Outcomes of the arterial switch for transposition during infancy using a standardized approach over 30 years

Abstract OBJECTIVES The aim of this study was to describe the early and late outcomes of the arterial switch for transposition. METHODS A single-centre retrospective cohort study was conducted to assess the early and late outcomes of arterial switch performed during infancy using a standardized institutional approach between 1988 and 2018, compared by morphological groups. RESULTS A total of 749 consecutive patients undergoing arterial switch during infancy were included, 464 (61.9%) with intact septum, 163 (21.8%) with isolated ventricular septal defect and 122 (16.3%) with complex transposition with associated lesions, including 67 (8.9%) with Taussig–Bing anomaly. There were 34 early deaths [4.5%, 95% confidence interval (CI) 3.1–6.1] with only 10 (2.6%) early deaths since 2000. Complex morphology (odds ratio 11.44, 95% CI 4.76–27.43) and intramural coronary artery (odds ratio 5.17, 95% CI 1.61–15.91) were identified as the most important risk factors for 90-day mortality. Overall survival was 92.7% (95% CI 90.8–94.6) at 5 years and 91.9% (95% CI 89.9–94.1) at 20 years; in hospital survivors, there were 15 (2.1%) late deaths during a median follow-up of 13.7 years. Cumulative incidence of surgical or catheter reintervention was 16.0% (95% CI 14.5–17.5) at 5 years and 22.7% (95% CI 21.0–24.0) at 20 years; early and late reinterventions were more common in the complex group, with no difference between the other groups. CONCLUSIONS Using a standardized approach, the arterial switch can be performed with low early mortality, moderate rates of reintervention and excellent long-term survival. Concomitant lesions were the most important risk factor for early death and were associated with increased risk of late reintervention.


SUPPLEMENTARY MATERIALS Expanded Methods
All patients undergoing arterial switch during infancy at Birmingham Children's Hospital, UK between January 1988 and December 2018 were included. Patients over one year of age at arterial switch were excluded as they were predominately referred early in the series, mostly following banding and/or shunt formation elsewhere and are not representative of our current practice of neonatal anatomical repair.
Preoperative diagnosis and assessment were made by echocardiography and ductal patency was maintained with intravenous prostaglandin as required. Patients with inadequate mixing underwent emergency balloon atrial septostomy prior to repair. Initial palliative procedures were performed to relieve coarctation, increase or restrict pulmonary blood flow, or train the subpulmonary ventricle, as indicated. Arterial switch was used in all patients who were deemed suitable for anatomical correction and was not influenced by preoperative assessment of coronary anatomy. Additional procedures were performed at the time of arterial switch to treat residual associated lesions.

Operative technique
The surgical technique for arterial switch at our institution remained remarkably consistent throughout the study period. Under general anaesthesia and following median sternotomy, an autologous pericardial patch was harvested, and the proximal great vessels mobilized. Bicaval cannulation was used with cooling to a nasopharyngeal temperature of 28-32°C, although in the early years of this series, a single right atrial cannula was used in infants <3kg or without a VSD, with cooling to 22°C and a brief period of circulatory arrest for closure of the atrial communication. Alphastat pH strategy was employed during cardiopulmonary bypass (CPB). The arterial duct was double ligated, divided and both ends oversewn. Further mobilization of the pulmonary arteries to the hilar vessels was performed. The ascending aorta was clamped and cold crystalloid cardioplegia administered via the aortic root.
Additional procedures were performed to correct concomitant lesions, as required. Transatrial approach to a VSD was used routinely and patched with a bovine pericardium or double velour Dacron patch and interrupted pledgetted polypropylene sutures. The ventricular outflow tracts were carefully inspected and, when necessary, resection performed to ensure unobstructed blood flow. Any aortic repair was performed under deep hypothermic circulatory arrest at 18°C or selective antegrade cerebral perfusion with the arterial cannula advanced into the brachiocephalic artery. Isolated coarctation was treated by resection and extended end-to-end anastomosis, whilst aortic arch hypoplasia underwent resection of the coarctation and either extended end-to-end anastomosis or homograft patch augmentation of the underside of the arch. Following arch repair, the vascular clamp and arterial cannula were repositioned enabling re-perfusion of the aortic arch and descending aorta.
The ascending aorta was then transected 3-5mm above the aortic valve commissures and the coronary arteries were excised with a generous cuff of aortic sinus tissue and mobilized. The resultant aortic defects were repaired with untreated autologous pericardium as a single patch. The main pulmonary artery was transected at the same level as the aorta and the coronary artery buttons were relocated to medially hinged trapdoor incisions to construct the proximal neoaorta. Intramural coronary arteries were mobilized with generous cuffs of aortic wall, taking down the valve commissure or laying opening the ostia as required. If unable to transfer as described above, a homograft tissue patch was used to augment the receiving aortic sinus, incorporating the intramural cuff with minimal mobilization. The Lecompte manoeuvre was performed whenever possible; in those with side-by-side great vessels, the pulmonary artery bifurcation was relocated as necessary by incising the branch pulmonary artery to accommodate the main pulmonary artery, oversewing the native opening. The neoaorta was reconstructed, the heart re-perfused and reconstruction of the neopulmonary artery was completed while rewarming and with the heart beating.
CPB was discontinued with inotropic support and routine left atrial pressure line monitoring. Primary sternal closure was undertaken whenever feasible, or on the ICU as a delayed procedure, usually within 24-48 hours. The use of delayed sternal closure was determined by the degree of cardiac swelling and ventricular function post-CPB, along with the cumulative ischaemic time.

Data collection
Detailed morphological data were collected from the operative records. Coronary arteries were classified as intramural if one or more passed obliquely through the aortic wall. Single coronary orifice was defined as the left anterior descending, circumflex and right coronary arteries arising from a common origin in the ascending aorta, discounting minor accessory arteries.
Ethnicity was obtained from routine UK National Health Service data collection, collapsed to the five groups used by the National Institute for Cardiovascular Outcomes Research (NICOR). For missing data, the most likely ethnicity was derived by assessing the ethnic group most associated with the patient's surname in our hospital database. Where this data was missing, it was imputed based on the patient's surname from a regional database of almost 60,000 surnames with confirmed ethnicity. Data on late deaths were obtained from the UK Office for National Statistics (ONS) using the Demographics Batch Service Bureau (DBSB) service. Date of latest follow-up was the last documented clinic review by the patient's cardiologist or death. Surgical and catheter reinterventions were defined as any cardiovascular surgery or procedure performed at cardiac catheterisation, respectively, following completion of arterial switch.

Statistical analysis
Event rates for reinterventions were estimated using cumulative incidence function [S1] with death as the competing risk [S2]. To estimate risk factors for all-cause mortality, a logistic regression model was developed using a Bayesian method in the R BRMS package [S3], with 90-day all-cause mortality as the outcome. Variables were selected based on expert opinion and previous studies alongside the requirement for data availability because of the duration of patient selection. The following variables were included in the model: year of surgery, age at surgery, morphological group, non-usual coronary pattern, intramural coronary, single coronary orifice, aorta-pulmonary artery arrangement, and balloon atrial septostomy. A number of prior distributions were assessed. The published model used a sceptical prior with a Student t distribution, with mean zero and 3 degrees of freedom. Checks were made for colinearity of variables. The model's development diagnostics were reviewed using summary statistics and graphical output to ensure that the model had converged. Model performance was assessed using leave-one-out cross-validation. 95% compatibility intervals were derived from the posterior distribution and are presented.

Expanded Results
Since 1988, arterial switch has been our procedure of choice for the anatomical correction of transposition and other operations have only been performed when this approach was deemed unsuitable: 62 patients with transposition, VSD and left ventricular outflow tract obstruction underwent Rastelli operation (49), REV (Réparation à l'Etage Ventriculaire) procedure (12) or Nikaidoh procedure (1); 2 patients with transposition, VSD and subaortic stenosis underwent Damus-Rastelli procedure; and 96 patients with complex transposition, including with double inlet left ventricle, ventricular hypoplasia, tricuspid atresia, atrial isomerism, unbalanced atrioventricular septal defect or straddling atrioventricular valve, who were deemed unsuitable for biventricular repair underwent univentricular palliation. No isolated Mustard or Senning procedures for transposition or Kawashima intraventricular repair for Taussig-Bing were performed during this period. In addition, arterial switch was performed in 10 patients who underwent Mustard (5)
Detailed coronary artery patterns and associated mortality following arterial switch are provided in Supplementary Table 2. In the 36 patients with an intramural pattern, the left coronary artery was intramural in 19 (53%), the left anterior descending artery in 8 (22%), the right coronary artery in 6 (17%), and both the left and right coronary arteries in 3 (8%). An intramural course was uncommon in patients with usual sinuses of origin (5, 1.0%) but was frequently associated with an interarterial course in those originating from sinus 2 only (24, 52.2%). In 26 (3.5%) patients, the anatomy could not be categorised according to the Yacoub classification All deaths in those with coronary patterns other than the usual arrangement (1LCx-2R), circumflex from the right (1L-2CxR), and inverted circumflex/right (1RL-2Cx), occurred within 30 days of surgery. Both patients with the combination of a single coronary orifice and an intramural pattern died. The origin, course and branching pattern of coronary arteries originating only from sinus 2 is highly variable due to the greater number of possible arrangements, and the variants observed in this series are shown in Supplementary Figure 4.

Early outcomes
Nine (1.2%) patients required ECLS in the early postoperative period for poor ventricular function. Four failed to wean from CPB: three with complex transposition (two with Taussig-Bing, one with mesocardia and arch hypoplasia), and one with TGA-IVS and an intramural/interarterial left coronary artery. Five patients developed low cardiac output on ICU and required ECLS, four of whom underwent emergency reintervention for coronary revision: three with TGA-IVS (one intramural) and one with complex transposition. Of these nine patients, five were successfully weaned from ECLS and remain alive, two were weaned but died (one during the index admission, one whilst awaiting heart transplant), and two were unable to be weaned and care was withdrawn.

Survival by group
Of 706 hospital survivors, 504 remained under local follow-up, 189 at other UK centres, and 13 abroad; late outcomes were available for 621 (88.0%).