137x Filetype PDF File size 1.00 MB Source: academic.oup.com
November 2001: 349-357 zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA Lead Review Article zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA Physical Activity and Cancer: Lessons Learned from Nutritional Epidemiology zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA Christine M. Friedenreich, Ph.D. Downloaded from https://academic.oup.com/nutritionreviews/article/59/11/349/1827811 by guest on 04 January 2023 State of Scientific Evidence for Physical Activity Convincing epidemiologic evidence currently ex- in Cancer Etiology ists for an association between physical activity and the prevention of colon and breast cancer. A comprehensive review of the epidemiologic literature Physical activity may also reduce the risk of can- on physical activity and cancer prevention was recently cer at several other sites. With increasing research conducted.' That review used the classifications for sci- on this topic, it is apparent that studies of physical entific evidence that had been developed by the panel of activity and cancer have numerous methodologi- scientists who wrote the report of the World Cancer Re- cal similarities with studies of nutrition and can- search Fund/American Cancer Institute for Research on zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA cer. Lessons learned from nutritional epidemiol- Food, Nutrition and the Prevention zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBAof Cancer: A zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBAGlobal ogy that can be applied to studies of physical ac- Perspective.2 Briefly, four categories for scientific evidence tivity and cancer prevention and recommendations were defined. Convincing evidence exists when consis- for future research are discussed in this review. zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA tent associations are reported in at least 20 studies con- ducted worldwide that used different study designs, that showed dose-response relationships, that are biologically Introduction plausible, and for which laboratory evidence is also sup- Whereas evidence of a role of physical activity in cancer portive of the association. Probable evidence exists when etiology has accumulated rapidly, particularly in the last consistent associations are found in less than 20 studies decade, there are numerous methodological limitations to conducted worldwide, using different study designs, studies in this area. The field of physical activity and can- when biologic plausibility is likely, and some supportive cer epidemiology has many similarities to nutrition and laboratory evidence exists. Possible evidence exists when cancer epidemiology, and numerous lessons learned in there are generally supportive studies; however, they are nutrition and cancer research are directly applicable to limited in quantity, quality, or consistency, and laboratory physical activity and cancer. During the past 30 years, evidence may not exist. Insufficient evidence exists when considerable advances in dietary assessment methods and there are only a few studies that are generally consistent studies of nutrition and cancer have been achieved through but more research is needed because there is simply not the collaboration of nutritionists, epidemiologists, and enough evidence to make any conclusions about the as- biostatisticians. For comparable advances to be achieved sociation. in physical activity and cancer research, similar Applying these definitions to the literature on physi- multidisciplinary collaborations are needed between exer- cal activity and cancer prevention, it was concluded that cise physiologists, epidemiologists, cancer biologists, and there is convincing evidence that physical activity re- biostatisticians. The objective of this review is to high- duces colon and breast cancer, whereas the evidence for light the state of evidence for physical activity and can- prostate cancer is probable, evidence is possible for lung cer, outline the methodological issues and gaps in knowl- and endometrium cancers, and evidence is still insufi- edge that remain, and suggest how advances can be cient to make any judgement for the remaining cancer achieved in this field by adapting methods and knowl- sites.' Table 1 summarizes the literature by listing the main edge gained in nutrition and cancer epidemiology. zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA findings from this body of literature. The key features listed include: 1) the number of studies that showed a risk reduction among the most physically active study sub- jects as compared with the least active participants (i.e., Dr. Friedenreich is with the Division of Epidemiology, Prevention and Screening, Alberta the consistency of the association); 2) the average risk Cancer Board, Calgary, Alberta, Canada, T2N 4N2. reduction found in these studies (i.e., the strength of the Nutrition Reviews@, Vol. 59, No. 11 349 Table 1. Summary zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBAof Epidemiologic Evidence on the Association between Physical Activity and Cancer by Criteria for Causalitv zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA Consistency of Strength of Overall Association Association Dose-ResDonse Level of Number of Average Risk Number Hypothesized Scientific Biologic Mechanisms Evidence zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA Cancer Site Studies Reduction of Studies Colon 39 of46 4&50% 23 of 29 Decreased gastrointestinal Convincing transit time, decreased ratio of prostaglandins (PGEPGF), lowered bile acid secretion or enhanced metabolism, decreased percent body fat,* decreased circulating insulin and glucose, Downloaded from https://academic.oup.com/nutritionreviews/article/59/11/349/1827811 by guest on 04 January 2023 * exercise-induced increase in antitumor immune defences, * improved antioxidant defence systems,* genetic predisposition of habitually active* Breast 25 of36 3040% 15 of22 Decreased lifetime exposure Convincing to estrogen Prostate 14of25 1 &30% lOofl8 Reduced exposure to testosterone Probable 3wo% zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA5 of7 Possible decreased exposure to Possible Lung 60f8 inhaled carcinogens with increased pulmonary hnction Endometrium 8 of 12 3040% 4of7 Decreased lifetime exposure to Possible estrogen, decreased circulating insulin and glucose Testicles 3 of8 20% 3 of5 Trauma to testes? Influence on Insufficient androgens and gonadotrophins? 2 of3 Decreased number of lifetime Insufficient Ovaries 20f5 zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBAo?? ovulatory cycles, decreased ratio of prostaglandins (PGEPGF) * These mechanisms are equally applicable to all cancer sites. association); 3) the number of studies that found that risk of the identified methodologic weaknesses in the extant of cancer decreased linearly as the level of physical activ- literature. These weaknesses include the crude and impre- cise measures of physical activity, the lack of full assess- ity increased (i.e., the dose-response zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBArelation); 4) the un- derlying hypotheses that might explain the reasons for ment and control for confounding factors including di- the association (i.e., the biologic plausibility of the asso- etary intake, the lack of consideration of effect modifica- ciation); and 5) the overall classification for the evidence tion by other factors, and the lack of understanding of the for this association. underlying biologic mechanisms that may be operative in Based on this review of the literature,' it was argued these associations.' Hence, more research is needed to that sufficient evidence has accumulated from observa- confirm the etiologic role for physical activity in cancer tional epidemiologic studies regarding the association prevention in most cancer sites before intervention stud- between physical activity and colon and breast cancer to ies can be conducted. warrant commencing intervention studies of physical ac- For faster progress to be made in the field of physical tivity for those two cancer sites. The outcome of these activity and cancer prevention, it is also prudent to learn intervention studies would be more precise information from progress made in the field of nutrition and cancer on the exact timing and dose of physical activity required prevention because of the similarities between the disci- for cancer risk reduction. At present, sufficient evidence plines and the common and overlapping biologic models exists to have general public health recommendations to for the influence of these lifestyle factors on carcinogen- increase physical activity in order to reduce cancer risk; esis. The next section highlights methodological issues however, insufficient evidence exists to formulate more that are common to both fields and areas for hture progress precise activity prescriptions for the general population. and collaboration. For the remaining cancer sites, there is insufficient Methodological Issues and Solutions in evidence at present to justify cancer prevention interven- Nutritional Epidemiology tion trials of physical activity. Rather, additional observa- tional epidemiologic studies are needed that address some In both nutritional and physical activity epidemiologic zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA 350 Nutrition Reviews@, Vol. 59, No. zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA11 studies, the complexity of the exposures and the correla- and that are of heterogeneous origins. One example is the tions between the exposures and other behaviors must be European Investigation in Cancer and Nutrition prospec- considered in the collection and analysis of data.3 Be- tive cohort study5 that has as one of its main objectives to cause each type of epidemiologic study available for the overcome the problem of homogeneous exposures and investigation of diet or physical activity and cancer asso- study populations. This study will compare and combine ciations has both strengths and some major limitations, data from ten European countries where the populations no one study design or method of data collection is con- have very different dietary habits. Future multicentered sidered definitive. Observational epidemiologic studies cohort studies of physical activity are needed that in- of diet and physical activity have several advantages when clude diverse populations with heterogeneous activity compared with intervention trials, including the ability to levels and different ethnic and racial origins. Such large- examine exposures that would be difficult to assign ex- scale studies that have sufficient statistical power could perimentally and the ability to assess effects of exposures then compare and contrast the effects of physical activity Downloaded from https://academic.oup.com/nutritionreviews/article/59/11/349/1827811 by guest on 04 January 2023 in free-living populations over a long time period. Obser- on disease risk for these different ethnic and racial groups. vational studies of diet and physical activity also have a number of key limitations that need to be considered when Data Collection Methods assessing the results of these studies. Measurement of The second methodological issue is related to the type of dietary intake and physical activity share numerous meth- data collection methods and their validity and reliability. odological challenges. Both are complex exposures that Many dietary assessment methods have been developed have several components that change during a study for different purposes. These methods include 24-hour participant’s lifetime, and that require current databases recalls, 7-day weighed food records, food-frequency ques- to estimate accurately either the energy and nutrient in- tionnaires, and diet history questionnaires. Dietary as- take or the energy expenditure. sessment methods are constantly undergoing improve- Nine of the main methodological issues of observa- ment and many of the methods used to develop and im- tional epidemiologic studies that have been encountered prove these instruments can be directly applied to the in nutritional epidemiology and that have direct relevance development of physical activity assessment methods. to physical activity epidemiology are homogeneity in popu- For example, food-frequency questionnaires were devel- lation sampling, validity and reliability of the data collec- oped from the analysis of 24-hour recalls, weighed food tion instruments, measurement error problems, accuracy records, and dietary survey data. On this basis, the most and completeness of databases, lack of consideration of commonly consumed food items were selected and in- dietary and lifestyle patterns, confounding, effect modifi- cluded in the food-frequency questionnaires. A compa- cation, appropriate statistical methods, and elucidation of rable approach can be applied to the development of physi- biologic mechanisms. zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBAcal activity questionnaires. Data on the most commonly reported activities could be compiled and these activities Population Sampling Methods used when preparing lists of activities that study respon- The first methodological issue is the range of exposure dents report. In addition, substantial research has been within the study population. Nutritional epidemiologic undertaken in nutritional epidemiology to estimate the studies have often been conducted within populations validity and reliability of dietary assessment methods, that have very homogeneous exposures. Furthermore, whereas far less research has determined the psychomet- there have frequently been insufficient numbers of study ric properties of physical activity assessment methods. subjects with very low or very high intakes to investigate Clearly, an important area of hture consideration in physi- the influence of exposures at these levels. It is possible, cal activity epidemiology is to establish and ensure the for example, that a threshold exists for the effect of certain reliability and validity of the data collection instruments macronutrients (e.g., fat) on tumor promotion and that the being used. lowest percentage of fat intake among individuals or groups in any particular study population is higher than the thresh- Measurement Error Problems old.4 The range of intakes within study populations may The third methodological issue addressed in nutritional therefore have been too narrow to detect associations epidemiologic studies is the systematic and random er- that do exist. A zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBArelated issue is the ethnic and racial com- rors that occur in dietary asse~sment.~.~ One source of position of the study sample. Epidemiologic studies have random error in dietary intake is the natural variation in too frequently been conducted within one ethnic or racial daily eating patterns.* Dietary assessment methods that group, thereby limiting the generalizability of the results. inquire about intake over a relatively extended time period These limitations are being overcome in nutritional attempt to decrease to a minimum the random error attrib- epidemiology by conducting studies in several centers or utable to daily fluctuations in intake and activity. These among populations that have heterogeneous exposures assessment methods can also introduce systematic error, zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA Nutrition Reviews@, Vol. 59, No. 11 zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA 35 1 however, because study subjects may under- or over-re- the sources of random and systematic error and by ana- port their intake or activity. Some incorrect reporting may lyzing measurement errors as has been done in nutritional be intentional (e.g., overweight study subjects tend to epidemiology. Some progress has already been made in under-report their dietary intake9J0 and over-report their the development of questionnaires that assess longer- activity”) and some reporting may be inaccurate simply term physical activity habits (to decrease the random er- because respondents are not sufficiently aware of their ror attributable to daily changes in activity patterns), and dietary intake to report it reliably. Substantial measure- that use cognitive methods to develop and administer the ment error and consequent misclassification are known to questionnaire (to decrease random and systematic errors exist in nutritional epidemiologic studies, as do conse- attributable to recall errors and difficulties associated with quent attenuation of the risk estimates and a reduction in estimation of past activity levels). One such example is the probability of finding an existing diet-cancer relation- the Lifetime Total Physical Activity Questionnaire,2’ which ship.12J3 Flegal14 has argued, however, that measurement was developed to measure all types of activity performed Downloaded from https://academic.oup.com/nutritionreviews/article/59/11/349/1827811 by guest on 04 January 2023 errors in dietary assessment are not always random and by respondents during their entire lifetimes. An additional that it is erroneous to assume that random measurement improvement for physical activity questionnaires is to error biases estimate only toward the null value. avoid the use of grouped activities and to have respon- Careful analysis of dietary data has revealed the diffi- dents report the actual activities that they have performed culties that respondents have when reporting their usual without the use of any predefined lists. Such an approach dietary intake. These problems include omitting or incom- has been developed for interview-administered question- pletely specifying foods, reporting foods that are not naires; self-administered versions are under development.21 actually consumed, or estimating portion sizes incor- This method is analogous to the interview-administered rectly.I5-I7 Dietary assessment methods assume a certain diet history questionnaire in which respondents report all level of knowledge about dietary constituents and the foods that they normally consume either by meals during ability to remember and estimate intake.18 New dietary as- the day or overall during the day. zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA sessment methods are being developed that take into con- sideration the cognitive processes that study subjects Dietary and Lifestyle Patterns undergo when responding to these types of questions. The fourth methodological issue is the lack of consider- Improvements in epidemiologic data collection methods ation of patterns of exposure. Dietary assessment meth- through the use of cognitive interviewing has been advo- ods to date have been restricted to capturing the intake of cated by cognitive psychologists and survey research- individual food items or groups of food items rather than er~.~~ These methods have been used in the development dietary patterns. The main reason for the focus on the of dietary assessment methods (e.g., Dietary History Ques- assessment of individual foods rather than patterns of tionnaire created by the intake in nutritional epidemiology relates to the feasibility U.S. NC120), although much more research in these methods is still needed. Assessments of and logistics of data collection methods. A few investiga- dietary intake require that the respondents have adequate tors have attempted to analyze dietary patterns of intake cognitive ability to estimate and quantify their intake and to identify dietary habits that are associated with increased report the frequency and portion size of their regular di- or decreased risks of cancer or m~rtality.~~.~~ An “integra- etary intake. tive,” rather than a “decompositional,” approach to study- Additional measurement errors arise when individual ing the association between diet and cancer is receiving foods are combined into one group in questionnaires zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBAmore recognition as a valid and more realistic method for .I4 Respondents must then estimate how frequently they con- assessing the total impact of dietary habits on risk.24 The sume a number of different foods and combine their esti- integrative approach emphasizes patterns and linkages of mates to produce an overall intake value. Measurement dietary lifestyle choices. Physical activity patterns over errors result from the difficulty in estimating the overall lifetimes have only rarely been e~amined,2~3~~ but may be frequency of the intake in addition to assigning the same more etiologically relevant than measures of single types nutrient content to all foods within a food group. Hence, of activities performed for restricted time periods in life. A in nutritional epidemiologic research, the recommended clear need exists to broaden the definition of physical approach is to have respondents report individual foods, activity to examine entire lifetime activity patterns. rather than grouped food items, wherever possible and to develop and use nutrient databases that are more com- Databases plete and comprehensive. The fifth methodological issue in nutritional epidemiol- The sources of random and systematic error found in ogy is the lack of current, complete, and accurate informa- nutritional epidemiologic studies are comparable with those tion on the composition of food products in the nutrient that exist in physical activity studies. Improvements in databases used to estimate nutrient intake.27 These data- physical activity assessment can be made by identifying bases are subject to changes in nutrient content of foods zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA 352 zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA Nutrition Reviews@, Vol. 59, No. 11
no reviews yet
Please Login to review.