Long-term outcome of hemiarch replacement in a proximal aortic aneurysm repair: analysis of over 1000 patients

Abstract

OBJECTIVES

The aim of this study was to investigate the impact of hemiarch replacement in patients undergoing an open repair of proximal thoracic aortic aneurysm without arch aneurysm.

METHODS

A retrospective review was performed on 1132 patients undergoing proximal aortic aneurysm repair at our Aortic Center between 2005 and 2019. Inclusion criteria were all patients undergoing root or ascending aortic aneurysm repair with or without hemiarch replacement. Exclusion criteria were age <18 years, aortic arch diameter ≥4.5 cm, type A aortic dissection, previous ascending aortic replacement, ruptured aneurysm and endocarditis. Propensity score matching in a 2:1 ratio (573 non-hemiarch: 288 hemiarch) on 19 baseline characteristics was performed. The median follow-up time was 46.8 months (range 0.1–170.4 months).

RESULTS

Hemiarch patients had significantly lower 10-year survival in the matched cohort (hemiarch 73.8%; 66.9–81.4%; vs non-hemiarch 86.5%; 81.1–92.3%; P < 0.001), driven by higher in-hospital mortality rate (4% vs 1%; P < 0.001). Cumulative incidence of aortic arch reintervention rates at 10 years was similarly low (hemiarch 1.0%; 0–2.5% vs non-hemiarch 1.3%; 0–2.6%, P = 0.615). Multivariate analysis with hazard ratios of the overall cohort showed hemiarch as an independent factor associated with long-term mortality (2.16; 1.42–3.27; P < 0.001) but not with aortic arch reintervention (0.76; 0.14–4.07, P = 0.750).

CONCLUSIONS

Hemiarch repair may be associated with higher short-term mortality compared to non-hemiarch. Arch reintervention was rare after a repair of proximal thoracic aortic aneurysm without arch aneurysm. Our data call for larger and prospective studies to further delineate the utility of hemiarch repair in proximal aortic surgery.

Indications for surgical repair of a proximal thoracic aortic aneurysm, defined as a repair of the aortic root or ascending aorta, include size, growth rate and the presence of symptoms.

INTRODUCTION

Indications for surgical repair of a proximal thoracic aortic aneurysm, defined as a repair of the aortic root or ascending aorta, include size, growth rate and the presence of symptoms. Uncertainty remains for the surgical management of a minimally or non-dilated adjacent aortic arch segment at the time of proximal aortic repair [1–4]. Consequently, clinical equipoise exists regarding whether to extend the proximal aortic repair to include the proximal aortic arch, by performing a hemiarch replacement, to mitigate the future risk of aortic aneurysm-related events in that zone. Indeed, a more aggressive approach of extending a proximal aneurysm repair to include the hemiarch, even without significant dilatation at the level of the proximal arch, has anecdotally been advocated by experienced aortic centres [5–8]. However, the increased surgical complexity with longer cardiopulmonary bypass (CPB) times and the additional requirement of circulatory arrest associated with the hemiarch procedure may confer increased risk of perioperative complications [9–15]. In addition, the natural history of minimally dilated proximal aortic arch tissue is not well known [3]. Nevertheless, the scarcity of large cohort studies with sufficient follow-up information permits no definitive conclusions on this subject [3, 5–8]. We hypothesized that hemiarch replacement to excise a non- or minimally aneurysmal proximal aortic arch does not have clinical benefit in patients undergoing proximal aortic aneurysm repair. The aim of this study was to compare the long-term survival and aortic arch reintervention rates in patients who had proximal aortic aneurysm repair with and without hemiarch replacement.

PATIENTS AND METHODS

This study was approved on 31 March 2020 by the Institutional Review Board at Columbia University (AAAR2949). We performed a single-centre retrospective review of the Columbia University Aortic Center Database at New York Presbyterian Hospital. All patients undergoing proximal aortic aneurysm repair from 2005 to 2019 were included (n = 1132). The proximal aorta was defined as the aortic segments from the aortic valve to the take-off of the innominate artery. Baseline demographics, comorbidities, concomitant procedures, complications and survival data, as well as operative details, were recorded. Aortic size data were measured via echocardiogram, computed tomography scan, MRI or operative record from the hospital electronic medical record. Follow-up data were acquired retrospectively using electronic medical record review between 2005 and 2016 and prospectively after 2017 per our Aortic Center protocol. The Aortic Center Database was queried to obtain follow-up information through scheduled postoperative visits at 1 month and then every 6 months until 3 years. Additional information was gained through periodic contact with patients and referring cardiologists per protocol to help capture events, such as mortality and reintervention, occurring at an outside institution. National Death Index database, which provides mortality data in the USA based on state death certificate information, was queried when a patient survival status is unknown due to the loss of contact. National Death Index Final file search was up to 2019; thus, long-term mortality follow-up rate in our Aortic Center is 100% complete at the end of 2019. Reintervention information remains unknown for ∼13% of the entire patient cohort due to the loss of contact. The median follow-up time was 46.8 months (range 0.1–170.4 months). Exclusion criteria were: age <18 years, aortic arch aneurysm ≥4.5 cm or diagnosed arch aneurysm in the operative report, type A aortic dissection, dissection in the arch, previous ascending aortic replacement, ruptured aneurysm and endocarditis. Hemiarch replacement was defined as a need for an open anastomosis between distal end of the replacement graft and the proximal aortic arch at the level of the brachiocephalic artery and anastomosing oblique incision to the aortic arch without replacing the supra-aortic arteries, while non-hemiarch is defined as a proximal aortic aneurysm repair (ascending and/or root replacement) with the aorta clamped and no hemiarch or circulatory arrest. The final study cohort consisted of 861 patients: 573 non-hemiarch and 288 hemiarch (Fig. 1). Our primary end point was long-term survival, the secondary end point was long-term freedom from aortic arch reintervention and the tertiary end point was in-hospital complications.

Figure 1:

Consort diagram of patients with proximal aortic aneurysm stratified by the presence and absence of hemiarch repair.

Patient management

While the surgical indication was determined by each attending surgeon, it generally followed published societal guidelines [3, 4, 16]. At the time of the surgery, all aneurysmal segments ≥4.5 cm were concomitantly replaced. Smaller aortic segments were replaced according to the surgeon’s discretion, taking a number of factors into consideration, such as patient preference and overall status (age, comorbidities and complexity of planned procedure). For patients undergoing an aortic root replacement, the decision whether to spare the valve included the integrity of the aortic valve tissue along with patient preference and comorbidities. Patients receiving root and valve replacement received either mechanical or biological prosthesis based on similar considerations. For patients receiving hemiarch replacement, the cerebral protection strategy was at the discretion of the operating surgeon. Typically, it included systemic moderate hypothermia (28°C) with anterograde cerebral perfusion either with axillary cannulation or direct innominate artery cannulation. When the former was chosen, the systemic circulation was also provided through the axillary cannulation and then switched to the side arm of the arch graft. For the latter, aortic cannulation was used until the arch procedure, and the side arm was used after the distal anastomosis. Left carotid cannulation was added according to the cerebral oximetry. Our standard bypass management included mild hypothermia (32°C) with a pump flow rate of 2.5 ml/cm² to a goal MAP of 60–80 mmHg.

Statistical analysis

A 2:1 nearest neighbour propensity score match was performed to achieve a well-balanced sample in which standardized mean differences are <0.10 for the following baseline characteristics (Table 1), which were chosen from commonly used parameters across aortic surgery literature: age, sex, body surface area, diabetes mellitus, hypertension, dyslipidaemia, cerebrovascular disease, chronic obstructive pulmonary disease, chronic kidney disease, coronary artery disease, peripheral artery disease, atrial fibrillation, Marfan syndrome, previous open cardiac surgery, elective surgery status (elective versus urgent/emergent), ejection fraction, aortic insufficiency ≥ moderate, concomitant coronary artery bypass and concomitant aortic root replacement [1–3, 5–7]. A calliper width of 0.025 standard deviation (SD) of the logit of the propensity score was chosen to achieve the optimal quality of the matching. Missing data were imputed via Random Forest based algorithm of predictive meaning matching using the ‘missRanger’ R package.

Table 1:

Preoperative and perioperative characteristics of overall and propensity score-matched groups

Variable	All patients (n = 1132)				Propensity score-matched patients (n = 861)
	Non-hemiarch (n = 825)	Hemiarch (n = 307)	P-Value	SMD	Non-hemiarch (n = 573)	Hemiarch (n = 288)	P-Value	SMD
Age, years, median (IQR)	61.0 (50.0, 70.0)	61.0 (52.0, 71.0)	0.26	0.097	62.0 (50.0, 71.0)	61.0 (52.0, 70.0)	0.93	0.017
Sex, male, n (%)	653 (79)	235 (77)	0.39	0.063	450 (79)	224 (78)	0.87	0.018
BSA, m2, median (IQR)	2.00 (1.85, 2.18)	1.99 (1.84, 2.14)	0.31	0.051	2.00 (1.85, 2.14)	2.00 (1.84, 2.16)	0.95	0.009
BMI, kg/m2, median (IQR)	27.3 (24.5, 30.6)	27.5 (24.9, 31.4)	0.32	0.005	27.3 (24.4, 30.4)	27.5 (24.7, 31.1)	0.36	0.014
DM, n (%)	92 (11)	31 (10)	0.69	0.035	59 (10)	30 (10)	>0.99	0.004
Hypertension, n (%)	578 (70)	211 (69)	0.72	0.029	393 (69)	199 (69)	0.94	0.011
Dyslipidaemia, n (%)	437 (53)	146 (48)	0.12	0.108	273 (48)	138 (48)	0.9974	0.005
CVD, n (%)	55 (7)	19 (6)	0.88	0.019	38 (7)	18 (6)	0.95	0.016
COPD, n (%)	64 (8)	21 (7)	0.69	0.035	36 (6)	19 (7)	0.98	0.013
* Chronic kidney disease (eGFR < 60 ml/min/1.73 m2), n (%)	311 (38)	144 (47)	0.007	0.184	258 (45)	134 (47)	0.73	0.030
CAD, n (%)	211 (26)	80 (26)	0.93	0.011	153 (27)	75 (26)	0.90	0.015
Dialysis, n (%)	4 (0)	0 (0)	0.51	0.099	3 (1)	0 (0)	0.54	0.103
NYHA ≥III, n (%)	80 (31)	21 (27)	0.067	0.075	48 (32)	18 (26)	0.47	0.131
PAD, n (%)	81 (10)	25 (8)	0.46	0.059	50 (9)	23 (8)	0.81	0.027
Afib, n (%)	87 (11)	27 (9)	0.41	0.064	54 (9)	25 (9)	0.82	0.026
BAV, n (%)	311 (38)	122 (40)	0.605	0.039	207 (36)	115 (40)	0.311	0.078
* Marfan syndrome, n (%)	24 (3)	1 (0)	0.016	0.206	2 (0)	1 (0)	>0.99	<0.001
Previous cardiac intervention, n (%)	141 (17)	57 (19)	0.64	0.037	109 (19)	54 (19)	0.9967	0.007
Previous MI, n (%)	27 (3)	11 (4)	0.94	0.017	20 (3)	11 (4)	0.96	0.018
Elective versus urgent/emergent, n (%)	801 (97)	298 (97)	0.97	0.016	561 (98)	281 (98)	0.94	0.023
* LVEF, %, median (IQR)	55.0 (50.0, 60.0)	55.0 (50.0, 55.0)	0.003	0.086	55 (50, 58)	55.0 (50.0, 55.5)	0.092	0.013
Aortic stenosis ≥ moderate, n (%)	216 (27)	91 (30)	0.45	0.089	147 (26)	82 (29)	0.44	0.062
* Aortic insufficiency ≥ moderate, n (%)	427 (52)	184 (60)	0.017	0.165	336 (59)	173 (60)	0.74	0.029
Concomitant CABG, n (%)	138 (17)	55 (9)	0.72	0.030	100 (17)	51 (18)	>0.99	0.007
Aortic root replacement, n (%)	567 (69)	230 (75)	0.051	0.138	427 (75)	219 (76)	0.69	0.035
* CPB time, min,** median (IQR)	123.0 (97.3, 160.0)	138.0 (107.0, 172.0)	<0.001	0.248	124.0 (98.0, 162.0)	137.0 (106.3, 171.5)	0.003	0.193
AoX time, min, median (IQR)	95.0 (75.0, 125.0)	100.0 (76.0, 129.0)	0.38	0.033	97.0 (75.8, 125.0)	100.0 (75.3, 129.0)	0.70	0.013
* Circulatory arrest,**n (%)	0 (0)	307 (100)	<0.001	5.208	0 (0)	288 (100)	<0.001	3.216
* Circulatory arrest time,** min, median (IQR)	0.0 (0.0, 0.0)	11.0 (9.0, 15.0)	<0.001	1.831	0 (0, 0)	11.0 (8.8, 15.3)	<0.001	1.814
* Lowest temp, °C,** median (IQR)	32.0 (32.0, 32.7)	28.0 (28.0, 28.0)	<0.001	1.891	32.0 (32.0, 32.5)	28.0 (28.0, 28.0)	<0.001	2.068

^*Statistically significant difference in comparisons between groups before propensity matching (P < 0.05).

^**Statistically significant difference in comparisons between groups after propensity matching (P < 0.05).

CPB: cardiopulmonary bypass; IQR: interquartile range; SMD: standardized mean differences; MRI: magnetic resonance imaging; MAP: mean arterial pressure; PPM: post-operative pacemaker; HR: hazard ratio; sHR: hazard ratio sub-distribution; LL: lower limit; BSA: body surface area; BMI: body mass index; DM: diabetes mellitus; CVD: cerebrovascular disease; COPD: chronic obstructive pulmonary disease; eGFR: estimated glomerular filtration rate; CAD: coronary artery disease; NYHA: New York Heart Association; PAD: peripheral artery disease; Afib: atrial fibrillation; BAV: bicuspid aortic valve; MI: myocardial infarction; LVEF: left ventricular ejection fraction; CABG: coronary artery bypass graft; AoX: aortic cross-clamp.

Kolmogorov–Smirnov test was used to assess the normality of distribution of continuous data. When normally distributed, continuous variables were reported with a mean (SD) and compared with Student’s t-test. When normally distributed, continuous variables were reported with a mean (SD) and compared with Student’s t-test. Otherwise, they were described with medians and interquartile ranges and compared using the Mann–Whitney U-test. Categorical variables, displayed with numbers and percentages of the total, were compared using the Pearson’s chi-square test or Fisher’s exact test, when appropriate.

The Kaplan–Meier method was used to analyse the respective survival for all groups, with comparisons made using the log-rank test. The strength of the association between hemiarch procedure and death was further analysed via the calculation of an E-value to assess the weight of unmeasured confounders [17]. Risk factors for long-term mortality and reintervention were determined using Cox-proportional hazards model and the Fine and Gray sub-distribution hazard model considering death as a competing risk, respectively. The proportional hazard assumption that hazard ratios do not depend on time was not violated, as there were no obvious trends in the flat and smoothed scatterplot of scaled Schoenfeld residuals over time across all variables. We also performed univariable and multivariable logistic regressions to examine variables that are independently associated with in-hospital mortality. A P < 0.1 from univariate logistic regression was selected for the retention of variables in the final logistic regression model. Variables with P < 0.1 in the univariable analysis, as well as significant variables from stepwise regression models, were entered for final multivariate analysis. Statistical significance for all tests was set at a two-sided P-value <0.05. All statistical analyses were performed with R version 4.1 (The R Foundation Vienna, Austria).

RESULTS

After 2:1 nearest neighbour propensity score matching, the original cohort of 1132 patients was reduced to 816 patients (573 non-hemiarch:288 hemiarch).

Preoperative and perioperative patient characteristics

Unmatched groups

Of the 1132 patients who met inclusion criteria, 307 (27%) underwent concomitant hemiarch replacement. The median follow-up time was 46.8 months (range 0.1–170.4 months). The preoperative and perioperative characteristics of the matched and unmatched groups are listed in Table 1. There was no significant difference in the median age between the groups (61 non-hemiarch vs 61 hemiarch, P = 0.26). The procedures were performed on an elective basis in 97% of patients in both groups (801/825 non-hemiarch vs 298/307 hemiarch, P = 0.97), and both were predominantly male (79% non-hemiarch vs 77% hemiarch; P = 0.39). The non-hemiarch group had higher rates of Marfan syndrome (3% in non-hemiarch vs 0.3% in hemiarch; P = 0.016) and lower rates of preoperative CKD (38% vs 47%, P = 0.007). LVEF was significantly different between groups [55.0; interquartile range (IQR), 50.0–60.0 non-hemiarch vs 55.0; IQR, 50.0–55.0 hemiarch; P = 0.003], as was the prevalence of aortic insufficiency ≥ moderate (52% non-hemiarch vs 60% hemiarch, P = 0.017). The non-hemiarch group had a smaller median ascending aorta diameter than the hemiarch group (48.0 vs 53.0 mm, P < 0.001). The aortic root size was not significantly different between non-hemiarch (47.0 mm) and hemiarch (45.0 mm) groups (P = 0.94). The non-hemiarch group had a significantly shorter median CPB time (123.0 vs 138.0 min hemiarch, P < 0.001). The median time of circulatory arrest in the hemiarch group was 11.0 min (IQR, 9.0–15.0 min).

Matched groups

The results of propensity score matching are shown in Table 1. The matching resulted in well-balanced groups for 19 matched preoperative variables.

Postoperative outcomes in matched cohort

Postoperative outcomes in the matched and overall cohort are displayed in Table 2. Aetiologies for in-hospital mortality included cardiogenic shock (3 non-hemiarch vs 1 hemiarch), bleeding (3 hemiarch), respiratory failure (5 hemiarch), intracerebral haemorrhage (1 hemiarch) and multisystem organ failure (1 hemiarch). The non-hemiarch group experienced significantly lower in-hospital mortality (1% vs 4% hemiarch, P < 0.001), 30-day mortality (0.5% vs 2.8% hemiarch, P = 0.014) and combined mortality, defined as either in-hospital or 30-day mortality (0.7% vs 4.2%, P < 0.001). Non-hemiarch patients also experienced lower rates of stroke (3% vs 6% hemiarch, P = 0.047), reintervention for bleeding (4% vs 9% hemiarch, P = 0.011) and respiratory failure (7% vs 13% hemiarch, P = 0.006). No significant differences were observed for hospital length of stay, aortic dissection, acute renal failure, pleural effusion requiring drainage, rhythm disturbance requiring PPM, deep sternal wound infection/mediastinitis and surgical site infection.

Table 2:

Short- and long-term outcomes for matched and overall cohorts

Variable	All patients (n = 1132)			Propensity score-matched patients (n = 861)
	Non-hemiarch (n = 825)	Hemiarch (n = 307)	P-Value	Non-hemiarch (n = 573)	Hemiarch (n = 288)	P-Value
* In-hospital mortality,** n (%)	8 (1)	13 (4)	<0.001	3 (1)	11 (4)	<0.001
* 30-Day mortality,** n (%)	8 (1.0)	10 (3.3)	0.014	3 (0.5)	8 (2.8)	0.014
* Combined mortality, n (%)	9 (1.1)	14 (4.6)	<0.001	4 (0.7)	12 (4.2)	0.0011
Hospital length of stay, days, median (IQR)	7.0 (5.0, 9.0)	7.00 (5.0, 10.0)	0.57	7.0 (5.0, 99.0)	7.0 (5.0, 10.0)	0.93
Stroke,** n (%)	28 (3)	19 (6)	0.058	15 (3)	16 (6)	0.047
* Reoperation for bleed,** n (%)	35 (4)	26 (8)	0.005	24 (4)	25 (9)	0.011
Reoperation for infection, n (%)	8 (1)	3 (1)	0.26	5 (1)	3 (1)	>0.99
* Respiratory failure,** n (%)	52 (6)	39 (13)	0.001	38 (7)	36 (13)	0.006
Aortic dissection, n (%)	1 (0)	1 (0)	0.20	1 (0)	1 (0)	>0.99
Acute renal failure, n (%)	62 (8)	26 (8)	0.34	52 (9)	23 (8)	0.68
Pleural effusion requiring drainage, n (%)	36 (4)	16 (5)	0.22	24 (4)	16 (6)	0.47
Rhythmic disturbance requiring pacemaker, n (%)	47 (6)	18 (6)	0.26	37 (6)	18 (6)	>0.99
Deep sternal infection/mediastinitis, n (%)	8 (1)	2 (1)	0.23	5 (1)	2 (1)	>0.99
Surgical site infection, n (%)	10 (1)	5 (2)	0.22	7 (1)	5 (2)	0.76
* 10-Year survival rate,** median (IQR)	86.7% (82.3, 91.4)	74.2% (67.6, 81.4)	<0.001	86.5% (81.1, 92.3)	73.8% (66.9, 81.4)	<0.001
10-Year landmark analysis of survival, median (IQR)	88.5% (84.1, 93.2)	81.0% (74.4, 88.2)	0.266	87.9% (82.5, 93.7)	79.9% (73.0, 87.5)	0.22
Arch size sensitivity analysis of 10-year survival, median (IQR)	–	–	–	90.4%; (86.0, 95.0)	75.4% (68.0, 83.7)	0.0024
10-Year cumulative incidence of arch reintervention (95% CI)	1.2% (0.1–2.2)	1.0% (0–2.3)	0.707	1.3% (0–2.6)	1.0%(0–2.5)	0.615

* Statistically significant difference in comparisons between groups before propensity matching (P < 0.05).

** Statistically significant difference in comparisons between groups after propensity matching (P < 0.05).

CI: confidence interval; IQR: interquartile range.

The long-term survival was significantly higher in the non-hemiarch group compared to the hemiarch group [86.5%; 95% confidence interval (CI), 81.1–92.3% in non-hemiarch vs 73.8%; 95% CI, 66.9–81.4% in hemiarch; P < 0.001, Fig. 2 and Table 2]. The 10-year cumulative incidence of aortic arch reintervention was not significantly different between groups [1.3% (0–2.6%) in non-hemiarch vs 1.0% (0–2.5%) in hemiarch; P = 0.615, Fig. 3 and Table 2].

Figure 2:

Kaplan–Meier survival curve of the matched cohort.

Figure 3:

Cumulative incidence curve of aortic arch reintervention in the matched cohort.

During follow-up, aortic reintervention was required in 62 patients (43 non-hemiarch vs 19 hemiarch), which included the following types of reinterventions: surgical aortic valve replacement (12 hemiarch vs 27 non-hemiarch), transcatheter aortic valve replacement (2 hemiarch vs 4 non-hemiarch), aortic root or ascending replacement (10 hemiarch vs 13 non-hemiarch), aortic arch replacement (2 hemiarch vs 6 non-hemiarch), abdominal aortic surgery (0 hemiarch vs 1 non-hemiarch), CABG (1 hemiarch vs 7 non-hemiarch), open mitral valve surgery (2 hemiarch vs 1 non-hemiarch) and transcatheter mitral valve surgery (0 hemiarch vs 1 non-hemiarch).

Unmatched cohort

Postoperative outcomes for the unmatched cohort were similar to those reported for the matched cohort (Table 2). Notably, the hemiarch group had higher in-hospital mortality (4% vs 1% in non-hemiarch, P < 0.001), reoperation for bleeding (4% vs 8%, P = 0.005), respiratory failure (6% vs 13%, P = 0.001) and lower 10-year survival rate (74.2% vs 86.7%, P < 0.001). The cumulative arch reintervention rate was similar (1.2% vs 1.0%, P = 0.707) (Table 2).

Long-term mortality and freedom from aortic arch reintervention predictors

In Cox regression multivariable analysis in the unmatched overall cohort, hemiarch replacement (HR, 2.16; 95% CI, 1.42–3.27, P < 0.001), age (HR, 1.06; 95% CI, 1.04–1.08, P < 0.001), previous cardiac intervention (HR, 1.99; 95% CI, 1.23–3.21, P = 0.005), CPB time (HR 1.01; 95% CI, 1.00–1.01, P < 0.001) and elective status (HR, 0.36; 95% CI, 0.15–0.84, P = 0.02) were significantly associated with mortality at 10 years (Table 3). Multivariable analysis in the unmatched cohort found hemiarch replacement (HR, 4.24; 95% CI, 1.72–10.47, P = 0.002), previous cardiac intervention (HR, 5.39; 95% CI, 2.23–13.06, P < 0.001) and CKD (HR, 2.70; 95% CI, 1.07–6.85, P = 0.036) to be independent predictors for in-hospital mortality (Table 4).

Table 3:

Cox regression multivariable analysis for 10-year mortality

Variable	Multivariable analysis in unmatched cohort
	Univariable analysis of long-term mortality		Multivariable analysis of long-term mortality
	Hazard ratio	P-Value	Hazard ratio	P-Value
* Hemiarch replacement	2.10 (1.4, 3.2)	<0.001	2.16 (1.42, 3.27)	<0.001
* Age, years	1.05 (1.03, 1.07)	<0.001	1.06 (1.04, 1.08)	<0.001
Concomitant CABG	1.98 (1.26, 3.11)	0.0021	1.13 (0.67, 1.90)	0.64
* Previous cardiac intervention	2.00 (1.28, 3.10)	0.0021	1.99 (1.23, 3.21)	0.005
* Elective versus urgent/emergent	0.33 (0.15, 0.76)	0.009	0.44 (0.15, 0.84)	0.02
* CPB time, min	1.01 (1.00, 1.01)	<0.001	1.01 (1.00, 1.01)	<0.001

^*Statistically significant difference in multivariable analysis (P < 0.05).

CPB: cardiopulmonary bypass.

Table 4:

Multivariable analysis of in-hospital mortality predictors in unmatched cohort

Variable	Univariable analysis of in-hospital mortality		Multivariable analysis of in-hospital mortality
	Odds ratio	P-Value	Odds ratio	P-Value
* Hemiarch replacement	4.50 (1.85, 10.98)	<0.001	4.24 (1.72, 10.47)	0.002
* Previous cardiac intervention	5.42 (2.27, 12.9)	<0.001	5.39 (2.23, 13.06)	<0.001
* CKD, (eGFR < 60 ml/min/1.73 m2)	3.02 (1.21, 7.56)	0.018	2.70 (1.07, 6.85)	0.036

^*Statistically significant difference in multivariable analysis (P < 0.05).

Due to the very low number of arch-related reinterventions [8], multivariable analysis found no significant predictors for aortic arch reintervention, including hemiarch (sHR, 0.76; 95% CI, 0.14–4.07, P = 0.750).

The E-value measures the strength of association between unmeasured confounders and the exposure-outcome relationship in the presence of covariates that are controlled for. On average, the observed effect estimates for hemiarch replacement and a worse long-term survival may be negated by an unmeasured confounder that was associated with both hemiarch replacement and the long-term survival by a risk ratio of >3.74 (lower confidence bound 2.19).

Landmark analyses

Landmark survival analysis of the matched and unmatched cohorts was performed to separate the influence of short-term mortality. In the matched cohort, this demonstrated no survival difference between the groups (87.9%; 95% CI, 82.5–93.7% in non-hemiarch vs 79.9%; 95% CI, 73.0–87.5% in hemiarch; P = 0.22; Fig. 4). No survival difference was demonstrated in the unmatched group either (88.5%; 95% CI, 84.1–93.2% in non-hemiarch vs 81.0%; 95% CI, 74.4–88.2% in hemiarch; P = 0.25; Supplementary Material, Fig. S3).

Figure 4:

Landmark survival analysis in the matched cohort.

Sensitivity analysis

While we carefully excluded patients with arch aneurysm with multiple layers of reviews, the actual size measurement was missing in 25% in the non-hemiarch group and 18% in the hemiarch group. To understand the impact of the missing size measurements, a subgroup survival analysis was performed which adjusted for arch size and included patients with arch size available (Supplementary Material, Table S1 and Supplementary Material, Fig. S1). The same matching algorithm with the addition of arch size was performed, constructing matched cohorts of 455 patients in the non-hemiarch group and 236 patients in the hemiarch group (Table 2 and Supplementary Material, Fig. S4). It demonstrated higher survival in the non-hemiarch group (90.4%; 95% CI, 86.0–95.0 non-hemiarch vs 75.4%; 95% CI, 68.0–83.7 hemiarch, P = 0.0024).

DISCUSSION

Using one of the largest cohorts in the current literature, our matching analysis found that: (1) patients who underwent hemiarch replacement demonstrated reduced survival due to the higher in-hospital mortality and (2) arch reintervention was rare after a repair of proximal thoracic aortic aneurysm without arch aneurysm.

While the safety of extending as aneurysm repair to include the proximal arch has been suggested by some of the expert aortic centres, literature on this practice in patients with proximal aortic aneurysms is scarce. In a propensity score-matched comparison of 232 patients undergoing ascending aneurysm repair with (n = 116) and without (n = 116) hemiarch, Sultan et al. [18] found no difference in stroke (1.7% non-hemiarch vs 3.4% hemiarch; P = 0.408), in-hospital mortality (1.7% non-hemiarch vs 3.4% hemiarch; P = 0.408) or 1-year mortality (5.3% in both groups; P = 1), supporting a liberal approach to identifying patients eligible for hemiarch repair when possible. In a cohort of 151 patients undergoing ascending aneurysm repair with (n = 40) and without (n = 111) hemiarch, Kozlov et al. [19] found no significant differences between the groups in mortality, freedom from reoperation or rates of any postoperative complication. A study of 266 propensity score-matched patients (133 non-hemiarch vs 133 hemiarch) by Malaisrie et al. [5] found the hemiarch group to have decreased survival compared to the non-hemiarch group, without statistical significance, at 5 years. These studies might be underpowered to identify a difference. Furthermore, these studies did not apply detailed anatomical information in the outcome analysis, including patients who did need hemiarch replacement for a significant proximal aortic arch aneurysm.

Our study, unique with the largest cohort of its kind and the exclusion of patients with a significant proximal aortic arch aneurysm, unexpectedly demonstrated that patients who underwent hemiarch repair demonstrated increased mortality, primarily in the short term. The difference in long-term survival appeared to originate from higher in-hospital mortality in the hemiarch group, which is supported by 2 analyses: the landmark survival analysis, which demonstrated no survival differences in the matched cohort, and the multivariable analysis including the time period up to hospital discharge (or death), which found hemiarch to be a significant predictor of in-hospital mortality. Hemiarch procedures may increase operative risk as a result of the increased CPB time and addition of circulatory arrest, both of which are associated with increased postoperative complication rates [20, 21]. Our patients with hemiarch required a median circulatory arrest time of 11.0 min (IQR, 9.0, 15.0) and significantly longer CPB time (137.0 min vs 124.0 min non-hemiarch, P = 0.003). While these differences might be regarded as clinically benign, our study with appropriate statistical power might have unmasked their detrimental effect.

While hemiarch replacement may offer a similar long-term advantage of avoiding distal aortic reintervention in patients with aortic aneurysm as well, our findings suggest that arch reintervention is rare after a repair of proximal thoracic aortic aneurysm without arch aneurysm, and furthermore, a hemiarch repair does not reduce future aortic arch-related reinterventions. These observations add substantial affirmatory data to the mounting literature [7, 22, 23]. A study of 168 patients undergoing ascending aneurysm repair found no significant growth of the aortic arch over a median follow-up period of 5.9 years [22]. Iribarne et al. [24] found a low rate of reintervention (<4%) due to aneurysm formation in patients undergoing proximal aortic surgery (n = 869), suggesting that aggressive surveillance for aneurysm development over time may not be necessary. Similarly, a recent study of 702 patients with bicuspid aortic valves undergoing aneurysm repair found no significant difference in reintervention rates or death between patients with (n = 225) and without (n = 477) hemiarch replacement during a median follow-up of 6 years [7]. A conservative approach, therefore, does not seem to increase the risk of reoperations secondary to aortic growth over time.

Limitations

This is a single-centre retrospective study with a high volume of referrals for aortic surgery, which may limit generalizability, and there may be bias in the patients selected for surgery as well. As a retrospective study, biases and unmeasured confounders could not be excluded even after the matching. Our study could not account for the patient selection bias due to a difference in selection for patients undergoing hemiarch replacement at the time of repair, which was performed at the discretion of the surgeons. The higher in-hospital mortality rate in the hemiarch group might be attributable to such unmeasured variables and biases. We believe that our rigorous propensity score match on over 19 variables mitigated confounders. Moreover, robustness of the observed association against unmeasured confounders was assessed; in specific, the E-value for the hemiarch procedure (E-value, 3.74; 95% LL, 2.19) suggests that an unmeasured confounder would need an effect of at least 3.74-fold to negate the significance of the current association between hemiarch replacement and long-term mortality. Together, the association between hemiarch replacement and mortality appears to exist [17]. The small number of aortic arch reinterventions (8 total) also limits the ability to draw conclusion regarding the association between undergoing hemiarch repair and reintervention. The reintervention rate might be underestimated due to the loss of follow-up in a small subset of our patients. Moreover, the outcome of reintervention can be influenced by patient selection and thus may vary depending on the institution. Finally, clinical benefit of hemiarch replacement might manifest if the patients were observed longer period.

CONCLUSION

In conclusion, our data suggest that repairing the proximal arch in proximal aortic aneurysms does not decrease the need for long-term aortic arch reinterventions and, instead, unexpectedly show increased mortality, particularly in-hospital mortality. Larger, prospective multicentre studies, ideally a randomized trial, are warranted to further address the validity of this practice and help guide surgical decision-making.

SUPPLEMENTARY MATERIAL

Supplementary material is available at EJCTS online.

Funding

This work was supported by the National Institutes of Health [5T35HL007616-40 to Christian Pearsall].

Conflicts of interest: None of the authors have any conflicts of interest to disclose and have attested to their impartiality in order to be approved to participate in this project as per Institutional Review Board protocol at Columbia University.

ETHICS STATEMENT

This study was approved 31 March 2020 by the Institutional Review Board at Columbia University (AAAR2949).

Data availability statement

All relevant data within manuscript were obtained through the New York Presbyterian hospital electronic medical record and cannot be disseminated per HIPPA compliance.

Glossary

ABBREVIATIONS

CI

Confidence interval

CPB

Cardiopulmonary bypass

IQR

Interquartile range

SD

Standard deviation

 

Presented at the ESC 2021 Congress—The Digital Experience, 30 August 2021, European Society of Cardiology, Brussels, Belgium.

Author contributions

Christian Pearsall: Conceptualization; Data curation; Formal analysis; Investigation; Methodology; Project administration; Resources; Software; Visualization; Writing—original draft; Writing—review & editing. David Blitzer: Supervision; Writing—review & editing. Yanling Zhao: Data curation; Formal analysis; Resources; Software. Tsuyoshi Yamabe: Data curation; Supervision; Validation; Visualization; Writing—review & editing. Kavya Rajesh: Data curation; Resources; Extensive chart review on >1000 patients for manuscript revisions. Ilya Kim: Data curation; Investigation. Casidhe Bethancourt: Data curation; Formal analysis. Diane Hu: Data curation; Formal analysis; Methodology. Josh Bergsohn: Data curation; Formal analysis. Paul Kurlanksy: Conceptualization; Investigation; Methodology; Supervision. Isaac George: Conceptualization; Methodology; Project administration. Craig Smith: Formal analysis; Methodology; Project administration. Hiroo Takayama: Conceptualization; Formal analysis; Investigation; Resources.

Reviewer information

European Journal of Cardio-Thoracic Surgery thanks Ikuo Fukuda, Christian Olsson, Roman Gottardi and the other, anonymous reviewer(s) for their contribution to the peer review process of this article.