Leveraging Big Data for Small Patients: Identifying Maternal and Neonatal Risk Factors for Pediatric Organ Transplant

There is growing evidence that certain in-utero and neonatal exposures can increase the risk of a broad range of chronic diseases, such as hypertension, diabetes, and obesity.^1-4 However, an explicit link between these maternal and neonatal exposures and the need for pediatric transplants has not been firmly established. In part this is due to the challenges of studying rare exposures and rare outcomes—it is hard to identify enough patients to detect meaningful associations. Not surprisingly, most prior work trying to quantify the association between maternal and neonatal exposures and the need for pediatric transplant has been limited to small, single-center cohorts generally consisting of merely dozens of patients.^5-7 Moreover, these studies usually focus on one specific exposure (eg, gestational diabetes) or one specific type of transplant. A large population-based study could overcome these weaknesses and strengthen our understanding of the link between these perinatal exposures and the need for pediatric transplants.

In this issue of Transplantation, Cote-Corriveau et al leveraged a near-universal database from Quebec, Canada to identify a longitudinal cohort of 1 038 375 children born from 2006 to 2019.⁸ These children were followed for a total of 7.7 million person-years to identify those who received a transplant by age 14. In this cohort, 436 children ultimately received a transplant (0.042%), of which 148 required a solid organ transplant (0.014%), and 198 received a bone marrow or stem cell transplant (0.02%). A number of neonatal exposures were associated with a higher likelihood of pediatric transplant across a number of different transplant types. For example, preterm birth was associated with a 3.9-fold higher likelihood of liver transplant and a 6.3-fold higher likelihood of kidney transplant. Similarly, receipt of a neonatal blood transfusion was associated with a higher likelihood of heart or lung (51.2-fold), liver (6.7-fold), kidney (24.4-fold), and bone marrow or stem cell transplant (3.5-fold). A qualitatively similar higher likelihood of most types of pediatric transplant was found for a number of other neonatal exposures including low birth weight, neonatal sepsis, neonatal jaundice, and congenital anomalies. In contrast to neonatal exposures, the vast majority of maternal exposures (eg, gestational diabetes, preeclampsia, placental abruption, etc) were not associated with a higher likelihood of transplant, with the exception of oligohydramnios, which was associated with a 4.7-fold higher likelihood of heart or lung transplant, and a 16.8-fold higher likelihood of kidney transplant.

This large population-based study successfully leveraged the power of big data to study the impact of a number of relatively rare neonatal and maternal exposures on the need for pediatric transplant, with a statistical power not possible on the scale of typical single-center cohort studies. Nevertheless, several important issues remain. First, despite the inclusion of more than one million children in the study, only a handful ultimately required transplants (n = 436, 0.042%). So few outcomes likely limit the ability to detect some clinically meaningful associations despite a large number of patients in the study population, particularly for extremely rare exposures. Second, it is not immediately clear exactly how some of these exposures would lead to the need for a transplant. For example, how does a blood transfusion during the neonatal period make a child more likely to need a transplant? Is the need for a transfusion simply a marker for some other condition more clearly linked to the need for a transplant—such as a congenital anomaly? Or is there a biological mechanism that might explain this link? Observational studies are inherently unable to answer these types of questions. Third, are there variables not captured in the chosen database that might explain these associations instead (ie, unmeasured confounding)? Observational studies are particularly susceptible to this issue, and there are characteristics not available in the study dataset that could potentially confound the study results, such as race/ethnicity and the presence of certain genetic syndromes. However, despite these issues, this large population-based study extends the work of prior single-center studies by virtue of its large sample size and its inclusion of a wide variety of both maternal and neonatal exposures.

Overall, this was a well-designed cohort study that utilized a large provincial database to identify novel maternal and neonatal exposures that represent risk factors for the subsequent need for a pediatric transplant. These could, for example, lead to enhanced surveillance strategies for children with these risk factors, resulting in earlier diagnosis and treatment before the consequences of end-stage organ disease occur. Ultimately, this study expands our understanding of perinatal risk factors for pediatric transplant and highlights additional areas of research that could be pursued to further develop our understanding of the population-level epidemiology of pediatric organ transplantation.