Cesarean Section Is Associated with an Increased Risk of Acute Lymphoblastic Leukemia and Hepatoblastoma in Children from Minnesota

Caesarean section (C-section) has been examined as a risk factor for some childhood cancers as it has been shown that C-section can modify immune system function in the offspring (1). This is of particular concern as the rates of C-section have increased over recent decades to now include nearly one-third of live births in the United States, annually (2). Data from both the United States and international sites have shown a decrease in the vaginal birth after cesarean rate (3), an increase in prelabor C-sections (4), and an increase in C-sections following a previous C-section delivery over the past 30 years (2, 5–7). Mechanisms of action for C-section in childhood cancer development may be due to the neonate not experiencing the necessary stress of vaginal delivery that is thought to properly activate the hypothalamic–pituitary–adrenal axis and prime the immune system for future function (8). The absence of this priming event in children born via C-section, particularly prelabor C-section, is hypothesized to (i) contribute to poor innate and adaptive immune system formation demonstrated by higher rates of asthma and other conditions in these children (1, 9); (ii) lead to variation in gene regulation modified by epigenetic changes conferred via C-section (8); and (iii) result in failure to seed the neonatal gut microbiome with necessary bacterial species from the maternal vaginal canal (8, 10–12). These C-section–driven changes to the neonatal immune system may create a permissive environment for malignancies to develop.

We and others have reported an approximate 20% increased risk of B-cell acute lymphoblastic leukemia (ALL) among children born via C-section (13–16), particularly prelabor C-section (5, 13). Other studies have reported null associations between C-section and ALL (5, 17); however, older studies where C-section, particularly that without trial of labor, was less common may be driving the reported lack of association. In addition, C-section has been found to be associated with some embryonal tumors (16), including neuroblastoma (18, 19), though results have been inconsistent and are often based on small sample sizes and older studies where C-section was not as common as recent decades.

The incidence of C-section has increased substantially in the United States and around the world in recent decades (4, 20). Simultaneously, the incidence of some childhood cancers, including leukemias (21), hepatoblastoma (22), and malignant brain tumors, is also increasing (23). Therefore, using linked birth and cancer registry data spanning nearly four decades of cancer diagnoses in the state of Minnesota, we examined the association between C-section and childhood cancer risk overall and for 19 different cancer types. We performed sex-stratified analyses to elucidate the role of C-section in the observed sex differences in childhood cancer risk (24). C-section is a potentially modifiable risk factor; therefore, it is important to understand the contribution of C-section to childhood cancer development as a potential means to lessen the burden of pediatric malignancies.

This analysis includes incident primary cancers diagnosed in children ages 0–14 years identified from the population-based cancer registry in Minnesota from 1976 to 2014. These data are the result of a linkage between the Department of Health, Minnesota Cancer Surveillance System and the Minnesota birth registry for all cancers diagnosed, as described previously (25). Cases and controls were frequency matched approximately 1:5 on birth year (1989–2010). Approval for the study was obtained from the Minnesota Department of Health and from the University of Minnesota Institutional Review Board. Informed consent was waived.

Cancer type was classified using the International Classification of Childhood Cancer, Third Edition (26). Cancers included were ALL, acute myeloid leukemia (AML), Hodgkin lymphoma, non-Hodgkin lymphoma, Burkitt lymphoma, ependymomas, astrocytomas, intracranial/intraspinal embryonal tumors, neuroblastoma, retinoblastoma, Wilms tumor, hepatoblastoma, osteosarcoma, Ewing sarcoma, rhabdomyosarcoma, fibrosarcoma, intracranial/intraspinal germ cell tumors (GCT), extracranial/extragonadal GCTs, and thyroid carcinoma. Tumor types with fewer than five cases born via C-section or vaginal delivery were excluded from the analyses. Children with down syndrome listed on their birth certificate (n = 25 cases and n = 7 controls) were also excluded.

The exposure of interest was mode of delivery defined as vaginal (referent) or C-section as listed on the birth certificate. Covariates considered in adjusted models were selected a priori on the basis of established associations with childhood cancer (27–30) and in consideration of the study design of matching by birth year. Covariates include sex (male, female), birth weight category (g; <2,500, 2,500–4,000, >4,000), gestational age (weeks; <38, 38–40, 41–42; gestational ages ≥43 were excluded), maternal race (non-Hispanic white, non-Hispanic black, Hispanic, other), maternal age (years; <25, 25–<35, ≥35), maternal education [<high school (HS) grad, HS grad/GED/some college, college grad], plurality (single, multiples), and birth year (1976–1989, 1990–1995, 1996–2001, 2002–2014).

We used logistic regression to estimate ORs and 95% confidence intervals (95% CI) as the measure of association between C-section and childhood cancer overall and by type. Crude models were adjusted for the matching variable, birth year. Fully adjusted models accounted for sex, birth weight, maternal race, maternal age, maternal education, plurality, and birth year. Gestational age was not included in the fully adjusted models as it was missing for 8% of cases and is largely correlated to birth weight, which was included in the models. To account for potential reverse causation related to cancers diagnosed prenatally, we conducted sensitivity analyses restricting to cases 2 months and older at the age of diagnosis. An interaction between C-section and sex was evaluated in logistic regression models for each cancer and a likelihood ratio test P_interaction value was estimated. The association between C-section and cancer was estimated for males and females separately using the fully adjusted models and only including cases diagnosed at 2 months of age or older. To further examine the associations between C-section and ALL, we conducted additional analyses stratified by birth year, age at diagnosis, and maternal race. All statistical analyses were conducted in SAS version 9.4 (SAS, Cary, NC).

From 1976 to 2014, there were 3,166 cancers in children <14 years of age at diagnosis in Minnesota that met the inclusion criteria for this study. There were 20,598 controls frequency matched by birth year included in this analysis. A higher percentage of cases were male (cases 56.5%, controls 51.1%), weighed more than 4,000 g at birth (cases 14.5%, controls 13.0%), born to mothers with college education (cases 58.9%, controls 55.8%), and were non-Hispanic white (cases 87.6%, controls 84.9%; Table 1). Cases were more frequently delivered by C-section than controls, 21.3% compared with 19.1%, respectively. In 2002–2014, C-section was more common overall and was more frequent among cases (cases 42.2%, controls 30.7%). There was much variation in the percentage of births born via C-section when considering cancer type with osteosarcoma having the lowest at 11.9% and extracranial/extragonadal GCTs having the highest at 42.2%.

When considering the association between C-section and childhood cancer overall, in the fully adjusted model the estimate was elevated but included the null (OR_adj: 1.08; 95% CI: 0.98–1.20). There was much variation in the association between C-section and cancer risk when considering cancer type. While many estimates had CIs that spanned the null, elevated estimates were observed for tumors occurring in various anatomic locations (Table 2).

We observed a statistically significant association between C-section and ALL (OR_adj: 1.20; 95% CI: 1.01–1.42) and hepatoblastoma (OR_adj: 1.90; 95% CI: 1.05–3.44) in the crude and fully adjusted models (Supplementary Table S1). These results remained statistically significant in analyses that excluded cases diagnosed at <2 months of age (Table 2). Elevated, though nonstatistically significant, associations were observed for C-section and intracranial/intraspinal embryonal tumors (OR_adj: 1.30; 95% CI: 0.88–1.91), neuroblastoma (full OR: 1.17; 95% CI: 0.87–1.58), Wilms tumor (OR_adj: 1.30; 95% CI: 0.95–1.79), intracranial/intraspinal GCT (OR_adj: 2.07; 95% CI: 0.94–4.54), and extracranial/extragonadal GCTs (OR_adj: 1.99; 95% CI: 0.99–3.98). Results were similar when excluding children diagnosed at less than 2 months of age for all cancers except extracranial/extragonadal GCTs where the association was attenuated (Table 2).

There was evidence of a statistical interaction between C-section and sex for ALL, AML, ependymomas, neuroblastoma, and hepatoblastoma (all P_interaction <0.20; Table 3). In sex-stratified analyses, we observed a statistically significant association between C-section and ALL (OR_adj: 1.34; 95% CI: 1.04–1.72) and hepatoblastoma (OR_adj: 3.87; 95% CI: 1.30–11.57) among females. There was an elevated, though nonstatistically significant, association between C-section and neuroblastoma (OR_adj: 1.59; 95% CI: 0.99–2.55) and AML among females as well (OR_adj: 1.30; 95% CI: 0.71–2.40). In males, there was an elevated association between C-section and ependymomas (OR_adj: 1.53; 95% CI: 0.84–2.82).

To better understand the observed association between C-section and ALL, we conducted analyses stratified by birth year, age at diagnosis, and maternal race/ethnicity. When considering year of birth, there was a statistically significant association between C-section and ALL for 2005–2014 (OR_adj: 1.62; 95% CI: 1.11–2.34), which had the highest percentages of controls (25.4%) and ALL cases born via C-section at 34.0% (Table 4). When we considered age at diagnosis, the ALL cases diagnosed from 1–5 years of age were most strongly associated with C-section (OR_adj: 1.28; 95% CI: 1.02–1.61). Again, sample size was too small to estimate an association between C-section and ALL for infants less than 1 year of age at diagnosis. We observed an association between C-section and ALL among children born to mothers classified as non-Hispanic white (OR_adj: 1.24; 95% CI: 1.03–1.48), but this is the only racial/ethnic group for which there was adequate sample size to detect an association.

In our analysis of 3,166 children diagnosed with childhood cancers in Minnesota from 1976–2014, we observed a statistically significant, positive association between C-section and ALL and hepatoblastoma after adjusting for suspected confounders. In crude models, we observed significant associations between C-section and Wilms tumor, neuroblastoma, and extracranial/extragonadal GCTs, but these associations were no longer statistically significant when excluding cases diagnosed prior to 2 months of age (neuroblastoma and extracranial/extragonadal GCTs) and upon adjustment for suspected confounders (Wilms tumor). For neuroblastoma and extracranial/extragonadal GCTs, the findings align with these tumors forming and oftentimes diagnosed in utero requiring a C-section for safe delivery so cancer treatment can begin (31, 32). We did not observe an association between C-section and AML, and this finding is consistent with a previous report (13). Similarly, we also did not observe an association between C-section and lymphomas, which is consistent with our previous report (33) but contrary to the increased risk for non-Hodgkin lymphoma among children born via C-section that was reported in another small study (34). We also did not observe an association between C-section and sarcomas, which are generally diagnosed during adolescence and likely depend on the growth changes accompanying puberty, particularly bone sarcomas (24, 35–37). There were suggested statistical interactions between sex and C-section for ALL, AML, ependymomas, neuroblastoma, and hepatoblastoma with females having a statistically significant increased risk of ALL and hepatoblastoma when born via C-section. We also observed an increased, though nonstatistically significant, risk of neuroblastoma for females born via C-section while a suggested increased risk of ependymoma among males was observed. Collectively, the findings of our study suggest that C-section may be associated with increased risk for some childhood cancers, namely, ALL and hepatoblastoma, and that this increase may be stronger among females, among children born in the early 2000s for ALL, and among young children diagnosed with ALL.

Our findings of an increased risk of ALL among children born via C-section agrees with many prior reports on this association (13, 14, 38–43). We also observed that this increased risk was strongest in females, which has not been reported, and among children diagnosed from ages 1–5 years (14, 39). The finding of a higher risk of ALL among females born via C-section is counter to the increased risk of ALL among males we have reported previously (24, 44); however, Black and colleagues (2015) have reported that female neonates had higher proportion of prelabor C-sections (1), which are more strongly associated with ALL risk (13). Therefore, a female excess in planned C-sections may underlie our observation, but as this is registry data we cannot investigate this hypothesis. C-section rates have increased in recent decades, as shown in our controls, and we observed a nearly 1.6-fold increase in ALL risk among children born via C-section during 2005–2014. These findings suggest that ALL incidence may continue to increase, particularly in young children, if C-section rates remain constant or continue to rise.

Previous reports have also described an elevated, though often nonstatistically significant, risk of embryonal tumors, including neuroblastoma and some central nervous system (CNS) tumors, among children born via C-section (16, 18, 19). In our study, we did not observe a statistically significant association between C-section and neuroblastoma and the size of our effect estimate was smaller than others have reported, 1.17 in our study and 1.40 elsewhere; however, we were able to account for suspected confounders including age, maternal education and race, and we restricted to cases 2 months of age and older, all of which were not accounted for in other studies (18, 19). Importantly, we did observe a higher neuroblastoma risk in females born via C-section, which has not been demonstrated to date. In our study, we observed largely null associations for CNS tumors including ependymomas and astrocytomas overall with an elevated, but nonstatistically significant increased effect estimate observed for intracranial/intraspinal embryonal tumors, which are presumably largely medulloblastomas. In the literature, all CNS tumors are generally combined prohibiting us from making direct comparisons between tumor types. In general, there is an elevated, though nonsignificant risk for CNS tumors when the mode delivery is C-section (6, 16, 19).

Our finding of an increased risk of hepatoblastoma for children born via C-section disagrees with a previous report where a null association was observed (45) and requires further investigation. As this is population-based, birth registry data, we were unable to ascertain detailed information on birth defects beyond Trisomy 21; however, based on a recent report by Lupo and colleagues (2019; ref. 46), a number of birth defects including Trisomy 18, congenital heart anomalies, and genitourinary anomalies are all strongly associated with hepatoblastoma risk with HRs ranging from 8 to 80. Therefore, it is likely that some cases of hepatoblastoma may due to the presence of a predisposing birth defect that necessitated a C-section delivery representing confounding by indication in our study (47). However, this should be investigated in other studies of hepatoblastoma cases with information on birth defects and mode of delivery. It is unlikely that our findings of an increased risk of hepatoblastoma are largely due to reverse causation of a prenatal diagnosis as only one case was diagnosed with hepatoblastoma at less than 2 months of age in our study. Should C-section act as a biologic risk factor for hepatoblastoma this will be important information as hepatoblastoma is the fastest rising cancer around the globe and the increased C-section rates may add to this rise in diagnoses in the future (22).

Mechanisms of action for C-section in childhood cancer development may depend on neonatal immune system dysfunction that is the result of not experiencing the necessary stress of a vaginal delivery whereby activation of the hypothalamic–pituitary–adrenal axis occurs leading to normal immune system development, which is a multifaceted process (8). This missed experience in children born via C-section, namely, prelabor C-section, may lead to a poorly formed innate and adaptive immune system, epigenetic changes in gene regulation, and/or underlie variation in the gut microbiota (8, 10, 11). The responsible mechanisms may of course vary by cancer type in children such that hematologic malignancies, namely, ALL, may depend more heavily on the aberrant development of innate and adaptive immune responses, which are known to differ by sex (48). In a study of neonatal inflammatory markers in children that developed B-ALL, there were sex differences in markers including C-reactive protein (49); however, these markers were not associated with mode of delivery, but there were only 16 children born via C-section in this study. Our findings of a higher risk of ALL in females born via C-section taken together with these observations suggests a larger study of inflammatory markers in C-section born children is necessary to better understand the biologic mechanisms underlying the sex differences we have observed. Immune system function has been hypothesized to depend heavily on gut microbiota transferred to the newborn from the mother's vaginal canal, which is absent in C-section deliveries (8, 12). This theory draws on the hygiene hypothesis that the lack of exposure to the vaginal microflora leads to colonization of pathogenic bacterial species from the hospital environment in the neonatal gut that can lead to dysregulation of the immune system and the development of immune-related diseases, including cancer (20). In animal models, it has been observed that mice with an intact gut microbiome did not develop ALL, further highlighting the importance of a properly seeded microbiome in immune function and leukemogenesis prevention (50). The altered immune system function after a C-section is observed through the higher risk for asthma and other respiratory conditions, type 1 diabetes, celiac, and other gastrointestinal conditions in children born in this manner (1, 9). Furthermore, it has been shown that C-section is often associated with a reduced rate of breastfeeding initiation compared with vaginal delivery and the epidemiologic literature, especially for ALL (51), has shown that breastfeeding reduces risk for developing malignancies in the child. The clear association with C-section and ALL and other cancers, including hepatoblastoma in our study, suggests mode of delivery plays an important role in the development of some childhood cancers and may represent an avenue for prevention intervention (52).

Although we present a large, population-based study of the association between C-section and 19 types of childhood cancer, our study findings should be interpreted with the following limitations in mind. As our data arrived from birth certificates, we are unable to account for type of C-section, prelabor or emergency, which we have found to vary in association with ALL (13). We also do not have information on immunophenotypic or cytogenetic subtypes of ALL, which have varying associations with C-section (13, 53). To that end, our study lacks information on molecular subtypes for other childhood cancers and we do not have data on predisposition syndromes outside of Trisomy 21. Similarly, we are unable to account for all types of birth defects and therefore cannot disentangle the relationship between birth defects, C-section, and hepatoblastoma or other cancers. Because of sample size limitations, we are unable to examine more types of cancer, including additional histologies of CNS tumors, sarcomas, and GCTs. Because of the nature of registry data, we are unable to account for childhood cancer cases among children who were born in Minnesota and left prior to diagnosis and are thus captured in another state's cancer registry. We do not have access to information on other risk factors that may impact C-section rates, such as previous C-section, breastfeeding duration, or maternal preeclampsia rates, which would inform the interpretation of our findings. Our study also lacks information on environmental exposures. Previous reports have found the association between C-section and ALL to be higher among children born to Hispanic mothers; unfortunately, we had a very low proportion of Hispanic mothers in our study resulting in an imprecise estimate of this association.

In conclusion, our population-based, linked birth and cancer registry study in Minnesota confirmed the previously reported statistically significant increased risk of childhood ALL among children born by C-section. We also observed an increased risk of hepatoblastoma for children born via C-section. These associations were strongest among females suggesting an interplay between sex and C-section that may moderate the development of ALL and hepatoblastoma, which warrants further investigations in other studies. We demonstrated that as C-section has increased in recent decades the association between C-section and ALL is higher in magnitude for more recent years. These findings suggest C-section is a strong risk factor for ALL that may impact future incidence rates if C-section rates remain high or increase. If the association between C-section and hepatoblastoma is independent of birth defects, which needs to be verified in other studies, the increase in C-section rates may also contribute to the increased incidence of hepatoblastoma that we are observing globally (22). C-section is a potentially modifiable risk factor for childhood ALL and other cancers that may serve as a future avenue of intervention as we work to reduce the incidence of childhood cancers.

No disclosures were reported.

L.A. Williams: Conceptualization, data curation, formal analysis, writing–original draft, writing–review and editing. M. Richardson: Data curation, formal analysis. L.G. Spector: Resources, funding acquisition, writing–review and editing. E.L. Marcotte: Conceptualization, funding acquisition, investigation, methodology, writing–review and editing.

This work was funded by the Children's Cancer Research Fund (to L.A. Williams and E.L. Marcotte).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.