Discussion

This is the first multicentre prospective study aimed to evaluate the association between impaired foetal environment and postoperative mortality in neonates with CHD. While the overall cohort did not demonstrate a statistically significant relationship between impaired foetal environment and mortality, a notable trend emerged among preterm infants with single-ventricle physiology, suggesting increased mortality associated with more severe impaired foetal environment. The absence of statistical significance in this subgroup is likely attributable to the limited sample size and the substantial proportion of missing data.

The impact of impaired foetal environment on postoperative outcomes in neonates with CHD was first described by Gaynor et al. in 2018. In this single-centre study, impaired foetal environment was commonly identified in pregnancies in which the foetus had CHD. Among neonates undergoing cardiac surgery, the presence of impaired foetal environment was associated with lower survival at 36 months of age for the entire cohort and for the subgroup with hypoplastic left heart syndrome. ^{Reference Gaynor, Parry and Moldenhauer10} However, prematurity was included in their definition of impaired foetal environment, potentially contributing to this large effect. In a follow-up study of 273 single ventricle infants after stage 1 Norwood palliation which did not include prematurity in the definition of impaired foetal environment (instead used it as a mediator), the hazard ratio of death at 36 months remained significantly elevated among infants exposed to impaired foetal environment (hazard ratio 6.05; 95% confidence interval, 3.59–10.21). ^{Reference Savla, Putt and Huang6} Population-based studies have shown conflicting results. The study using an administrative California-wide database confirmed the association of impaired foetal environment and 1-year mortality in infants with CHD, although the effect size was attenuated with a hazard ratio of 1.55 (95% confidence interval 1.34–1.78). ^{Reference Steurer, Peyvandi and Baer8} This decrease in effect size is not surprising, as administrative datasets rely on ICD codes and non-differential misclassification that biases the effect size towards the null is common. The same study demonstrated that small for gestational age status and placental abruption are the main driver behind the increased mortality rates in neonates with impaired foetal environment and CHD. ^{Reference Steurer, Peyvandi and Baer8} Interestingly, using the same database impaired foetal environment was shown to be a significant mediator explaining some of the racial and ethnic disparities. ^{Reference Santana, Peyvandi and Costello11} In contrast, Kristensen et al. showed no association of impaired foetal environment and mortality in children with CHD undergoing cardiac operations in Denmark from 1994 to 2018. ^{Reference Kristensen, Omann and Ekelund9} Possibly, these findings may be reflective of the marked difference in maternal health and prenatal health service of that country, as well as secondary to differences in the composition of CHD and age at surgery. In our prospective multicentre study, we found a trend towards higher mortality in preterm infants with single ventricle physiology exposed to impaired foetal environment. This is in accordance with the studies by Gaynor ^{Reference Gaynor, Parry and Moldenhauer10} and Savla ^{Reference Savla, Putt and Huang6} who also found the biggest impact on infants undergoing Stage 1 Norwood palliation.

The lack of statistical significance in our study may be attributed to decreased power due to the degree of absent data. Collecting prenatal maternal data in the neonatal cardiac population is a unique challenge, even above typical neonatal studies. In the past, these patients would often be admitted to a neonatal ICU and then transferred a cardiac ICU post-operatively. However, over the past decade we have seen a shift toward admission directly to a cardiac ICU from the delivery hospital ^{Reference Chaudhry, Sen and Steurer12} with variable involvement of a neonatal team. This leads to a lower likelihood of maternal data being recorded at the time of admission, as cardiac ICU staff are less likely (especially in the case of term neonates) to include such factors on initial documentation. Impaired foetal environment a neonate with CHD is born preterm and remains in the neonatal ICU prior to admission to the cardiac ICU, this information may be more easily collected, but only if the neonatal ICU is within the same electronic medical record system. Dealing with two medical records that need to be connected or linked is not always straightforward and depends on the electronic medical record setup of each individual hospital system. If a patient is born at a different hospital and then transferred, loss of access to maternal data is almost certain. We found that impaired foetal environment data were missing in more than 10% of patients. Maternal smoking, for example, was missing in >50% of records and thus was not used for data analysis but is known to be an important contributor to impaired foetal environment and therefore likely would have had a significant impact on our data. ^{Reference Savla, Putt and Huang6} The Cardiovascular Round Table of the European Society of Cardiology held a workshop among international experts in cardiovascular clinical research and health informatics to explore how electronic medical record could advance cardiovascular clinical research. ^{Reference Cowie, Blomster and Curtis13} Cowie et al. summarised their findings, and while acknowledging multiple limitations regarding electronic medical record in research, they don’t report on challenges associated with linking mother and baby records. It may be that European national health systems provide an improved infrastructure over private healthcare in the United States. However, the consistent ability to retrieve maternal, perinatal, and neonatal data would facilitate powerful research opportunities. ^{Reference Cowie, Blomster and Curtis13} Chaudhry et al. have previously discussed the need for the creation impaired foetal environment a specific database for neonates with CHD, acknowledging both—the data gaps that occur in neonatal databases regarding cardiac and surgical specific data and those that occur in cardiac/surgical databases regarding maternal, prenatal, and perinatal data. ^{Reference Chaudhry, Ball and Hamrick14} Without a centralised database that includes all levels of this data, abstracted by skilled neonatal AND cardiac champions, we will continue to miss opportunities for important clinical advances.

Factors that were clearly associated with increased missingness were prematurity, post-natal cardiac diagnosis, older age at surgery, lower STAT category, and biventricular circulation. Premature infants might have been born without much warning and thus without complete maternal work up. For example, gestational diabetes is commonly first tested for between 24 and 28 weeks gestation, and therefore, mothers who deliver prior to this point may not have a “diagnosis” of gestational diabetes even if they do have an impaired glucose tolerance environment. Antenatal diagnosis also leads to higher likelihood of delivery at a high-risk obstetric hospital with early transfer to the receiving paediatric critical care consortium centre. These patients were often already being followed by cardiology and therefore have more robust records even within the prenatal period. Older age at surgery, lower STAT category, and biventricular circulation were all associated with an increased degree of data gaps; these variables point towards less severe forms of CHD which may reflect lower rates of prenatal diagnosis and later transfer of neonates to the cardiac ICUs, leading to a lower likelihood of accompanying maternal data.

The extent of missingness of impaired foetal environment data is a serious limitation and was not anticipated in this study. It led to decreased power, and our negative results need to be assessed acknowledging this. The study per se was not powered for a certain effect size, and the enrolment was time limited. However, looking at the preterm single ventricle group, the absolute mortality difference in this study was 24.6% between neonates without impaired foetal environment and neonates with severe impaired foetal environment; this did not reach statistical significance given our small numbers (n = 23 in the no impaired foetal environment group and n = 15 in the significantly impaired foetal environment group). With a higher sample size (i.e., n = 42 in each group), this result would reach statistical significance with a power of 0.8 and a p-value < 0.05. It is important to state that despite the lack of statistical significance in this study, it would be wrong and potentially harmful, to deem impaired foetal environment as a non-important predictor. Further studies should focus on the preterm single ventricle group and aim for complete data and appropriate sample size.

The additional concern with missing data is non-randomness. For example, many Pediatric Cardiac Critical Care Consortium centres have associated high-risk delivery units with easily accessible maternal records. We performed several sensitivity analyses to demonstrate that missing data were most likely random. A second limitation is that the patient population was limited to those in a cardiac ICU. One must consider that the study population has an inherent selection bias: only neonates who survived to a CICU at a tertiary care centre were included. Neonates who were not referred because they were not deemed a surgical candidate, or whose families chose comfort care pathways were not included in our cohort, or those who suffered significant complications leading to mortality prior to transfer (e.g., a preterm neonate with hypoplastic left heart syndrome who developed necrotising enterocolitis and died during their neonatal intensive care nursery stay) were not included in our cohort. It is safe to assume that many of these babies had a component of impaired foetal environment and would have thus increased the effect of impaired foetal environment on this patient group. A third limitation in our study is that point evaluations for impaired foetal environment scoring may not appropriately reflect the relative magnitude of adversity attributed to an individual risk factor.

The evidence and the biological plausibility that impaired foetal environment impacts outcomes later in life in neonates with CHD is strong. What has started as an epidemiological observation over 3 decades ago—the Barker Hypothesis ^{Reference Barker15} —has reached in depth molecular understanding; metabolic imprinting and epigenetic influences have been found in many conditions confirming their foetal origin. ^{Reference Dyer and Rosenfeld16–Reference Monk, Mackay and Eggermann19} In this study, we identified the group of preterm infants with single ventricle physiology as the group that might be most vulnerable to an adverse impaired foetal environment.

Our findings underscore the need for improved data collection methods to accurately assess the impact of impaired foetal environment on neonatal CHD outcomes. Establishing integrated databases that link maternal, foetal, and neonatal records could enhance research capabilities. Future studies should focus on elucidating the mechanistic pathways through which impaired foetal environment affects neonatal outcomes and explores interventions aimed at mitigating these effects. Understanding the interplay between maternal health and neonatal CHD outcomes is crucial for developing comprehensive care strategies for this vulnerable population.