Filtering Scientific Noise: Colorectal Cancer and Red/Processed Meat

While the 2015 Dietary Guidelines for Americans provided a pass on red and processed meat, two recent articles advanced the position that consuming 100 g or 50 g a day of these foods, respectively, increased the relative risk of developing colorectal cancer by 17%–18% (Chan et al. 2011; Bouvard et al. 2015). These findings were based on dose-response meta-analyses evaluated in an International Agency for Research on Cancer (IARC) document yet to be published.

Selecting for attributes without an understanding of the importance or role of those parameters in the outcome clouds the credibility of the conclusion. It appears that the meta-analyses were weighted without regard to biologic plausibility. It is interesting to note that upon a review of the forest plots by Chan et al. (2011), out of the 11 studies germane to red and processed meat and colorectal cancer, only three studies presented a significant association, and only one of eight studies that focused on red meat suggested a significant relationship. In addition, the generated dose-response curves with a 95% confidence interval were curvilinear. These curves suggested that daily consumption of red and processed meat above 140 g actually decreases the relative risk of developing colorectal cancer. Total daily consumption of red meat in 2014 was about 130 g per capita (USDA 2015).

It is important to remember that although IARC evaluated results from the meta-analyses and calculated relative risk, the agency’s findings are limited as to whether an agent is capable of causing cancer (technically defined as “hazard”). Hazard does not measure the likelihood that cancer will occur (defined as “risk”) as a result of exposure to the agent. The distinction between hazard and risk is critical (Clemens et al. 2015). An agent is considered a cancer hazard if it is capable of causing cancer under some circumstances. Risk measures the probability that cancer will occur, taking into account factors including genetic predisposition, other environmental factors, and the level of exposure. Cancer hazards are identified even when risks are very low with known patterns of use or exposure.

In an effort to identify the hazard that red meat and processed meat might pose in relation to colon and rectal cancers, IARC identified a database of 800 statistical studies and screened them for methodologic consistency and focus. Limited effort was made to assess cell biology or oncologic work. Of the 800 studies originally screened according to inclusion criteria, only 47, or about 6%, were deemed eligible or informative in looking at the putative association of red meat and the dependent variable, colorectal cancer. These 47 studies were further categorized and differentially weighted by cohort, case control, and processed versus red meat independent variables. Finally, the presence or absence of a statistical association between the variables was determined. The results across all study categories yielded a positive association in 26 studies and no association in 21 studies. On the basis of an arguably slight edge for the positive association, biological explanations were then marshaled to explain what the investigators implied was a causal inference (Chan et al. 2011).

The National Cancer Institute (NCI) of the National Institutes of Health defines cancer as a term that refers to a very large spectrum of disorders consisting of hundreds of different diseases that occur in virtually every organ system and at every stage of life. This spectrum of diverse pathology shares only a small number of very broad but frightening characteristics, such as abnormal cell growth with the capacity for invasion and metastasis.

With the advent of molecular biology, a better understanding of the underlying mechanisms of cancer in general and specific genetic hits is possible through seemingly small events and the windows between those events (Knudson 2001). Despite the tremendous advances in cancer therapy in reducing the risks associated with cancer development, oncologists and scientists now acknowledge the startling revelation that cancer etiology is staggeringly more complex than was imagined, even by Judah Folkman, the discoverer of angiogenesis inhibitors. In fact, each lesion subtype and site is the result of an exquisitely complex interaction of host and environmental factors. The prevalence and steadily increasing incidence of cancers globally despite modern science and modern medicine attests to the challenge that is presented by this group of loosely related diseases.

A review of the incidence and prevalence data provided by NCI and the World Health Organization is quite sobering and includes the following points.

• Cancer is among the leading causes of death worldwide. In 2012, there were 14 million new cases and 8.2 million cancer-related deaths worldwide.

• The number of new cancer cases is expected to reach 22 million within the next two decades. More than 60% of the world’s new cancer cases occur in Africa, Asia, and Central and South America; 70% of the world’s cancer deaths also occur in these economically depressed regions where red meat consumption is lowest and cigarette smoking is high.

• Cancer-causing viral infections, such as Hepatitis B virus, Hepatitis C virus, and Human Papilloma virus, are responsible for up to 20% of cancer deaths in low- and middle-income countries (de Martel et al. 2012).

--- PAGE BREAK ---

What clearly emerges is that nothing about cancer is simple or easy, especially its etiology and epidemiology. The number of flawed epidemiologic studies that suggests this is, in fact, the case serves to highlight a global sociopolitical frustration and even desperation about the failure to effectively meet the challenge this group of diseases continues to mount.

Let us consider the most current arguments that have arisen from groups such as IARC. It has been proposed that heme iron, heterocyclic aromatic amines, and N-nitroso compounds comprise potentially significant dietary sources of carcinogenicity. However, none of these compounds is unique to red or processed meat. The IARC Working Group itself concluded that in animal models, there is inadequate evidence for the hypothesis.

The Nurse’s Health Study, the Health Professional Follow-up Study, and the Multiethnic Cohort Study all report nonsignificant or inverse associations with red meat and colorectal cancer (Ollberding et al. 2012; Bernstein et al. 2015). Ollberding et al. (2012) concluded that for two large groups exceeding 165,000 and 130,000 subjects, respectively, there was no association between risk of colorectal cancer and density-adjusted total meat, red meat, or processed meat intake or for total heterocyclic amines intake whether comparing quintiles of dietary exposure or using continuous variables. Bernstein et al. (2015) validated these results in looking at two independent cohorts of health professionals (of >87,000 and >47,000 subjects) and found little evidence that higher intake of unprocessed red meat substantially increased the risk of colorectal cancer.

Among heavy smokers (a significant confound), high red meat consumption and low adherence to a Mediterranean diet are associated with increased risk of lung cancer (Gnagnarella et al. 2013). Melina et al. (2013) reported a higher burden of stomach cancer in indigenous populations globally and a rising incidence in some indigenous groups in stark contrast to the decreasing global trends. Policies that address initiatives such as improving nutrition and sanitation and programs such as Helicobacter pylori eradication, a potentially significant confounder, have the potential to reduce inequalities in stomach cancer rates. Interestingly, the Mediterranean-style diet includes red meat and processed meat, yet the preponderance of dietary guidances suggest dietary patterns consistent with this style reduce the risk of several noncommunicable diseases, including some cancers (Verberne et al. 2010; Trichopoulou et al. 2010).

The foregoing is not meant to suggest that any possible association of meat and cancer is to be ignored; on the contrary, statistically significant associations must be explored to determine if causality can be proven. Experimental work must minimize the layers of threats to validity in the context of human nutrition. Laboratory work must account for the limitations of in vitro models of disease and the typically acute and exaggerated exposure of cell lines to putative procarcinogens dosed independently of the food matrix. Investigations using animal models must note genomic differences between those models and humans. Additionally, studies need to control for variabilities in meat processing, preparation, cuts, sources, livestock feeds, and genetics. Finally, in following human subjects, a very high level of resolution is required in order to rigorously identify and quantify such factors as biometrics, lifestyle variables, and genetic and medical histories.

It is important to remember that without identification of the causal link, any series of epidemiologic investigations where effects are shown some of the time and not shown other times indicate that conflicting and inconsistent results are due to as yet unclassified confounders in the methodology. Observational epidemiology is simply hypothesis generating. Prospective studies allow investigators to define the importance of exposure to the outcome, but unlike randomized controlled trials, these kinds of studies cannot be used to establish causality.

Scientists and investigators must exercise vigilance in not developing a policy path amid deceptively clean mathematical associations of variables that may, in fact, be related, but in a nonlinear and noncausal fashion.

Roger Clemens, DrPH, CFS, Contributing Editor
Adjunct Professor, Univ. of Southern California School of Pharmacy, Los Angeles, Calif.
[email protected]

A. Wallace Hayes, PhD, is visiting scientist
Harvard University, Boston
([email protected]).

Claire Kruger, PhD, is President,
Spherix Consulting, Rockville, Md.
([email protected]).