Title: Epidemiology and Animal Models
Key words: animal studies, population studies, case-control studies, cohort studies, carcinogenesis,
Date: June 1999
Category: 14. Measurement
Type: Article
Author: Dr van Rhijn
Introduction
Sources and quality of research data must be sound before issuing any recommendations regarding a preventative, balanced diet with the lowest cancer risk to the public. As most of the current data is obtained from animal models and epidemiological studies, their strength and weaknesses will be discussed below1.
Animal Models
Animal models are designed to mimic human cancer states in order to allow experiments. This, of course, is unethical and impractical for human research. Budget and time constraints have led researchers to place rodents in artificially induced cancer states (model sporadic cancers), or to breed special strains with a high cancer risk (model dominant genetic predisposition). Adhering to strict, unnatural diets for rats enlarges the artificial equation. They may be helpful in explaining specific mechanisms of interaction between a carcinogen and target cells, but cannot be used to extrapolate on the nature of the relationship between diet and cancer in an uncontrolled, multivariable human population.
Epidemiological Studies
Epidemiological research is concerned with the frequency of cancer in the general population related to the intake of putative risk factors. The utility and comparability depends upon an acceptance of strict diagnostic criteria, but can provide useful clinical information2 about prevalence & relapse rates, possible pathogenic factors, the course, natural history and potential prophylactic measures of cancer development. Three different study designs have been devised:
These descriptive studies predominantly examine geographic relationships of indices of dietary intake, nutritional status and health3. Comparing groups with widely different cancer risks (mortality/morbidity) to a dietary factor can provide strong correlations but other uncontrolled variables4 in the diverse population groups can make interpretation difficult.
Past history of exposure to suspect dietary risk factors from cancer cases are compared against carefully selected healthy controls from the general population5. These studies are relatively quick and require fewer controls but are prone to selection, response, interview and recall bias6.
Retrospective diet-recall methods only reflect recent profiles, are notoriously unreliable7,8, therefore weakening the correlation between dietary intake and cancer risk. Carcinogenesis is a prolonged, multi-staged process and numerous confounding factors, including dietary changes, are not addressed in such studies.
Healthy individuals are selected on the basis that they may be exposed to a specific dietary factor9 and a longitudinal study10 identifies the characteristics of those who develop cancer and those who don't. Several outcomes from exposure to the same variable11 (baseline diet) can be studied, which is less prone to bias and observer error. It allows more rigorous testing of aetiological hypotheses, but is costly and requires large numbers (> 100), followed up for years (10) to yield each case. Selecting a high-risk subgroup, with a shorter follow up period may cut costs, but makes the sample less representative and only considers promotional stages of carcinogenesis.
Conclusion
Results from animal models explain mechanisms, but cannot be extrapolated to cause and effect. Any associated relationship between food and cancer has to be confirmed by human epidemiology, which provide weak data due to the limitations of measuring a variable dietary intake over time. Many cohort studies under different conditions are required to yield reliable data.
References