Exclusive Breastfeeding Duration and Risk of Childhood Cancers

This cohort study investigates whether the duration of exclusive breastfeeding is associated with risk of developing acute lymphoblastic leukemia and other childhood cancers.


Introduction
In Europe, cancer is diagnosed in 1 in 350 children before age 15 years and is the leading diseaserelated cause of death in childhood after infancy. 1,2At least 10% of cancers in childhood are attributable to rare germline mutations, 3,4 yet the etiology of most childhood cancers remains obscure. 5Consequently, there are currently no established preventive measures.
In this void, emerging research suggests that breastfeeding is associated with reduced risk of childhood cancers, such as acute lymphoblastic leukemia (ALL), the most common cancer in childhood.][8][9][10][11] Additionally, meta-analyses considering various durations of breastfeeding have found associations of breastfeeding with reduced risks of childhood acute myeloid leukemia (AML), 6,7,11 Hodgkin lymphoma, and neuroblastoma. 7,103][14] In line with this, aberrant immune responses to infectious stimuli are believed to play a central role in the development of B-cell precursor (BCP) ALL, the most common ALL subtype, in childhood. 15,16[18] The suggested protective effect of breastfeeding against childhood cancer is noteworthy.It not only points to potential biologic pathways that could modulate childhood cancer risk but also suggests a simple preventive measure.It is therefore crucial to revisit and corroborate previous observations of reduced childhood cancer risk in breastfed children, [6][7][8][9][10][11] as existing evidence stems from case-control studies, which are inherently vulnerable to recall and selection biases.
In this study, we leveraged the administrative records from the Danish National Child Health Register (DNCHR) to conduct a register-based cohort study.We aimed to investigate the association between exclusive breastfeeding duration and risk of childhood cancers among children born in Denmark between 2005 and 2018.

Methods
According to the European Union General Data Protection Regulation (Article 30), this cohort study was listed in the record of processing activities for research projects in and was approved by the Danish Cancer Society.As per Danish law, purely register-based studies such as the present one do not require ethics approval or informed consent.The study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cohort studies.
We identified all children born in Denmark from January 2005 to December 2018 in the Danish Civil Registration System. 19For each child, the information retrieved from the register included sex, date of birth, vital status (ie, dates of death or emigration), residence, birth dates of parents and siblings, and parents' place of birth.Furthermore, using the unique identification number issued to all individuals living in Denmark as the key, we linked with the Medical Birth Register to obtain information on birth characteristics and with the Population Education Register to ascertain the mother's highest educational level. 20,21Children missing information on birth weight, gestational age, mother's age, and mother's educational level were excluded from the analyses.We further excluded children with Down syndrome (International Statistical Classification of Diseases and Related Health Problems, Tenth Revision code Q90) through linkage with the Danish National Patient Register due to their increased risk of leukemia with distinct biology. 22,23bsequently, we linked the cohort with the DNCHR to obtain information on breastfeeding. 24is database holds information collected by health care nurses at regular home visits during the

JAMA Network Open | Pediatrics
Exclusive Breastfeeding Duration and Risk of Childhood Cancers child's first years of life.These visits are offered to all parents of newborns in Denmark to monitor the child's health and to provide guidance to parents regarding their child's development, feeding, and other health-related factors.
In the database, breastfeeding data specifically pertain to the duration of exclusive breastfeeding.This period was defined as when lactation was the child's primary nutrition supplemented only by water and/or, at most, 1 formula milk meal weekly. 25We extracted dates of exclusive breastfeeding cessation from the database, that is, when the child first received multiple formula meals weekly or solid foods.
Between 2005 and 2011, only a select number of Danish municipalities used the DNCHR.
Beginning in 2012, reporting to the database became mandatory for all municipalities, and most complied.Accordingly, we excluded children from the cohort if their breastfeeding information was missing in the database.We calculated exclusive breastfeeding duration as the time from birth to the date of exclusive breastfeeding cessation.We classified children as "never breastfed" if the date of exclusive breastfeeding cessation was less than 14 days after birth, assuming that the cumulative exposure to breastfeeding was minimal among these children.In addition, we identified children diagnosed with cancer at ages 1 to 14 years through linkage with the Danish Cancer Register 26 using morphology and topography codes defined in the International Classification of Childhood Cancer (Third Edition) (ICCC-3). 27

Statistical Analysis
The children in the cohort were followed up from age 1 year until the date of childhood cancer diagnosis, loss to follow-up or emigration, death, age 15 years, or December 31, 2020, whichever occurred first.We used Cox proportional hazards regression models stratified by sex and birth order (1, 2, or Ն3) with age as the underlying time scale to estimate hazard ratios (HRs) and 95% CIs for the association between exclusive breastfeeding duration and childhood cancer risk.Analyses were conducted from March to October 2023 using R, version 4.2.2 (R Project for Statistical Computing) on a dedicated Statistics Denmark server.P values and 95% CIs were derived using likelihood ratios.Two-sided P < .05 was considered significant.
Based on their potential association with both breastfeeding duration and childhood cancer risk, we adjusted for several potential confounders: mother's age at delivery (linearly, 1-year intervals), birth weight (linearly, 1-g intervals), gestational age (linearly, 1-day intervals), mode of birth (vaginal or cesarean delivery), birth year (linearly), and mother's highest achieved educational level (low: mandatory school, Յ9 years; medium: upper secondary or high school or vocational education, 10-12 years; and high: Ն13 years).Analyses were carried out for childhood cancers combined and for the 3 major ICCC-3 subtypes: hematologic cancers, central nervous system (CNS) tumors, and solid tumors.Additionally, we conducted separate analyses for ALL and BCP-ALL (morphology code 9835) and neuroblastoma (morphology codes 9490 and 9500).
In analyses of childhood hematologic cancer, ALL, and BCP-ALL, we divided the duration of exclusive breastfeeding into approximately 3-month intervals: never or less than 14 days, 14 days to 2 months, 3 to 5 months, and 6 or more months.Additionally, to increase statistical power, breastfeeding duration was grouped into 2 categories (0-2 months and Ն3 months) to compare longer and shorter breastfeeding durations while also considering the number of events.For CNS and solid tumors, we analyzed exclusive breastfeeding duration in 2 categories only (0-2 months and Ն3 months) due to few events.For childhood cancers combined and for the aforementioned cancer subtypes, we also assessed log-linear trends in HRs for the association with exclusive breastfeeding duration per month excluding children who were exclusively breastfed never or for less than 14 days.
In supplementary analyses, we assessed the risk of childhood BCP-ALL associated with exclusive breastfeeding duration according to birth cohort with voluntary vs mandatory reporting to the DNCHR (2005-2011 vs 2012-2018), the attained age at which BCP-ALL incidence peaks (2-6 years) vs older ages (7-14 years), and birth mode (vaginal vs cesarean delivery).The proportional hazards assumption was evaluated by visual inspection of scaled Schoenfeld residuals.No obvious violations were detected.

Results
Information on duration of exclusive breastfeeding was available for 318 034 children.a Some individuals were missing information in more than 1 category.

JAMA Network Open | Pediatrics
Exclusive Breastfeeding Duration and Risk of Childhood Cancers Compared with the entire cohort, children diagnosed with cancer were less likely to be the first born in their families and to have a higher weight at birth.In addition, children with ALL were more likely to have a mother aged 35 years or older at the time of delivery compared with children in the entire cohort (Table 1).
In the full cohort, 104 132 children (33.6%) were exclusively breastfed for less than 3 months.
Log-linear trends in HRs for the association of cancer risk with 1 additional month of exclusive breastfeeding are presented in Table 2.While the HR was close to 1 for most of the investigated cancers, log-linear trend adjusted HRs (AHRs) were 0.91 (95% CI, 0.83-0.99)for hematologic cancers overall and 0.90 (95% CI, 0.80-1.01)for BCP-ALL.In dichotomized analyses, compared with exclusive breastfeeding for shorter periods, exclusive breastfeeding for at least 3 months was associated with a reduced risk of hematologic cancers (AHR, 0.66; 95% CI, 0.46-0.95).This risk reduction was largely attributable to a decreased risk of BCP-ALL (AHR, 0.62; 95% CI, 0.39-0.99).In contrast, risk of CNS tumors (AHR, 0.96; 95% CI, 0.51-1.88),solid tumors (AHR, 0.87; 95% CI, 0.55-1.41),and neuroblastomas (AHR, 0.98; 95% CI, 0.32-3.06)did not vary by exclusive breastfeeding duration (Table 2).The small number of children diagnosed with AML or Hodgkin lymphoma in the cohort precluded estimation of the association of exclusive breastfeeding with these cancers (eTable 1 in Supplement 1).In the supplementary analyses stratified by attained age, birth cohort, and birth mode, we found that children exclusively breastfed for at least 3 months had a decreased risk of BCP-ALL at ages 2 to 6 years and 7 to 14 years (P = .58for interaction), among children born between 2005 and 2011 and children born between 2012 and 2018 (P = .20for interaction), and among children delivered vaginally and those born by cesarean delivery (P = .92for interaction) (eTable 2 in Supplement 1).

Discussion
In this prospective cohort study, we used unique register data to assess the association between breastfeeding duration and risk of childhood cancer.Our analyses showed that children exclusively breastfed for at least a lower risk of BCP-ALL at ages 1 to 14 years than did children breastfed exclusively for less than 3 months or never at all.Also, we observed no association between exclusive breastfeeding duration and risk of CNS tumors or solid tumors at ages 1 to 14 years.][8][9][10][11] Notably, our results for BCP-ALL align with the approximately 30% reduced ALL risk in children breastfed exclusively for at least 4 months vs never breastfed in recent pooled analyses of international case-control studies including more than 10 000 children with ALL. 11rrent models posit that childhood BCP-ALL development often begins before birth, when for unknown reasons, an initial genetic event gives rise to a preleukemic clone (such as ETV6-RUNX1). 15,16e prevalence of preleukemia in neonates is still debated, 28,29 but most likely, BCP-ALL develops in only a small proportion of children born with preleukemia. 15It has long been speculated that in these children, the malignant transformation of a preleukemic clone into ALL is triggered by a dysregulated immune response to infections. 15,16e association between breastfeeding and childhood BCP-ALL risk, if determined to be causal, could be mediated by preventing this immunologic dysregulation. 30]31 The role of gut microbiome maturation in childhood ALL pathogenesis and the potential for gut microbiome-targeted preemptive interventions has recently gained increasing attention. 16,30veral studies have shown that children with ALL have alterations of their gut microbiome at diagnosis compared with healthy peers. 30,32,33While the sequence of events is difficult to disentangle from such studies, murine models of BCP-ALL have found that in genetically predisposed mice (eg, Pax5 +/− ), gut microbiome alterations precede leukemia onset. 17,18Moreover, in these predisposed mice, leukemia development could be triggered by the destruction of the gut microbiome with antibiotics, even in the absence of infectious stimuli. 17It has therefore been suggested that suboptimal enrichment and delayed maturation of the gut microbiome may increase BCP-ALL risk in children born with preleukemic cells. 16,17,30 our study, we could not ascertain whether the association between breastfeeding and childhood BCP-ALL risk primarily reflects benefits of breast milk itself, delayed introduction to formula milk, or a combination of both.In the aforementioned international investigation, 11 introduction to formula milk within the first week of life was associated with an increased risk of childhood ALL even among children who were breastfed for at least 6 months.In the current study, we did not have data to assess the precise age at formula milk introduction.Still, because the Danish definition of exclusive breastfeeding includes supplementation with up to 1 formula meal weekly, our results suggest that breastfeeding is associated with decreased risk of childhood BCP-ALL despite infrequent formula meals.
During the neonatal period, the mode of birth is the predominant factor associated with the composition of the gut microbiota. 34Children born by cesarean delivery display reduced overall gut microbiome stability during the first months of life and delayed maturation of the gut microbiome by the second year of life. 31,35Furthermore, cesarean delivery is a suspected risk factor for childhood BCP-ALL. 36In this view, it is notable that we found longer duration of exclusive breastfeeding to be associated with decreased risk of BCP-ALL similarly across birth modes (ie, both among children born by vaginal delivery and by cesarean delivery).No association between mode of birth and risk of childhood ALL was found in the present study or in previous, larger register-based studies from our group. 37,38me case-control investigations previously showed that longer duration of breastfeeding was associated with decreased risk of rare childhood cancers, including AML, 6,7,11 Hodgkin lymphoma, 7 and neuroblastoma. 7,10While the limited number of cases in the present study precluded such assessments for AML and Hodgkin lymphoma, our findings did not support an association between exclusive breastfeeding duration and risk of neuroblastoma or solid tumors overall.Moreover, in keeping with our findings, breastfeeding duration was not associated with risk of CNS tumors in a recent pooled analysis of case-control studies including more than 2600 cases 39 or in previous meta-analyses. 7,10

Strengths and Limitations
Strengths of our study include its prospective, register-based design with independent ascertainment of exclusive breastfeeding duration and childhood cancer diagnoses and inclusion of a representative sample of the Danish childhood population across geographic regions and socioeconomic groups.Through linkage between the Danish nationwide registers, we could adjust This represented 44.9% of the 708 521 children born between 2005 and 2018 in municipalities that contributed to the DNCHR.Database coverage varied over time, from 63 066 of 296 871 children (21.2%) born between 2005 and 2011, when reporting was voluntary, to 254 968 of 411 650 children (61.9%) born between 2012 and 2018, when reporting was mandatory.The Figure provides an overview of inclusion in the study, with criteria for exclusion, case numbers, and reasons forcensoring.The final study cohort included 309 473 children (51.3% boys; 48.7% girls).3%) had other and unspecified malignant neoplasms.Among the children diagnosed with hematologic cancers, 81 (65.3%) had ALL, of whom 74 (91.4%) had BCP-ALL; 7 (5.6%) were diagnosed with AML, and fewer than 5 were diagnosed with Hodgkin lymphoma.The most frequently diagnosed CNS tumors were astrocytoma, ependymoma, and intracranial and intraspinal embryonal tumors.Kidney tumors, neuroblastoma, and soft tissue sarcomas were the most common solid tumors (eTable 1 in Supplement 1).

Table 1 .
Baseline Characteristics of the Study Population and Duration of Exclusive Breastfeeding in Infancy According to Childhood Cancer Overall and ALL Abbreviation: ALL, acute lymphoblastic leukemia.a Exact numbers are not presented to blind numbers less than 5 (directly or by calculation through group totals) in accordance with the interpretation of the General Data Protection Regulation by Statistics Denmark.

Table 2 .
Hazard Ratios of Childhood Cancer Associated With Exclusive Breastfeeding Duration by Childhood Cancer Groups and Diagnoses

Table 2 .
Hazard Ratios of Childhood Cancer Associated With Exclusive Breastfeeding Duration by Childhood Cancer Groups and Diagnoses (continued) Adjusted for year of birth (linearly), birth weight (linearly, in 1-g intervals), gestational age (linearly, in 1-day intervals), mother's age at delivery (linearly, in 1-year intervals), mode of birth (vaginal or cesarean delivery), and mother's highest educational level and stratified by sex and birth order.P values and 95% CIs were based on the likelihood ratio.Children exclusively breastfed for less than 14 days or never were excluded.Exact number is not presented to blind numbers less than 5 (directly or by calculation through group totals), in accordance with the interpretation of the General Data Protection Regulation by Statistics Denmark.