Obesity and Body Composition
Total Page:16
File Type:pdf, Size:1020Kb
22 Obesityand Body Composition RACHELBALLARD-BARBASH, CHRISTINE FRIEDENREICH, MARTHA SLATTERY, AND INGERTHUNE An" explanationfor the increasingincidence of cancersin devel- papersor books were also identified and reviewed. Summary figures rr.-/oping countries has been the hypothesis that a chronic state of of risk estimatesfor the highestcompared to lowest quantileof BMI positive energy balance promotes tumor growth. (Tannenbaum and the major cancer sites were generated.Because of the very large l940a,b) first explored this hypothesisin animal models and demon- number of studiesin breastcancer, studies with at least 100 cases strated that both increasedcalories and fat increasedthe occurrence each of pre- and postmenopausalbreast cancer casesare included and number of breastcancer tumors in mice. Similarly, epidemiologic in the breast cancer figures. Becauseof differencesin risk for colon studiesin humansin the 1970sdemonstrated that heavierwomen were and rectal cancer and the limited number of studies on rectal cancer at increasedrisk of breast and endometrial cancer (De Waard et al., alone, only data on colon cancer were summarized in the colon 1974;Blitzer e[ al., 1976).Other population-basedstudies in the early cancer figure. Biological mechanismspotentially involved in the 1980ssuggested that risk increasedwith increasingweight for several associationbetween obesity and specific cancers were summaized other cancer sites, specifically colon, prostate, gallbladder (among in a table. In addition, population attributable risks estimated for women), and kidney (Lew and Garfinkel, 1979: Hartz et al., 1984; the EuropeanUnion by Bergstromet al. (2001) were comparedto Garfinkel, 1986).Since then, numerousepidemiologic studies have currentestimates for the United Statesand summarizedin a table.A examined the associationof weight and other anthropometric meas- summarystatement for eachcancer site was developedfollowing the ures with cancer incidenceand have explored potential biological criteria for convincing, probable, possible, and no associationas mechanismsto explain observedassociations. The majority of these summarizedin the 1997World CancerResearch Fund and American studies have focused on breast, colon, endometrial, renal cell, Institue for Cancer Research(World Cancer ResearchFund, 1997) esophageaUgastric,and prostatecancer; limited data are available on report on Food, Nutition and the Prevention of Concer: A Global body size associationswith other cancer sites. Perspective. The aim of this review is to provide a comprehensiveoverview of the stateof scientificevidence for.the associationbetween obesity- related risk factors and cancer, with a focus on those cancers with METHODOLOGICAND MEASUREMENTISSUES sufficientevidence for review: thoseof the colon, rectum,esophagus/ gastriccardia, renal cell, endometrium,ovary, breast,prostate, thyroid, Existingmethods for assessingthe etiologicand prognosticinfluences lung, and head and neck. The specific objectives of this review of energybalance and body size on cancerhave limitations that should are to: be consideredwhen evaluatingthe epidemiologic evidenceand plan- ning future research.Briefly summarized,these limitations fall into . Summarizethe epidemiologic literature for the associationbetween threeareas: validity of exposureassessment, difficulty of comparisons cancer risk and obesity-relatedmeasures of weight or body mass acrossstudies, and insufficientsample sizes to explorefully the many index (BMI; definedmost commonly as weight in kilograms [kg] factorsthat relateto body size and fat massand that could potentially divided by height in meterssquared), fat depositionpatterns (com- modify observedassociations. monly approximatedby waist circumferenceor by the ratio of waist- The measurementof anthropometricindices is generallywell stan- to-hip circumferences[WHR] or by ratios of truncal to extremity dardizedand documented(Lohman et al., 1988;WHO, 1995).In fact, skinfolds), and weight change; the standardization,accuracy, and reliability of these measuresare . Highlight gaps in scientific knowledge for the association of betterthan for many other cancer-relatedexposures. Weight and height obesity-relatedfactors and cancerrisk; are the most standardizedmeasurements and least subject to variabil- . Identify the possibleunderlying biological mechanismsfor these ity. More recent investigationshave used interviewersto measure associations: variousanthropometric indices from study participants,hence reduc- . Comment on areasfor future research;and ing the possibility of random and systematic error. However, many . Estimate the public health impact of obesity on cancer risk at the of cancer have relied on weight and internationallevel. studies etiology self-reported height.Studies on self-reportedheight suggest a reasonabledegree of Becauseof spacelimitations, the epidemiologicevidence on height accuracywith a bias in overreportingheight that is somewhatgreater and cancerrisk is not includedin this review but has beenwell sum- in men comparedto women and that increaseswith age (Rowland, marizedin severalprevious reviews (Ballard-Barbash, 1999; Gunnell 1990)or is limited to peopleover age 60 (Kuczmarskiet a1.,2001), et al..2001). presumablydue to age-relatedloss in height.Studies on self-reported weight indicate the presence of significant misreporting at the extremesof weight, with heavier people underreportingand lighter METHODS people overreportingtheir weight (Rowland, 1990; Stevenset al., 1990;Must et al., 1993).Therefore, studies of chronicdisease that rely A search was conductedon MEDLINE and PUBMED for all publi- on self-reportedheight and weight will underestimatethe risk associ- cations on weight, body mass index, anthropometric factors, and ated with thesemeasures (Gunnell et al., 2000). Nonetheless,com- specific cancersin human populationsand was supplementedby a pared to estimatesof correlationbetween reported versus measured manual search of all major relevant journals. The literature search assessmentof otherexposures, such as dietary intake, correlation coef- includedall publicationsup to February2003. Studies included in rhis ficientsbetween recalled and measuredweight suggestthat weight is review focused on some aspect of anthropometric risk factors in recalledwith a reasonabledegree of accuracy.For example,in one relation to cancer risk for the cancersnoted above. Maior review U.S. study.correlations between reporled and measuredweishts fbr . =.; 422 '.;:ti ri; ,F*::- Obesity,end BodyCcmposition 423 elderly subjectsasked to recall weight currently and from 4 and 28 COLORECTALCANCER years previously were reponed to be 0.98 for current,0.94 for 4-year, and 0.82 for 28-year recall (Stevenset al., 1990).Measurement of Summaryof Findings skinfoldsand circumferencesis lessreliable than the measurementof Data from case-control and cohort studies provide convincing evi- weightand height (Lohman et al., 1988;WHO,1995). However, reli- dence of an approximately twofold increasein risk of colon cancer ability is improvedwith standardizationin measurementtechnique. It among men with a BMI of 30 or more. Risk estimatesfor men are is not feasible to obtain these anthropometric measuresfrom self- most commonly reported to be about 2. Among women, less exten- report.Waist circumference is currentlyconsidered to be the mostcon- sivedata provide evidence of a probablyincreased risk of coloncancer venient and simple measureof abdominal or central adiposity for from a high BMI, with risk estimatesusually between 1.0 and 1.5. epidemiologicresearch (WHO, 1995,2000). Percentbody fat can be Among obese women who are estrogen positive (defined as pre- estimatedfrom bioelectricimpedance but only has been reportedin menopausal or postmenopausaland taking hormone replacement one studyof canceretiology (Lahmann,2003). Bioelectric impedance therapy [HRT]), a twofold increasein risk has been reported.Most of is currently the most promising field method for estimating body the literaturehas found no associationbetween BMI and rectal cancer. composition in large epidemiologic studies becauseit is portable, Meta-analysesthat provide a more quantitativestatement of the asso- inexpensive,easy to use, and highly precise.Issues related to use of ciation acrossseveral studies have not beenpublished. different methodsfor estimatingbody compositionfor chronic disease epidemiology are well summaized by Baumgartner and colleagues (l99s). Overview Comparisonacross studies is difficult becausemost studieshave Many aspectsof the associationsbetween body weight, obesity, and examined risk by quantile distributions(most commonly tertile or colorectal cancer have been examined. Colorectal cancer has been quartile) for the BMI or other anthropometricmeasures used. As the studied by examining risk associatedwith colon and rectal cancers distributionof BMI varies acrosspopulations, these quantile groups combined, as well as for each site separately.Although many studies are not comparableacross studies, making comparisondifficult. With have examined associationswith colon cancer,few have considered increasinginterest in understandingthe risk for many chronicdiseases rectal cancer specifically.Additionally, unlike many cancers,associa- by standard WHO BMI categories,investigators have begun to tions betweenbody size and the precursorlesion, adenomatouspolyps, examinerisk by thesecategories: underweight as BMI of less than have been reported. Several consistentdifferences in associationsby 18.50,normal weight