Diet and Cancer: Title and subTitle BreakAn Evolving Picture

In a 1981 landmark report, Doll and Peto¹ estimated that 35% of US cancer deaths were attributable to dietary factors. This estimate was primarily based on the large differences in rates of specific cancers among countries and observations that these rates were strongly correlated with aspects of national food supplies. However, these authors acknowledged major uncertainties in responsible aspects of diet and the magnitude of impact.

Since the early 1980s many detailed investigations including mechanistic studies, animal experiments, epidemiological observations, and clinical trials have addressed the potential effects of diet on cancer incidence. Although much has been learned, progress has been slower and more difficult than was anticipated. At the beginning of this period, high total fat consumption was widely believed to be the primary reason for the high rates of breast, colon, prostate, and several other cancers.² National dietary recommendations and policy were largely driven by this presumed relation. However, little relation between total fat intake and risk of breast cancer and risk of colon cancer has been found in large prospective studies.^{3 - 4} Although a weak effect is impossible to exclude, evidence is strong that simply reducing the percentage of energy intake from fat during midlife will not have a major impact on breast and colon cancer.⁵ Prospective studies of prostate cancer are far fewer, but no clear role for dietary fat per se has been seen. The Women’s Health Initiative trial, primarily justified to test the hypothesis that reducing total fat would decrease breast cancer incidence, will be concluding soon, but even if a modest effect is seen, this would be difficult to interpret because many aspects of diet have been changed simultaneously.⁶

As evidence to support the dietary fat hypothesis has waned, enthusiasm increased during the 1990s for increasing fruit and vegetable consumption to prevent cancer; the national Five-a Day program was launched on this basis. Although inverse associations between intakes of fruits or vegetables and incidence of various cancers were reported in numerous case-control studies,⁷ the findings from more recent large prospective studies have been far less supportive of a benefit. In a recent report combining 2 large cohorts, no relation was observed between total fruit and vegetable consumption and overall cancer incidence.⁸ Randomized trials using high doses of single constituents of fruits and vegetables, and beta carotene in particular, also failed to show benefits and some even suggested harm.⁹

One lesson from this experience has been that case-control studies of diet, in which patients with cancer and a control group are asked about their diet years in the past, can be misleading. Recall of diet might be biased by the diagnosis of cancer, but in typical case-control studies selection bias may be even more problematic. Participation rates of patients with cancer are usually high, but participation of population controls is often only 50% or 60%.¹⁰ Those who participate are likely to be more health conscious and therefore consume more fruits and vegetables and less fat than those who do not. When cases and controls are compared, this would lead to apparent inverse associations with fruits and vegetables and positive associations with fat.

The 2 reports in this issue of JAMA from large and well-conducted prospective studies provide valuable increments to current knowledge about diet and cancer.^{11 - 12} The lack of association between fruit and vegetable intake and incidence of breast cancer seen in the European Prospective Investigation Into Cancer and Nutrition (EPIC) study¹¹ is consistent with the pooled data from other large prospective studies that included 351 825 women and 7377 cases of breast cancer.¹³ The authors provide the caveats that necessarily attend any negative study: a small benefit can never be excluded and a modest benefit could exist for a subgroup of women (perhaps defined by genetic factors) or a subset of cases (for example, defined by estrogen receptor status). Measurement of diet will always be imperfect, which will tend to attenuate a true association. However, assessments of fruits and vegetables are informative because they predict blood levels of micronutrients provided by these foods and correlate reasonably well with intakes measured by more detailed approaches.¹⁴

The most important limitation of these studies is the lack of data on diet during childhood. Studies on cancer risk among survivors of the American bombing of Japan suggest that radiation exposure during childhood most strongly increases risk of breast cancer and little effect is seen for exposure after age 40 years.¹⁵ Thus, if constituents of fruits and vegetables are acting to protect DNA from damage, published studies could have entirely missed the critical periods. Despite the caveats, the current body of evidence clearly indicates that increasing fruit and vegetable consumption during midlife will not have a major effect on overall incidence of breast cancer. Such results are valuable because they prevent a false sense of security and stimulate the exploration of alternative means for prevention; for breast cancer these alternatives will probably need to include pharmacological methods.

The positive findings of Chao et al¹² relating intake of red meat to risk of colon cancer contrast with the results for breast cancer. Among the international correlations between dietary factors and various cancers, the relation between meat consumption and colon cancer has been the strongest. There is no shortage of plausible mechanisms; heterocyclic amines formed in cooking, nitroso compounds in processed meats, and heme iron have been among the proposed agents,¹¹ and these could act in combination. Positive associations with intake of red meat have been seen consistently in case-control studies, and a similar relationship was reported earlier in the prospective Nurses’ Health Study.¹⁶ In subsequent cohort studies, this relationship has been observed only inconsistently, but in a recent meta-analysis of cohort studies an overall positive association was seen; this was strongest for processed meats.¹⁷ The report by Chao et al¹² adds further support for this relationship and emphasizes the value of long follow-up and multiple measurements of diet.

The rapid changes in colon cancer risk among migrants from low to high incidence regions (or vice versa) indicate an important role of exposures during adult life in colon carcinogenesis. However, the relevant period between the assessment of diet and diagnosis could still range from a few years to several decades; for smoking this appears to be roughly 40 years.¹⁸ A single assessment of diet would not be problematic if individuals’ diets remained constant over time, but intake of red meat has fluctuated greatly during the last 25 years in the United States.¹⁹ Thus, only one measure of intake could seriously misrepresent an individual’s intake during the critical period. Added to the differences among studies in the population characteristics, the amounts of red meat consumed, and methods of cooking, some inconsistencies in findings are not surprising. Although the overall data for red meat and colon cancer are strongly suggestive of an important relation, they are not conclusive. Further studies with long follow-up, repeated measures of diet, genetic markers of susceptibility, more detailed measures of cooking methods, and molecular characterization of colon cancer cases may be helpful.

Although recent findings on fruit and vegetable consumption and cancer may be disappointing, reductions in blood pressure and epidemiological evidence for lower risks of cardiovascular disease provide sufficient reason to consume these foods in abundance. The relation between red meat consumption and colorectal cancer may not be conclusive, but prudence would suggest that red meat, and processed meats in particular, should be eaten sparingly to minimize risk. When combined with other healthful diet and lifestyle factors, it appears that approximately 70% of colon cancer can potentially be avoided.²⁰ Replacing red meat with a combination of fish, nuts, poultry, and legumes will also reduce risk of coronary heart disease, in part, because some of these foods have positive benefits.²¹ This substitution is an important part of the Mediterranean dietary pattern, which improves blood lipids and other metabolic parameters²² and has been related to lower rates of total mortality.²³ Thus, keeping red meat consumption low is best viewed, not as an isolated goal, but as part of an overall dietary and lifestyle strategy to optimize health and well-being. Fortunately, substituting pistachio-encrusted salmon and gingered brown basmati pilaf for roast beef with mashed potatoes and gravy is not a culinary sacrifice.