• There are now sufficient studies of the relationship between red meat consumption and cancer of the colon and rectum, using the preferred cohort design, to make meta-analysis possible.
• Few of these cohort studies have reported individually significant associations of high red meat consumption with increased colorectal cancer, but summary estimates of risks obtained from three meta-analytic studies have all found a modest, but significant elevation in risk among the highest consumers of red, or red + processed, meat.
• Two of three analyses of data pooled across cohort studies have found no association between red meat consumption and colorectal cancer; a third reported a relative risk of 1.22 per 100 g of red and processed meat per day.
• While the combined relative risk estimates are modest (ranging from 1.0 to 1.3 for highest vs lowest consumption categories, or per 100 g of meat/day), the very high proportion of omnivores in the population means that even a modest association, if proven to be causal, could have a significant impact on public health.
• In the absence of randomised controlled trials, it is important to establish credible mechanisms by which red meat might cause colorectal cancer.
• Polyaromatic hydrocarbons (PAHs) and heterocyclic amines (HCAs) are carcinogens in the food supply, but as food items not implicated in the aetiology of colorectal cancer may actually be greater contributors to the total dietary intake of these compounds than red meat, they do not provide a plausible basis for implicating red meat in carcinogenesis.
• N-nitroso compounds (NOCs) found in some processed meat products and produced endogenously in the gut, are another class of carcinogens. Although the epidemiological evidence implicating NOCs in colorectal cancer is weak, the propensity of haem iron in meat to facilitate the endogenous production of NOCs is currently under investigation.
• Despite the lack of consistent evidence that red meat is carcinogenic, and despite some obvious epidemiological inconsistencies associated with possible mechanisms currently under scrutiny, it may be prudent to minimise the use of cooking and preparation techniques that are responsible for introducing all three classes of carcinogens into meat dishes.
An early ecological study1 found strong associations between the consumption of animal protein and the incidence of several cancers, and ignited a huge research effort to confirm and refine these findings. Given the apparent strength of these correlations, it is perhaps surprising that some 30 odd years later, the picture is still both confused and confusing. The evidence for an association of high meat consumption with an elevated risk of cancer, especially colorectal cancer, is neither strong nor consistent. While most observers would agree that the association between high consumption of white meat (fish and poultry) and colorectal cancer is either null or inverse (in causal terms, ‘protective’), they would be divided on what these studies say with respect to a high consumption of red meat (beef, lamb and pork).
The reasons for this confusion are manifold. The question cannot be readily addressed using randomised controlled trials, and the observational studies published so far have varied markedly with respect to:
• study design (case–control vs cohort studies)
• the definitions of the various types of meat (despite compelling reasons to consider processed meats separately from fresh meat, several studies have failed to make that distinction)
• the instruments used to assess red meat intake
• the measurement and adjustment for potential confounding variables
• the outcomes measured (mortality vs incidence, colon vs rectum and other subsites)
• the range of meat consumption in the study population
• the number of intake groupings chosen
• the cultural and geographical context
But despite this obvious lack of comparability, three groups have applied meta-analytic techniques to those studies which have used the more robust cohort study design.2–4 This paper summarises the literature on relevant studies to address the question: Does eating red meat increase the risk of colorectal cancer?
Figure 1 summarises a meta-analysis of cohort studies undertaken by Sandhu et al.,2 who reported their findings as odds ratios of contracting colorectal cancer per 100 g of meat consumed. The only two studies that individually reportedsignificant associations with increasing red meat intake were the Health Professionals' Study5 and the Nurses' Health Study,8 but it should be noted that subsequently, Wei et al.11 revised these estimates downward (and they were no longer significantly greater than one) following more extensive adjustment for potential confounding factors (see Figure 3). Nevertheless, all the cohort studies considered worthy of inclusion by Sandhu et al. did report odds ratios greater than one, and together their weighted odds ratio estimate of 1.17 per 100 g of meat consumed per day (95% CI 1.05–1.31) was significantly greater than one.
Results of a meta-analysis of Norat et al.3 are presented in Figure 2. Although they chose to report relative risks of highest versus lowest consumption categories (and it is worth noting that some studies divide red meat consumption into quarters while others used fifths), their findings are qualitatively similar to those of Sandhu et al.2 The overall relative risk estimate for ‘highest’ versus ‘lowest’ was 1.27 (95% CI 1.11–1.45).
The most recent meta-analysis, undertaken by Larsson and Wolk,4 and including several new studies, is summarised in Figure 2. The combined estimate of relative risk (highest vs lowest consumption categories) of 1.28 (95% CI 1.15–1.42) is similar in magnitude to the estimates of the other meta-analyses.
Not included in these meta-analyses because of insufficient detail, or because of a lack of adjustment for important confounders, were at least four further cohort studies:
• Hirayama's census cohort of Japanese, in which there was a very significantly decreased risk of colorectal cancer among those who consumed meat (mainly pork) on a daily basis when compared with those who ate meat less frequently.23
• Phillips and Snowdon's cohort of white Seventh Day Adventists, in which there were non-significant changes in the risk of fatal colorectal cancer in subjects eating the most meat and poultry.24
• Two further studies25,26 in which dietary fibre was the principal focus, but for which high meat consumers had relative risks very close to 1.0.
None of these studies reported a significant increase in the risk of colorectal cancer.
Pooled analyses provide a unique opportunity to reduce heterogeneity by applying a single analytic procedure to a more standardised coding of the individual contributing studies.
Key et al.27 conducted a pooled analysis of colorectal cancer mortality data from five prospective studies of 76 000 men and women, comparing vegetarians with non-vegetarians, and concluded that there was no support whatever for claims that a vegetarian lifestyle provides any protection from fatal colorectal cancer (Figure 4).
The Pooling Project28 at Brigham and Women's Hospital and Harvard University, in Boston, USA, combined individual records from 14 prospective studies (725 258 study subjects) conducted in Europe and North America, and in the only publicly available document reporting on their findings (an abstract), two key sentences are as follows:
The pooled multivariate relative risks (RRs) of colorectal cancer were 1.00 (95% CI 0.92–1.08) for each 90 g (approximately 3 oz)/d increase of red meat and 1.05 (95% CI 0.96–1.15) for each 30 g/d increase of processed meat. . . . In conclusion, these prospective data do not support a positive association between higher red meat and fat intake and colorectal cancer risk.
Findings from 12 cohorts (478 040 study subjects) participating in the European Prospective Investigation into Cancer and Nutrition are shown in Figure 5. The authors chose to present their findings for red and processed meats combined. The estimated hazards ratio per 100 g of meat consumed per day in Figure 5 were inferred from the hazards ratios per 10 g of meat per day reported in their Figure 3. The pooled estimate was 1.22 (95% CI 0.90–1.79) per 100 g, and the hazard ratio for the highest versus the lowest consumption (five categories) of red and processed meat was 1.35 (95% CI 0.96–1.88) (Figure 5).
Conclusions from population research
About all we can currently conclude about the association between red meat and colorectal cancer is that it is relatively weak. Relative risk estimates of ≤1.3 would normally command little attention in epidemiological circles unless they were generated by strictly comparable randomised controlled trials, but ‘exposure’ to red meat is such a common event that the public health consequences of even a modest association, should it prove to be causal, would still be considerable. If the relative risk of colorectal cancer in the top 25% of red meat consumers is 1.30 compared with the lowest 25% and the dose–response is linear, the overall proportion of colorectal cancer attributable to red meat might be around 15%. In Australia, a causal association of this magnitude might amount to some 1950 new cases per year being attributable to red meat consumption. From an individual's perspective, however, reducing one's red meat consumption to below the current first quartile in Australia (57 g per day in the Melbourne Collaborative Cohort Study19) might, at most, reduce one's risk of colorectal cancer over an entire lifetime from 4.8%29 to around 4.2%.
An important step towards determining whether a relationship between an exposure and a disease is causal is the identification of plausible mechanisms. A substantial research effort is currently focusing on carcinogenic compounds and their precursors in meat dishes.
FOOD-BASED RESEARCH: CARCINOGENIC COMPOUNDS IN FOODS
That there are substances with mutagenic activity in some cooked foods has been demonstrated repeatedly in many laboratories.30 There are at least three classes of such compounds that may be found in some cooked meats—but importantly it is now very clear that these compounds are introduced by some of the methods used for preserving and cooking—and that they are not present in detectable levels in fresh uncooked meat.
The three classes are:
There is now little support for the notion that the protein and fat in meat are promoters of carcinogenesis.
These compounds are generated by the reaction of the muscle compounds creatine or creatinine, with amino acids, naturally occurring sugars in the meat, during high-temperature cooking. At least seven such HCAs have been identified and measured in cooked foods to date. High-temperature cooking methods such as grilling (sometimes referred to as ‘broiling’ in the US literature), pan-frying and barbecuing are consistently associated with higher levels of HCA in the final product as consumed. Microwave heating, boiling and casseroling do not generate the temperatures required to produce HCAs.31 While roasting temperatures may be higher, Thomson and Lake were unable to detect HCAs in roasted beef, chicken or hogget (young sheep) in New Zealand.32
Maximal generation of HCAs appears to occur only over a specific range of sugar concentrations (glucose and glucose-6-phosphate)—and at both low levels, such as in meat from animals whose muscle glycogen has been exhausted—and at high levels created by the use of sweet marinades or the addition of starch, the synthesis of HCAs is inhibited.33,34
Layton et al. have shown that beef steak, lamb, chicken and fish contain quite comparable amounts of HCAs when cooked by similar methods.35 As there is general agreement that high intakes of fish and poultry are negatively associated with the risk of colorectal cancer, the HCA hypothesis is not consistent or ‘coherent with existing knowledge and data’—one of Hill's criteria for assessing causality from observational data.36
Polycyclic aromatic hydrocarbons
Polycyclic aromatic hydrocarbons (PAHs) are ubiquitous compounds arising from the incomplete combustion of organic material. PAHs and/or their metabolites can intercalate with DNA to form DNA adducts. They are therefore carcinogens or procarcinogens. They can also induce the synthesis of specific enzyme catalysts,37 thereby significantly altering ‘metabolic flow’ along competing chemical pathways, and influencing the ratio of harmful to harmless metabolites. PAHs may also interfere with the lymphocyte-programmed cell death/apoptosis machinery (e.g. induction of pre-B cell apoptosis).38
In order to assess claims that PAHs in meat may be responsible for colorectal cancer, we need to appreciate that PAHs are widely distributed throughout the food supply. When foods are cooked, dried or flavoured by processes which expose those foods to combustion products, or when they are cooked at temperatures sufficiently high such that the foods themselves are charred or burned, then much higher levels of PAHs may be found in the final product as consumed. The cooking of meat above open fires, including charcoal fires and on barbecues where there is not a metal plate separating the flames from the meat, is undeniably problematic in this regard, with molten fat dripping from the meat and igniting in the fire below, adding further to the PAH load.39,40
Dennis et al. found that in the UK, meats contributed only around 4% of the total dietary exposure to PAH, with around 35% coming from cereal foods, and a further 34% from fats and oils.41 Subsequently, the Opinion of the Scientific Committee on Food (European Commission, Health And Consumer Protection Directorate-General) noted that in the UK and Holland, cereal foods and oils and fats accounted for around 60% of all dietary PAH, vegetables and fruit together accounted for a further 10–20% (with nuts adding a further 14% in Holland), while meat and fish accounted for only 4–6%.42 In Sweden, cereals were found to be the highest contributors of PAH (about 34%), followed by vegetables (about 18%), and oils and fats (about 16%).43,44 Significant intakes were also found for a ‘fruit and sugar’ group and for smoked meat products. In good agreement with the European studies, Kazerouni et al. found that around 29% of the intake of a marker PAH, benzo[a]pyrene (BαP), in 228 US subjects living in Washington, DC, came from cereals, breads and grain products, but they estimated a higher figure of around 21% coming from barbecued meat.45
These data point reasonably consistently to meat and meat products being only modest contributors to population exposures to PAHs and, therefore, that the hypothesis that reduction of red meat intake may reduce exposure to PAHs may be far too simplistic. It might also be noted that the additional intake of the PAH benzo[a]pyrene for a person smoking 20 cigarettes/day is estimated to be around 210 ng, which is of the same order of magnitude as the mean intake by ingestion of food.46 Cigarette smoking is not a recognised risk factor for colorectal cancer, although it must be remembered that exposure of the lower gut to PAHs from smoking would be substantially less than that of the lung.
Nitrates, nitrites and NOCs
N-nitroso compounds are capable of forming DNA adducts and hence are carcinogenic.47,48 NOCs are present in the food supply,48–50 especially in preserved meats, and they are also synthesised in the colon by the reaction of nitrite (generated under anaerobic conditions from nitrates) with amines and amides produced by bacterial decarboxylation of amino acids. In turn, nitrate and nitrite may be present as such in the food supply (typically in vegetables!), or they may be produced endogenously by macrophages, which use dietary arginine to generate nitric oxide to kill target cells, and by NO synthase (endothelium-derived relaxing factor). Protein, peptide fragments and amino acids are nitrogen-rich substrates arriving in the colon. Bacterial deamination of proteins generates ammonia and short-chain fatty acids. Depending on the availability of other fermentation products, the ammonia may be rendered harmless by incorporation into glutamic acid and then other amino acids. Free ammonia is undesirable, given that it can promote the carcinogenic effects of NOC in rodent models.51 The most studied NOC is N-nitrosodimethylamine (NDMA), and it is sometimes found in cured meats and fish.
The potential role of processed meat consumption in colorectal cancer prompted Knekt et al. to look for an association of nitrates, nitrites and nitrosamines in a Finnish cohort of 9985 individuals enrolled during 1966–1972 and followed for up to 24 years.52 They estimated that 51.9% of dietary NDMA was provided in smoked and salted fish, and 48.1% came from cured meats and sausages. The 25% with the highest estimated NDMA exposure were at 2.12 times (95% CI 1.04–4.33) the risk of the quarter with the lowest intake, but there was no evidence of a linear trend (P = 0.47), and no significant associations were found for dietary nitrate and nitrite intakes.
Given that processed meat may be a major source of NOCs in some communities, it is regrettable that some of the studies discussed in this review have not distinguished adequately between cooked fresh meat and processed meat, which may be raw (fermented) or cooked. Although the role of nitrates, nitrites and nitroso compounds in the aetiology of colorectal cancer remains equivocal, a clear distinction between fresh and processed meats would be highly desirable for developing appropriate public health recommendations with respect to the role of meat in healthy diets, and (if necessary) for regulating the types and amounts of preservatives used in manufactured meat products.
A research group in Cambridge, UK, has published experimental/human metabolic data suggesting that haem iron may catalyse the formation of NOCs from natural precursors in the gut.53–55 Although haem is present in both red and white meats, red meats owe their colour to the fact that they are a richer source of haem iron (around twice as much as white meats), so such an effect might theoretically explain why associations of colorectal cancer are stronger for red meat than for white meat. It would not, however, explain why high poultry and fish consumption is often observed to be negatively associated with (i.e. apparently protective against) colorectal cancer, and further data on whether this mechanism is relevant to human health will be eagerly awaited.
The evidence that eating red meat increases the risk of colorectal cancer remains weak and inconsistent. Although it cannot be claimed with any certainty that the dosage levels of carcinogens in meat as consumed are dangerous, it may be prudent for people to minimise their exposure to NOCs, HCAs and PAHs. Limiting consumption of processed meats, and limiting high-temperature cooking methods and charring of meat, are practical means for lowering these exposures.