Abdominal aortic aneurysm (AAA) is a degenerative disease that becomes more prevalent with age. Due to the increasing overall life expectancy in developed countries, growing numbers of elderly patients require AAA repair [1]. Most patients with a ruptured AAA die immediately, and even those who survive long enough to reach the hospital have an operative mortality rate of 30%–50% [2]. Elective AAA repair may prevent rupture in patients at risk, and it is associated with much lower operative mortality: approximately 5% for open surgery and 1%–2% for endovascular aneurysm repair (EVAR) [3]. In the management of AAA, predicting the mortality risk of patients prior to open or endovascular repair of a ruptured or intact aneurysm is difficult. Therefore, a risk-scoring model could be very useful, particularly for surgeons who rarely deal with this problem. There are many risk estimation models already available across a range of surgical specialties; these are often used to calculate risk-adjusted, surgeon-specific mortality data [45]. Although these models may help to inform surgeons and patients about the individual risks of intervention, most are much too complex to be practical, and some are outdated and no longer able to predict outcomes accurately in the era of endovascular procedures [56789].

The aim of our study was to evaluate retrospectively the clinical outcomes and risk factors associated with open and endovascular AAA repair, and to propose a new, simplified, multivariable risk-scoring model to enable easy prediction of 30-day mortality in individual patients.

The clinical characteristics of the patients are summarized in Table 1. Of the 485 patients, 408 (84.1%) were men, and 77 (15.9%) were women, with a mean age of 69 years. Of the 485 AAA repairs, 160 (33.0%) were EVARs, and 325 (67.0%) were open surgical repairs. Elective AAA repair was performed in 420 (86.6%) of the patients and emergency repair in 65 patients (13.4%). Coronary artery disease was diagnosed in 72 patients (14.9%); 35 underwent coronary revascularization (22 percutaneous coronary interventions and 13 coronary artery bypass grafts), and 37 were treated conservatively. There were 27 mortalities (5.6%) within 30 days of the AAA repair, and the causes of death are listed in Table 2. The most common cause of death was multiorgan failure (9 patients, 33.3%) followed by myocardial infarction (8 patients, 29.6%).

The relationship between clinical characteristics and 30-day mortality post-repair, as determined by univariate analyses, is shown in Table 1. Patient age (odds ratio [OR], 1.07; 95% confidence interval [CI], 1.02–1.13; P = 0.010) was the only demographic variable significantly associated with 30-day mortality. Among features related to AAA anatomy and procedure timing, the followings were significant predictors of 30-day mortality: maximal diameter of the AAA (OR, 1.29; 95% CI, 1.07–1.56; P = 0.008), location of the AAA (supra-/juxta-renal) (OR, 4.22; 95% CI, 1.90–9.38; P < 0.001), rupture of AAA (OR, 6.98; 95% CI, 3.04–16.03; P < 0.001), and emergency repair (OR, 7.23; 95% CI, 3.22–16.23; P < 0.001). Among the preoperative clinical characteristics, the followings were statistically significant: pulmonary dysfunction (OR, 7.70; 95% CI, 2.51–23.64; P < 0.001) and serum hemoglobin (OR, 0.75; 95% CI, 0.62–0.90; P = 0.002). Preoperative renal dysfunction was biologically significant (OR, 2.41; 95% CI, 0.98–5.96; P = 0.056) without statistical significance. Univariate analysis of intra- and postoperative variables showed that intraoperative mean blood pressure (OR, 0.95; 95% CI, 0.91–0.99; P = 0.011), intraoperative transfusion (OR, 1.14; 95% CI, 1.07–1.21; P < 0.001), and postoperative cardiac events (OR, 10.20; 95% CI, 4.47–23.25; P < 0.001), postoperative renal failure (OR, 11.85; 95% CI, 4.98–28.21; P < 0.001), and postoperative ischemic colitis (OR, 5.85; 95% CI, 1.50–22.78; P = 0.011) were significantly associated with 30-day mortality after AAA repair.

Factors showing a significant association with 30-day mortality (i.e., with P-values < 0.1 in univariate analyses) were included in the final multivariate models. The final model identified three independent predictors of 30-day mortality after AAA repair. The constant of the scoring system was defined as an increase in 30-day mortality in regression units associated with the location of the AAA (supra- or juxta-renal location: 1 point; OR, 5.02; 95% CI, 2.10–12.00; P < 0.001), rupture of AAA (ruptured AAA: 1 point; OR, 6.62; 95% CI, 2.71–16.19; P < 0.001), and preoperative pulmonary dysfunction (pulmonary dysfunction: 1 point; OR, 4.36; 95% CI, 1.79–10.64; P = 0.001) (Table 3). For each predictive factor, the distance from the base category in regression coefficient units was divided by this constant and rounded to the nearest integer to determine a point value. The final multivariate model was assessed for discrimination (c statistic, 0.811) and calibration (the Hosmer-Lemeshow test, P = 0.793).

The risk for each patient was calculated by adding the points assigned to each risk factor identified in the patient, thereby giving the total score for the patient. Table 4 shows the observed and predicted mortality rates corresponding to different total risk scores. When we compared the observed 30-day mortality rate with that predicted by our new risk score, we found that they did not differ significantly. The predicted and observed 30-day mortality rates were, respectively, 1.10% and 0.87% for patients with 0 points, 4.64% and 6.06% for patients with one point, 17.49% and 17.65% for patients with two points, and 48.02% and 80.00% for patients with 3 points, suggesting that this model was a good predictor of risk across all risk groups, except for patients with three points (Table 4). Additionally, this model was also a good predictor of risk, according to the treatment modalities (open or endovascular AAA repair) (Table 5).

Open repair of an infrarenal AAA is associated with considerably higher perioperative mortality than EVAR, which has increased the popularity of the endovascular approach in these aneurysms. However, most surgeons still perform open AAA repairs when the anatomy does not favor the use of EVAR, and occasionally EVAR may be converted to open repair during or after the repair due to technical failure on deployment, inappropriate stent-graft placement, vessel rupture, or graft thrombosis [91112]. Overall rates of open conversion of up to 6% have been reported in large EVAR series, and the incidence of early conversion (<30 days after stent-graft implantation) varies from 0.3% to 5.9% in the English-language literature [1112131415]. More recent studies have reported lower rates of early conversion to open surgery, which is probably due to improvements in endograft technology and endovascular technique, better patient selection, and a tendency to treat most complications via the endovascular approach. However, the average mortality rate associated with early conversion was still as high as 12.4% (0%–28.5%) in a total of 12,236 endovascular AAA repairs [11].

For any repair procedure, whether open or endovascular, the process of informed consent requires a thorough explanation of the risks and benefits to the patient, and the risks presented by vascular surgeons to patients undergoing AAA repair are based on anecdotal experience or on currently available risk estimation tools. In the past two decades, many studies have attempted to identify factors capable of predicting death in patients undergoing open repair of AAAs, whether intact or ruptured, and various risk assessment tools are currently used to predict perioperative mortality; each model has its strengths and weaknesses [56716171819202122232425]. The existing risk-scoring models vary greatly in the nature and quality of their results, and the methods used for reporting. Although several studies have modeled statistically the predictive variables in order to design scoring systems that can forecast outcomes, many of these did not use a sound methodology, and only a few having undergone robust testing, let alone prospective validation.

A new risk-scoring model should be able to predict the outcomes prior to the AAA repair, rather than based on operative and postoperative findings. Furthermore, given the increasing popularity of endovascular procedures, and the moderate incidence of early conversion-related mortality, it is reasonable that a new risk-scoring model should include patients undergoing both open and endovascular AAA repair. A new model should also be easy to use to enable rapid application by any grade of medical staff, and generate the most accurate risk score from the smallest number of variables, in an elective and urgent setting. To date, the quality and level of available evidence has been insufficient to make such a model a reality.

The goal of our model was to predict final outcomes based on preoperative findings, to do fairly straightforwardly, and to reflect the changes in preferred treatment modality. With this in mind, we incorporated the type of repair (open or endovascular) and the available preoperative patient characteristics into our model, and assigned an independent weight to each to calculate the overall risk. We found that supra-/juxta-renal location of the AAA, rupture of AAA, and preoperative pulmonary dysfunction were independent predictors of 30-day mortality. When we compared the observed 30-day mortality rates with those expected from our new risk score, we found that they did not differ significantly, which showed that our model had good predictive power across all risk groups, except for patients with three points.

Our model has several limitations, including the retrospective assessment of patients and inclusion of limited information; other factors, e.g., data that are not collected routinely, may also affect patient outcomes. Another limitation is the single center nature and a long period of data collection of this study that were factors likely to limit external validity of this model. Also, the numbers of patients who underwent endovascular repair (33.0%) were relatively small, and this may not have been sufficient to make accurate predictions. Furthermore, because of extremely small number of patients with three points, we cannot validate predictive power of our model for these patients. The model also requires validation, preferably using larger data sets from throughout the world.

In conclusion, we have developed a new risk estimation tool to predict postoperative mortality associated with open and endovascular AAA repair. With its high discriminative ability, this model is a step toward improving preoperative decision-making and the informed consent process in an era when endovascular procedures are increasingly common. Further studies are needed to validate this tool using larger and more varied samples.