Instruction manuals

Exposures in childhood, adolescence and early adulthood and breast cancer risk: a systematic review of the literature

Exposures in childhood, adolescence and early adulthood and breast cancer risk: a systematic review of the literature
of 54
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
   Breast Cancer Research and Treatment   78:  223–276, 2003.© 2003  Kluwer Academic Publishers. Printed in the Netherlands.  Review Exposures in childhood, adolescence and early adulthood and breastcancer risk: a systematic review of the literature Mona Okasha 1 , Peter McCarron 2 , David Gunnell 1 , and George Davey Smith 1 1  Department of Social Medicine, University of Bristol, Bristol;  2  N. Ireland Cancer Registry, Department of  Epidemiology & Public Health, The Queen’s University of Belfast, Belfast, UK Key words:  alcohol, birthweight, body mass index, breast cancer, height, life course epidemiology, passivesmoking, physical activity, pre-eclampsia, smoking, twins Summary A growing body of work indicates that exposures over the life course have important roles to play in the aetiologyof breast cancer. This review synthesises the literature that has been published in the area of early life eventsand female breast cancer risk. The review finds some evidence, primarily from cohort studies on the relationshipbetween birthweight and breast cancer, to suggest that  in utero  events are related to breast cancer risk in adulthood.Strong evidence to support a positive association between height and breast cancer exists. Postulated mechanismsforthisrelationshipincludetheroleofearlydietinsubsequentdisease risk, andtheinfluenceofendogenousgrowthfactors mediating the relationship. There is some evidence to suggest that leg length is the component of heightwhich is generating the observed associations between height and breast cancer. There is no consistent pattern of association between relative weight in childhood or adolescence and risk of breast cancer. The evidence to suggestan association between physical activity in early life and breast cancer risk is convincingfrom case-control studies,but is not fully substantiated by the results of three cohort studies. There are inconsistent results regarding theassociation between smoking at a young age and breast cancer risk. There is little evidence for an associationbetween passive smoking in early life and breast cancer risk. No clear association between early drinking andbreast cancer risk exists. These results are discussed in relation to possible underlying mechanisms and healthpromotion strategies which could reduce breast cancer risk. Introduction Accumulating evidence provides convincing supportfor the view that events occurring throughout the lifecourse may influence health in later life [1]. It hasbeen proposed that a model which considers criticalperiods of exposure, risk accumulation and interac-tions between early and adult life exposures is likelyto be the most useful in understanding disease risk [2]. A growing body of work indicates that exposuresover the life course have important roles to play in theaetiology of breast cancer. Markers of pre-natal andchildhood exposures which have been associated withbreast cancer risk include birthweight [3], growth [4],age at menarche [5], peak height velocity [6], attainedadult height [7], diet [8] and overweight [9] in child-hood and adolescence. The aim of this review is tosynthesise the literature that has been published in thearea of early life events and female breast cancer risk.Understanding exposures which influence breastcancer risk at different ages is important for two mainreasons. Firstly, a better understandingof predictorsof riskallows a clearercomprehensionoftheaetiologyof disease. Secondly, identifying risk factors which maybe amenableto changepaves the way for public healthinterventions to reduce disease risk. This may be of particular importance for risk factors which are identi-fied in pre-adult life, since behaviour in young peoplemay be more amenable to change than behaviour inlater life.Two mainstages ofpre-adultlifehavebeenstudiedin relation to breast cancer. Firstly, exposures which  224  M Okasha et al. occur  in utero  and during infancy, which are the fo-cus of a recent comprehensive literature review [10].Secondly, adolescence, during which breast tissue un-dergoes substantial cellular proliferation [11], may beof importance. The high rate of cell division duringadolescencediminishesthe time availableforDNA re-pair. Cellular differentiation occurs during a woman’sfirst pregnancy. Thus the time between menarche andfirst birth is thought to be the time at which the breastis at its most vulnerable. Exposure to potential car-cinogenic factors before cellular differentiation mayincrease later cancer risk [12].Epidemiological evidence adds to the importanceof consideringpre-adult exposures in the developmentof breast cancer risk. Firstly, early age at menarcheincreases the risk of breast cancer, with women inwhom menarche occurs before 12 years having a 30%higher risk than women whose menarche occurred atage 15 years or later [5]. Secondly, migrant studiesshow that the age at which a woman moves from onesocio-cultural setting to another is related to their risk of breast cancer. Among women who moved fromsouthern to northern Italy and from Japan to the US,those who migrated at an early age experienced a risk of breast cancer similar to their host nation, whereasthose who migrated later had risks similar to theirregion of srcin [13, 14].These findings suggest that certain exposureswhich occur in early life may be related to breastcancer risk. This review identifies and summarisesthe results of studies which have reported the rela-tionship between childhood and adolescent exposuresand breast cancer risk. Exposures up to age 25 areincluded. For completeness, the role of   in utero  andinfant exposures, which have recently been compre-hensively reviewed [10], are summarised and updated. Methods The electronic database MEDLINE was searched sys-tematically by one reviewer (MO) for articles pub-lished between 1966 and March 2002. The searcheswere restricted to articles published in English. Forarticles published in 2000–2002, two reviewers per-formed the searches independently (MO and PMcC).Each search included the Medical Subject Headingsunder ‘breast neoplasms’ and the keyword ‘breastcancer’. The search terms which were used for eachrelevant exposure are listed in the Appendix.If the title appeared relevant, the abstract was re-viewed. If no relevant results were mentioned in theabstract, the full article was reviewed to determinewhether it should be included or not. Relevant pa-pers cited in the retrieved articles were also obtained.References cited by the authors of past reviews wereobtained (see Appendix). Data were abstracted fromthe srcinal papers and tabulated by one author (MO).No quality assessment was made for inclusion pur-poses. However, details of study design and analysisthat may have affected the results were noted at thetime of data extraction. In the narrative synthesis,more weight was put on larger or prospective studiescompared to smaller or retrospective investigations.No attempt was made to quantitatively synthesise theresults, because of the potential mis-interpretationof results from meta-analyses of observational data[15].Imperial measures were converted to metric forease of comparison across studies. Relative risks werepresented in relation to an increase in the relevant ex-posure, which required taking the reciprocal of theresults reported in some papers. Results were abstrac-ted according to the number of decimal places usedin the srcinal paper. Where more than one paper hasbeen published from the same cohort, only the mostrecent results are presented. In some of the includedcase-control studies, hospital and population controlswere analysed together. In these studies, the resultsare includedwith the hospital-basedstudies, since sig-nificant deviation from population norms are likely toexist in hospital controls. Where both a comparisonof means and odds ratios (OR) are reported, only thelatter were abstracted. Studies nested in a cohort witha follow-up component were considered with cohortstudies; those based on cross-sectional data were in-cluded with case-control studies. Of this latter group,those based on women attending screening were con-sidered population-based; those based on outpatientattendees were considered as hospital-based. Results Peri-natal events: a summary and update Pre-eclampsia is a condition of pregnancy character-ised by hypertension and proteinuria. Although theaetiology of the condition is unclear, pre-eclampsia isassociated with low levels of maternal oestrogen [16].It has been suggested that breast cancer risk may be  Pre-adult exposures and breast cancer risk   225partly “programmed” by early exposure to oestrogenlevels [17], and that girls born of pre-eclamptic preg-nancies may be at reduced risk. A recent follow-upof women born from pre-eclamptic pregnancies sug-gests that if pre-eclampsia is related to breast cancer,the mechanistic pathway is unlikely to involve age atmenarche or anthropometry[18]. Of four case-controlstudies identified in the review by Potischman andTroisi [10], one reporteda marked increased risk (OR:3.5, 95% confidence interval (CI): 0.9–14), thoughthis was based on only eight pre-eclamptic pregnan-cies. Two others found a marked decreased risk (OR0.2, 95% CI: 0.1 to 0.7 and 0.4 (95% CI: 0.2–0.8),respectively), and one found no relationship of breastcancer with pre-eclampsia (OR: 0.8, 95% CI: 0.4–1.5). A more recent case-control study found resultsconsistent with an inverse relationship between pre-eclampsia and breast cancer, although the results didnot reach conventionallevels of statistical significance(OR: 0.4, 95% CI: 0.1–1.4) [19].However, recent evidence suggests that umbilicalcord oestrogen is similar in eclamptic and normalpregnancies, whereas cord plasma levels of alphafetoprotein ( α FP) were substantially higher in severeeclamptic compared to normal pregnancies. A re-cent US study reported reduced risks of breast canceramong women with small placentae or placental in-farction. There is some evidence that such placentalcharacteristics are associated with lower oestrogenlevels and possibly pre-eclampsia [20]. It is knownthat in animals, fetoproteins bind estradiol and sup-press the estrogen-dependent growth of breast cancercells [21]. Thus,  α FP appears to possess biologi-cally important anti-carcinogenic properties [22]. Ithas been reported that a high level of   α FP in mater-nal serum during any pregnancy is associated with alow overall incidence of breast cancer and, in particu-lar, with a low incidence of advanced breast cancer atdiagnosis. This association appears particularly strongfor a pregnancy occurring at a young age [23]. Whilethese findings are provocative, the scant evidence todate coupled with the lack of understanding of thepathogenesis of pre-eclampsia invites further work inthis area.The presence of two placentas generally resultsin higher oestrogen levels in dizygotic (DZ) com-pared to monozygotic(MZ) twin pregnancies. Of fourcase-control studies of twinship identified in the 1999review [10], only one classified twins as DZ or MZ.In that study DZ twins were at increased risk rela-tive to singletons, whereas MZ twins were not. Twofurther studies used sex of the other twin as a proxymarker for zygosity. Although this is accurate for as-signing dizygosity to different sex twins it could resultin a degree of mis-classification for same-sex twins,since not all same-sex twins are MZ. One of thesefound no relationship with twinship, and the otherreported higher risk of breast cancer in women witha twin brother compared to a twin sister. The fourthcase-control study considered only twin births versusnon-twin births, and detected no association.One cohort study found no difference in risk among women with twin brothers compared to twinsisters. Three studies using twin registries comparingrisks to the general population have found a higherrisk among all twins (Denmark), a lower risk amongsame-sex twins (Finland), and no association withtwinship (Sweden). These studies have now been ad-ded to by a large US cohort study examining theassociation between twinship and post-menopausalbreast cancer risk [24]. There was an increased risk of breast cancer in twins compared to singletons, 1.72(95% CI: 1.22–2.42). The risk in twins was raisedwhere both twins were female, 1.82 (95% CI: 1.20–2.75), but not significantly higher in unlike-sex twins,1.49 (95% CI: 0.80–2.78). The risk was particularlyraised where the female twins were DZ, 2.14 (95%CI: 1.21–3.79). While not all studies support thepositive association with twinship, and particularlyDZ twinship, the evidence for such an associationis accumulating. Further cohort studies comparingDZ to MZ twins, adequately controlled for the num-ber of placentas and other twinship factors, are stillrequired.An interest in birthweight as a marker of intra-uterine exposures has recently developed. Based onfive studies, Potischman concluded that high birth-weight may be related to an increased risk of breastcancer, although possibly only for pre-menopausaldisease [10]. This conclusion was based on positivefindings from two case-control studies. A third case-control study found no association. In the Nurses’Health Study, women of highest birthweight wereat highest risk, and the same pattern was seen ina Swedish birth cohort. However using additionaldata from other hospitals, the same authors couldnot replicate these findings. Since the 1999 review, anumber of studies have investigated the relationshipbetween birthweight and breast cancer. Four furthercase-control studies have all provided evidence of a positive association with birthweight [25–28], al-thoughallbutoneofthese[25]weresmallstudieswith  226  M Okasha et al. imprecise estimates. Two further case-control studiesreported no association between self-reported birth-weight and breast cancer [29, 30]. Among membersof the UK 1946 birth cohort followed up until 1992,high birthweight ( ≥ 4kg v.s.  < 3kg) was associatedwith higher breast cancer risk (RR: 2.02 (95% CI:0.59–6.90),  P  (trend): 0.13) [31]. In this paper therewas an interaction between birth weight and height atage 7, suggesting that the effect of fetal growth/birthweight maybe modifiedbychildhoodgrowthvelocity.A Swedish study linking hospital birth records to thecancer registry found a non-significant higher risk of breast cancer among women in the top compared tothe bottom quintile of birthweight (OR: 1.57, 95%CI: 0.67–3.64,  P (trend): 0.23) [32]. In conclusion, itappears that there is a modest, positive associationbetween birthweight and breast cancer.It has been suggested that maternal oestrogens arehigher in older pregnant women, and that this mayconfer an increased risk of breast cancer on girlsborn to older mothers. Potischman reviewed 13 ar-ticles which report associations between maternal ageand breast cancer risk [10]. Six of these report mod-estly increased risks and seven report no association.Maternal age may be related to birth order, whichitself may be related to cancer risk, although mech-anisms underlying this are uncertain. The crude effectof birth order and the possible modification of thematernal age–breast cancer relationship by birth or-der remain unclear. A recent study from Denmark,which confirmed that increasing age at first birth isa risk factor for breast cancer, also showed that ageat any birth is associated with maternal breast can-cer risk [33]. Furthermore, maternal oestrogen levelsmay not adequately reflect umbilical levels. A clearerunderstanding of the determinants of fetal oestro-gen exposure is required to understand the observedrelationships.Beingbreastfedas aninfantconferswell-describedadvantages on child health, though whether the ben-efit persists to adulthood is not clear. Early hypoth-eses that breast cancer was caused by a viral agenttransmitted from mother to child were not upheld byresults from an ecological study or three individual-based studies. These, and six more recent studies of the relationship between infant diet and breast can-cer, are detailed in the review by Potischman andTroisi [10]. Three of the recent studies were basedon self-reported early diet and found a protective ef-fect of having been breastfed. Similar results werefound in one of the three studies based on ma-ternal reports. These divergent results suggest thatself-recall of early diet may be inaccurate. A re-cent US study found that there was no associationbetween self-report of having been breastfed and thedevelopment of breast cancer later in life among pre-menopausal women, OR 0.97 (95% CI: 0.78, 1.20)or post-menopausal women, OR 1.12 (95% CI: 0.92,1.37) [34]. In summary, evidence of an associationbetween being breast-fed and breast cancer is incon-sistent.In conclusion, there is some evidence, primarilyfrom cohort studies on the relationship between birth-weight and breast cancer, to suggest that  in utero events are related to breast cancer risk in adulthood.However associations with pre-eclampsia, twinship,early diet and maternal age as well as the inter-relationships between these factors and birthweightand breast cancer are far from clear.  Diet and nutritional status Although it has been estimated that between a thirdand a half of all breast cancers could be prevented bychanges in dietary practice [35], the epidemiologicalevidence to support this is inconsistent. Differencesbetween studies may be explained by the accuracywith or the age at which the dietary intake was mea-sured. Speculations regarding associations betweenadolescent diet and breast cancer risk have convinc-ing biological rationale [36, 37], though few studieshave quantitatively described any such relationship.This may be due to methodological complexities of dietary evaluation, difficulties associated with the re-call necessitated by case-controlstudies, or long latentperiods required by cohort studies. Childhood diet  There have been few studies able to directly studychildhood diet and subsequent breast cancer risk. An-imal models suggest that pre-pubertal diet rich inphyto-oestrogensmay reduce breast cancer incidence,although it is unclear whether the results of suchexperiments are applicable to humans [38]. One un-published study based on the Boyd Orr cohort, whichhas measures of diet from early childhood, reportedno associations between fruit, vegetables and vitaminC and breast cancer. Higher vitamin E intake was as-sociated with decreased risk of breast cancer, whilsthigher intake of retinol was associated with increasedrisk (Maynard, 2002, personal communication).  Pre-adult exposures and breast cancer risk   227  Adolescent diet  Six groups of investigators studied the relationshipbetween recalled adolescent diet and breast cancerrisk. Three case-control studies collected adolescentdietary intake from a food frequency questionnaire(FFQ) [8, 39, 40]. In one, inverse associationswere found between adolescent dairy fat intake andcancer risk (OR for the top to bottom quartile inpre-menopausal women: 0.4 (95% CI: 0.1–1.1); inpost-menopausal women: 0.2 (95% CI: 0.0–0.8) [40].Similar relationships were not evident for total fatintake or for fat from other sources. In the secondstudy, numerous associations were investigated andthe only statistically significant finding was betweenconsumption of the visible fat on meat and the risk of pre-menopausal breast cancer [8]. The third studyfound a reduced risk of breast cancer among womenwho consumed a high proportion of soy foods in theirdiet when aged 13–15 years (OR for the top to bottomquintile of consumption: 0.51; 95% CI: 0.40–0.65)[39].Results weresimilarinpre-andpost-menopausalwomen. A fourth study asked pre-menopausal womenabout consumption of 29 items at ages 12–13 years[41]. High intake of fruit and vegetables appearedprotective, though the results were of borderline sta-tistical significance andnumerousstatistical tests wereperformed.A Norwegian cohort study of pre-menopausalbreast cancer found lowered risks of breast canceramong women who consumed high quantities of milk as a child and as an adult [42]. Evidence of an as-sociation between childhood milk consumption andbreast cancer was not strong (greater than six glassesv.s. none: 0.64 (95% CI: 0.22–1.87)). The final study,only published in abstract form, is an analysis of re-called diet duringhigh school from the Nurses’ HealthStudy [43]. Protective effects were seen for fibre, me-thionine,eggsandmonounsaturatedfat, amongstpost-menopausal women only. Although of major interestbecause of the potential for preventativeinterventions,overall these results do not provide a consistent pic-ture of the relationships between dietary variables andbreast cancer, much less, evidence on which dietaryrecommendations could be based. Calorie restriction Natural experiments of under-nutrition, primarilythose enforced by rationing during the Second WorldWar (WWII, 1939–1945), have allowed opportunisticinvestigation of the relationship between nutrition andcancer risk. By comparing trends in incidence datafrom the Norwegian cancer registry, it appears thatwomen who experiencedpuberty during the war had adecreased risk of breast cancer in subsequent life [44].Although not the only explanation, these results areconsistent with the hypothesis that dietary restrictionduring adolescence may have favourable long-termconsequences on breast cancer risk. In contrast, aDutch study found no association between area of res-idence during WWII – a proxy for caloric intake –and breast cancer risk [45]. Such crude classificationof exposure of these women, by area of residence,clearly has limitations. Further studies exploiting dataavailable from times of enforced changes in nutritionwould contribute to our understanding of this area.Results fromanimalexperimentssuggestthatover-nutritionmayberelatedtocancerrisk. Astudyofmicedemonstrated a positive relationship between caloricintake and mammary tumour incidence, independentof fat intake [46]. In an ecological study, calorie re-striction in early life was related to lower breast cancermortality [47]. The only study with prospective mea-sures of energy intake in childhood is the Boyd Orrstudy [48]. In that cohort, energy intake was weaklyassociated with breast cancer mortality (a 10% in-creased risk per 1MJ increase; 95% CI: − 11 to 36%).This analysis is limited by the fact that only 26 womendied of breast cancer.  Indirect measures of exposure The use of anthropometric measures is an indirectmethod by which associations between childhood en-ergy balance and subsequent disease risk may beevaluated. In the following sections, the associationsbetween breast cancer and height, leg length andrelative weight are considered.  Height  Constrained by a genetically determined maximumpotential,adultheightreflects, amongotherexposures,nutritional status during early life [49] while heightdeficit is a marker of chronic material deprivation[50]. An interest in the relationship between heightand breast cancer has existed for over 30 years sincethe association was first highlighted by MacMahonet al. [51]. The results of four ecological studies aredetailed in Table 1. These studies show strong correla-tions between height and breast cancer incidence andmortality on an international and national (UK) level[52–55]. However, difficulties in the interpretation of these data, which include confounding and ecological
Similar documents
View more...
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks

We need your sign to support Project to invent "SMART AND CONTROLLABLE REFLECTIVE BALLOONS" to cover the Sun and Save Our Earth.

More details...

Sign Now!

We are very appreciated for your Prompt Action!