Background
The incidence and mortality associated with childhood cancer is increasing sharply in developed and developing countries [
1]. The most common childhood cancers include acute leukemia at 26.3%, especial for acute lymphoblastic leukemia (ALL) [
2], central nervous system tumors at 17.6%, and lymphoma at 14.6% of all cancers [
3].
Breastfeeding is the major food for newborn babies, who receive almost all essential needs to meet the requirements of growth and development [
4]. It has become a universal phenomenon for babies to be fed with formula in recent years; nevertheless, components of formula are different from those of breastfeeding. Breastfeeding has components not found in formula, including active hormones and peptides, all of which play essential roles in development during the newborn period and infancy. A researcher suggested that breastfeeding might help prevent childhood cancer by stimulating the immune system. To date, several lines of evidence suggest that breastfed babies are healthier, with benefits of increased immunity and intelligence, lower incidence of sudden infant death, childhood obesity, and allergies [
5‐
7]. There was a hypothesis that breastfeeding might help protect against several childhood diseases, not limited to ALL [
8].
There were eight studies reporting the results of meta-analyses on the relationship between breastfeeding and childhood cancer [
4,
9‐
15]. The relationship of breastfeeding and hematological malignancy-related diseases was confirmed by several studies of these studies [
4,
9‐
12,
14,
15]. Two studies expanded the scope of research to the relationship between breastfeeding and central nervous system diseases [
13,
14]. There was little evidence to suggest that breastfeeding was significantly associated with acute non-lymphoblastic leukemia (ANLL) [
16‐
18], non-Hodgkin’s lymphoma (NHL) [
9,
16,
17,
19,
20], central nervous system cancers, malignant germ cell tumors, juvenile bone tumors and other solid cancers. None of these studies mentioned breastfeeding and other systemic diseases and the influence of breastfeeding on childhood cancer. Moreover, a clear association between modes of breastfeeding (or formula) and risk of childhood cancer had not been explored in detail in previous studies. The World Health Organization (WHO) suggested that breastfeeding had additional health benefits that extend into adulthood, and it recommended maintaining breastfeeding for 2 years or longer [
21]. With respect to the duration of breastfeeding, there were controversy among researchers and no consensus had yet been reached. Given the lack of comprehensive and systematic research confirmed by reviewing the literature, this meta-analysis was conducted to comprehensively explore the association of breastfeeding and childhood cancer, involving several countries, modes of breastfeeding, various feeding durations, and systemic diseases of childhood.
Methods
Search strategy
PubMed and Embase were systematically searched for relevant studies that met our eligibility criteria. The literature search was carried out on January 10, 2021 to identify published studies on the relationship of breastfeeding and cancer in children. Two authors were responsible for screening the studies to obtain full manuscripts, as well as the titles and abstracts. To ensure completeness and accuracy of this meta-analysis and systematic review, two reviewers participated in the entire literature search process without interfering with one another. The search terms for this study were as follows: “Childhood”, “Children”, “Child”, “Neoplasms”, “Cancer”, “Tumour”, “Neoplastic”, “Malignancy”, “Bottle-feeding”, “Breastfeeding”, “Infant nutrition”, “Perinatal”, and “Milk”, with language and publication status restrictions. The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) was followed [
22].
Eligibility criteria
Studies were included in the meta-analysis if they met the following criteria: (1) the study was published in English; (2) the exposure of interest was breastfeeding; (3) the outcome was childhood cancer; (4) the age of included population was less than 18 years old; and (5) the estimates of the relative odds ratio (OR) with 95% confidence interval (CI) were reported.
Exclusion criteria were as follows: (1) duplicated reports; (2) defective study designs; (3) incomplete data and uncertain outcome effects; and (4) incorrect statistical methods or those that could not be amended, could not be provided, or could not be converted into OR.
The following data were extracted from each study: country of the patients, individual ages and average ages, types of cancer (e.g., hematological malignancies including lymphoma and leukemia, and cancers of the nervous, motor, urinary, reproductive, and sensory systems), the modes of breastfeeding (e.g., exclusive breastfeeding, mixed breastfeeding, and bottle feeding or formula feeding; non-exclusive breastfeeding including mixed and formula feeding; the same classification applied for non-mixed and -formula feeding), feeding duration, and the number of patients in the breastfeeding and control. Adjustment factors and other directly extractable data were first extracted for each study. The durations of breastfeeding were classified as either past breastfeeding or never breastfed, greater than or equal to 6 months or less than 6 months, and greater than or equal to 12 months or less than 12 months.
Validity assessment
The Newcastle-Ottawa Scale (NOS) was used to assess quality of included studies in this meta-analysis and systematic review [
23]. NOS includes case definition, representativeness of the cases, selection of controls, definition of controls, comparability of cases and controls, ascertainment of exposure, same method of ascertainment for cases and controls, and no response rate. Each asterisk in the table represents a point, and the final score is the sum of the points.
Statistical analysis
The effect of breastfeeding on childhood cancer was analyzed using OR and 95%CI as the effect measure. Between-study heterogeneity [
22] was assessed using Cochran’s Q and I
2 statistics. Initial analyses with I
2 < 40% were performed using a fixed-effects model; otherwise, a random effects model was adopted. Potential confounders, including the modes of breastfeeding (exclusive, mixed, and formula with different durations of breastfeeding), duration of breastfeeding (including ≥1 month vs. < 1 month, ≥6 months vs. < 6 months, and ≥ 12 months vs. < 12 months), different countries, and cancers of different systems were regarded as the main sources of heterogeneity. Subgroup and sensitivity analyses were employed. We used the one-stage robust error meta-regression model to establish the potential dose-response relationship between duration of breastfeeding and risk of cancer [
24,
25]. This was a one-stage method that treated each study as a cluster combined with robust error estimation as a solution to deal with potential correlations within each study [
26]. The restricted cubic spline function with three auto-generated knots was used to fit the potential non-linearity trends [
27]. The remr command of Stata software was used to run the dose-response meta-analyses [
28]. All statistical analyses were performed using RevMan 5.3 and STATA 15.0.
Discussion
Childhood cancer is a major cause for death of children and adolescents in many countries. The relationship of breastfeeding and the incidence of childhood cancer had been addressed by several researchers. This meta-analysis and systematic review was conducted to explore the relationships between breastfeeding and childhood cancer, and a total of 46 studies were included in this study. To the best of our knowledge, there have been several meta-analyses and systematic reviews reporting these associations. We further investigated some other aspects of breastfeeding, involving modes and duration of breastfeeding and systemic cancers such as leukemia and cancers of the skeletal system, but it found that several meta-analyses and original studies had returned inconsistent results or suggestions regarding this topic. Mammas et al. [
13] discussed the association between breastfeeding and viral infections, and stated that breastfeeding may help prevent infections during the first years of life. Two studies [
11,
14] were more comprehensive than that of Mammas et al. Davis et al. [
70], a review of nine published case-control published in 1998, drew the conclusion that breastfeeding for 6 months and beyond was more effective than short-term breastfeeding in terms of reducing the incidence of Hodgkin’s disease. However, there was no evidence to show an association between infant feeding and any other cancer. Rodent et al. [
15], Amitay et al. [
4], Darcy et al. [
10] and Kwan et al. [
12] found that there was an inverse correlation between any breastfeeding and childhood acute lymphoblastic leukemia, and that breastfeeding for 6 months and beyond was superior to short-term feeding in terms of reducing leukemia morbidity and mortality. They found no evidence to suggest that long-term breastfeeding was protective against AML.
The modes of breastfeeding were classified as exclusive or non-exclusive, and various modes of breastfeeding were subjected to subgroup analysis. In addition, we investigated the relative effects of three types, including differences between exclusive breastfeeding and non-exclusive breastfeeding, differences between mixed breastfeeding and non-mixed breastfeeding, and differences between bottle feeding and non-formula feeding, for cancer in children. The findings of subgroup analysis suggested that mixed and exclusive breastfeeding were superior to only bottle feeding and that the combination of two modes of breastfeeding was more beneficial than one alone. These findings suggested that the inclusion of breastfeeding might reduce the incidence of childhood cancer, regardless of the modes of breastfeeding. In other words, mixed-breastfeeding was better for children than exclusive bottle-feeding and exclusive breastfeeding.
There were some differences of results in this meta-analysis compared to the WHO suggestions regarding modes of breastfeeding. The latter recommended that exclusive breastfeeding was needed for the first 6 months of life and should be continued breastfeeding until 2 years of age or longer with the addition of appropriate supplements. However, no significant difference was found in this meta-analysis regarding the protective effect of exclusive breastfeeding and mixed feeding for 0–6 months, specifically for childhood cancer. The difference might derive from the small study populations due to there being only three studies [
33,
39,
47] in this meta-analysis.
According to the results of duration of breastfeeding in this meta-analysis, there was protective effect for any duration of breastfeeding. However, there was no significant difference between ≥12 months and < 12 months of breastfeeding, suggesting that more than 6 months of breastfeeding could reduce the incidence of childhood cancer even more. Based on this dose-response relationship, it observed that, as the duration of breastfeeding increased, the odds of cancer significantly decreased. This suggested that extending the duration of breastfeeding as long as possible would reduce the incidence of cancer.
We removed the three largest and high-quality studies for separate analysis. These studies suggested weak significant inverse associations between any breastfeeding and long-term breastfeeding and childhood cancer, suggesting a very small influence of these studies on the results of this meta-analysis and systematic review.
Twenty-four studies [
8,
16,
18‐
20,
29,
31‐
33,
36,
39,
40,
42,
43,
47,
52‐
56,
60,
62,
64,
66] reported hematological malignancies and were included in this meta-analysis. Common hematological malignancies, such as ALL, AML, ANLL, HL and NHL, were included in a subgroup analysis. Four meta-analyses [
4,
12,
14,
15] probed the association between breastfeeding and the incidence of childhood leukemia. Based on subgroup analysis of leukemia, we found that breastfeeding indeed protected children by reducing the incidence of leukemia in childhood. This finding was consistent with conclusions of previous meta-analyses and systematic reviews. Undoubtedly, for ALL and AML, the functions of breastfeeding were identical because most leukemia cases were acute leukemia; nevertheless, there was no evidence to suggest that breastfeeding was associated with ANLL or lymphoma (HL and NHL).
The most comprehensive systemic diseases to date were analyzed in this meta-analysis and systematic review with respect to the relationship of breastfeeding with childhood cancer. According to the available evidence, breastfeeding reduced the prevalence of childhood hematological malignancies and cancers of the nervous and urinary systems. There were no remarkable associations of breastfeeding with cancers of the skeletal, reproductive, or sensory systems. Nevertheless, we cannot rule out that there may be variation because of the small sample size. Moreover, we cannot draw a conclusion regarding the relationship of breastfeeding with the incidence of cancers of other systems (i.e., respiratory, digestive, and endocrine systems) because of the absence of data.
The data from developed countries were considered separately from those of countries accounting for the majority of the collected data in this meta-analysis. Therefore, the conclusions needed to be confirmed further and compared with those of the total sample. The protective effect in the total sample was better than that of the developed countries. The hypothesis was that developing countries may do better than developed countries with respect to breastfeeding. The relationship of breastfeeding to childhood cancer in developed and developing countries needed to be further explored in the future.
There were several limitations of this meta-analysis. Firstly, the data collected from single studies were not complete and available, and some negative results might be missed, this might produce certain biases in the meta-analysis. In addition, unpublished literature was not included in terms of eligibility criteria. Secondly, the definitions of some studies did not clearly specify mixed breastfeeding, exclusive breastfeeding or bottle feeding, and the data were insufficient. Furthermore, the sample sizes of studies regarding the relationship between breastfeeding and cancers of certain systems of childhood were small, and this may affected the reliability of the meta-analysis. Thirdly, Most of populations included in this study come from high-income countries, and there was a lack of evidence support from low-income countries. Therefore, the conclusion of this study may only be applicable to population in high-income countries, and evidence applicable to population in low-income countries needs to be further produced and verified. In the course of cancer development, children may be exposed to more uncertain potential cancer risk factors. However, the original study did not provide relevant information about the potential cancer risk of children, which may have some effect on the accuracy of results. Finally, there was no information on covariates that may serve as mediators or confounders, including parental smoking behavior and family history of cancer.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.