Review of Case-Control Studies Related to Breastfeeding and Reduced Risk of Childhood Leukemia
http://www.100md.com
《小儿科》
Departments of Obstetrics and Gynecology
Department of Medical Informatics and Clinical Epidemiology
Department of Preventive Medicine and Public Health
Evidence-Based Practice Center, Oregon Health and Science University, Portland, Oregon
ABSTRACT
Objective. To conduct a systematic review to evaluate the evidence for the effect of breastfeeding on the risk of developing childhood leukemia.
Review Methods. We sought studies providing data regarding the association of breastfeeding and occurrence of childhood leukemia. Studies were identified by using Medline, HHS Blueprint for Action on Breastfeeding, US Department of Health and Human Services Office on Women’s Health, Cochrane Database of Systematic Reviews, National Centre for Reviews and Dissemination, reference lists, and national experts. Methodologic quality was evaluated for each study by using criteria from the US Preventive Services Task Force and the National Health Service Centre for Reviews and Dissemination.
Results. We reviewed 111 citations to identify 32 potentially eligible full-text articles. Of the 10 studies reviewed, only 4 were sufficient to provide at least fair-quality evidence regarding the association between maternal breastfeeding and childhood leukemia. Studies conflicted regarding the protective effect of breastfeeding on childhood leukemia. In the 2 largest and highest quality studies, breastfeeding was associated with a significant risk reduction in one study with longer breastfeeding duration, reflecting greater protection, and a nonsignificant but suggestive difference in the other. Taken together, half of the studies associated breastfeeding with a lower risk of acute lymphocytic leukemia.
Conclusions. There are few high-quality studies that examine the potential for a protective effect of breastfeeding for childhood leukemia. Furthermore, the few studies that exist disagree regarding the association. It is estimated that the United States spends $1.4 billion annually on the treatment of childhood leukemia. Patients, clinicians, and policy makers do not have the data that they need to make decisions regarding this important potential preventive measure.
Key Words: breastfeeding human milk leukemia case-control studies childhood malignancy evidence review review evidence-based medicine risk assessment human infant female pregnancy
Abbreviations: ALL, acute lymphocytic leukemia SES, socioeconomic status OR, odds ratio CI, confidence interval UKCCS, United Kingdom Childhood Cancer Study AML, acute myelogenous leukemia
Leukemia is the most common malignancy in childhood, with an average annual incidence of 3.8 to 4.8 cases per 100000 children aged 0 to 14 years.1 Leukemia accounts for 30% of all childhood malignancies, with 75% being diagnosed as acute lymphocytic leukemia (ALL). The incidence is slightly higher among males than females and 50% higher among white children than among black children.2 Although the etiology of childhood leukemia has been studied for a half of a century, causal factors associated with leukemia are largely unidentified. Certain conditions such as Down syndrome, other genetic abnormalities, and ionizing radiation exposures are thought to explain a small percentage of cases of leukemia.1,3
The infectious etiology of leukemia continues to be explored, with potential for bacterial and viral sources. After viruses were identified as the cause of some leukemias in animals,4 1 popular etiological hypothesis for human leukemia, especially that occurring in childhood, has been the result of a viral infection. Two additional observations suggest that immunologic factors may play a role in the development of childhood leukemia. The first factor originates from epidemiologic observations of geographic clusters of leukemia cases, suggesting that a common exposure (such as infectious exposure) may play a role.5 The second factor arises from observations that children with leukemia are less commonly found to be exposed very early in infancy to common infections compared with children without leukemia, promoting an aberrant response to future exposures.6 Because breastfeeding is noted for providing the newborn with passive immunity, protecting from some early infections, investigators have hypothesized that breastfeeding could reduce the risk of childhood leukemia.7 This review was conducted for the US Department of Health and Human Services Office on Women’s Health to evaluate the evidence for the effect of breastfeeding on the risk of developing childhood leukemia.
METHODS
Searching
Pertinent articles were identified from a larger literature search performed by the Office on Women’s Health. Briefly, it included a Medline search from 1990 through March 2004, articles identified from the HHS Blueprint for Action on Breastfeeding,5 and additional articles identified from experts at the Breastfeeding Experts Panel Meeting in 2003. Additionally, 1 of the investigators (C.D.M.) and a medical librarian searched Medline from 1990 to January 4, 2004, by using search terms "leukemia," "breastfeeding," or "human milk" to identify additional articles.
Selection
We sought studies providing data regarding the association of breastfeeding and occurrence of childhood leukemia. Inclusion criteria were established by the Office on Women’s Health. For inclusion, a study must have been in full text, been published after 1990, contained >100 participants, had concurrent comparison groups, been conducted in a developed country, and been published in the English language. Two investigators (J.-M.G. and C.D.M.) independently reviewed all titles and abstracts for inclusion; there were no disagreements on inclusion. Full-text articles were then reviewed for eligibility. A study was considered relevant if it met all inclusion criteria and contained data for both breastfeeding exposure and outcome of childhood ALL or all childhood leukemias.
Validity Assessment
Three investigators independently rated study quality by using criteria from the US Preventive Services Task Force8 and the National Health Service Centre for Reviews and Dissemination.9 Studies received a poor rating if the cases were not assessed reliably (potentially unreliable method for diagnosing disease), if the groups assembled were not comparable, if there was considerable attrition or differences in nonrespondents between cases and controls, or if there was not adequate consideration given to confounding. Studies were rated as poor if their case definition included infants diagnosed under the age of 6 months and did not allow for separate analysis for children older than 6 months of age, because the time from exposure to breastfeeding and diagnosis of leukemia is crucial for biological plausibility of association. Regarding accurate ascertainment of cases and controls, studies were examined for the response rate within first-choice controls and were rated as poor if there was potentially biased ascertainment of second- or higher-order controls. Additionally, studies that used hospital or nonconcurrent controls were considered inferior to studies that used population-based controls. Regarding consideration of confounders, examining the role of socioeconomic status (SES) was considered important, because it is associated with both leukemia and the likelihood of breastfeeding.
Data Abstraction
each study, 3 reviewers independently abstracted study design, setting, demographics, leukemia definition, breastfeeding initiation and duration methods of assessment and results, and methods for assessing or adjusting for confounders.
RESULTS
Quality Assessment and Study Characteristics
We identified 111 citations and reviewed 32 full-text articles. Ten studies10–19 met all eligibility criteria and were reviewed for study quality to identify 2 good-quality case-control studies,10,17 2 fair-quality studies,11,16 and 6 poor-quality studies.12–15,18,19 Three case-control studies met all inclusion criteria except that they were conducted in developing countries: Russia, United Arab Emirates, and China.20–22 Table 1 describes the population and study characteristics of all case-control studies of breastfeeding and childhood leukemia. The presented results focus on details and findings of the 2 good-quality studies,10,17 comparisons to the 2 fair-quality studies,11,16 factors that distinguish them from studies rated as "poor" in quality, and implications for future research.
The descriptions of the studies included in this systematic review are shown in Table 1. Of the 10 studies, 6 were conducted in European countries. All studies but 1 focused solely on childhood leukemia. The majority included <1000 cases. Notably, 6 of the studies explicitly sought to characterize the relationship between breastfeeding and leukemia as the primary objective, whereas the others included breastfeeding measures from the perspective of measuring broader characteristics of the immune system and early infections in the etiology.
The quality ratings of the studies are shown in Table 1. The 3 reviewers varied minimally in their ratings: at least 2 reviewers were in complete agreement 100% of the time, and all 3 reviewers were in agreement after discussion. Of the 10 studies, 2 received a consensus rating of "good," 2 "fair," and 6 "poor." Two of the largest studies with the largest number of cases included, from the United Kingdom Childhood Cancer Study (UKCCS) and the Children’s Cancer Group, were both rated as good.10,17 Both studies satisfactorily met all criteria for accurate assessment of cases, response rate of cases and controls, and accurate measurement of exposure. The 2 studies rated as fair were smaller in size; cases were assembled from national or regional cancer registries in New Zealand and Quebec, Canada.11,13 Both merited fair ratings because each included infants diagnosed with ALL at <1 year of age without subanalysis. In all but 2 of the 10 case-control studies,12,15 there was a potentially important difference in response rate between cases and all controls or first-choice controls as noted. Both articles that were rated as good, however, attempted to identify the degree of response bias associated with replacement controls and adjusted for resulting bias in the analysis.
Of the 6 studies rated as poor, 4 had unclear or biased selection of cases or controls. These studies used nonconcurrent controls,18 2 used hospital-based controls,15,19 and another had an unclear case or control definition including only fatal leukemia.14 Two studies12,16 failed to account for potential important confounding in their analysis. Two of these articles failed to record how mothers were asked about breastfeeding. Because of the potential bias introduced by selection of cases and controls, the lack of control of confounding, and the potential misclassification introduced by the lack of specificity in exposure definition, these studies will not be discussed further.
Study Findings
The 2 good-quality studies present conflicting results regarding the association between breastfeeding and leukemia. As 1 of the 2 studies rated as good, the Children’s Cancer Group17 included 1914 children with ALL as cases from the United States, Canada, and Australia; controls were recruited by random-digit dialing. These patients were participating in cooperative cancer treatment trials at cancer centers. After adjustment for maternal race, education, and family income, breastfeeding was associated with a significantly reduced risk of ALL (odds ratio [OR]: 0.80; 95% confidence interval [CI]: 0.69–0.93). When considering duration, there was a graded effect, with breastfeeding 6 months conferring somewhat less protection against ALL (OR: 0.86; 95% CI: 0.73–1.01) than >6 months of breastfeeding (OR: 0.72; 95% CI: 0.60–0.87), as compared with no breastfeeding. Considering durations from none to >12 months, there was a statistically significant trend for decreased risk for ALL with increasing duration of breastfeeding (P = .0034).
The other study rated as good by all 3 reviewers included the largest population of all reviewed. In the UKCCS,10 1637 population-based leukemia cases were selected with 6964 controls selected from health services records. Considering all types of leukemia, breastfeeding had a nonsignificant relationship with risk (OR: 0.89; 95% CI: 0.80–1.00). When only cases with ALL were considered, breastfeeding conferred a nonsignificant reduction in risk (OR: 0.91; 95% CI: 0.81–1.04) after adjustment for age at diagnosis, gender, region, birth order, and a measure of SES. Duration of breastfeeding was not significantly related to risk (2 for trend: P = .48): duration <1 month (OR: 0.96; 95% CI: 0.82–1.17), duration 1 to 6 months (OR: 0.90; 95% CI: 0.77–1.04), and duration 7 months (OR: 0.89; 95% CI: 0.75–1.05). Risk was also not related to subcategories of ALL.
Similarly, the 2 fair-quality studies disagreed on the protective effect of breastfeeding. In the study by Infante-Rivard et al13 conducted in the Province of Quebec, the authors recruited cases of ALL including children <1 year of age. As compared with population-based controls, breastfeeding was associated with protection from ALL comparing a duration of >3 months with a duration 3 months (matched OR: 0.67; 95% CI: 0.47–0.94). However, in the other study rated as fair, Dockerty et al11 similarly recruited children aged 0 to 14 years with leukemia in New Zealand and controls selected from birth records, matched on age and gender. After adjustment for relevant confounders, breastfeeding was not associated with ALL (OR: 0.98; 95% CI: 0.39–2.47).
DISCUSSION
Of the 10 studies reviewed, only 4 were sufficient to provide at least fair-quality evidence regarding the association between maternal breastfeeding and childhood leukemia. In the 2 studies rated as good, breastfeeding was associated with a significant risk reduction in 1 study17 and a nonsignificant but suggestive difference in the other.10 Similarly, there was a relationship with duration of breastfeeding in the former study but not in the latter. In the 2 studies rated as fair, 1 was associated with risk reduction.13 Taken together, half of the studies associated breastfeeding with a lower risk of ALL.
Our review differs in methodology from previous meta-analyses performed by Beral et al10 and Kwan et al.23 Neither of these groups examined studies for quality or used quality ratings as a determinate for inclusion in analysis. The review performed by Beral et al was a meta-analysis of 15 studies regarding breastfeeding and leukemia or childhood cancer. In that analysis, having ever been breastfed was associated with a small risk reduction for childhood leukemia (OR: 0.86; 95% CI: 0.81–0.92). Breastfeeding for >6 months seemed to confer somewhat greater protection (OR: 0.78; 95% CI: 0.71–0.85) than for a duration of <6 months (OR: 0.91; 95% CI: 0.85–0.99).10 However, this protective effect was uniform for all pediatric cancers. This would either indicate a universal effect of an immunologic influence across all cancers or, more likely, an inherent bias in the control ascertainment, indicative of unresolved confounding. Although they seemed to have performed a comprehensive search, they did not report following standard steps required for a systematic review, including explicit search criteria, inclusion and exclusion criteria, and quality assessment. Their methodologic rigor and findings differ from this review. No quality rating was considered for the studies included in their meta-analysis. Of the 15 studies included in their analysis, 8 were also included in our systematic review. Four were excluded because they were published before 1990, and 3 were conducted in developing countries. Notably, of the 8 studies that both Beral et al and our group considered eligible for review, 4 were judged of poor quality and were not included in this review.12,15,16,18
The meta-analysis by Kwan et al23 used similar methods to include 14 case-control studies. These authors concluded that short-term (6 months) and long-term (>6 months) breastfeeding was associated with a reduced odds of both ALL (OR: 0.76; 95% CI: 0.68–0.84) and acute myelogenous leukemia (AML) (OR: 0.85; 95% CI: 0.73–0.98). No quality rating of studies was included. This lack of a relationship specific to ALL and the lack of a duration effect of breastfeeding may be indicative of bias, particularly confounding by SES. Many included studies that failed to adjust for SES, and the imbalance in this factor between cases and controls as well as participation bias may be contributory given the strong relationship with breastfeeding. Mothers who breastfeed 6 months differ from those who breastfeed less, in more ways than just having a different SES. Taveras et al24 report that mothers who feed their infants longer have less restrictive behavior regarding child feeding at 1 year. These groups of mothers also differ significantly by age, prepregnancy BMI, income, education, race, and gravidity. Risk for ALL may be associated with 1 of these factors, and the effect may not be removed entirely by adjusting for SES.
The certainty that one has in making clinical recommendations is guided by the confidence that the effect observed in the literature is likely to be true. Leading organizations agree on the importance of explicitly evaluating the quality of the evidence to provide guidance regarding the certainty of recommendations.8,25–35
This is the first review of the literature to evaluate the quality of each study and use it to guide the analysis and presentation of results. All peer-reviewed full-text studies mentioned in each of the prior reviews were evaluated in this review; however, after formal quality assessment, only 4 were rated as at least fair in quality and they form the foundation for our results.
The primary findings of our review indicate that there are few high-quality studies to inform an important question for parents as to whether it is possible to reduce the risk of childhood leukemia by breastfeeding, and those few studies disagree. The studies frequently failed to measure important factors such as breastfeeding exclusivity (reporting ever breastfed rather than quantifying breastfeeding by exclusivity combined with duration) and consideration of other important confounders such as SES and other infectious exposures (such as household or school contacts). Because childhood leukemia is a rare disease, a case-control study is the only study design that will allow calculation of the risk that breastfeeding plays in the development of this disease. However, case-control studies are notoriously among the most difficult to conduct given the influence of bias and confounding. None of the studies included in this systematic review were without flaw. An optimal study might be conducted within the framework of a large population-based registry or cohort with full access to medical records, pathology, and demographic data that would be able to accurately identify all cases of ALL diagnosed (with pathologic confirmation) at >1 year of age within a defined time period. Access to original records, pathology, and clinical examination is essential to ascertain diagnostic accuracy. Ideally, this same registry or cohort would allow for identification of controls on the basis of the same sampling frame, with equal access to medical care and the ability to accurately match for important influential factors. Breastfeeding duration and exclusivity would be measured practically by maternal survey, and ideally, duration and cessation would be confirmed with infant health records. This would eliminate a strong influence of recall bias given the ability to verify the answers for the principal risk factor in the context of an interview of mothers of cases and controls. Last, SES remains an important confounder and should be assessed by multiple measures such as parental education, employment, and income, if possible.
Childhood leukemia remains the most common type of cancer diagnosed in children. It is estimated that the United States spends $1.4 billion dollars annually for the treatment of childhood leukemia.1 There are 2400 children diagnosed annually with ALL in the United States.36 An ability to prevent 10% to 20% of these cases through breastfeeding would constitute an obvious health and fiscal benefit. Despite the public health importance of identifying potential interventions that could prevent the onset of childhood leukemia, the current literature regarding the preventive role of breastfeeding has substantial limitations. Given the burden of disease and potential cost-effectiveness, conducting high-quality research should be a high priority.
ACKNOWLEDGMENTS
This study was conducted under contract to the Department of Health and Human Services Office on Women’s Health, Washington, DC (contract 0990-0115), with the support of Suzanne Haynes, PhD, senior science advisor. Support for Dr Guise was also provided by the Agency for Healthcare Research and Quality grant 1 K08 HS11338-01.
FOOTNOTES
Accepted May 23, 2005.
No conflict of interest declared.
REFERENCES
Boice J, Land C, Preston D. Ion radiation. In: Schottenfeld D, Fraumeni J, eds. Cancer Epidemiology and Prevention. 2nd ed. New York, NY: Oxford University Press; 1996
Evans AS, Mueller NE. Viruses and cancer. Causal associations. Ann Epidemiol. 1990;1 :71 –92
Greaves M. Childhood leukaemia. BMJ. 2002;324 :283 –287
Greaves MF. Speculations on the cause of childhood acute lymphoblastic leukemia. Leukemia. 1988;2 :120 –125
Harris RP, Helfand M, Woolf SH, et al. Current methods of the US Preventive Services Task Force: a review of the process. Am J Prev Med. 2001;20(3 suppl) :21 –35
Beral V, Fear NT, Alexander F, Appleby P. Breastfeeding and childhood cancer. Br J Cancer. 2001;85 :1685 –1694
Dockerty JD, Skegg DC, Elwood JM, Herbison GP, Becroft DM, Lewis ME. Infections, vaccinations, and the risk of childhood leukaemia. Br J Cancer. 1999;80 :1483 –1489
Hardell L, Dreifaldt AC. Breast-feeding duration and the risk of malignant diseases in childhood in Sweden. Eur J Clin Nutr. 2001;55 :179 –185
Infante-Rivard C, Fortier I, Olson E. Markers of infection, breast-feeding and childhood acute lymphoblastic leukaemia. Br J Cancer. 2000;83 :1559 –1564
Lancashire RJ, Sorahan T; OSCC. Breastfeeding and childhood cancer risks: OSCC data. Br J Cancer. 2003;88 :1035 –1037
Petridou E, Trichopoulos D, Kalapothaki V, et al. The risk profile of childhood leukaemia in Greece: a nationwide case-control study. Br J Cancer. 1997;76 :1241 –1247
Rosenbaum PF, Buck GM, Brecher ML. Early child-care and preschool experiences and the risk of childhood acute lymphoblastic leukemia. Am J Epidemiol. 2000;152 :1136 –1144
Shu XO, Linet MS, Steinbuch M, et al. Breast-feeding and risk of childhood acute leukemia. J Natl Cancer Inst. 1999;91 :1765 –1772
Schuz J, Kaletsch U, Meinert R, Kaatsch P, Michaelis J. Association of childhood leukaemia with factors related to the immune system. Br J Cancer. 1999;80 :585 –590
Perrillat F, Clavel J, Auclerc MF, et al. Day-care, early common infections and childhood acute leukaemia: a multicentre French case-control study. Br J Cancer. 2002;86 :1064 –1069
Smulevich VB, Solionova LG, Belyakova SV. Parental occupation and other factors and cancer risk in children: I. Study methodology and non-occupational factors. Int J Cancer. 1999;83 :712 –717
Bener A, Denic S, Galadari S. Longer breast-feeding and protection against childhood leukaemia and lymphomas. Eur J Cancer. 2001;37 :234 –238
Shu XO, Clemens J, Zheng W, Ying DM, Ji BT, Jin F. Infant breastfeeding and the risk of childhood lymphoma and leukaemia. Int J Epidemiol. 1995;24 :27 –32
Kwan ML, Buffler PA, Abrams B, Kiley VA. Breastfeeding and the risk of childhood leukemia: a meta-analysis. Public Health Rep. 2004;119 :521 –535
Woolf SH, Atkins D. The evolving role of prevention in health care: contributions of the U.S. Preventive Services Task Force. Am J Prev Med. 2001;20(3 suppl) :13 –20
Briss PA, Zaza S, Pappaioanou M, et al. Developing an evidence-based guide to community preventive services—methods. The Task Force on Community Preventive Services. Am J Prev Med. 2000;18(1 suppl) :35 –43
Zaza S, Wright-De Aguero LK, Briss PA, et al. Data collection instrument and procedure for systematic reviews in the Guide to Community Preventive Services. Task Force on Community Preventive Services. Am J Prev Med. 2000;18(1 suppl) :44 –74
Greer N, Mosser G, Logan G, Halaas GW. A practical approach to evidence grading. Jt Comm J Qual Improv. 2000;26 :700 –712
West S, King V, Carey TS, et al. Systems to rate the strength of scientific evidence. Evid Rep Technol Assess (Summ). 2002;(47) :1 –11
Guyatt GH, Sackett DL, Sinclair JC, Hayward R, Cook DJ, Cook RJ. Users’ guides to the medical literature. IX. A method for grading health care recommendations. Evidence-Based Medicine Working Group. JAMA. 1995;274 :1800 –1804
Atkins D, Best D, Briss PA, et al. Grading quality of evidence and strength of recommendations. BMJ. 2004;328 :1490
Atkins D, Best D, Shapiro EN, eds. Third US Preventive Services Task Force: background, methods and first recommendations. Am J Prev Med. 2001;20(3 suppl) :1 –108
Smith MA, Gloeckler Ries LA, Gurney JG, Ross JA. Leukemia. Available at: http://seer.cancer.gov/publications/childhood/leukemia.pdf. Accessed November 15, 2004(Jeanne-Marie Guise, MD, M)
Department of Medical Informatics and Clinical Epidemiology
Department of Preventive Medicine and Public Health
Evidence-Based Practice Center, Oregon Health and Science University, Portland, Oregon
ABSTRACT
Objective. To conduct a systematic review to evaluate the evidence for the effect of breastfeeding on the risk of developing childhood leukemia.
Review Methods. We sought studies providing data regarding the association of breastfeeding and occurrence of childhood leukemia. Studies were identified by using Medline, HHS Blueprint for Action on Breastfeeding, US Department of Health and Human Services Office on Women’s Health, Cochrane Database of Systematic Reviews, National Centre for Reviews and Dissemination, reference lists, and national experts. Methodologic quality was evaluated for each study by using criteria from the US Preventive Services Task Force and the National Health Service Centre for Reviews and Dissemination.
Results. We reviewed 111 citations to identify 32 potentially eligible full-text articles. Of the 10 studies reviewed, only 4 were sufficient to provide at least fair-quality evidence regarding the association between maternal breastfeeding and childhood leukemia. Studies conflicted regarding the protective effect of breastfeeding on childhood leukemia. In the 2 largest and highest quality studies, breastfeeding was associated with a significant risk reduction in one study with longer breastfeeding duration, reflecting greater protection, and a nonsignificant but suggestive difference in the other. Taken together, half of the studies associated breastfeeding with a lower risk of acute lymphocytic leukemia.
Conclusions. There are few high-quality studies that examine the potential for a protective effect of breastfeeding for childhood leukemia. Furthermore, the few studies that exist disagree regarding the association. It is estimated that the United States spends $1.4 billion annually on the treatment of childhood leukemia. Patients, clinicians, and policy makers do not have the data that they need to make decisions regarding this important potential preventive measure.
Key Words: breastfeeding human milk leukemia case-control studies childhood malignancy evidence review review evidence-based medicine risk assessment human infant female pregnancy
Abbreviations: ALL, acute lymphocytic leukemia SES, socioeconomic status OR, odds ratio CI, confidence interval UKCCS, United Kingdom Childhood Cancer Study AML, acute myelogenous leukemia
Leukemia is the most common malignancy in childhood, with an average annual incidence of 3.8 to 4.8 cases per 100000 children aged 0 to 14 years.1 Leukemia accounts for 30% of all childhood malignancies, with 75% being diagnosed as acute lymphocytic leukemia (ALL). The incidence is slightly higher among males than females and 50% higher among white children than among black children.2 Although the etiology of childhood leukemia has been studied for a half of a century, causal factors associated with leukemia are largely unidentified. Certain conditions such as Down syndrome, other genetic abnormalities, and ionizing radiation exposures are thought to explain a small percentage of cases of leukemia.1,3
The infectious etiology of leukemia continues to be explored, with potential for bacterial and viral sources. After viruses were identified as the cause of some leukemias in animals,4 1 popular etiological hypothesis for human leukemia, especially that occurring in childhood, has been the result of a viral infection. Two additional observations suggest that immunologic factors may play a role in the development of childhood leukemia. The first factor originates from epidemiologic observations of geographic clusters of leukemia cases, suggesting that a common exposure (such as infectious exposure) may play a role.5 The second factor arises from observations that children with leukemia are less commonly found to be exposed very early in infancy to common infections compared with children without leukemia, promoting an aberrant response to future exposures.6 Because breastfeeding is noted for providing the newborn with passive immunity, protecting from some early infections, investigators have hypothesized that breastfeeding could reduce the risk of childhood leukemia.7 This review was conducted for the US Department of Health and Human Services Office on Women’s Health to evaluate the evidence for the effect of breastfeeding on the risk of developing childhood leukemia.
METHODS
Searching
Pertinent articles were identified from a larger literature search performed by the Office on Women’s Health. Briefly, it included a Medline search from 1990 through March 2004, articles identified from the HHS Blueprint for Action on Breastfeeding,5 and additional articles identified from experts at the Breastfeeding Experts Panel Meeting in 2003. Additionally, 1 of the investigators (C.D.M.) and a medical librarian searched Medline from 1990 to January 4, 2004, by using search terms "leukemia," "breastfeeding," or "human milk" to identify additional articles.
Selection
We sought studies providing data regarding the association of breastfeeding and occurrence of childhood leukemia. Inclusion criteria were established by the Office on Women’s Health. For inclusion, a study must have been in full text, been published after 1990, contained >100 participants, had concurrent comparison groups, been conducted in a developed country, and been published in the English language. Two investigators (J.-M.G. and C.D.M.) independently reviewed all titles and abstracts for inclusion; there were no disagreements on inclusion. Full-text articles were then reviewed for eligibility. A study was considered relevant if it met all inclusion criteria and contained data for both breastfeeding exposure and outcome of childhood ALL or all childhood leukemias.
Validity Assessment
Three investigators independently rated study quality by using criteria from the US Preventive Services Task Force8 and the National Health Service Centre for Reviews and Dissemination.9 Studies received a poor rating if the cases were not assessed reliably (potentially unreliable method for diagnosing disease), if the groups assembled were not comparable, if there was considerable attrition or differences in nonrespondents between cases and controls, or if there was not adequate consideration given to confounding. Studies were rated as poor if their case definition included infants diagnosed under the age of 6 months and did not allow for separate analysis for children older than 6 months of age, because the time from exposure to breastfeeding and diagnosis of leukemia is crucial for biological plausibility of association. Regarding accurate ascertainment of cases and controls, studies were examined for the response rate within first-choice controls and were rated as poor if there was potentially biased ascertainment of second- or higher-order controls. Additionally, studies that used hospital or nonconcurrent controls were considered inferior to studies that used population-based controls. Regarding consideration of confounders, examining the role of socioeconomic status (SES) was considered important, because it is associated with both leukemia and the likelihood of breastfeeding.
Data Abstraction
each study, 3 reviewers independently abstracted study design, setting, demographics, leukemia definition, breastfeeding initiation and duration methods of assessment and results, and methods for assessing or adjusting for confounders.
RESULTS
Quality Assessment and Study Characteristics
We identified 111 citations and reviewed 32 full-text articles. Ten studies10–19 met all eligibility criteria and were reviewed for study quality to identify 2 good-quality case-control studies,10,17 2 fair-quality studies,11,16 and 6 poor-quality studies.12–15,18,19 Three case-control studies met all inclusion criteria except that they were conducted in developing countries: Russia, United Arab Emirates, and China.20–22 Table 1 describes the population and study characteristics of all case-control studies of breastfeeding and childhood leukemia. The presented results focus on details and findings of the 2 good-quality studies,10,17 comparisons to the 2 fair-quality studies,11,16 factors that distinguish them from studies rated as "poor" in quality, and implications for future research.
The descriptions of the studies included in this systematic review are shown in Table 1. Of the 10 studies, 6 were conducted in European countries. All studies but 1 focused solely on childhood leukemia. The majority included <1000 cases. Notably, 6 of the studies explicitly sought to characterize the relationship between breastfeeding and leukemia as the primary objective, whereas the others included breastfeeding measures from the perspective of measuring broader characteristics of the immune system and early infections in the etiology.
The quality ratings of the studies are shown in Table 1. The 3 reviewers varied minimally in their ratings: at least 2 reviewers were in complete agreement 100% of the time, and all 3 reviewers were in agreement after discussion. Of the 10 studies, 2 received a consensus rating of "good," 2 "fair," and 6 "poor." Two of the largest studies with the largest number of cases included, from the United Kingdom Childhood Cancer Study (UKCCS) and the Children’s Cancer Group, were both rated as good.10,17 Both studies satisfactorily met all criteria for accurate assessment of cases, response rate of cases and controls, and accurate measurement of exposure. The 2 studies rated as fair were smaller in size; cases were assembled from national or regional cancer registries in New Zealand and Quebec, Canada.11,13 Both merited fair ratings because each included infants diagnosed with ALL at <1 year of age without subanalysis. In all but 2 of the 10 case-control studies,12,15 there was a potentially important difference in response rate between cases and all controls or first-choice controls as noted. Both articles that were rated as good, however, attempted to identify the degree of response bias associated with replacement controls and adjusted for resulting bias in the analysis.
Of the 6 studies rated as poor, 4 had unclear or biased selection of cases or controls. These studies used nonconcurrent controls,18 2 used hospital-based controls,15,19 and another had an unclear case or control definition including only fatal leukemia.14 Two studies12,16 failed to account for potential important confounding in their analysis. Two of these articles failed to record how mothers were asked about breastfeeding. Because of the potential bias introduced by selection of cases and controls, the lack of control of confounding, and the potential misclassification introduced by the lack of specificity in exposure definition, these studies will not be discussed further.
Study Findings
The 2 good-quality studies present conflicting results regarding the association between breastfeeding and leukemia. As 1 of the 2 studies rated as good, the Children’s Cancer Group17 included 1914 children with ALL as cases from the United States, Canada, and Australia; controls were recruited by random-digit dialing. These patients were participating in cooperative cancer treatment trials at cancer centers. After adjustment for maternal race, education, and family income, breastfeeding was associated with a significantly reduced risk of ALL (odds ratio [OR]: 0.80; 95% confidence interval [CI]: 0.69–0.93). When considering duration, there was a graded effect, with breastfeeding 6 months conferring somewhat less protection against ALL (OR: 0.86; 95% CI: 0.73–1.01) than >6 months of breastfeeding (OR: 0.72; 95% CI: 0.60–0.87), as compared with no breastfeeding. Considering durations from none to >12 months, there was a statistically significant trend for decreased risk for ALL with increasing duration of breastfeeding (P = .0034).
The other study rated as good by all 3 reviewers included the largest population of all reviewed. In the UKCCS,10 1637 population-based leukemia cases were selected with 6964 controls selected from health services records. Considering all types of leukemia, breastfeeding had a nonsignificant relationship with risk (OR: 0.89; 95% CI: 0.80–1.00). When only cases with ALL were considered, breastfeeding conferred a nonsignificant reduction in risk (OR: 0.91; 95% CI: 0.81–1.04) after adjustment for age at diagnosis, gender, region, birth order, and a measure of SES. Duration of breastfeeding was not significantly related to risk (2 for trend: P = .48): duration <1 month (OR: 0.96; 95% CI: 0.82–1.17), duration 1 to 6 months (OR: 0.90; 95% CI: 0.77–1.04), and duration 7 months (OR: 0.89; 95% CI: 0.75–1.05). Risk was also not related to subcategories of ALL.
Similarly, the 2 fair-quality studies disagreed on the protective effect of breastfeeding. In the study by Infante-Rivard et al13 conducted in the Province of Quebec, the authors recruited cases of ALL including children <1 year of age. As compared with population-based controls, breastfeeding was associated with protection from ALL comparing a duration of >3 months with a duration 3 months (matched OR: 0.67; 95% CI: 0.47–0.94). However, in the other study rated as fair, Dockerty et al11 similarly recruited children aged 0 to 14 years with leukemia in New Zealand and controls selected from birth records, matched on age and gender. After adjustment for relevant confounders, breastfeeding was not associated with ALL (OR: 0.98; 95% CI: 0.39–2.47).
DISCUSSION
Of the 10 studies reviewed, only 4 were sufficient to provide at least fair-quality evidence regarding the association between maternal breastfeeding and childhood leukemia. In the 2 studies rated as good, breastfeeding was associated with a significant risk reduction in 1 study17 and a nonsignificant but suggestive difference in the other.10 Similarly, there was a relationship with duration of breastfeeding in the former study but not in the latter. In the 2 studies rated as fair, 1 was associated with risk reduction.13 Taken together, half of the studies associated breastfeeding with a lower risk of ALL.
Our review differs in methodology from previous meta-analyses performed by Beral et al10 and Kwan et al.23 Neither of these groups examined studies for quality or used quality ratings as a determinate for inclusion in analysis. The review performed by Beral et al was a meta-analysis of 15 studies regarding breastfeeding and leukemia or childhood cancer. In that analysis, having ever been breastfed was associated with a small risk reduction for childhood leukemia (OR: 0.86; 95% CI: 0.81–0.92). Breastfeeding for >6 months seemed to confer somewhat greater protection (OR: 0.78; 95% CI: 0.71–0.85) than for a duration of <6 months (OR: 0.91; 95% CI: 0.85–0.99).10 However, this protective effect was uniform for all pediatric cancers. This would either indicate a universal effect of an immunologic influence across all cancers or, more likely, an inherent bias in the control ascertainment, indicative of unresolved confounding. Although they seemed to have performed a comprehensive search, they did not report following standard steps required for a systematic review, including explicit search criteria, inclusion and exclusion criteria, and quality assessment. Their methodologic rigor and findings differ from this review. No quality rating was considered for the studies included in their meta-analysis. Of the 15 studies included in their analysis, 8 were also included in our systematic review. Four were excluded because they were published before 1990, and 3 were conducted in developing countries. Notably, of the 8 studies that both Beral et al and our group considered eligible for review, 4 were judged of poor quality and were not included in this review.12,15,16,18
The meta-analysis by Kwan et al23 used similar methods to include 14 case-control studies. These authors concluded that short-term (6 months) and long-term (>6 months) breastfeeding was associated with a reduced odds of both ALL (OR: 0.76; 95% CI: 0.68–0.84) and acute myelogenous leukemia (AML) (OR: 0.85; 95% CI: 0.73–0.98). No quality rating of studies was included. This lack of a relationship specific to ALL and the lack of a duration effect of breastfeeding may be indicative of bias, particularly confounding by SES. Many included studies that failed to adjust for SES, and the imbalance in this factor between cases and controls as well as participation bias may be contributory given the strong relationship with breastfeeding. Mothers who breastfeed 6 months differ from those who breastfeed less, in more ways than just having a different SES. Taveras et al24 report that mothers who feed their infants longer have less restrictive behavior regarding child feeding at 1 year. These groups of mothers also differ significantly by age, prepregnancy BMI, income, education, race, and gravidity. Risk for ALL may be associated with 1 of these factors, and the effect may not be removed entirely by adjusting for SES.
The certainty that one has in making clinical recommendations is guided by the confidence that the effect observed in the literature is likely to be true. Leading organizations agree on the importance of explicitly evaluating the quality of the evidence to provide guidance regarding the certainty of recommendations.8,25–35
This is the first review of the literature to evaluate the quality of each study and use it to guide the analysis and presentation of results. All peer-reviewed full-text studies mentioned in each of the prior reviews were evaluated in this review; however, after formal quality assessment, only 4 were rated as at least fair in quality and they form the foundation for our results.
The primary findings of our review indicate that there are few high-quality studies to inform an important question for parents as to whether it is possible to reduce the risk of childhood leukemia by breastfeeding, and those few studies disagree. The studies frequently failed to measure important factors such as breastfeeding exclusivity (reporting ever breastfed rather than quantifying breastfeeding by exclusivity combined with duration) and consideration of other important confounders such as SES and other infectious exposures (such as household or school contacts). Because childhood leukemia is a rare disease, a case-control study is the only study design that will allow calculation of the risk that breastfeeding plays in the development of this disease. However, case-control studies are notoriously among the most difficult to conduct given the influence of bias and confounding. None of the studies included in this systematic review were without flaw. An optimal study might be conducted within the framework of a large population-based registry or cohort with full access to medical records, pathology, and demographic data that would be able to accurately identify all cases of ALL diagnosed (with pathologic confirmation) at >1 year of age within a defined time period. Access to original records, pathology, and clinical examination is essential to ascertain diagnostic accuracy. Ideally, this same registry or cohort would allow for identification of controls on the basis of the same sampling frame, with equal access to medical care and the ability to accurately match for important influential factors. Breastfeeding duration and exclusivity would be measured practically by maternal survey, and ideally, duration and cessation would be confirmed with infant health records. This would eliminate a strong influence of recall bias given the ability to verify the answers for the principal risk factor in the context of an interview of mothers of cases and controls. Last, SES remains an important confounder and should be assessed by multiple measures such as parental education, employment, and income, if possible.
Childhood leukemia remains the most common type of cancer diagnosed in children. It is estimated that the United States spends $1.4 billion dollars annually for the treatment of childhood leukemia.1 There are 2400 children diagnosed annually with ALL in the United States.36 An ability to prevent 10% to 20% of these cases through breastfeeding would constitute an obvious health and fiscal benefit. Despite the public health importance of identifying potential interventions that could prevent the onset of childhood leukemia, the current literature regarding the preventive role of breastfeeding has substantial limitations. Given the burden of disease and potential cost-effectiveness, conducting high-quality research should be a high priority.
ACKNOWLEDGMENTS
This study was conducted under contract to the Department of Health and Human Services Office on Women’s Health, Washington, DC (contract 0990-0115), with the support of Suzanne Haynes, PhD, senior science advisor. Support for Dr Guise was also provided by the Agency for Healthcare Research and Quality grant 1 K08 HS11338-01.
FOOTNOTES
Accepted May 23, 2005.
No conflict of interest declared.
REFERENCES
Boice J, Land C, Preston D. Ion radiation. In: Schottenfeld D, Fraumeni J, eds. Cancer Epidemiology and Prevention. 2nd ed. New York, NY: Oxford University Press; 1996
Evans AS, Mueller NE. Viruses and cancer. Causal associations. Ann Epidemiol. 1990;1 :71 –92
Greaves M. Childhood leukaemia. BMJ. 2002;324 :283 –287
Greaves MF. Speculations on the cause of childhood acute lymphoblastic leukemia. Leukemia. 1988;2 :120 –125
Harris RP, Helfand M, Woolf SH, et al. Current methods of the US Preventive Services Task Force: a review of the process. Am J Prev Med. 2001;20(3 suppl) :21 –35
Beral V, Fear NT, Alexander F, Appleby P. Breastfeeding and childhood cancer. Br J Cancer. 2001;85 :1685 –1694
Dockerty JD, Skegg DC, Elwood JM, Herbison GP, Becroft DM, Lewis ME. Infections, vaccinations, and the risk of childhood leukaemia. Br J Cancer. 1999;80 :1483 –1489
Hardell L, Dreifaldt AC. Breast-feeding duration and the risk of malignant diseases in childhood in Sweden. Eur J Clin Nutr. 2001;55 :179 –185
Infante-Rivard C, Fortier I, Olson E. Markers of infection, breast-feeding and childhood acute lymphoblastic leukaemia. Br J Cancer. 2000;83 :1559 –1564
Lancashire RJ, Sorahan T; OSCC. Breastfeeding and childhood cancer risks: OSCC data. Br J Cancer. 2003;88 :1035 –1037
Petridou E, Trichopoulos D, Kalapothaki V, et al. The risk profile of childhood leukaemia in Greece: a nationwide case-control study. Br J Cancer. 1997;76 :1241 –1247
Rosenbaum PF, Buck GM, Brecher ML. Early child-care and preschool experiences and the risk of childhood acute lymphoblastic leukemia. Am J Epidemiol. 2000;152 :1136 –1144
Shu XO, Linet MS, Steinbuch M, et al. Breast-feeding and risk of childhood acute leukemia. J Natl Cancer Inst. 1999;91 :1765 –1772
Schuz J, Kaletsch U, Meinert R, Kaatsch P, Michaelis J. Association of childhood leukaemia with factors related to the immune system. Br J Cancer. 1999;80 :585 –590
Perrillat F, Clavel J, Auclerc MF, et al. Day-care, early common infections and childhood acute leukaemia: a multicentre French case-control study. Br J Cancer. 2002;86 :1064 –1069
Smulevich VB, Solionova LG, Belyakova SV. Parental occupation and other factors and cancer risk in children: I. Study methodology and non-occupational factors. Int J Cancer. 1999;83 :712 –717
Bener A, Denic S, Galadari S. Longer breast-feeding and protection against childhood leukaemia and lymphomas. Eur J Cancer. 2001;37 :234 –238
Shu XO, Clemens J, Zheng W, Ying DM, Ji BT, Jin F. Infant breastfeeding and the risk of childhood lymphoma and leukaemia. Int J Epidemiol. 1995;24 :27 –32
Kwan ML, Buffler PA, Abrams B, Kiley VA. Breastfeeding and the risk of childhood leukemia: a meta-analysis. Public Health Rep. 2004;119 :521 –535
Woolf SH, Atkins D. The evolving role of prevention in health care: contributions of the U.S. Preventive Services Task Force. Am J Prev Med. 2001;20(3 suppl) :13 –20
Briss PA, Zaza S, Pappaioanou M, et al. Developing an evidence-based guide to community preventive services—methods. The Task Force on Community Preventive Services. Am J Prev Med. 2000;18(1 suppl) :35 –43
Zaza S, Wright-De Aguero LK, Briss PA, et al. Data collection instrument and procedure for systematic reviews in the Guide to Community Preventive Services. Task Force on Community Preventive Services. Am J Prev Med. 2000;18(1 suppl) :44 –74
Greer N, Mosser G, Logan G, Halaas GW. A practical approach to evidence grading. Jt Comm J Qual Improv. 2000;26 :700 –712
West S, King V, Carey TS, et al. Systems to rate the strength of scientific evidence. Evid Rep Technol Assess (Summ). 2002;(47) :1 –11
Guyatt GH, Sackett DL, Sinclair JC, Hayward R, Cook DJ, Cook RJ. Users’ guides to the medical literature. IX. A method for grading health care recommendations. Evidence-Based Medicine Working Group. JAMA. 1995;274 :1800 –1804
Atkins D, Best D, Briss PA, et al. Grading quality of evidence and strength of recommendations. BMJ. 2004;328 :1490
Atkins D, Best D, Shapiro EN, eds. Third US Preventive Services Task Force: background, methods and first recommendations. Am J Prev Med. 2001;20(3 suppl) :1 –108
Smith MA, Gloeckler Ries LA, Gurney JG, Ross JA. Leukemia. Available at: http://seer.cancer.gov/publications/childhood/leukemia.pdf. Accessed November 15, 2004(Jeanne-Marie Guise, MD, M)