Effect of previous cardiac surgery on the outcomes of acute type A aortic dissection repair: a nationwide cohort study

Abstract

OBJECTIVES

The effect of previous cardiac surgery (PCS) on the outcomes of acute type A aortic dissection (ATAAD) repair remains controversial. This study compared the primary outcome (in-hospital mortality) and secondary outcomes (postoperative complications and post-discharge mortality/aortic reoperation rates up to 5 years) of patients who underwent ATAAD repair with and without PCS through a nationwide cohort analysis.

METHODS

We used Taiwan’s National Health Insurance Research Database to enroll patients who underwent ATAAD repair with and without PCS (94 and 4532, respectively) between 1 July 2004 and 31 March 2017. A 1:4 propensity score matching (PSM) was used to create well-balanced PCS (n = 74) and non-PCS (n = 296) groups. Results before and after PSM were aligned to determine the role of PCS in primary and secondary outcomes.

RESULTS

Before PSM, the PCS group was older and had more comorbidities, including diabetes mellitus, heart failure, atrial fibrillation and malignancy, but less stroke history than the non-PCS group. More patients in the PCS group received coronary artery bypass grafting during the repair than in the non-PCS group. The PCS group had a higher in-hospital mortality than the non-PCS group. After PSM, the in-hospital mortality was similar between the 2 groups (27.0% vs 26.7%; P = 0.953). For patients who survived to discharge, the all-cause mortality up to 5 years for PCS and non-PCS groups were 29.7% and 18.4% (P = 0.015) before and 29.6% and 23.1% (P = 0.313) after PSM, respectively. The 2 groups had similar aortic reoperation rates up to 5 years before (13.5% vs 11.7%; P = 0.727) and after PSM (13.8% vs 11.6%; P = 0.776).

CONCLUSIONS

Patients with PCS who underwent ATAAD repair showed higher in-hospital and long-term mortality rates than those without PCS when not controlled for confounding factors. However, the disparity disappeared after PSM, indicating that worse outcomes might be due to its old age and different perioperative characteristics.

Open in new tabDownload slide

INTRODUCTION

Acute type A aortic dissection (ATAAD) is a cardiovascular emergency associated with high morbidity and mortality, and prompt surgical treatment is necessitated for lifesaving [1, 2]. Surgical procedures can be further complicated in patients with previous cardiac surgery (PCS) because of extensive mediastinal adhesions, anatomic distortion, manipulation of patent bypass grafts and management of previously inserted valve prostheses [3]. The incidence of PCS in the ATAAD population ranges from 2.3 to 15%, varying with different international aortic dissection research registries [3–7]. As reported by Estrera et al. [3, 5] in the 2010s, open surgical repair of ATAAD after PCS was associated with higher in-hospital mortality and complication rates compared to first-time surgery. However, in the 2020s, Norton et al. [6, 8] reported compatible results in patients with PCS and first-time surgery. Therefore, whether PCS is associated with elevated postoperative mortality and complication rates after ATAAD repair remains controversial [3, 5–8]. Furthermore, disparities in clinical features and surgical procedures commonly exist between patients with and without PCS but were not adjusted [5–9]. In this nationwide cohort study, we aimed to investigate whether PCS was associated with inferior in-hospital and 5-year outcomes among patients who underwent ATAAD repair using propensity score matching (PSM) to adjust the confounding factors.

PATIENTS AND METHODS

Data sources

This national cohort study used the Taiwan National Health Insurance Research Database (NHIRD) and Taiwan Death Registry (TDR). The National Health Insurance programme was implemented by the government in 1995 and provides healthcare services for 99.7% of the 23 million residents in Taiwan. The NHIRD contains registration and comprehensive claim information. All beneficiaries’ status (change working unit, withdraw, die) was updated monthly and stored in the Registry for Beneficiaries, a subset of NHIRD. This registry file allows us to compute the follow-up period accurately. Disease diagnoses are coded on the International Classification of Diseases, Ninth and Tenth Revision, Clinical Modification (ICD-9-CM between 2000 and 2015; ICD-10-CM since 2016), and procedure codes of the National Health Insurance Administration. The accuracy of the diagnoses of major cardiovascular diseases in the NHIRD has been validated [10–12]. We used TDR to assess mortality, including patients who died in hospital and those who died without hospital visits.

Ethics statement

The Institutional Review Board of Chang Gung Medical Foundation approved the study (202202133B1). The requirement for informed consent was waived due to the retrospective and delinked nature of the dataset. Researchers perform on-site data analyses at the Health and Welfare Data Centre to ensure health data protection. No electronic devices are permitted in the data centre. All analysis results before release are reviewed to ensure patient confidentiality [10].

Study design

From the NHIRD, we enrolled patients admitted with newly diagnosed ATAAD (ICD-9-CM, 441.0; ICD-10-CM, I71.0) and underwent repair surgery between 1 July 2004 and 31 March 2017. We stopped enrolling patients after 31 March 2017 because ATAAD repair was consolidated into a single code (68043 B) without a specific classification of the extent of the aortic repair procedure. We excluded patients who received medical therapy instead of surgical intervention, those treated for type B aortic dissection and those with missing dates of birth, sex or aberrant data. Note that the proportion of patients undergoing ATAAD repair, surgical repair for type B aortic dissection and medical therapy without surgical intervention was similar to the results of the previous study [13]. A total of 4626 patients underwent surgical intervention for ATAAD, which targeted the ascending aorta and/or aortic root replacement with cardiopulmonary bypass (CPB) according to the procedure codes, in coordination with aortic arch replacement and thoracic endovascular aortic repair, were eligible for analysis. Patients were dichotomized into non-PCS and PCS groups based on whether they received cardiac surgeries before the index admission. The PCS group comprised 4 types of previous surgeries: valve repair/replacement, coronary artery bypass grafting (CABG), open aortic replacement and other cardiac surgeries, including cardiotomy, intracardiac tumour resection, congenital heart defect repair, ventricular aneurysm repair, maze procedure and pulmonary embolectomy. To reduce the confounding, PSM with a 1:4 ratio between the PCS and non-PCS groups was made, resulting in a matched cohort of 74 and 296 patients in the PCS and non-PCS groups, respectively. Clinical outcomes during hospitalization and after discharge were compared between the 2 groups (Fig. 1).

Study outcomes and covariates

The primary outcome was in-hospital mortality. Secondary outcomes were postoperative complications (blood transfusion volumes, re-exploration for bleeding, length of hospital stay, implementing mechanical support, acute respiratory failure, acute renal failure, new stroke, sepsis, and wound infection) and post-discharge mortality/aortic reoperation rates up to 5 years.

The covariates of interest were demographics (age, sex and index year), comorbidities (hypertension, diabetes mellitus, heart failure, myocardial infarction, atrial fibrillation, peripheral arterial disease, stroke, chronic kidney disease, end-stage renal disease, chronic liver disease, Marfan syndrome, coagulopathy and malignancy) and preoperative emergency management (Supplementary Material, Table S1). To reduce misclassification of comorbidities, the relevant ICD appeared at least twice in outpatient visits or once in hospitalization within 1 year before the surgical repair date.

Statistical analyses

Statistical analyses were performed using SAS software (version 9.4; SAS Institute Inc, Cary, NC, USA). We computed 3 outcome rates (in-hospital mortality, mortality rate up to 5 years and reoperation up to 5 years), simply the number of outcomes divided by the number of patients at risk. The non-PCS and PCS groups were matched in a 4:1 ratio without replacement, within a calliper width equal to 0.2 of the standard deviation of the logit of the estimated propensity score to balance the covariates (Tables 1 and 2) between the 2 groups [14]. The reason for the 1:4 ratio is that the gain of statistical power sharply increases until the ratio is 1:4 and then slowly increases [15]. We used the generalized boosted model to obtain the propensity score by including demographic data, comorbidities, and surgical variables as independent variables and study grouping as a dependent variable. Generalized boosted model estimation involves an iterative process with multiple regression trees to capture the complex and nonlinear relationships between dependent and independent variables without overfitting the data [16]. It is less affected by extreme weights [17].

For variables not involving time, mean ± standard deviation for continuous variables and frequency (%) for categorical data were used. Before PSM, we compared variables between the PCS and non-PCS groups using independent t-tests, chi-square or Fisher’s exact tests where appropriate. Following PSM, we employed conditional logistic regression to assess the differences in variables between the PCS and matched non-PCS groups. Conditional logistic regression is well-suited for matched sets with varying cases and controls, as it incorporates matched sets as strata within the model. The regression coefficients in this model represent the differences in independent variables between cases and controls within each matched set. Importantly, conditional logistic regression can accommodate continuous and categorical independent variables [18]. Given the relatively small sample size (74 PCS and 296 non-PCS), we used P-values to assess baseline data balance after PSM instead of absolute standard mean difference. The absolute standard mean difference is generally considered a more robust statistic for assessing balance, as it is less influenced by sample size. For variables involving time, i.e., all-cause mortality and aortic reoperation rates after discharge, we plotted them versus their follow-up time up to 5 years using the Kaplan–Meier method. When comparing the all-cause mortality after discharge between the PCS and non-PCS groups, Cox regression with hazard ratio (HR) was made for the data without PSM, and the robust sandwich estimate was added to Cox regression to solve the data independence violation after PSM [19]. Because the 2 study groups were well-balanced in baseline covariates after PSM, only the PCS group variable was included in Cox’s model [20]. For the aortic reoperation rates after discharge, we used the Fine-Gray analysis with the sub-distribution hazard ratio and treated mortality as a competing risk [21]. Statistical significance for all analyses, both before and after PSM, was determined at a threshold of P < 0.05.

RESULTS

Demographics, comorbidities and preoperative management

Before PSM, the PCS group was older and presented with a higher proportion of diabetes mellitus, heart failure, atrial fibrillation and malignancy, and a lower proportion of stroke history than the non-PCS group. The preoperative management, including cardiopulmonary resuscitation and endotracheal intubation, was not significantly different between the 2 groups. Among the PCS group, valve repair/replacement was the most common previously performed cardiac surgery, followed by CABG, aortic replacement, and others. After PSM, the 2 groups showed no significant differences in index year of surgery, age, sex, chronic comorbidities and preoperative management (Table 1).

Surgical information

Before PSM, except for a higher proportion of CABG in the PCS group, the 2 groups were similar in ascending aorta replacement, aortic arch replacement, aortic root replacement, thoracic endovascular aortic repair and concomitant valve replacement. After PSM, all surgical variables were balanced between the 2 groups (Table 2).

Postoperative mortality and complications

Before PSM, except for the higher in-hospital mortality rate in the PCS group than in the non-PCS group, the length of hospital stay, blood transfusion volume and postoperative complications, including re-exploration for bleeding, implementation of mechanical support, respiratory failure, renal failure, sepsis and wound infection, were similar between the 2 groups. After PSM, the in-hospital mortality rate and all postoperative complications were similar between the 2 groups (Table 3). The subgroup analysis of the PSM cohort according to different types of PCS is illustrated in Fig. 2. It revealed no significant differences in in-hospital mortality rates between the non-PCS group and each PCS subgroup.

All-cause mortality and aortic reoperation rates after discharge

Among those who survived the ATAAD repair, all patients were followed up for up to 5 years (average 4.4 ± 1.4; median, 5.0; range, 0.1–5.0). A total of 335 patients died in hospital, and 351 died without hospital visits. Before PSM, the PCS group had a higher all-cause mortality rate than the non-PCS group (29.7% vs 18.4%; P = 0.015; Fig. 3A). After PSM, the all-cause mortality rate was similar between the 2 groups (29.6% vs 23.1%; P = 0.313; Fig. 3B). There was no significance in the aortic reoperation rates between the 2 groups before [13.5% vs 11.7%; hazard ratio 1.14; 95% confidence interval 0.54–2.41; P = 0.728; Fig. 4A) and after PSM (13.8% vs 11.6%; hazard ratio 1.14; 95% confidence interval 0.46–2.81; P = 0.776; Fig. 4B]. The subgroup analyses showed no significant differences in different types of aortic reoperation between the 2 groups before and after PSM (Supplementary Material, Table S2). After considering the competing risk of death, the sub-distribution hazard ratio was 1.09 (95% confidence interval 0.45–2.66; P = 0.853) of the PCS group compared to the non-PCS group.

DISCUSSION

Reoperative cardiac surgery was reportedly a risk factor associated with increased postoperative mortality and adverse events, including treating acute aortic dissection [3, 5, 22, 23]. However, with advancements in surgical techniques and perioperative care over the last decade, some recent studies have reported that patients undergoing ATAAD repair with PCS could achieve outcomes comparable to those undergoing first-time operations [6, 8]. Our findings provide new evidence and contribute to the body of knowledge regarding the effect of PCS on the outcomes of ATAAD repair. This study has several clinical implications. First, we observed different clinical features and operative characteristics between the non-PCS and PCS groups, including age, chronic comorbidities and concomitant CABG. Second, the PCS group had significantly higher mortality rates during hospitalization and after discharge than the non-PCS group before PSM. However, the non-PCS and PCS groups revealed similar short- and long-term outcomes after PSM. Third, we observed similar findings for different types of PCS procedures in subgroup analyses after PSM.

As reported by Estrera et al. [3, 24] in the 2010s, open surgical repair of ATAAD after PCS was associated with higher in-hospital mortality and complication rates and a more frequent need for postoperative cardiac support compared to first-time surgery. Therefore, it was previously accepted that PCS was a risk factor for poor outcomes in ATAAD treatment. Several factors could contribute to this issue. First, excessive blood loss resulting from sternal re-entry and mediastinal dissection injuries is a major complication of reoperation, which can lead to haemodynamic instability, re-exploration for haemostasis, renal failure, lung injury, infection and sternal dehiscence. These surgical morbidities associated with bleeding tendencies can be aggravated in the context of ATAAD repair surgery, which is commonly complicated with consumption coagulopathy induced by systemic activation of the coagulation system and consumption of clotting factors from intravascular thrombosis of the dissected false lumen [25]. Moreover, prior CPB established by peripheral vessel cannulation was commonly used to prevent major bleeding from the ruptured aorta or cardiac chamber during resternotomy and the division of mediastinal adhesions. It results in prolonged CPB time and increases the risk of subsequent bleeding complications. Second, for patients with ATAAD, all management strategies were performed on an emergency basis, and preoperative angiographic evaluation of the coronary arteries was usually not conducted. Inadequate coronary assessment would increase the risk of graft injuries and insufficient myocardial protection during reoperative cardiac surgery, which requires prolonged CPB and cardiac arrest times for complex procedures. Furthermore, a significantly higher incidence of heart failure history, a recognized adverse prognostic factor in ATAAD repair, was found in the PCS group [26]. Third, lateral thoracotomy was introduced as a less invasive approach. It could provide benefits compared with redo sternotomy in reducing CPB time and perioperative bleeding in patients undergoing reoperative CABG and valvular surgeries [27, 28]. However, this alternative approach is usually unsuitable for ATAAD repair surgeries because of the limited exposure of the aortic arch and root. Additionally, endovascular intervention is a developing modality utilizing a minimally invasive treatment for selected patients with proximal aortic diseases, including ATAAD [29, 30]. However, this may not be feasible for every patient because of individual variations in the anatomic morphology of aortic dissection. In this study, all investigated patients underwent open aortic repair surgery with CPB. By dividing the study period into 3 consecutive intervals (2004–2009; 2010–2014; 2015–2017) before PSM, patients had trends of increasing rates of aortic arch replacement and concomitant thoracic endovascular aortic repair with progress of time (Supplementary Material, Table S3).

It is commonly accepted that cardiac reoperation is a technically demanding procedure performed in patients with more underlying health conditions, which requires high surgical expertise [31]. According to Krebs et al. [5] and Klodell et al. [9], patients with PCS were older with more comorbidities, including heart failure, hypertension, diabetes mellitus, chronic kidney disease, chronic lung disease and arrhythmia. These prognostic factors could result in poorer outcomes based on different operative risk-estimating systems [32, 33]. Moreover, PCS could require more complex surgeries and an extended operative process. Similar outcomes were observed in the present study; patients in the PCS group presented with older age and higher rates of comorbid conditions, including diabetes mellitus, heart failure, atrial fibrillation, malignant disease and more concomitant CABG. These risk factors are possibly the major causes contributing to the high mortality rate of the PCS group. We found that the difference in mortality rates during hospitalization and after discharge between PCS and non-PCS groups was diminished by adjusting the perioperative disparities with PSM. Notably, the association between PCS and inferior outcomes has been challenged in several recent studies demonstrating compatible results in patients with PCS and first-time operations [6, 8], which is consistent with the primary findings of our study. Therefore, if ATAAD occurs in patients with PCS without multiple comorbid conditions, an aggressive strategy managed by emergency open repair surgery should still be justifiable because acceptable survival could be expected.

Limitations

This study has a few limitations that should be acknowledged. First, the haemodynamic data, laboratory profiles, medication dosages and cause of death were unavailable in the NHIRD. Furthermore, the NHIRD did not contain data such as CPB profiles and details of aortic repair techniques because they did not refer to specific procedure codes. Besides, the NHIRD did not have information on imaging studies, which were used to evaluate the complexity of aortic dissection. Second, as this was a retrospective and non-randomized controlled study, a potential bias influencing the homogeneity of the non-PCS and PCS groups might have existed; however, PSM reduced the inter-group heterogeneity. Moreover, unmeasured confounders may still exist. Therefore, we interpreted the results as an association instead of causality. Third, because this retrospective cohort spanned ∼13 years, modifications and advancements in CPB technology, myocardial protection, cerebral perfusion strategies and postoperative care protocols may have been developed. Fourth, further investigation into other populations and settings should be warranted, as the study was conducted in Taiwan. Furthermore, well-controlled randomized trials focusing on strategies and techniques to improve outcomes in the PCS population undergoing ATAAD repair with long-term follow-up are suggested.

CONCLUSIONS

Based on real-world data, patients who underwent ATAAD repair with PCS presented with high comorbidity and surgical mortality. However, the worse outcomes of PCS in ATAAD repair might be due to its old age and different perioperative characteristics. After adjusting for inter-group baseline disparities by PSM, patients with PCS were comparable to those without PCS in terms of in-hospital and 5-year after-discharge outcomes.

SUPPLEMENTARY MATERIAL

Supplementary material is available at EJCTS online.

FUNDING

The Chang Gung Memorial Hospital research project (grant number CMRPD1N0451) supported this study.

Conflict of interest: none declared.

ACKNOWLEDGEMENTS

None declared.

DATA AVAILABILITY

The data for this study were sourced from the NHIRD, housed exclusively in the Health and Welfare Data Science Centre, and not publicly available.

Author contributions

Chun-Yu Lin: Conceptualization; Methodology; Writing—original draft. Wei-Min Chen: Data curation; Formal analysis; Visualization. Shu-Hao Chang: Data curation; Formal analysis; Visualization. Sheng-Yueh Yu: provided clinical opinions and revised the manuscript. Lai-Chu See: Data curation; Methodology; Visualization; Writing—original draft.

Reviewer information

European Journal of Cardio-Thoracic Surgery thanks Ari Mennander, Sven Peterss and the other anonymous reviewers for their contribution to the peer review process of this article.

REFERENCES

ABBREVIATIONS

ABBREVIATIONS
 
• ATAAD

  Acute type A aortic dissection

 
• CABG

  Coronary artery bypass grafting

 
• CPB

  Cardiopulmonary bypass

 
• ICD-9-CM

  International Classification of Diseases, Ninth Revision, Clinical Modification

 
• ICD-10-CM

  International Classification of Diseases, Tenth Revision, Clinical Modification

 
• NHIRD

  National Health Insurance Research Database

 
• PCS

  Previous cardiac surgery

 
• PSM

  Propensity score matching

 
• TDR

  Taiwan Death Registry