The arterial switch operation (ASO) has become the procedure of choice for treatment of infants born with transposition of the great arteries (TGA) and certain forms of double outlet right ventricle (DORV) (1- 5). Over the past 15 years, the results of the ASO have progressively improved as a result of steady refinements in diagnostic, surgical and perioperative management techniques (5- 9). Several studies published in the early to mid 1990s reported on the risk factors for morbidity and mortality of the ASO (10- 14). These studies were based on experience with the ASO from the 1980s through the early 1990s. In the most recently published cohort of patients with dextrotransposition of the great arteries (d-TGA) who underwent ASO between 1984 and 1991 at Children’s Hospital, Boston, Wernovsky and colleagues (12) found that independent risk factors for early death included several of the less common types of coronary patterns, including single right coronary and inverted coronary patterns. They also noted that aortic arch repair, the year of surgery and circulatory arrest time were independent risk factors for increased mortality. An increase in the risk of death associated with abnormal coronary artery patterns and with aortic arch obstruction was also reported in a multicenter study of the Congenital Heart Surgeons Society (10), as well as by individual centers (3- 4,13).

With increasing experience, improving perioperative outcomes and reassuring results of medium-term follow-up (1- 2,11- 15), patients with increasingly complex cardiovascular anatomy are undergoing ASO (16- 20). Included are patients with DORV, left and right ventricular outflow obstruction, multiple ventricular septal defects (VSD), anomalies of the atrioventricular canal and valves and aortic arch anomalies. As the application of the ASO to patients with increasingly complex anatomy and hemodynamics broadens and as methods for diagnosis and management of these infants evolve, it is likely that the risk factors for this operation also change. Specifically, our clinical observations suggest that in recent years, coronary artery pattern may not be associated with increased risk of early mortality after the ASO. Mayer and colleagues (21) speculated that although coronary artery pattern was an overall risk factor for early mortality, its impact over time would diminish. The present study, therefore, was designed to examine the hypothesis that coronary artery pattern is no longer an independent risk factor of early mortality, and to determine the independent risk factors in the current era for early mortality after ASO.

Patients were identified by a search of the prospectively recorded clinical databases of the Cardiovascular Program at Children’s Hospital, Boston. All patients who underwent ASO for TGA or DORV between January 1, 1992 and December 31, 1996 were included. Exclusion criteria included segmental anatomy of {S,L,L} (i.e., solitus atria, L-ventricular loop and L-transposition) prior atrial repair (Mustard or Senning operations) and concomitant atrial and arterial switch operation (“double switch”). Hospital charts, echocardiographic and cardiac catheterization data and operative reports were reviewed. During the study period, no patient was refused the ASO because of coronary anatomy. The study protocol was approved by the Committee on Clinical Investigations at Children’s Hospital.

Patient variables included demographic information, prematurity (gestational age <37 weeks), birth weight, age, weight and body surface area at surgery and associated noncardiac malformations. The following anatomic variables were recorded: presence or absence of VSD—patients were categorized as having an intact ventricular septum if no VSD or a tiny defect was documented by a preoperative echocardiogram and/or catheterization, and VSD was defined as a defect judged to be sufficiently large to warrant surgical closure at the time of ASO; the presence of multiple VSDs; qualitative assessment of the position of the great arteries relative to each other; right ventricular hypoplasia (defined as tricuspid valve annulus z-score <−2 or a non–apex-forming right ventricle), and aortic arch anomalies including interruption, coarctation or arch hypoplasia. Coronary anatomy was determined by direct visualization at the time of surgery. Coronary artery patterns were classified as previously published (21- 23) and are outlined in (Figure 1). Both the individual and grouped patterns were analyzed. Preoperative data included balloon atrial septostomy, prior cardiac surgical procedures and the use of extracorporeal support. Intraoperative procedural data recorded included surgical techniques such as Lecompte maneuver, VSD closure, aortic arch repair and right ventriculotomy, as well as takedown of a prior Blalock-Taussig shunt and pulmonary artery (PA) band. Total cardiopulmonary bypass (CPB) support times, circulatory arrest times, cross-clamp times, revision of the coronary anastomosis and delayed sternal closure were recorded. Postoperative data included length of mechanical support and ventilation, cardiac intensive care unit (ICU) and hospital lengths of stay, the need for early reoperation and mortality. In this study, early mortality was defined as death before hospital discharge or within 30 days of ASO. This, therefore, included one child who was discharged from the hospital on day 18 after surgery, and died at home five days later from a presumed arrhythmia, and one child who died 79 days postoperatively before discharge.

The primary outcome variable in the study was early mortality as described above. The relationships with perioperative risk factors were assessed by means of contingency table methods and logistic regression analysis. To explore the simultaneous effects of perioperative characteristics on early death, variables that were significant at the 0.1 level in univariate analysis were included in a multivariate logistic regression model. A significance level of 0.05 was required for retention in the multivariate model. Survival rates were estimated using the Kaplan-Meier method. Relationships between the preoperative variables, coronary artery pattern and diagnosis group, and selected intraoperative and postoperative variables were evaluated using the chi-square test, the Wilcoxon rank-sum test or the Kruskal-Walls test, as appropriate.

There were 223 patients who underwent ASO at a median age of 6 days and median weight of 3.5 kg. Patient characteristics are summarized in (Table le1). Of the 97 patients with VSDs (44% of the study population), 12 (12%) had multiple VSDs and six (6%) had conoventricular malalignment-type defects. Twenty-six patients (12%) had aortic arch obstruction requiring surgery; 13 underwent aortic arch repair before ASO, 11 patients had arch repair at the time of ASO and 2 had coarctation repair within the first year after ASO. Eight patients (4%) had right ventricular hypoplasia, according to the defined criteria. The position of the great arteries (aorta in relation to the PA) was anterior and rightward in 57% of patients, anterior/posterior in 15%, side-by-side in 8% and anterior and leftward in 5%, and in 1% of the patients, the aortic valve was posterior and rightward relative to the pulmonary valve. Other associated lesions included abnormal pulmonary valve morphology (n = 11), juxtaposed atrial appendages (n = 5), abnormal tricuspid valve attachments (n = 4), cleft mitral valve (n = 2), aberrant origin of the right subclavian artery (n = 2), straddling mitral valve (n = 1), right aortic arch (n = 1), primum-type atrial septal defect (n = 1), bicuspid aortic valve (n = 1), dextrocardia (n = 1) and situs inversus (n = 1). There were 12 patients (5%) in this series with Taussig-Bing anomaly, defined as a double outlet right ventricle with a subpulmonary ventricular septal defect, mitral-to-pulmonary valve fibrous discontinuity and side-by-side great vessels (24). Patients with Taussig-Bing anomaly differed from the d-TGA group in that they were significantly older at the time of surgery (median age at surgery for d-TGA 6 days, and for Taussig-Bing 70 days, p < 0.002). Repair beyond three months of age was performed in 17 patients (8%), 6 (35%) of whom underwent two-stage repair (25- 26) with a PA band placement before ASO. Most patients (n = 194, 87%) underwent preoperative balloon atrial septostomy, either in the ICU under echocardiographic guidance or in the cardiac catheterization laboratory. The distribution of coronary anatomy pattern and the associated mortality in each type are shown in (Figure 1). This distribution is comparable to the coronary distribution of the previous cohort (12). Of the 4 patients with intramural coronary anatomy, 2 had intramural left coronary arteries, 1 had an intramural right coronary and 1 had an intramural left anterior descending artery. The prior operations performed are summarized in (Table le2).

The salient intraoperative and postoperative variables are summarized in (Table le3). The surgical and perfusion management methods at our institution have been described elsewhere (27- 29). Revision of the coronary anastomosis at the time of the ASO was required in 17 patients (8%) due to post-CPB myocardial ischemia or inability to wean from CPB. Other procedures performed in addition to the ASO are summarized in (Table le2).

Details of postoperative management in the cardiac intensive care unit have been previously described (28). Fifty-four patients (25%) returned to the cardiac ICU with an open sternum. Delayed sternal closure was subsequently performed in the cardiac ICU in all survivors within three days. The median duration of mechanical ventilation was three days (range 1 to 26 days), with a median length of stay in the cardiac ICU of five days (range 2 to 28). The median time from surgery to hospital discharge for the survivors was 10 days (range 5 to 74 days). Duration of mechanical ventilation (4 vs. 3 days, p = 0.003) and length of postoperative hospital stay (11 vs. 9 days, p = 0.006) were significantly longer for patients with an associated VSD.

There were 16 early deaths (7%) in this cohort. Of the 109 patients who, according to criteria published in previous reports (10,14- 15), would be considered “low risk” candidates for the ASO (i.e., full-term infants with d-TGA and intact ventricular septum repaired in the first 2 weeks of life), there were three early deaths (2.7%). The first patient was a 4.4-kg full-term infant with an intramural right coronary artery who had an Aubert-type coronary conduit (30) and died on the 8th postoperative day of myocardial ischemia. Patient 2 (with “usual” coronary anatomy) had ongoing left ventricular anterior wall ischemia postoperatively requiring a second period of CPB and a revision of the coronary anastomosis. He died six days postoperatively with multisystem organ failure from persistent low cardiac output. Patient 3 had a massive capillary leak syndrome and severe pulmonary edema immediately postoperatively, a large right ventricular thrombus requiring reoperation for thrombus removal, subsequent severe right ventricular failure leading to extracorporeal membrane oxygenation and multisystem organ failure. He died on postoperative day 9. When mortality data for the entire cohort are examined with respect to time after ASO, the probabilities of survival at 1 week, 1 month and 6 months are 97%, 94% and 92%, respectively (Figure 2). The median time from surgery to early death was 9 days (range: 0 to 79 days).

Several risk factors were associated with early mortality by univariate analysis and are summarized in (Table le4). Patient characteristics included lower birth weight and prematurity. Anatomic variables included Taussig-Bing type DORV, right ventricular hypoplasia, aortic arch obstruction requiring surgical repair and side-by-side great vessel anatomy. The procedural risk factors for early death included takedown of PA band, longer intraoperative CPB times, reintervention at the coronary anastomosis site and delayed sternal closure. In the group as a whole, aortic arch obstruction requiring surgical intervention increases the risk of early mortality (23% vs. 5%, p = 0.005). However, eliminating the patients who had a preoperative aortic arch repair, the risk of death in patients who undergo arch repair at the time of arterial switch operation is not significantly different from the population as a whole (15% vs. 5%, p = 0.2). Therefore, aortic arch repair before the ASO was a significant risk factor for early mortality in the univariate analysis. Presence and repair of a ventricular septal defect, age at surgery and surgeon were not associated with increased early mortality. Calendar year of surgery was not associated with increased mortality.

The summary of coronary patterns and the distribution of deaths is shown in (Figure 1). None of the individual patterns nor any combination of patterns was significantly associated with early mortality by univariate analysis. Although coronary anatomy was not associated with increased mortality, there was an association with increased morbidity. Compared with usual coronary pattern, inverted coronary patterns and single right coronary artery were associated with delayed sternal closure (48% and 54% vs. 19%, p = 0.003). Inverted coronary arteries were also associated with longer duration of mechanical ventilation (median 5 days vs. 3 days, p = 0.008) and a trend toward longer length of postoperative hospital stay (12 days vs. 9 days, p = 0.08). Although circulatory arrest time did not differ among coronary groups, total cardiopulmonary bypass time was significantly longer in patients with inverted coronary anatomy (p = 0.02), with a trend toward longer CPB times among those with single right coronary origins (p = 0.06). These complex coronary patterns were also noted to have an increased risk of coronary reintervention in the operating room (p = 0.006).

Using risk factors associated with early mortality after ASO in univariate analysis, two multivariate models were constructed (Table le5). The first model employed only those variables available to clinicians preoperatively (preoperative model). Stated differently, this model assesses the risk of mortality before performing the ASO. The second model employs both preoperative and intraoperative variables and aims to assess independent risk factors at the conclusion of surgery before arrival in the cardiac ICU (postoperative model). The preoperative model indicates that lower birth weight (odds ratio 2.6, p = 0.02) and aortic arch repair before ASO (odds ratio 12.3, p = 0.007) are both risk factors for early mortality. The odds ratio for birth weight is that associated with a 1-kg decrease in weight. The risk of early mortality in patients with a birthweight of ≤2.8 kg was significantly increased compared with those with a birthweight >2.8 kg (17% vs. 4%, p = 0.004). Pulmonary artery banding before ASO was also strongly associated with early mortality. However, its strong association with aortic arch repair before ASO made it difficult to separate the individual effects of these variables and precluded the inclusion of both in the same model.

In the postoperative model, longer circulatory arrest time (odds ratio 1.8, p = 0.004) and total CPB time (odds ratio 1.3, p < 0.001) are associated with an increased risk of early mortality. The odds ratios reflect the risk of death associated with 10-min increments in support time. In addition, right ventricular hypoplasia was an independent risk factor of early mortality (odds ratio 8.3, p = 0.05).

During the past 15 years, the results of the arterial switch operation have improved substantially. The mortality rate has decreased from approximately 20% in the early 1980s to 3% in the full-term infants with d-TGA who are operated on in the first 2 weeks of life (1- 15). With continued improvements in the prompt noninvasive diagnosis, preoperative management, refinement of surgical techniques and intra- and postoperative management strategies, patients with increasingly complex anatomy are being referred for ASO. Hence, the risk factors for adverse outcome of the ASO are also likely to evolve. Comparison of patient characteristics in the present cohort with the previous cohort from our institution (12), as well as with other series, indicates that a higher proportion of patients with complex anatomy underwent ASO from 1992 through 1996. For example, compared with many of the initial reports on the ASO (11- 14,19), the current cohort has a higher incidence of patients with aortic arch anomalies (12% compared with 8% in the Wernovsky et al. series (12) and 5% in the series of Serraf et al. 19). More than 12% of this cohort had multiple VSDs, compared with 7% in the earlier cohort. The increasing complexity of patients undergoing ASO reported here may reflect an institutional referral pattern. It is also possible that with continued publication of favorable reports on the long-term outcome of the ASO, its application is being expanded to more complex congenital heart disease. The early mortality rate of the “low risk” infant with d-TGA continues to be low (2.7%), with a somewhat increased mortality rate when all patients undergoing ASO are included (7%). Analysis of candidate risk factors in this cohort revealed that factors affecting early mortality have evolved since the previous cohort.

Although previous studies found that complex coronary artery patterns were associated with increased mortality (10- 15,31), the results of this study show that the impact of coronary artery anatomy has diminished with surgical experience. Wernovsky et al. (12) found that inverted origins of the right and circumflex coronary arteries as well as the pattern of a single origin of the right and left anterior descending coronary arteries were risk factors for early mortality. Yamaguchi et al. (13) found an overall higher mortality rate in patients with unusual coronary anatomy. Mayer et al. (21) found that coronary anatomy impact over time seemed to decrease, despite an overall correlation of inverted coronary patterns with increased mortality rates. Similar to the results of this study, Quaegebeur and colleagues (1) found no effect of coronary anatomy pattern on early death in an early series of 66 patients. Although this study shows that coronary artery pattern was no longer associated with an increased risk of early mortality, there is an impact of coronary pattern on intraoperative variables and postoperative morbidity. Patients with inverted coronary origins and those with a single right coronary artery pattern had longer duration of cardiopulmonary bypass, a higher incidence of reintervention at the coronary anastomosis site and an increased risk of delayed sternal closure. Postoperatively, patients with inverted coronary anatomy had a longer duration of mechanical ventilation and a trend toward longer postoperative hospital stay. These findings likely reflect an evolution in progress of surgical management of complex coronary patterns. Although the techniques used to transfer these coronary arteries have evolved to such a degree that they are not associated with excess mortality, the time required to perform these procedures is still longer compared with uncomplicated coronary patterns.

Lower birth weight infants continue to be at increased risk for both mortality and morbidity after cardiopulmonary bypass and complex congenital cardiac surgery. These data corroborate prior studies (10,12,19) showing that repair of aortic arch obstruction before ASO both with and without right ventricular hypoplasia are important independent risk factors for death in this population. Further investigation of the right ventricular structure is needed to more accurately evaluate the ability of a small right side to accommodate an ASO in these patients.

Because of the strong association of pulmonary artery band with early aortic arch repair, it was not possible to evaluate the independent effects of these covariables. Intraoperative circulatory arrest times as well as overall cardiopulmonary bypass times also continue to be independent risk factors for early mortality.

The surgical approach in patients with d-TGA or DORV and associated aortic arch anomalies is still under debate. Some centers have advocated a staged approach, with early repair of the aortic arch (either coarctation or interrupted arch repair) with or without placement of a PA band to restrict pulmonary blood flow, followed by ASO and PA-plasty at a later date. Although this approach affords a low initial mortality and a left thoracotomy access for the aortic arch, the potential disadvantages include branch pulmonary artery stenosis secondary to PA band manipulation, neoaortic valve insufficiency and adhesions as a result of prior surgical intervention. Other studies have advocated a single-stage repair for this lesion (32- 35). Planche et al. (35) reported a mortality rate of 14% with primary repair compared with 30.7% mortality with two-stage repair. These results, however, are difficult to interpret, since the two series were from different time periods. Several centers have reported the feasibility and success of the anterior approach for aortic arch repair (36- 37). The data from this cohort suggest that the risks of repair of the aortic arch as a separate operation preceding the ASO exceeds the risk of the slightly longer intraoperative CPB times of the single-stage complete repair. The present study supports the feasibility of the single-stage approach, and more importantly, found that the two-stage approach for repair of aortic arch anomalies was an independent risk factor for early death. Single-stage repair with ASO and arch repair is now the procedure of choice for most patients with d-TGA and arch obstruction or interruption in our institution.

The procedure of choice for patients with Taussig-Bing type DORV remains controversial. Mavroudis et al. (18) compared the results of ASO (n = 16) with those of the Kawashima intraventricular repair (n = 4) in patients with Taussig-Bing anatomy and side-by-side great vessels. Similar to our data, they found a mortality in the ASO population of approximately 7%. None of the four patients who underwent the Kawashima operation had died. There is also evidence that the ASO can also be used successfully for Taussig-Bing anomaly with anterior–posterior great vessel anatomy (4,12,17- 20). The present study found that although the age at operation for this group was significantly older, the morbidity and mortality outcomes did not differ from the group of patients with d-TGA and VSD. The ASO continues to be the procedure of choice at this institution for patients with Taussig-Bing type DORV in which a biventricular repair can be performed.

Because the number of patients with intramural coronary arteries was small (n = 4), there was limited power for risk analysis of intramural coronary patterns. Another potential concern is the ability of the sample size in this cohort to detect an increase in risk of mortality secondary to abnormal coronary patterns. However, analysis of outcome data shows that the mortality rate in patients with usual coronary pattern and origin of the circumflex coronary artery from the right coronary artery (types I and II, Figure 1) was 7.18%, and the mortality rate in patients with unusual patterns was 7.14%. Therefore, any difference in risk appears to be negligible. In addition to that limitation, comparison to the previous cohort, even within the same institution, is complicated by differing patient characteristics, shifting referral patterns and changing management strategies. The limitation of highly associated variables is inherent to our study population.

Results of this study indicate that the arterial switch operation continues to offer an excellent surgical option for infants with d-TGA and certain types of DORV, even in the presence of complex associated lesions. Complex coronary branching patterns were no longer associated with an increased risk of early death but continue to be associated with longer ICU and hospital stays. Lower birth weight, right ventricular hypoplasia and staged repair of aortic arch obstruction continue to be independent risk factors for mortality after the ASO. Ongoing assessment of risk factors and long-term follow-up of the outcome of these children are imperative for continued evaluation of the ASO.

The authors thank David L. Wessel, MD, FACC for his useful comments and Ms. Emily Flynn-MacIntosh for artwork.