The question of whether prenatal diagnosis of transposition of the great arteries (TGA) affects patient outcomes is controversial. The rate of prenatal diagnosis of all congenital heart disease has increased over the past 3 decades with improved ultrasound technology and availability (Hill et al., 2015). It seems intuitive that prenatal diagnosis would further improve outcomes of patients with TGA by allowing more rapid resuscitation, yet most studies have not confirmed this hypothesis (Copel et al., 1997; Kumar et al., 1999; Mahle et al., 2001; Verheijen et al., 2001; Bartlett et al., 2004; Brown et al., 2006; Blyth et al., 2008; Atz et al., 2010; Levey et al., 2010; Landis et al., 2013; Morris et al., 2014; Oster et al., 2014; Wright et al., 2014). Three studies found a decrease in hemodynamic compromise in infants with a prenatal diagnosis of TGA with no change in mortality (Kumar et al., 1999; Verheijen et al., 2001; Brown et al., 2006). By contrast, three European studies have found significantly lower mortality in prenatally diagnosed patients with TGA (Bonnet et al., 1999; Khoshnood et al., 2005; van Velzen et al., 2015).

A major outcome of prenatal diagnosis is planned delivery at a cardiac surgical center (CSC). In a previous study, our group reported that increased distance between a birth hospital and a CSC in patients with hypoplastic left heart syndrome (HLHS) was associated with increased neonatal mortality compared with those born close to a CSC (Morris et al., 2014) and also found a significant difference in mortality based on CSC HLHS volume, which is a phenomenon that has previously been described (Scott and Fixler, 2001; Checchia et al., 2005; Hirsch et al., 2008; Welke et al., 2009; Pasquali et al., 2012; Karamlou et al., 2014). Recent studies have suggested that delivery location is critical for mothers with fetuses affected by TGA given the potential need for urgent intervention (Donofrio et al., 2014).

The purpose of this study was to use data from the Texas Birth Defects Registry (TBDR) to determine the associations between prenatal diagnosis, distance from birthplace to a CSC, CSC TGA volume and neonatal mortality in patients with TGA in a large, ethnically and socioeconomically diverse group of such patients over a 9-year period.

During the years 1999 to 2007, there were 3,401,057 live births to mothers who were residents of the state of Texas, of whom 468 met criteria for inclusion; birth prevalence of TGA was 1.4 cases per 10,000 live births. A VSD was present in 41.9%. Maternal and infant characteristics are listed in Table 1.

Forty-eight patients (10.3%) were prenatally diagnosed. Median driving time from birth hospital to a CSC was 17.3 min (interquartile range, 4.9–82.6 min). Thirty-five percent of the infants were born <10 min from a CSC, 41.4% were born 10 to 90 min away, and 23.6% were born >90 min away (Table 1). For the group with estimated driving times > 90 min, median driving time was 141 min, with range 91 to 598 min. During the study period, 11 CSCs performed the arterial switch procedure, and TGA volume ranged from 3 to 112 patients over the 9-year study period (Fig. 1). For 25 remaining patients, CSC was not documented in the registry, and these were excluded from the remainder of the analysis of hospital volume.

Two-hundred nine infants were documented as having undergone balloon atrial septostomy. Age at septostomy was available in 199 infants. For these infants, 57 (29%) underwent BAS on day of life 0, 73 (37%) on day of life 1, and the other 69 underwent BAS on day of life 2 or later. Sixty-four infants in the cohort (14%) met study criteria for HRAS.

Our findings show that in the overall Texas TGA population, prenatal diagnosis and decreased distance at birth from a CSC were not associated with improved neonatal survival. However, prenatal diagnosis may have the potential to modify outcomes at lower volume CSCs. The mechanism of this interaction could be twofold. First, prenatal diagnosis may allow lower volume centers to intervene rapidly with BAS in cases of an HRAS, where if postnatally diagnosed, deployment of resources may be slower than in institutions that perform these interventions more commonly. Indeed, neither of the prenatally diagnosed patients with an HRAS who delivered at a low-volume institution died in the neonatal period versus 3 of the 14 (21%) postnatally diagnosed patients. In the mid- and high-volume institutions, mortality in those prenatally diagnosed with an HRAS was 0/4 (0%) compared with 1/41 (2.4%) in those postnatally diagnosed. While these numbers are too low to prove the benefit of prenatal diagnosis in patients with HRAS at low-volume institutions, they are supportive of this theory. Second, it is possible that, for infants with TGA, prenatal diagnosis may in fact improve survival in patients cared for at low volume hospitals by prompting fetal cardiologists to refer more complex patients to higher volume centers and less complex patients to lower volume centers.

Prenatal diagnosis was not associated with overall lower mortality in our study as it was in some European studies (Bonnet et al., 1999; Khoshnood et al., 2005; van Velzen et al., 2015) (Table 4) possibly because of our lower rate of prenatal diagnosis and lower mortality limited our power to show a significant difference. However, even descriptive statistics and unadjusted analyses did not indicate a difference. Furthermore, there was enough power to demonstrate a significant interaction.

As shown in Table 4, the prenatal diagnosis rate in Texas during this period (10.3%) was much lower than some comparison studies. To determine whether this was due to under ascertainment of prenatal diagnoses by the TBDR, we conducted an independent study on a cohort of patients at Texas Children’s Hospital with TGA during the same time period and found a prenatal diagnosis rate of 15.4%. Given that Texas Children’s Hospital is a large referral center with its own fetal cardiology program, it is unsurprising that its rate would be higher; however, this rate is still relatively close to the statewide prenatal diagnosis rate obtained from the TBDR. Furthermore, while there are many population-based studies of prenatal detection of TGA, with rates ranging from 0% to 72.5% (Montana et al., 1996; Garne et al., 2001; Jaeggi et al., 2001; Stoll et al., 2001, 2002; Khoshnood et al., 2005; Khoo et al., 2008; Friedberg et al., 2009; Marek et al., 2011; Nielsen et al., 2014), a study by Pinto and colleagues using data from the Utah Birth Defects Network, a state with similar demographics to Texas, reported a prenatal detection rate of 14% (Pinto et al., 2012).

Despite the overall low mortality rate, there was a significantly higher mortality in the lower volume CSCs. Our findings are similar to those of Hirsch and colleagues who found an inverse relationship between surgical volume and postoperative mortality (Hirsch et al., 2008). They demonstrated a mortality rate in CSCs performing ≤2 ASOs per year of 9.4%, compared with 27.8% in our study, and a mortality rate of 3.2% in CSCs performing over 10 ASOs per year, compared with 2.3% in our study. Other studies have found an overall in-hospital mortality of 1.4 to 11.4% (Blume et al., 1999; Sarris et al., 2006; Lalezari et al., 2011; Fricke et al., 2012; Khairy et al., 2013; Anderson et al., 2014; Karamlou et al., 2014). Karamlou et al found that CSC volume and surgeon experience matter more in ASO surgeries, compared with Norwood procedures (Karamlou et al., 2010). This finding was strengthened in another study by Karamlou and colleagues, which showed that surgeon volume plays a larger role than center volume in outcomes of the ASO (Karamlou et al., 2014). Unfortunately, surgeon data were not available in the TBDR.

The previously cited studies describing the association between low surgical volume and increased mortality in congenital heart surgery were from large national data-bases in the United States (Scott and Fixler, 2001; Checchia et al., 2005; Hirsch et al., 2008; Welke et al., 2009; Pasquali et al., 2012; Karamlou et al., 2014). Although the data were consistent, application of these findings has been difficult in that often the selection of represented institutions was not consistent with real world options available to families, in that these hospitals are from many different states. This highlights one of the unique aspects of this study: our dataset was a complete 9-year birth cohort of TGA in a single state. Given that over half of neonates in Texas are covered by Medicaid, analysis of the relationship between volume and mortality among all CSCs within the state allows for examination of disparities in care among a similar payer pool (State of Texas, 2012). Our findings suggest that surgical site is a modifiable risk factor of significant importance, but one that may be offset by increased prenatal diagnosis.

It is difficult to determine why there was a higher mortality rate at the larger volume CSCs compared with the medium volume centers. Only one of the five deaths at the highest volume centers was associated with an HRAS, versus none at the medium volume centers. Both high and low volume centers had 11% of infants that were born preterm. Furthermore, it is unclear why patients with VSDs at high volume center had higher mortality. It may be that the patients at the larger CSCs (especially those with VSDs) had greater complexity not otherwise captured in the dataset.

Distance from a CSC was not associated with neonatal mortality among neonates with TGA, in contrast to our previous study in HLHS (Morris et al., 2014). In the HLHS cohort, the majority of deaths that were associated with increased distance from a CSC were presurgical. In comparison, the presurgical mortality in this study was quite low. Part of this difference in presurgical mortality is likely due to the fact that TGA is easier to diagnose postnatally than HLHS, as patients with TGA tend to be more cyanotic initially. This may increase early detection and offset the distance to a CSC phenomenon seen with HLHS. Also, patients with HLHS who decompensate develop shock, whereas patients with TGA most often present with severe hypoxemia, which is often less fatal and more responsive to prostaglandin therapy.

It will be interesting to see if the new recommendations for universal postnatal screening for critical congenital heart disease change outcomes with respect to prenatal diagnosis in TGA (Martin, 2013). Pulse oximetry screening detects hypoxia in those for whom it was not previously recognized. However, the primary contributors to mortality in TGA appear to be either an HRAS, which presents very early with severe hypoxia and is only improved with rapid intervention, or surgical complications. Neither of these would expect to be altered by routine screening. Also, age at diagnosis was not associated with mortality. In contrast, pulse oximetry screening may be critical in predominantly ductal-dependent lesions, which may not clinically present until closure of the ductus arteriosus. Early studies of pulse oximetry screening in TGA suggest that screening may prevent hospital discharge before diagnosis for some patients with TGA, but prenatal screening would still be critical in preventing adverse neurological outcomes and circulatory instability present in the sickest neonates with TGA (Bartos et al., 2015; Eckersley et al., 2015).

While the total number of patients in this study is large, the primary outcome of neonatal mortality is rare. This limited our ability to develop a multivariable regression model for neonatal mortality that included all potential confounders and covariates. For example, four of the five patients who died in higher volume CSCs had a VSD (p = 0.08); a larger study sample or higher mortality would have increased our statistical power and may have resulted in this being statistically significant.

Also, we were unable to reliably determine the degree of atrial septal restriction for all patients. While evidence of an HRAS was apparent in the medical records for 64 of the infants as defined, we could not reliably determine who did not have an HRAS, so did not believe this was appropriate to include in the multivariable analysis. For example, in some patients, there is only a record of an echocardiogram with TGA and a patent foramen ovale, BAS on day of life 1 and an arterial switch operation some days later. Without clinical records suggesting hypoxia, acidosis, or changes on x-ray in patients like this, it is challenging to know whether the BAS was performed somewhat electively or for clinical deterioration. As almost half of the population underwent BAS before ASO, this is too nonspecific of an indicator of an HRAS. Given this, we thought it was most appropriate to perform stratified analyses in the population which had evidence of an HRAS, and believe this subanalysis adds to the study.

This study used abstracted registry data and not full medical charts and some of our variables are missing values; for example, age at surgery was available for only 78.9% and age at diagnosis for 96.7% of the patients. To use a uniform population, patients with extracardiac birth defect or chromosomal abnormalities were excluded from analysis; the outcomes of these patients would be an interesting future analysis.

Coronary artery anatomy, which has been shown to be a risk factor for mortality in TGA was not available in this dataset (Pasquali et al., 2002).