from five multi-center EVAR clinical trials to further characterize outcomes after EVAR.²² These authors reported a 30-day mortality of 2.9% in a population of 565 patients that they defined as high risk, based upon criteria derived from EVAR trial 2 that included one or more of the following comorbidities: severe valvular disease, significant arrhythmia, uncontrolled CHF, dyspnea with stair climbing, poor pulmonary function, hypoxemia, hypercarbnia, or a serum creatinine > 2.27mg/dL.²¹ The mortality observed by Sicard et al was dramatically less than that found in EVAR 2 (2% versus 9%). Although there are several possible explanations for the dramatic difference in findings of the two studies, the most likely is that the “high risk” population defined by Sicard is indeed different than the high-risk cohort studied in EVAR trial 2. Patients recruited into clinical trials are usually homogeneous and patients with poor longevity or those at extremely high risk are often excluded from pivotal investigations. There are also exclusion criteria in clinical trials that eliminate patients with unfavorable arterial anatomy. It has previously been demonstrated that endovascular repair in patients with favorable anatomy is less risky. Despite apparent differences in mortality outcomes in these two studies, one can conclude from both that there is indeed a cohort of patients who are high risk for EVAR.

Our analysis further addresses this issue by providing information about the factors that define patients at high risk for EVAR. As well, we have provided insight into the size of the cohort that has a prohibitively high mortality. Of the almost 67,000 patients evaluated in this study, a risk of perioperative mortality of 9% or greater (the EVAR2 outcome) was found in only 1.3% of the treated population. We have also identified preoperative characteristics that can determine this small but high-risk cohort. We found that renal failure, lower extremity vascular disease, liver disease, neurological disorders, female gender, age, hospital volume, surgeon experience, heart failure, and chronic pulmonary diseases all increased the potential of death within 30 days following EVAR.

Multiple similar analyses have been performed for patients undergoing open aneurysm repair.⁷, ⁸, ⁹, ¹⁰ Although the demographic factors and comorbidities that increase mortality are similar for open and endovascular repair, their relative importance appears to differ. For EVAR, we found risk factors in descending order of importance to be: renal failure with dialysis, lower extremity ischemia, age ≥85 years, liver disease, CHF, renal failure without dialysis, female gender, a neurological disorder, chronic obstructive pulmonary disease, and low hospital volume and surgeon experience with EVAR. For open repair, renal failure leads the list (similar to EVAR), but is followed by myocardial disorders, such as ischemia and CHF, then pulmonary disease, age, and female gender.¹⁰ One might predict that major heart and/or lung disease is of less relevance as a risk factor for endovascular aneurysm repair versus open repair and this appears to be the case. Perhaps it is not surprising that CHF or pulmonary disease are less important predictors of death in EVAR since the surgical intervention (groin cut-downs) is associated with a less profound physiologic demand on the heart and lungs.²³ In fact, the factors that lead to mortality following EVAR may be related more to complex arterial anatomy than to complex patient physiology.

Our multivariate analysis revealed that patients at highest risk are those with renal failure. Findings from the EUROSTAR registry were similar.²⁴ The increase in mortality associated with renal disease is possibly due to the high prevalence of multifocal atherosclerosis in these patients, including the heart and cerebrovascular circulation. Also of great significance is the strong association between renal failure and calcified and diseased iliac arteries.²⁵, ²⁶ When performing an EVAR in a patient with renal failure, the interventionalist may be faced with heavily calcified, tortuous, and narrowed iliac arteries that are difficult to navigate with an endovascular device. The consequence can be arterial rupture, occlusion, and the need for a conduit or a prolonged intervention. Thus, renal failure may be a surrogate for complex arterial anatomy. Unfortunately, one of the limitations of large datasets, such as Medicare, is the absence of information about anatomy; therefore, we are unable to verify this hypothesis.

A number of additional risk factors predicted perioperative mortality. Patients with lower extremity vascular disease are at increased risk presumably for the same reason as those with renal failure. Lower extremity vascular disease is also a marker for generalized atherosclerosis, including myocardial insufficiency. Patients with chronic liver disease experience greater morbidity and mortality following most elective surgeries.²⁷ In two small prospective studies, the influence of gender on outcome of EVAR was evaluated and no differences were found between men and women with respect to 30-day mortality.²⁸, ²⁹ In both studies, however, it was noted that women have a significantly higher rate of aborted procedures, less deployment success, and an increased risk of access-related complications. The lack of an association between gender and mortality in these smaller studies may be due to the small sample size, (n = 26)²⁹ and (n = 24).²⁸ The increased mortality that we observed in women may be largely related to anatomic issues. Anatomical characteristics inherent to women include shorter infrarenal necks, smaller proximal neck diameters, and smaller diameters of iliac (access) arteries.²⁹

Neurological disorders, including a prior history of cerebrovascular accident and transient ischemic attack, were present in 11.2% of our cohort, and were found to increase 30-day mortality by 59% in our multivariable regression model. Cerebrovascular disease has been observed to increase peri-procedural complications and mortality, and has, therefore, been included in several preoperative scoring systems, including the Revised Cardiac Risk Index,³⁰ Glasgow aneurysm score,³¹ and the Customized Probability Model.³² Additionally, we found surgeons just beginning their EVAR practice, as well as hospitals with lower annual volumes of EVARs, to have significantly higher perioperative mortality. We are not able to determine if surgeons at the early phase of their experience with EVAR are in mentored situations such as a group practice or an academic medical center. Possibly this might explain the low number of procedures necessary to gain expertise. Such volume-outcome relationships have long been recognized for open aneurysm repair³³ and it should be noted that despite the same opportunity for mentoring with open repair, the number of procedures necessary to achieve proficiency is significantly higher than for endovascular repair. We have composed a scoring system that can be used to assist interventionalists and patients. Assessing the surgical risk of a patient with multiple comorbidities can be remarkably difficult, yet it is these particular patients who benefit most from an accurate preoperative evaluation, as they are likely to have increased early and late mortality as a consequence of their associated illness. Preoperative risk stratification for noncardiac vascular surgery has been investigated by others and validated, using mathematical models to derive scoring systems and predict mortality.¹⁰, ³⁰, ³¹, ³², ³⁵, ³⁶, ³⁷ Analysis of a large Medicare dataset provides sufficient statistical power to accurately identify specific individual criteria predictive of 30-day mortality for EVAR. Our scoring system allows a comparison of the impact of individual factors on mortality, and importantly, a summation of their combined effects. It is our hope that this scoring system can be used by the practicing interventionalist to identify those who are indeed candidates for EVAR. For example, a patient with CHF and chronic pulmonary disease would be at relatively low risk for repair (total score of four). Alternatively, a female patient with renal failure would be of profoundly high risk (total score of nine). We realize that scoring systems have their limitations, and in the real world clinicians need to individualize therapeutic decisions for patients. However, the scoring system that we have devised does take into account the majority of pertinent risk factors and could potentially be used as a guide to assist clinicians in their evaluation of patients with aneurysmal disease. Admittedly, our scoring system does not directly assess the important effect that vascular anatomy may have on outcome.

Other risk stratification systems have been employed to predict outcome of endovascular aneurysm repair.³⁴, ³⁵ Patients included in a randomized trial such as Dutch Randomized Endovascular Aneurysm Management (DREAM)³⁶ are selected and often more homogenous than those treated in standard practice. For example, many of the very- high-risk patients that are included in the Medicare data base would not have passed screening criteria for a randomized trial. Moreover, in this analysis we have created a de novo scoring system from the data available rather than attempting to retrofit a scoring system previously designed for open repair for EVAR. That said, this is one of several proposed methodologies for risk-stratifying patients proposed for EVAR, and only future studies on new cohorts of patients will determine which of these systems has the greatest validity.

There are a number of limitations of administrative datasets that should be noted. First, knowledge of the severity of comorbidities is often lacking. Second, diagnosis codes are broad and vague and provide limited detail about the specific patient disease state. By accepting only comorbidities that are coded as chronic or that have been present on previous admissions, we may miss occasional comorbidites that have appeared between hospitalizations or that are not associated with a “chronic” code. By assuming this approach, we have likely increased our accuracy but may have also, to some extent, diminished our sensitivity. This is a common approach that is used in the evaluation of administrative data bases and overall the trade-off of accuracy for sensitivity is thought to be desirable. The third limitation, and possibly the most important for this analysis, is the lack of information regarding patient anatomy. We are not able to understand which of these 66,943 patients had diseased iliac arteries, nor do we know the size of the aneurysms treated. Lastly, as with all administrative datasets, there is the potential for coding inaccuracies and oversights. The effect of coding issues are likely diminished by the “randomization” of non-systematic errors that results when massive numbers of observations are statistically analyzed.³⁷ These limitations aside, the distinct advantage of administrative data bases such as Medicare, is the very large sample size. Our report of almost 67,000 EVARs is one of the largest ever published. This population-based dataset is rich in information regarding diagnoses, procedures, and demographics and is a true representation of clinical practice in the United States.

We determine high-risk EVAR patients by using a cohort of patients who have already undergone EVAR. This approach is frequently used to support clinician decision making on patient eligibility for various surgical procedures, including open repair of AAA.⁷, ⁹ However, it is important to note that clinical judgments made before patients receive EVAR may involved a different weighting of factors and may also take into account additional parameters versus those considered in this study. Our study stratifies operative risk only for patients already selected for EVAR. In the absence of compelling level I evidence (a randomized clinical trial of intervention versus no intervention), a retrospective analysis of surgical outcomes has proven to be a useful approach for operative risk stratification.

It should be noted that multiple studies have supported a considerably high incidence of death from rupture in high-risk AAA cohorts that are treated conservatively.²¹, ³⁸, ³⁹ Thus, a significant portion of the AAA population will likely die from rupture if they are excluded from endovascular repair. Our data reveal the existence of a subgroup of patients who are indeed high risk for endovascular repair. However, we also show that this cohort of patients is exceedingly small. In sum, we believe that EVAR is safe and effective in the majority of the elderly population, even those with multiple comorbidities. The proportion of patients truly unfit for EVAR is small. Moreover, we feel that the described scoring system can be a useful aid to preoperatively identify patients unfit for even minimally invasive treatment of their aneurysm.

Back to Article Outline

Author contributions 

Back to Article Outline

Appendix (online only) 

Appendix (online only). List of ICD-9-CM codes for comorbidities

Back to Article Outline

References 

1. Melton LJ, Bickerstaff LK, Hollier LH, Van Peenen HJ, Lie JT, Pairolero PC, et al. Changing incidence of abdominal aortic aneurysms: a population-based study. Am J Epidemiol. 1984;120:379–386
   • View In Article
   • MEDLINE
2. Schneider EL. Aging in the third millennium. Science. 1999;283:796–797
   • View In Article
   • MEDLINE
   • CrossRef
3. Parodi JC, Palmaz JC, Barone HD. Transfemoral intraluminal graft implantation for abdominal aortic aneurysms. Ann Vasc Surg. 1991;5:491–499
   • View In Article
   • Abstract
   • Full-Text PDF (2990 KB)
   • CrossRef
4. Anderson PL, Arons RR, Moskowitz AJ, Gelijns A, Magnell C, Faries PL, et al. A statewide experience with endovascular abdominal aortic aneurysm repair: rapid diffusion with excellent early results. J Vasc Surg. 2004;39:10–19
   • View In Article
   • Abstract
   • Full Text
   • Full-Text PDF (301 KB)
   • CrossRef
5. EVAR trial participants (Endovascular aneurysm repair and outcome in patients unfit for open repair of abdominal aortic aneurysm (EVAR trial 2): randomised controlled trial). Lancet. 2005;365:2187–2192
   • View In Article
   • Abstract
   • Full Text
   • Full-Text PDF (94 KB)
   • CrossRef
6. Brewster DC, Cronenwett JL, Hallett JW, Johnston KW, Krupski WC, Matsumura JSJoint Council of the American Association for Vascular SurgerySociety for Vascular Surgery. Guidelines for the treatment of abdominal aortic aneurysms (Report of a subcommittee of the Joint Council of the American Association for Vascular Surgery and Society for Vascular Surgery). J Vasc Surg. 2003;37:1106–1117
   • View In Article
   • Abstract
   • Full Text
   • Full-Text PDF (23 KB)
   • CrossRef
7. Dardik A, Lin JW, Gordon TA, Williams GM, Perler BA. Results of elective abdominal aortic aneurysm repair in the 1990s: A population-based analysis of 2335 cases. J Vasc Surg. 1999;30:985–995
   • View In Article
   • Abstract
   • Full Text
   • Full-Text PDF (85 KB)
   • CrossRef
8. Forbes TL, Steiner SH, Lawlor DK, DeRose G, Harris KA. Risk-adjusted analysis of outcomes following elective open abdominal aortic aneurysm repair. Ann Vasc Surg. 2005;19:142–148
   • View In Article
   • Abstract
   • Full Text
   • Full-Text PDF (686 KB)
   • CrossRef
9. Huber TS, Wang JG, Derrow AE, Dame DA, Ozaki CK, Zelenock GB, et al. Experience in the United States with intact abdominal aortic aneurysm repair. J Vasc Surg. 2001;33:304–310discussion 310-1
   • View In Article
   • Abstract
   • Full Text
   • Full-Text PDF (15 KB)
   • CrossRef
10. Steyerberg EW, Kievit J, de Mol Van Otterloo JC, van Bockel JH, Eijkemans MJ, Habbema JD. Perioperative mortality of elective abdominal aortic aneurysm surgery (A clinical prediction rule based on literature and individual patient data). Arch Intern Med. 1995;155:1998–2004
    • View In Article
    • MEDLINE
11. Kazmers A, Perkins AJ, Jacobs LA. Outcomes after abdominal aortic aneurysm repair in those > or = 80 years of age: recent Veterans Affairs experience. Ann Vasc Surg. 1998;12:106–112
    • View In Article
    • Abstract
    • Full-Text PDF (623 KB)
    • CrossRef
12. Sullivan LM, Massaro JM, D'Agostino RB. Presentation of multivariate data for clinical use: The Framingham Study risk score functions. Stat Med. 2004;23:1631–1660
    • View In Article
    • MEDLINE
    • CrossRef
13. Brewster DC, Geller SC, Kaufman JA, Cambria RP, Gertler JP, LaMuraglia GM, et al. Initial experience with endovascular aneurysm repair: comparison of early results with outcome of conventional open repair. J Vasc Surg. 1998;27:992–1003discussion 1004-5
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (4020 KB)
    • CrossRef
14. Chuter TA, Reilly LM, Faruqi RM, Kerlan RB, Sawhney R, Canto CJ, et al. Endovascular aneurysm repair in high-risk patients. J Vasc Surg. 2000;31(1 Pt 1):122–133
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (347 KB)
    • CrossRef
15. Finlayson SR, Birkmeyer JD, Fillinger MF, Cronenwett JL. Should endovascular surgery lower the threshold for repair of abdominal aortic aneurysms?. J Vasc Surg. 1999;29:973–985
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (139 KB)
    • CrossRef
16. Hua HT, Cambria RP, Chuang SK, Stoner MC, Kwolek CJ, Rowell KS, et al. Early outcomes of endovascular versus open abdominal aortic aneurysm repair in the National Surgical Quality Improvement Program-Private Sector (NSQIP-PS). J Vasc Surg. 2005;41:382–389
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (196 KB)
    • CrossRef
17. May J, White GH, Yu W, Ly CN, Waugh R, Stephen MS, et al. Concurrent comparison of endoluminal versus open repair in the treatment of abdominal aortic aneurysms: analysis of 303 patients by life table method. J Vasc Surg. 1998;27:213–220discussion 220-1
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (81 KB)
    • CrossRef
18. Minor ME, Ellozy S, Carroccio A, Oak J, Chae K, Agarwal G, et al. Endovascular aortic aneurysm repair in the octogenarian: is it worthwhile?. Arch Surg. 2004;139:308–314
    • View In Article
    • MEDLINE
    • CrossRef
19. Sicard GA, Rubin BG, Sanchez LA, Keller CA, Flye MW, Picus D, et al. Endoluminal graft repair for abdominal aortic aneurysms in high-risk patients and octogenarians: is it better than open repair?. Ann Surg. 2001;234:427–435discussion 435-7
    • View In Article
    • MEDLINE
    • CrossRef
20. Tefera G, Carr SC, Turnipseed WD. Endovascular aortic repair or minimal incision aortic surgery: Which procedure to choose for treatment of high-risk aneurysms?. Surgery. 2004;136:748–753
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (98 KB)
    • CrossRef
21. Brown LC, Epstein D, Manca A, Beard JD, Powell JT, Greenhalgh RM. The UK Endovascular Aneurysm Repair (EVAR) trials: design, methodology and progress. Eur J Vasc Endovasc Surg. 2004;27:372–381
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (781 KB)
    • CrossRef
22. Sicard GA, Zwolak RM, Sidawy AN, White RA, Siami FSSociety for Vascular Surgery Outcomes Committee. Endovascular abdominal aortic aneurysm repair: long-term outcome measures in patients at high-risk for open surgery. J Vasc Surg. 2006;44:229–236
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (205 KB)
    • CrossRef
23. Cuypers PW, Gardien M, Buth J, Charbon J, Peels CH, Hop W, et al. Cardiac response and complications during endovascular repair of abdominal aortic aneurysms: a concurrent comparison with open surgery. J Vasc Surg. 2001;33:353–360
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (82 KB)
    • CrossRef
24. Buth J, van Marrewijk CJ, Harris PL, Hop WC, Riambau V, Laheij RJ EUROSTAR Collaborators. Outcome of endovascular abdominal aortic aneurysm repair in patients with conditions considered unfit for an open procedure: a report on the EUROSTAR experience. J Vasc Surg. 2002;35:211–221
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (116 KB)
    • CrossRef
25. McCullough PA. Why is chronic kidney disease the “spoiler” for cardiovascular outcomes?. J Am Coll Cardiol. 2003;41:725–728
    • View In Article
    • Full Text
    • Full-Text PDF (158 KB)
    • CrossRef
26. Sarnak MJ, Levey AS, Schoolwerth AC, Coresh J, Culleton B, Hamm LL, et al. American Heart Association Councils on Kidney in Cardiovascular Disease, High Blood Pressure Research, Clinical Cardiology, and Epidemiology and Prevention Kidney disease as a risk factor for development of cardiovascular disease: a statement from the American Heart Association Councils on Kidney in Cardiovascular Disease, High Blood Pressure Research, Clinical Cardiology, and Epidemiology and Prevention. Circulation. 2003;108:2154–2169
    • View In Article
    • CrossRef
27. Friedman LS. The risk of surgery in patients with liver disease. Hepatology. 1999;29:1617–1623
    • View In Article
    • MEDLINE
    • CrossRef
28. Mathison M, Becker GJ, Katzen BT, Benenati JF, Zemel G, Powell A, et al. The influence of female gender on the outcome of endovascular abdominal aortic aneurysm repair. J Vasc Interv Radiol. 2001;12:1047–1051
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (119 KB)
    • CrossRef
29. Wolf YG, Arko FR, Hill BBOlcott C4th, Harris EJ, Fogarty TJ, Zarins CK. Gender differences in endovascular abdominal aortic aneurysm repair with the AneuRx stent graft. J Vasc Surg. 2002;35:882–886
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (142 KB)
    • CrossRef
30. Lee TH, Marcantonio ER, Mangione CM, Thomas EJ, Polanczyk CA, Cook EF, et al. Derivation and prospective validation of a simple index for prediction of cardiac risk of major noncardiac surgery. Circulation. 1999;100:1043–1049
    • View In Article
31. Samy AK, Murray G, MacBain G. Glasgow aneurysm score. Cardiovasc Surg. 1994;2:41–44
    • View In Article
    • MEDLINE
32. Kertai MD, Boersma E, Klein J, van Sambeek M, Schouten O, van Urk H, et al. Optimizing the prediction of perioperative mortality in vascular surgery by using a customized probability model. Arch Intern Med. 2005;165:898–904
    • View In Article
    • MEDLINE
    • CrossRef
33. Killeen SD, Andrews EJ, Redmond HP, Fulton GJ. Provider volume and outcomes for abdominal aortic aneurysm repair, carotid endarterectomy, and lower extremity revascularization procedures. J Vasc Surg. 2007;45:615–626
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (135 KB)
    • CrossRef
34. Patterson BO, Holt PJ, Hinchliffe R, Loftus IM, Thompson MM. Predicting risk in elective abdominal aortic aneurysm repair: a systematic review of current evidence. Eur J Vasc Endovasc Surg. 2008;36:637–645
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (193 KB)
    • CrossRef
35. Bohm N, Wales L, Dunckley M, Morgan R, Loftus I, Thompson M. Objective risk-scoring systems for repair of abdominal aortic aneurysms: applicability in endovascular repair?. Eur J Vasc Endovasc Surg. 2008;36:172–177
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (194 KB)
    • CrossRef
36. Prinssen M, Verhoeven EL, Buth J, Cuypers PW, van Sambeek MR, Balm R, et al. Dutch Randomized Endovascular Aneurysm Management (DREAM) Trial Group A randomized trial comparing conventional and endovascular repair of abdominal aortic aneurysms. N Engl J Med. 2004;351:1607–1618
    • View In Article
    • CrossRef
37. Lee WA, Carter JW, Upchurch G, Seeger JM, Huber TS. Perioperative outcomes after open and endovascular repair of intact abdominal aortic aneurysms in the United States during 2001. J Vasc Surg. 2004;39:491–496
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (116 KB)
    • CrossRef
38. Conway KP, Byrne J, Townsend M, Lane IF. Prognosis of patients turned down for conventional abdominal aortic aneurysm repair in the endovascular and sonographic era: Szilagyi revisited?. J Vasc Surg. 2001;33:752–757
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (78 KB)
    • CrossRef
39. O'Donnell TF, Darling RC, Linton RR. Is 80 years too old for aneurysmectomy?. Arch Surg. 1976;111:1250–1257
    • View In Article
    • MEDLINE