Clinical Research Open versus Endovascular Repair of Ruptured Abdominal Aortic Aneurysms Paul J. Speicher, Andrew S. Barbas, and Leila Mureebe, Durham, North Carolina

Background: Management of ruptured abdominal aortic aneurysms (rAAA) remains one of the most challenging situations in vascular surgery. The aim of our study was to examine national trends and compare contemporary outcomes of open and endovascular repair (EVAR) for rAAA across a wide spectrum of hospitals in the United States. Methods: Data from the 2005 to 2011 National Surgical Quality Improvement Project (NSQIP) Participant User File were used to identify patients with rAAA undergoing open or EVAR for this analysis. The primary outcome measures for our analysis were 30-day postoperative mortality, operative mortality, 30-day overall complication rate, and early return to the operating room. A nonparsimonious multiple logistic regression model was constructed to adjust for patient- and procedure-related confounding factors. As confirmatory analysis to address concerns of selection bias, we also conducted a separate propensity-matched analysis of the study cohort. Results: A total of 1,997 patients were identified who underwent repair of ruptured or dissecting AAA and met study inclusion criteria. Of these, 1,383 (69.3%) patients underwent an open procedure and 614 (30.7%) patients underwent EVAR. After adjusting for potential confounders, open repair had 1.7 times higher odds (95% confidence interval [CI] 1.24e2.35, P ¼ 0.001) of all-cause 30-day mortality; 2.06 times higher odds (95% CI 1.23e3.46, P ¼ 0.006) of operative mortality, and 1.9 times higher odds (95% CI 1.41e2.55, P < 0.001) of overall complications, compared with EVAR. Conclusions: EVAR is a safe and, in appropriately selected patients, superior approach to open surgery for the management of patients with rAAA. On multivariate analysis, patients who underwent open repair were at significantly increased risk of morbidity and mortality. In clinical settings where adequate resources, personnel, and surgical expertise are present, an endovascular approach should be strongly considered for all patients with acceptable anatomy.

INTRODUCTION Management of ruptured abdominal aortic aneurysms (rAAA) remains one of the most challenging situations in vascular surgery. These patients frequently present in extremis and rapid but judicious clinical decision-making is necessary. Historically, patients undergoing open repair of rAAA Department of Surgery, Duke University Medical Center, Durham, NC. Correspondence to: Leila Mureebe, Department of Surgery, Duke University Medical Center, DUMC Box 3467, Durham, NC 27710, USA; E-mail: [email protected] Ann Vasc Surg 2014; -: 1–9 http://dx.doi.org/10.1016/j.avsg.2013.12.025 Ó 2014 Elsevier Inc. All rights reserved. Manuscript received: October 30, 2013; manuscript accepted: December 16, 2013; published online: ---.

have experienced significant morbidity and mortality.1 With the continued evolution of endovascular devices and techniques over the last 20 years, there has been a significant increase in the frequency of endovascular approaches for rAAA. Multiple studies report significant improvements in perioperative morbidity and mortality with endovascular aneurysm repair (EVAR) compared with a traditional open approach.2e8 In light of these favorable outcomes, several centers with robust endovascular expertise have adopted an ‘‘EVAR-first’’ algorithm for managing patients with rAAA.7,9e12 While the advantages of the endovascular approach appear to be substantial, critics raise the question of patient selection bias in many of these studies. The decision-making processes used to determine which patients undergo open versus 1

2 Speicher et al.

Annals of Vascular Surgery

Fig. 1. Study design and patient inclusion criteria. y Other indications were any procedures without corresponding ICD-9 codes for abdominal aneurysm rupture

(441.3). AAA, abdominal aortic aneurysm; EVAR, endovascular aneurysm repair; ICD-9, International Classification of Diseases Ninth Revision.

EVAR vary with both surgeon and center. Multiple factors including patient stability, the anatomic features of the aneurysm, and surgical experience all contribute to the critical decision regarding which surgical approach to proceed with. Moreover, the favorable results reported for EVAR in the setting of rAAA frequently come from high-volume referral centers with considerable resources and expertise and it remains unclear if these superior outcomes are completely generalizable and representative of the national experience. In light of these questions, the aim of our study was to examine national trends and compare contemporary outcomes of open and EVAR for rAAA across a wide spectrum of hospitals in the United States, by examining outcomes reported in the ACS-NSQIP database. Our second aim was to then confirm the results of our primary analysis by performing a propensity-matched analysis, comparing patients undergoing open versus EVAR repair for rAAA. Propensity adjustment has gained substantial attention recently as a potential means of better controlling for selection bias and patient factors that may contribute to overall outcomes; as unlike traditional methods, propensity-based methods rely on a model of treatment decisions rather than a model of outcomes.13

independently audited and validated clinical data available in the United States. The database captures numerous clinically relevant perioperative variables using data from both community and academic medical centers. These variables collected include preoperative risk factors, intraoperative variables, and 30-day postoperative mortality and morbidity. This information is prospectively collected by an on-site surgical clinical reviewer at each participating institution and is subject to a validation and auditing process to ensure data integrity. Patients were included for analysis using Current Procedure Terminology (CPT) code and International Classification of Diseases ninth edition (ICD-9) diagnosis codes (Fig. 1). Open repair for rAAA was identified using either CPT codes for repair of AAA specifically for rupture (35082, 35092, 35103) or open repair of AAA (35081, 35091, 35102) with concomitant ICD-9 code indicating abdominal aortic aneurysm rupture (441.3). EVAR cases were identified using specific CPT codes for EVAR (34800, 34802e5) or for unsuccessful EVAR with conversion to open repair (34830e2), both with the same concomitant ICD9 code specifying rupture as used to define the open group. The primary outcome measures for our analysis were 30-day postoperative mortality, operative mortality, 30-day overall complication rate, and early return to the operating room. Secondary outcome measures included the 21 specific postoperative complications captured by ACS-NSQIP, as outlined in Table I. Preoperative patient characteristics and postoperative outcomes were compared for

METHODS Data from the 2005 to 2011 NSQIP Participant User File (PUF) were used for this analysis. The database provides one of the largest collections of

Vol.

-,

No.

-, -

2014

Open versus endovascular repair 3

Table I. Variables and data points available in NSQIP and included for analysis Preoperative characteristicsa

Postoperative end pointsb

Patient sex ASA class Preoperative sepsis Dyspnea Functional status Preoperative ventilator requirement Preoperative transfusion (>4 units) DNR status Tobacco use Alcohol use Diabetes COPD Coronary artery diseaseg Congestive heart failure Dialysis dependence Bleeding disorder

Mortality (30 days) Operative mortalityc Overall morbidityd Major complication ratee Early return to the ORf Length of stay (days)

Recent steroid use Recent weight loss Recent chemotherapy (30 days) Recent radiotherapy (90 days) Case contamination Trainee participation Preoperative albumin Preoperative hematocrit Preoperative creatinine

Operative time (min) Superficial SSI Deep SSI Organ space SSI Wound dehiscence Sepsis Septic shock Pneumonia Reintubation Prolonged (>48 hr) vent dependence Pulmonary embolism Acute kidney injury Renal failure Urinary tract infection Stroke Coma Cardiac arrest Myocardial infarction Postoperative bleeding Deep venous thrombosis

COPD, chronic obstructive pulmonary disease; OR, operation room; SSI, surgical site infection. a Preoperative characteristics available in NSQIP and included for analysis. b Postoperative end points available in NSQIP, either directly or indirectly, and reported. c Operative mortality defined as death on postoperative day 0. d Overall morbidity represents the sum of all available complications outlined herein and reported by NSQIP. e Major complications defined as any NSQIP-reported complication with the exception of superficial SSI, urinary tract infection, and deep venous thrombosis. f As defined by NSQIP, early return to the OR defined as within 30 days of index procedure. g Coronary artery disease defined as any of the following: history of myocardial infarction, history of angina in past 1 month, history of previous percutaneous coronary intervention or cardiac surgery, history of congestive heart failure, or history of hypertension requiring medication.

patients undergoing open versus EVAR of ruptured or dissecting AAA using Pearson’s chi-squared test for categorical variables and Student’s t-test for continuous variables. To determine whether

surgical approach had an independent effect on our primary outcome measures, a nonparsimonious multiple logistic regression model was constructed to adjust for patient- and procedure-related confounding factors. In addition to surgical approach (open versus EVAR), the predictor variables considered for inclusion in this model are described in Table I. Because of variability in rates of utilization of EVAR for rAAA over the study period (Fig. 2), the year of operation was also included in the model. Consistent with previous studies modeling ACS-NSQIP data, laboratory values were categorized as normal or abnormal using standard cutoffs and missing laboratory data were assigned a third categorical indicator variable. Missing data for other variables were handled with complete case analysis in light of the substantial completeness of NSQIP data for variables other than laboratory values. Specifically, all patient characteristic variables were >99% available with the exception of alcohol abuse, do-not-resuscitate (DNR) status, American Society of Anesthesiologists (ASA) class, and chemoradiation; patient data for these variables were 90% available. As there are likely nonrandom, fundamental differences between patients who are selected for EVAR and those who are not, we also conducted a separate confirmatory propensity-matched analysis of the study cohort. The purpose of this study was to create a cohort of EVAR patients who, based on known and possible confounders, would have a similar propensity to undergo open repair as patients who underwent a traditional open approach. Doing this addresses concerns that a selection bias led to specific patients, possibly those who appeared more stable or robust, being selected for EVAR over open. Variables chosen for the matching algorithm were those thought most likely to act as potential confounders.13e15 These potential confounders were defined a priori as any preoperative patient characteristic or comorbidity that could have potentially influenced the decision of whether to pursue open versus endovascular intervention and included age, sex, smoking, body mass index, alcohol use, dyspnea, diabetes, functional status, coronary artery disease, bleeding disorder, chronic obstructive pulmonary disease, ASA class, preoperative transfusion requirement, preoperative Systemic inflammatory response syndrome/sepsis, preoperative ventilator requirement, year of operation, trainee assistance, creatinine, hematocrit, and albumin. These variables were entered into a logistic regression model to calculate propensity scores and a nearest-neighbor algorithm was employed to find the most appropriate matched pairs. Missing

4 Speicher et al.

Annals of Vascular Surgery

Fig. 2. Trends in surgical approach to rAAA, by year of operation.

data were imputed using the methods described by D’Agostino and Rubin,16 whereby indicator variables were created for missing values and input into the regression model. After propensity score matching to create 2 identically sized groups with balanced characteristics, baseline preoperative variables between the 2 were compared using Pearson’s chi-squared tests or Fisher’s exact test for categorical variables and Student’s t-test for continuous variables. Similar comparisons were tested between the 2 matched groups with respect to the previously described postoperative end points. Model diagnostics and balance were assessed and no major model assumptions were violated. We made an affirmative decision to control for type I error at the level of the comparison. A P value of 0.05 was used to indicate statistical significance for all comparisons and analyses. Statistical analyses were performed using R version 3.0.1, Vienna, Austria.

RESULTS A total of 1,997 patients were identified who underwent repair of rAAA and met study inclusion criteria. Of these, 1,383 (69.3%) patients underwent an open procedure and 614 (30.7%) patients underwent EVAR. Conversion to an open approach was required in 17 (2.8%) endovascular cases and these were included in the EVAR group based on intent-to-treat. Patient- and procedure-related factors, stratified by procedure type, are shown in Table II. There were statistically significant

differences between open versus EVAR groups with respect to nonindependent functional status (43.2% vs. 28.3%, P < 0.001), preoperative ventilator dependence (18.5% vs. 12.4%, P ¼ 0.001), preoperative bleeding disorders (16.1% vs. 20.0%, P ¼ 0.035), case contamination of contaminated or dirty class (5.3% vs. 0.8%, P < 0.001), and resident participation in the operating room (61.3% vs. 53.4%, P ¼ 0.001). Unadjusted outcomes for open and endovascular approaches are shown in Table III. Regarding our primary end points, there were statistically significant differences between open and EVAR for overall 30-day mortality (38.5% vs. 26.2%, P < 0.001), operative mortality (16.1% vs. 9.9%, P < 0.001), and overall complication rate (73.1% vs. 64.2%, P < 0.001). There were statistically significant differences between groups for 12 of the other secondary postoperative end points. After adjusting for the previously described potential confounders using our logistic regression model, open repair had 1.7 times higher odds (95% confidence interval [CI] 1.24e2.35, P ¼ 0.001) than EVAR for risk of all-cause 30-day mortality (Table IV). Similarly, an open approach had 2.06 times higher odds (95% CI 1.23e3.46, P ¼ 0.006) of operative mortality than an endovascular approach. Multivariable logistic regression also identified an increased risk of overall complications (odds ratio [OR] 1.9, 95% CI 1.41e2.55, P < 0.001), renal failure (OR 1.61, 95% CI 1.07e 2.41, P ¼ 0.02), and respiratory complications (OR 1.84, 95% CI 1.4e2.42, P < 0.001) in the open group compared with an endovascular approach. We did not observe a statistically significant

Vol.

-,

No.

-, -

2014

Open versus endovascular repair 5

Table II. Baseline characteristics before and after propensity matching Total NSQIP cohort Characteristic

Total (n ¼ 1,997)

Open (n ¼ 1,383)

EVAR (n ¼ 614)

Age 4 units) DNR status 39 (2.2%) 26 (2.1%) 13 (2.5%) Tobacco use 727 (36.4%) 516 (37.3%) 211 (34.4%) Alcohol use 89 (5%) 63 (5%) 26 (5%) Diabetes 239 (12%) 152 (11%) 87 (14.2%) COPD 355 (17.8%) 237 (17.1%) 118 (19.2%) Coronary artery disease 446 (25.1%) 314 (25.1%) 132 (25.2%) Congestive heart failure 36 (1.8%) 28 (2%) 8 (1.3%) Dialysis dependence 25 (1.3%) 14 (1%) 11 (1.8%) Bleeding disorder 345 (17.3%) 222 (16.1%) 123 (20%) Recent steroid use 100 (5%) 63 (4.6%) 37 (6%) Recent weight loss 31 (1.6%) 21 (1.5%) 10 (1.6%) Recent chemotherapy 15 (0.8%) 10 (0.8%) 5 (1%) (30 days) Recent radiotherapy 5 (0.3%) 3 (0.2%) 2 (0.4%) (90 days) Contaminated/dirty case 78 (3.9%) 73 (5.3%) 5 (0.8%) Trainee participation 1,176 (58.9%) 848 (61.3%) 328 (53.4%) Preoperative albumin Abnormal 721 (36.1%) 477 (34.5%) 244 (39.7%) Missing 742 (37.2%) 540 (39%) 202 (32.9%) Normal 534 (26.7%) 366 (26.5%) 168 (27.4%) Preoperative hematocrit Abnormal 857 (42.9%) 588 (42.5%) 269 (43.8%) Missing 183 (9.2%) 132 (9.5%) 51 (8.3%) Normal 957 (47.9%) 663 (47.9%) 294 (47.9%) Preoperative creatinine Abnormal 985 (49.3%) 671 (48.5%) 314 (51.1%) Missing 217 (10.9%) 161 (11.6%) 56 (9.1%) Normal 795 (39.8%) 551 (39.8%) 244 (39.7%)

Propensity-matched cohort P value

Open (n ¼ 614)

EVAR (n ¼ 614)

115 194 223 82 141 609

(18.7%) (31.6%) (36.3%) (13.4%) (23%) (99.5%)

114 193 214 93 147 606

0.829 (18.6%) (31.4%) (34.9%) (15.1%) (23.9%) 0.760 (99.3%) 0.725 0.783

426 159 20 0.206 107

Open versus endovascular repair of ruptured abdominal aortic aneurysms.

Management of ruptured abdominal aortic aneurysms (rAAA) remains one of the most challenging situations in vascular surgery. The aim of our study was ...
665KB Sizes 0 Downloads 0 Views