Arthur Schatzkin, MD, DrPH
Chief, Nutritional Epidemiology Branch
National Cancer Institute
National Institutes of Health
Bethesda, Maryland
It’s hard to go into a supermarket these days and not be bombarded with claims—from the cash register magazine rack to the grocery shelves themselves—that some nutritional item will lower or raise your risk for this or that cancer.
The belief that nutrition--broadly conceived to include diet, body habits, and physical activity--plays an important role in the etiology of cancer is longstanding and widely held. To the extent that nutrition does affect cancer, however, most of these effects are likely to be modest, with at most a doubling or halving of risk for those at the highest, compared to lowest, categories of exposure. But, considering the many people who fall into such high- or low-exposure categories, even modest relative risks translate into a substantial impact on cancer morbidity and mortality.
There is no shortage of biologically credible nutrition and cancer hypotheses. This is not surprising given the variety and complexity of human nutritional exposures, as well as the many anatomic and histologic manifestations of malignant disease. Animal experiments have proven that nutritional change can alter tumorigenesis, even in transgenic models with genetic predispositions to develop neoplasms.(1) Moreover, nutritional factors plausibly influence a host of intermediate biologic processes, involving hormones, inflammatory and immune factors, reactive oxygen species, mutagens, mitogens, and so on, that could be important in carcinogenesis. 
Nevertheless, the scientific evaluation of these nutrition and cancer hypotheses is far from conclusive and considerable controversy prevails. For some seemingly established hypotheses (fruits and vegetables versus cancer at several sites, for example), the evidence has weakened in recent years.(2) Other hypotheses (fiber versus colorectal cancer), thrown into question by one set of findings, (3,4) have been ‘resurrected’, or at least rendered not yet dead, by new studies.(5,6) Important nutrition-cancer links with consistent observational epidemiologic findings have been contradicted by those from experimental epidemiologic studies (randomized trials)--carotenoids versus lung cancer, for example.(7,8) Conclusions from recent consensus panels are remarkable for how few nutrition-cancer hypotheses achieved a level of evidence considered ‘convincing’.(9)
To further advance our understanding of nutrition and cancer and thereby develop truly effective strategies for preventing malignant disease, we need to recognize and address several challenges to the field:
Recall bias. It is difficult to prove that recall bias in case-control studies is really a problem, though there are several instances now where associations seen consistently in earlier case-control studies are attenuated or absent in prospective cohort studies.(10) A few, but not all, studies in which diet is assessed prospectively and retrospectively (after cancer diagnosis) in the same population suggest that recall bias may be operative. The problem is now essentially moot. There are enough prospective studies around the world, including some very large ones, as well as pooling projects (efforts to combine several cohort studies together to increase sample sizes and statistical power), so even less frequently occurring cancers (pancreatic and ovarian, for example), can now be investigated prospectively.
Measurement error. Does the food frequency questionnaire, the typical method for assessing diet in epidemiologic studies, measure what people eat accurately enough to detect important (though modest) associations between diet and cancer? Some recent data suggest that this instrument may be less accurate than was previously thought, but this is currently a topic of considerable controversy and active research.(11) This is complicated further by the possibilities of measurement error in several nutritional and nonnutritional factors considered together in multivariate models.(12) The use of dietary patterns and various comprehensive indexes and scores may reduce misclassification of dietary exposure, but this is not certain. It is possible, but by no means proved, that the use of alternative (and generally more expensive) instruments such as dietary records and multiple 24-hour recalls will further clarify the role of nutrition in cancer—revealing, for example, certain dietary associations that were not revealed in food frequency questionnaire data.(13) Similar questions have been raised about the accuracy of standard physical activity questionnaires, and researchers are actively investigating new ‘high tech’ approaches (activity monitors, for example) for assessing activity in epidemiologic studies.
Inadequate intake range. It is important in epidemiologic studies that the intake range for various foods and nutrients is wide enough to make meaningful comparisons of persons with high and low consumption. If few people in a given study have high intake of fiber, or low intake of folate (especially postfortification in the US), then observed associations between ‘high’ and ‘low’ intake of these nutrients and cancer may not reflect the true biologic action of these nutrients. Because the extent to which true associations between nutrition and cancer are attenuated (biased toward the null) is a function of the ratio of inter-individual to intra-individual variability, a wide intake range compensates to some degree for inaccuracy in our dietary assessment. Investigators have found that the intake range for a number of foods and nutrients is wider in some populations than others, and that pooling of cohorts can generate more participants at high and low intake levels for a given study.
Specificity of exposure. Some researchers are recognizing that we may need to zero in more sharply on dietary exposures. Red meat intake, for example, has been linked to colorectal cancer, but the specific culprit has not yet been identified: red meat, processed meat, the products of cooking meat at high temperature (heterocyclic amines or aromatic hydrocarbons), heme iron, or nitrosamines? New data bases and assessment tools are being developed to allow investigators to appreciate these nuances of diet.(14)
Biological markers. Biological markers of intake and nutritional status complement data obtained from self-report. Some nutritional factors, such as selenium (which varies considerably throughout the food supply), can only be reasonably assessed through assays of blood levels. Combining information from self-report instruments and biological markers may improve the accuracy of dietary assessment.
A role for genes? Nutrition-gene studies may complement more traditional nutritional epidemiologic studies. These studies involve two strategies: 1) Elucidation of nutrition-gene interactions. These can reveal markedly enhanced relative risks within subgroups defined on the basis of allelic variation in nutrient metabolizing or transport genes; 2) ‘Mendelian randomization’(15), a strategy based on the associations with cancer of functional genetic polymorphisms mimicking high or low dietary exposure. Such associations largely preclude measurement error bias—and, because of the random assortment of alleles, confounding bias as well. Very large sample sizes, which may be achieved through efforts such as the newly established Consortium of Cohorts, are needed for these nutrition-genes studies, especially for detecting nutrition-gene interactions. A caveat, researchers have become increasingly aware, in the context of looking at many nutritional factors and many genes, of the dangers of false positive findings.(16)
Confounding. Confounding bias is a particularly worrisome problem in the nutrition-cancer field where we are likely dealing with weak to moderate effects. That is, people who tend to eat less fat, or more fiber, or consume more multivitamins may differ from their higher fat, lower fiber, fewer multivitamins counterparts on a variety of other biologic or lifestyle factors that are truly related to cancer. Observational epidemiologic studies attempt to deal with confounding by measuring and statistically adjusting for a variety of potentially confounding factors. The problem is that some true confounders may be unknown or inadequately measured. Randomized controlled trials (RCTs) avoid confounding bias because, if the study is large enough, randomization makes treatment groups similar with respect to both known and unknown confounders. The disparity between randomized trials and observational studies (as in the beta carotene-lung cancer studies, alluded to above) has been sobering for those working in the nutritional epidemiology of cancer. Positive results from RCTs are especially compelling, but it is clearly difficult to conduct RCTs of many nutrition-cancer hypotheses. A trial of alcohol intake versus breast cancer, for example, is simply not feasible, though studies of alcohol versus intermediate end points can be conducted. Ongoing trials of folate and calcium in colorectal neoplasia and vitamin E and selenium versus prostate cancer may be quite informative.
End point heterogeneity. Cancer end point heterogeneity may also represent a challenge for nutrition-cancer research. Certain nutrition-cancer relations may emerge only when we look within site-specific cancer subgroups defined by clinical, histologic, or molecular characteristics (disease stage or hormone receptor status, for example).
Given the current state of the science, the controversies, and the ongoing research efforts, what recommendations can we reasonably make at the present time (and what cancers would be affected)?
1. Avoid obesity (linked consistently to cancers of the large bowel, breast, endometrium, and kidney and esophageal adenocarcinoma; data are suggestive, but sparse, for liver, pancreas, nonHodgkin's lymphoma).
2. Maintain regular physical activity (linked to colorectal cancer, especially in men, and breast cancer in women).
3. Limit consumption of alcoholic beverages (cancers of the oral cavity, pharynx/larynx, esophagus, liver, and possibly breast in women).
And now the evidence gets somewhat weaker:
4. Limit consumption of red, preserved meat (colorectal cancer)
5. Consume at least four to five servings of fruits and vegetables daily (oral cavity, esophagus, stomach, colorectum)
There has been great interest in whether calcium/vitamin D or folate can prevent colorectal cancer, and whether vitamin E or lycopene can reduce the incidence of prostate cancer, but these hypotheses are not yet ready for ‘prime time’ public health recommendations. And the jury is still out on such long-standing important questions as fat versus breast cancer and fiber versus colorectal cancer.
There is very good reason to think that nutritional change will reduce the incidence and mortality from several cancers. But we have a lot of work to do to prove this and determine exactly what those dietary changes need to be.
References
- Mai V, Colbert LH, Berrigan D, Perkins SN, Pfeiffer R, Lavigne JA, Lanza E, Haines DC, Schatzkin A, Hursting SD. Calorie restriction and diet composition modulate spontaneous intestinal tumorigenesis in Apc(Min) mice through different mechanisms. Cancer Res 2003; 63:1752-1755.
- World Health Organization and International Agency for Research on Cancer. IARC Handbooks of Cancer Prevention. Fruit and Vegetables. Lyon: IARC Press, 2003.
- Fuchs CS, Giovannucci EL, Colditz GA, Hunter DJ, Stampfer MJ, Rosner B, Speizer FE, Willett WC. Dietary fiber and the risk of colorectal cancer and adenoma in women. New England Journal of Medicine 1999; 340:169-176.
- Schatzkin A, Lanza E, Corle D, Lance P, Iber F, Caan B, Shike M, Weissfeld J, Burt R, Cooper MR, Kikendall JW, Cahill J. Lack of effect of a low-fat, high-fiber diet on the recurrence of colorectal adenomas. Polyp Prevention Trial Study Group. N Engl J Med 2000; 342:1149-1155.
- Peters U, Sinha R, Chatterjee N, Subar AF, Ziegler RG, Kulldorff M, Bresalier R, Weissfeld JL, Flood A, Schatzkin A, Hayes RB. Dietary fibre and colorectal adenoma in a colorectal cancer early detection programme. Lancet 2003; 361:1491-1495.
- Bingham SA, Day NE, Luben R, Ferrari P, Slimani N, Norat T, Clavel-Chapelon F, Kesse E, Nieters A, Boeing H, Tjonneland A, Overvad K, Martinez C, Dorronsoro M, Gonzalez CA, Key TJ, Trichopoulou A, Naska A, Vineis P, Tumino R, Krogh V, Bueno-de-Mesquita HB, Peeters PHM, Berglund G, Hallmans G, Lund E, Skeie G, Kaaks R, Riboli E. Dietary fibre in food and protection against colorectal cancer in the European Prospective Investigation into Cancer and Nutrition (EPIC): an observational study. Lancet 2003; 361:1496-1501.
- The ATBC Study Group, Virtamo J, Pietinen P, Huttunen JK, Korhonen P, Malila N, Virtanen MJ, Albanes D, Taylor PR, Albert P. Incidence of cancer and mortality following alpha-tocopherol and beta-carotene supplementation: a postintervention follow-up. JAMA 2003; 290:476-485.
- Lawlor DA, Smith GD, Bruckdorfer KR, Kundu D, Ebrahim S. Those confounded vitamins: what can we learn from the differences between observational versus randomised trial evidence? Lancet 2004; 363:1724-1727.
- Diet, Nutrition and the Prevention of Chronic Diseases: Report of a joint WHO/FAO expert consultation. Geneva: World Health Organization, 2003.
- Willett, W. Nutritional Epidemiology. New York: Oxford University Press, 1998.
- Kipnis V, Subar AF, Midthune D, Freedman LS, Ballard-Barbash R, Troiano RP, Bingham S, Schoeller DA, Schatzkin A, Carroll RJ. Structure of dietary measurement error: results of the OPEN biomarker study. Am J Epidemiol 2003; 158:14-21.
- Schatzkin A and Kipnis V. Could exposure assessment problems give us wrong answers to nutrition and cancer questions? J Natl Cancer Inst 2004; 96:1564-1565.
- Bingham SA, Luben R, Welch A, Wareham N, Khaw KT, Day N. Are imprecise methods obscuring a relation between fat and breast cancer? Lancet 2003; 362:212-214.
- Sinha R, Knize MG, Salmon CP, Brown ED, Rhodes D, Felton JS, Levander OA, Rothman N. Heterocyclic amine content of pork products cooked by different methods and to varying degrees of doneness. Food Chem Toxicol 1998; 36:289-297.
- Smith GD and Ebrahim S. Mendelian randomization: prospects, potentials, and limitations. International Journal of Epidemiology 2004; 33:30-42.
- Wacholder S, Chanock S, Garcia-Closas M, El Ghormli L, Rothman N. Assessing the probability that a positive report is false: an approach for molecular epidemiology studies. J Natl Cancer Inst 2004; 96:434-442.
