Is surgical risk of aortic arch aneurysm repair underestimated? A novel perspective based on 30-day versus 1-year mortality

Abstract

Open in new tabDownload slide

OBJECTIVES

The decision to undergo aortic aneurysm repair balances the risk of operation with the risk of aortic complications. The surgical risk is typically represented by perioperative mortality, while the aneurysmal risk relates to the 1-year risk of aortic events. We investigate the difference in 30-day and 1-year mortality after total arch replacement for aortic aneurysm.

METHODS

This was an international two-centre study of 456 patients who underwent total aortic arch replacement for aneurysm between 2006 and 2020. Our primary end-point of interest was 1-year mortality. Our secondary analysis determined which variables were associated with 1-year mortality.

RESULTS

The median age of patients was 65.4 years (interquartile range 55.1–71.1) and 118 (25.9%) were female. Concomitantly, 91 (20.0%) patients had either an aortic root replacement or aortic valve procedure. There was a drop in 1-year (81%, 95% confidence interval (CI) 78–85%) survival probability compared to 30-day (92%, 95% CI 90–95%) survival probability. Risk hazards regression showed the greatest risk of mortality in the first 4 months after discharge. Stroke [hazard ratio (HR) 2.54, 95% CI (1.16–5.58)], renal failure [HR 3.59 (1.78–7.25)], respiratory failure [HR 3.65 (1.79–7.42)] and reoperation for bleeding [HR 2.97 (1.36–6.46)] were associated with 1-year mortality in patients who survived 30 days.

CONCLUSIONS

There is an increase in mortality up to 1 year after aortic arch replacement. This increase is prominent in the first 4 months and is associated with postoperative complications, implying the influence of surgical insult. Mortality beyond the short term may be considered in assessing surgical risk in patients who are undergoing total arch replacement.

INTRODUCTION

Thoracic aortic aneurysms (TAA) are asymptomatic in >95% of cases, with most detected incidentally through imaging [1]. Surgical intervention in asymptomatic aortic aneurysm is intended to avoid catastrophic complications including dissection and rupture, which lead to high rates of mortality [2]. The risk of continued monitoring is typically considered from a yearly perspective.

In the guidelines-recommended size criteria for surgical intervention, this risk is balanced against the risk of surgical repair, which is typically represented by perioperative mortality [3]. The outcome of arch aneurysm with a total arch replacement is improving, and yet remains associated with significantly higher perioperative mortality, compared to more proximal aneurysm repair [4, 5]. In addition to mortality, operations carry significant risk of morbidity including stroke, reoperation for bleeding, respiratory failure and renal failure, which might compromise subsequent life of the patients even if they survive the operation [6].

The difference in observation time between the monitoring (annual risk) and surgical repair (perioperative risk) might introduce a flaw in risk assessment. A study comparing 30-day and 1-year mortality for abdominal aortic aneurysm repair found an increase in 1-year mortality from 1% in patients with no risk factors to 67% in patients with multiple risk factors [7]. However, such data are lacking in repair of TAA, and such predictors of 1-year mortality can identify patients less likely to benefit from elective repair. We hypothesize that the current assessment of surgical risk underestimates the procedure-related mortality rate. The aim of this study was to determine the difference between perioperative and 1-year mortality rates of patients who underwent a total arch replacement for aortic arch aneurysm.

PATIENTS AND METHODS

Ethical statement

This study was approved by the Institutional Review Boards of the Columbia University Irving Medical Center and the University of Bologna (Columbia; number AAAU0575; most recent approval date 4 April 2022; Bologna, 121/2022/Disp/AUOBo). Written consent from patients was waived given the retrospective nature of the study.

Study design

This is a two-centre, retrospective, observational, cohort study of consecutive patients who underwent aortic arch repair for aortic aneurysm between 2006 and 2020 at the Columbia University Irving Medical Center/NewYork Presbyterian Hospital and the University of Bologna. The STROBE (strengthening the reporting of observational studies in epidemiology) checklist for retrospective studies was followed. The primary end-point of interest was 1-year mortality, which was compared to 30-day mortality that represented our perioperative mortality. Data were collected for all our patients from our Aortic Center Database and the electronic medical record from both our institutions. Definitions of the postoperative complications followed those of the Society of Thoracic Surgeons Adult Cardiac Surgery Database (STS ACSD) [8]. All-cause mortality during the follow-up period was collected through clinical encounters as well as phone call to patients and referring physicians. For the patients at Columbia, death information was supplemented through the Centers for Disease Control and Prevention National Death Index accessed on 26 May 2021 and complete to 31 December 2019 [9]. Patients at the University of Bologna were routinely followed up by outpatient clinical evaluation at 1, 6 and 12 months or by phone. For the entire cohort, median follow-up time was 6.25 [interquartile range (IQR) 5.71–6.89] years.

Study population

Patients who underwent aortic surgery at 2 institutions between 2006 and 2020 were selected, specifically those who underwent open total arch replacement for aortic aneurysm were included in the primary analysis (Supplementary Material, Fig. S1). Inclusion criteria included patients over 18 years of age who underwent open total aortic arch replacement whose primary indication for surgery was aneurysm. Patients who underwent ascending or hemiarch replacement were excluded. Acute type A aortic dissection and endocarditis for primary indication were also excluded. For the secondary analysis, patients who survived 30 days after their operation were included.

Patient management

Surgical management for aortic arch replacement at both institutions has been previously described [10–14]. Surgical indication was determined by the attending surgeon, based on most recent American Heart Association/American College of Cardiology and European Society of Cardiology Guidelines [15, 16]. The arterial cannulation site and method of cerebral protection were chosen based on the surgeon’s discretion [17, 18]. Management of cardiopulmonary bypass was standard for our study period. Standard bypass parameters were mild hypothermia (32°C) with a pump flow rate of 2.5 ml/cm²/min and goal mean arterial pressure of 60–80 mmHg. Distal aortic anastomosis for arch replacement was performed under moderate hypothermia (24–28°C, nasopharyngeal) and antegrade cerebral perfusion. The aortic valve was spared during root replacements with reimplantation technique whenever appropriate; when replacement was necessary, the prosthetic valve was chosen based on American Heart Association/American College of Cardiology and European Society of Cardiology Guidelines as well as patient preference [10–12].

Statistical analysis

Statistical analyses for the current study were performed with R software version 4.2.1 (R Foundation for Statistical Computing, Vienna, Austria). Categorical variables are reported as count (%). Normality of continuous variables was assessed via the Shapiro–Wilk test, and are reported as mean and standard deviation, if a normal distribution was confirmed, or median and interquartile range otherwise. For all analyses, a P-value of <0.05 was considered as statistically significant. All collected variables had no degree of missingness. Therefore, no data imputation was performed.

Univariable and multivariable mixed effect Cox regression analyses were performed to determine the relationship between baseline characteristics and 1-year mortality as the dependent variables for all patients, and those who survived 30 days following the operation. Variables for inclusion in multivariable analysis were chosen based on results from univariable analysis and clinical acumen. Univariable Cox regression was used to graph cubic splines and determine which continuous variables should be entered as categorical in the model and determine thresholds for each category. Centre was entered as a random effect to adjust for site of surgery in the model. Multivariable modelling was used to adjust for potential confounding. Variables were checked for multicollinearity and variance inflation factors (VIFs) of <5 were obtained in regression models, indicating minimum potential intercorrelation among variables. Schoenfeld residuals were checked to confirm the absence of violation of the proportional hazard assumption of all final models.

Kaplan–Meier (KM) curves were used to study the survival of the cohort over time among the entire cohort, 30-day survivors and hospital-stay survivors. A risk hazards function was plotted to determine when patients had the greatest risk of all-cause mortality. KM curve stratified based on the number of postoperative complications was plotted and a pairwise log-rank test was performed to examine the impact of these complications on 1-year mortality. Sensitivity analysis was performed by plotting KM curve with the start date as the date of discharge or 30 days after surgery, whichever comes later. Landmark analysis was performed at 30 days to the 1-year follow-up period and the restricted mean survival time was calculated at 30 days and 1 year.

RESULTS

Patient characteristics

The baseline characteristics of 456 patients who underwent open total aortic arch replacement are listed in Table 1. The median age was 65.4 (55.6–71.1) years with 118 (25.9%) being female. Of all 456 patients, 364 (79.8%) had hypertension, 27 (5.9%) had a bicuspid aortic valve and 23 (5.0%) had Marfan’s syndrome. The median left ventricular ejection fraction (LVEF) was 60 (55–63)%. The median EuroScore II of the patient cohort is 8 (IQR 5–9). The baseline characteristics of patients who survived 30 days are listed in Supplementary Material, Table S1.

Operative details

The operative details of all patients are described in Table 2. Of all 456 patients who underwent a total arch replacement, 91 (20.0%) had proximal extension (root replacement or aortic valve replacement/repair), 51 (11.2%) had a conventional elephant trunk and 224 (49.1%) had a frozen elephant trunk. In addition, 47 (10.3%) patients had a concomitant coronary artery bypass graft, while 68 (14.9%) had a concomitant aortic valve procedure. The median cardiopulmonary bypass time was 200 (171–244) min, while the median aortic cross-clamp time was 122 (92–168) min. Most patients had antegrade cerebral perfusion [428 (93.9%)], while the most common cannulation strategies were axillary [171 (37.5%)], aorta [105 (23%)] and brachiocephalic [122 (26.8%)]. The operative details of patients who survived 30 days are described in Supplementary Material, Table S2.

In-hospital outcomes

Overall, 35 (7.7%) patients died within 30 days of surgery and 53 (11.6%) patients died during their hospital stay, with combined in-hospital and 30-day mortality occurring in 54 (11.8%) patients. Of 403 patients who survived their hospital stay, 212 (53%) were discharged home and 191 (47%) were discharged to an inpatient rehabilitation facility. Regarding complications, 38 (8.3%) had a stroke, 76 (16.7%) had renal failure, 111 (24.3%) had respiratory failure and 41 (9%) underwent reoperation for bleeding.

One-year survival of patient cohort

The survival probability curve of the entire patient cohort is shown in Fig. 1a. The survival probability of patients 30 days after their operation is 92.8% [95% confidence interval (CI) 90.4–95.2%], which drops to 81.0% (95% CI 77.5–84.8%) 1 year after their operation. The risk hazards regression in Fig. 1b shows patients have the greatest risk of mortality during the first 4 months after their operation, which then decreases to a plateau and remains low during the rest of the year. The estimated restricted mean survival time is 28 days and 10 months, at 30 days and 1 year, respectively. This elevated risk hazard in the first 4 months is seen even when patients who died within 30 days are excluded, alongside a similar decrease in survival probability (Supplementary Material, Fig. S2). We also performed an analysis using a combined definition of in-hospital and 30-day mortality, with a KM curve starting at 30 days after the operation or the date of discharge, whichever is later (Fig. 2). Similarly, a substantial drop in survival probability at 1 year is seen—92% (95% CI 89–95%).

Association of postoperative complications and post-discharge mortality

Multivariable Cox regression in Table 3 shows stroke (HR 2.54, 95% CI 1.16–5.58, P = 0.02), renal failure (HR 3.59, 95% CI 1.78–7.25, P < 0.001), respiratory failure [hazard ratio (HR) 3.65, 95% CI 1.79–7.42, P < 0.001] and reoperation for bleeding (HR 2.97, 95% CI 1.36–6.46, P = 0.007) are associated with 1-year mortality in 421 patients who survived 30 days. A KM curve based on the number of these complications shows decreased survival probability with increasing numbers of complications with precipitous drop in survival within the first 4 months (Fig. 3), with a significant difference between 0 and 2 (P < 0.001), 0 and 3–4 (P < 0.001), 1 and 2 (P < 0.043), and 1 and 3–4 (P < 0.001) complications (Supplementary Material, Table S3).

Sensitivity analysis

We chose 30-day mortality to represent perioperative mortality to allow mortality comparison at 2 time points (30-day vs 1-year). Since perioperative mortality may be represented by in-hospital mortality, we also analysed 1-year survival of the patients who survived the hospital stay after the index operation. The KM curve showed a substantial drop in survival probability to 92% (95% CI 90–95%) in the 1st postoperative year (Supplementary Material, Fig. S3). Landmark analysis of survival probability at 30 days and 1 year is shown in Supplementary Material, Fig. S4.

In order to further study the impact of postoperative complications on 1-year mortality, we performed a multivariable Cox regression on all 456 patients who showed renal failure (HR 3.07, 95% CI 1.80–5.22, P-value < 0.001), respiratory failure (HR 4.00, 95% CI 2.33–6.87, P-value < 0.001) and LVEF (HR 0.97, 95% CI 0.95–0.99) were significant. We believe this difference, in which stroke and reoperation for bleeding are no longer significant when including patients who died within 30 days occurs since these complications are not as strongly associated with short-term death but drag on after 30 days.

DISCUSSION

The key findings of the present study, which included patients from 2 major high-volume aortic centres undergoing total arch replacement for TAA, are: patients who survive their procedure are at continued risk of mortality, particularly within 4 months after the procedure, and this risk is increased by postoperative complications, implying the influence of surgical insult.

These observations provide important data in the field of aortic arch repair, supporting our hypothesis that the current assessment of surgical risk in the short term underestimates the mortality rate associated with total arch replacement. In addition to the above main findings, when patients who survive 30 days or survive their entire hospital stay were analysed, there is a continued decrease in survival probability over 1 year (Supplementary Material, Figs S2–S4). Others have studied mid- to long-term mortality after total aortic arch replacement, but ours is the 1st to focus on 1-year outcomes and postulate the drop in survival is tied to the surgery itself rather than disease or patient characteristics. A study from 2 European institutes studied total aortic arch replacement with frozen elephant trunk and found an in-hospital mortality rate of 14.9%, including patients with aneurysm and acute aortic dissection. The presented data show a similar decrease in survival probability at 1-year post-discharge seen in our study albeit the authors did not particularly address this observation [19]. Di Eusanio et al. [20] studied long-term outcomes after aortic arch surgery, and found an 11.1% in-hospital mortality rate in patients with a chronic aortic pathology, while overall survival at 1 year was 92.1% in patients who survived their hospital stay. The decrease in survival at 1 year after discharge suggests patients are not ‘out of the woods’ even if they survive the perioperative period. While the operative decision of risk–benefit analysis driving the decision to operate is often based on short-term mortality, there is a clinically significant mortality that persists after 30 days.

This is the 1st study to investigate a risk hazards regression in the mid-term, showing this mortality is concentrated in the postoperative period. The timing of this mortality risk seems to be related to the procedure itself. Our findings that postoperative complications are associated with 1-year mortality in patients who survive 30 days support the influence of surgical insult. However, in a multivariable Cox regression with the entire patient cohort, postoperative renal failure and respiratory failure, together with decreased LVEF, are significant for 1-year mortality (Supplementary Material, Table S4). Other groups have found that postoperative permanent neurological dysfunction, chronic renal failure and intraoperative bleeding are associated with increased morbidity and mortality during arch repair, supporting our findings [19–21]. In cardiac surgery, overall patients with major postoperative complications had continued decrease in survival at 1-year follow-up, with a dose-dependent relationship based on the number of complications [22, 23]. While results that show these patients with complications are at increased risk of mortality at the mid-term may be too obvious and not be considered as an outstanding finding, they have a different implication on estimating surgical mortality rate, or surgical risk. It seems plausible to expect a continued decrease in survival after a successful hospital discharge, and this survival drop could be profound for procedures that have high morbidity rate, such as total arch replacement. Although this study lacks a comparison group in patients who underwent continued monitoring without surgical repair, these findings place an emphasis on considering mid-term outcomes when making operative decisions. Importantly, there does not appear to be a consensus on postoperative decrease in mortality 1 year following aortic arch replacement, highlighting the need for more studies on the topic. A study from Tanaka et al. [4] found no difference in predicted survival rate in hospital survivors with or without major morbidities compared to the general population.

These observations can apply beyond cardiac surgery. Other noncardiac studies have shown similar decreases in mortality in the 1-year period but focus on baseline conditions of the patient or natural disease progression rather than the surgery itself. In patients undergoing major noncardiac surgery, a prospective observational study found 1-year mortality of 5.5% in all patients and 10.3% in patients older than 65 years, associated with cumulative deep hypnotic time based on anaesthesia and intraoperative hypotension, focusing on patient characteristics [24]. Beck et al. [7] studied 1-year mortality after elective infra-renal abdominal aortic aneurysm (AAA) repair, focusing on the differences between open and endovascular repair, identifying patients at increased risk of mortality at 1 year due to patients’ pre-existing comorbidities and less likely to benefit from AAA repair, even in the form of endovascular repair. Future work should focus on invasive procedures with high rates of morbidity considering the impact of surgical insult itself on mortality, as a similar relationship seen in our study could be apparent. Furthermore, future studies that create a risk model for complications and mid-term mortality would help improve operative decision-making.

Limitations

Our study limitations are related to its retrospective nature. While this is a two-centre study with large volume aortic centres, the associations we have described between postoperative complications and mortality cannot be viewed as causation and may not be generalizable. Since no formal functional status testing was imbedded in the clinical care of either surgical team, we are unable to comment on the functional status of patients who died after discharge during the follow-up period. While the number of patients is considerable for a total arch replacement study, we are still limited by the available sample size and may be underpowered to detect certain differences. We also lack data on the cause of death of patients who died post-hospital discharge, and thus cannot confirm they are directly related to aortic intervention. We would propose that careful evaluation of larger databases explore the operation-attributable risk of mortality to determine more precisely the most appropriate time interval for mortality assessment. Finally, our primary analysis used 30-day mortality to represent perioperative mortality to allow mortality comparison at 2 time points; however, the perioperative mortality may be defined with in-hospital mortality or combination of 30-day and in-hospital mortality. In our sensitivity analyses, we observed a similar drop in survival even when these definitions were applied, supporting our hypothesis.

CONCLUSION

Patients who undergo total arch replacement for aneurysm have an increased 1-year mortality compared to 30 days. We claim this mortality is related to surgical insult rather than patient or disease characteristics. Our data implies the true operative mortality rate is higher than what is currently perceived; however, the discussed limitations preclude affirmation of the causation between the surgical procedures and subsequent 1-year mortality. Prospective studies in larger patient population would help to better inform practice for aortic arch replacement. Our study challenges the current surgical risk assessment.

SUPPLEMENTARY MATERIAL

Supplementary material is available at EJCTS online.

Conflict of interest: Hiroo Takayama is a speaker with Terumo and a consultant with Artivion and Edwards and received research grant from the Rudin Foundation. The other authors report no conflicts of interest.

DATA AVAILABILITY

The data underlying this article will be shared on reasonable request to the corresponding author.

Author contributions

Kavya Rajesh: Conceptualization; Data curation; Formal analysis; Investigation; Methodology; Writing—original draft. Dov Levine: Conceptualization; Data curation; Formal analysis; Investigation; Resources. Giacomo Murana: Conceptualization; Data curation; Investigation. Sabrina Castagnini: Data curation. Edoardo Bianco: Data curation. Patra Childress: Data curation. Yanling Zhao: Formal analysis. Paul Kurlansky: Conceptualization; Investigation; Methodology; Project administration; Resources; Supervision; Validation; Writing—original draft. Davide Pacini: Conceptualization; Investigation; Methodology; Project administration; Resources; Validation; Writing—original draft. Hiroo Takayama: Conceptualization; Investigation; Methodology; Project administration; Resources; Validation; Writing—original draft.

Reviewer information

European Journal of Cardio-Thoracic Surgery thanks Roman Gottardi, Woon Heo, Gabriele Piffaretti and the other anonymous reviewers for their contribution to the peer review process of this article.

Presented at the EACTS 37th Annual Meeting, Vienna, Austria, 4–7 October 2023.

REFERENCES

ABBREVIATIONS

ABBREVIATIONS
 
• CI

  Confidence interval

 
• HR

  Hazard ratio

 
• LVEF

  Left ventricular ejection fraction

 
• TAA

  Thoracic aortic aneurysm