EPI242: Cancer Epidemiology
2009
Nutritional Epidemiology Overview
Role of nutritional epidemiology is to:
- provide the best possible scientific evidence to
support an understanding of the role of nutrition
in the causes and prevention of ill health
- ensure that the information upon which public
health decisions are made is of the highest
quality
Focus of the study may be etiological, prevention,
intervention
Following is the fra mework provided by Willett8 for classifying the vast number of constituents of
foods consumed by humans. The categories are not necessarily exclusive.









Energy sources. Energy sources, including proteins, carbohydrates, fats, and alcohol, are
transformed into energy to sustain life .
Nutrients. Nutrients, including vitamins, minerals, lipids, and amino acids, are required for good
health.
Other natural compounds. Countless enzymes and enzyme inhibitors, specialized lipids, genetic
material, and other substances are required to sustain animal and plant cell structure and function.
Cholesterol, for example, is an important structural component of animal cell membranes.
Natural toxins. Although most people consider pesticides to be chemical food contaminants, some
pesticides are produced naturally by plants as a defen se
mechanism. The toxins that make some plants poisonous to humans provide an obvious example of
natural toxins.
Microbial contamination. Foods are easily contaminated, especially during processing or storage, by
fungi. Aflatoxins such as aspergillus have received much attention because of their etiologic link to
liver cancer.
Food additives. Substances are added to foods to preserve them or to improve their color, taste, or
consistency. These chemicals attract a great deal of public attention and concern, yet they are among
the best characterized and most intensively regulated.
Agricultural contaminants. T hese substances include fungicides, herbicides, pesticides, and growth
hormones.
Inorganic contaminants. Inorganic substances, including metals, such as lead and cadmium, and
synthetic compounds, such as polychlorinated biphenyls, can contaminate our food supply.
Chemicals formed by cooking. Numerous substances are created by cooking foods. Some specific
agents created by cooking or charring meats have been hypothesized to cause cancer of the
gastrointestinal system.
History
 Relatively new as a formal area of research
 Long history of understanding of how foods
relate to diseases



1600; Beri-beri
1753; Scurvy
1980; Keshan disease
Consideration of nutrients in balance
1.
Excess- western/affluent societies; long latent period;
multiple causes; penetrance variable; unknown critical
period; may not be reversible
2.
Deficiency- anemia, wasting, stunting; short latent
period; single/few causes; high penetrance
3.
Insufficiency- may have characteristics of either excess
or deficiency; osteoporosis
4.
Unknown- interactions among nutrients and other
factors
Complexities in studying nutrients and
disease
Relationship between food and health is affected by acute
or chronic illnesses; access to health care; lifestyle
Therefore, nutritional data needs to be collected with
purpose in mind and clear hypotheses to enable reliable
estimate of the relationship of interest
- clear and testable research questions
- consideration of the effects of other factors
- exposures and outcomes of interest are measured
with sufficient precision
The need for practical methods
to measure diet
Methods need to be:
Accurate
Relatively inexpensive
Have high compliance
Debate
Are useful measurements of the diet of individual
subjects within free-living populations possible to
collect?
Are diets of persons living within one country too
homogeneous to detect relationship with disease?
Development of nutritional epidemiology

Development and evaluation of methods for dietary
assessment in epidemiological applications is growing
 Many aspects of diet can be measured readily with
sufficient accuracy to provide useful information
 Finding that meaningful between person variations within
populations do exist
 Identification of ways to deal with errors/limitations
1. Identify the principle sources of error
2. Explore the impact of these errors on the results
3. Design studies which prevent or control these errors
Gains in the field









Defining patterns of food consumption vs.. individual foods
Discovery that the ways foods are stored and prepared influence
impact
Study of whole foods in addition to specific nutrients
Causal pathway of nutrients
Development of measurement instruments
Knowing the sources, characteristics and effects of errors in the
measurement of dietary exposures
Use of biomarkers to identify unreliable dietary intake
reporting
Need for improvement:
Determining the relevance of biomarkers or proxies for dietary
intake
Identification of susceptible subgroups
Study Designs: Correlation or Ecologic

The exposure variable is the population per capita
consumption of specific dietary factors
 Relies on disappearance data (food produced and
imported minus food exported and fed to animals)
 Some correlations strong (0.85 for meat intake and
colon cancer incidence)
 Strengths: contrasts in dietary intake are typically
large (fat intake between US and nonwestern nations);
diets at a regional level are stable over time; values are
derived from large populations and are subject to only
small random errors
Study Designs: Correlation or Ecologic

Weaknesses: highly correlated with other factors (GNP);
disappearance data may lead to erroneous conclusions
(alcohol consumption and breast cancer- who is drinking
the alcohol in the population?); cannot be independently
reproduced

Solutions: collect information on actual dietary intake in
a uniform manner from the population subgroups of
interest; China has 65 geographical areas with unusually
large variation in rates of many cancers; select men vs.
women; etc.
Study Designs: Correlation or Ecologic
Special Exposure Groups
 Seventh-day Adventists (SDA): vegetarian, have
half the expected colon cancer rates of non-SDA
in the same geographical region
 Subject to same strengths/weaknesses as
ecological studies, but useful in refuting
hypothesis; i.e., SDA abstain from alcohol
consumption yet do not have significantly
different rates of breast cancer mortality than
non-SDA
Study Designs: Correlation or Ecologic
Migrant Studies and Secular Trends
 Migrant studies show influence of environmental
exposures in contrast to genetic factors
 Secular trends likewise may reveal non-genetic
factors that are influential
Study Designs: Case-Control

Control for confounding by design (matching, restriction)
or in the analysis (multivariate methods) if confounding
variables are known
 However, for dietary studies, bias may seriously effect
outcome
 Due to the limited range of variation in diet, the relative
risks are modest (0.5-2) and even a small error in
measuring intake (3-4%), which is difficult to avoid (recall
bias), can distort relative relationship
 Inconsistent findings between studies would be predicted
and are indeed found in the literature
Study Designs: Case-Control
 Selection of controls with medical conditions
may potentially have diet related conditions
 Inherent biological complexity
 Nutrient-nutrient interactions may cause
inconsistent results, but are nonetheless accurate
 Even consistent findings may be wrong; i.e.,
positive association between total energy intake
and risk of colon cancer, that in prospective
studies found either non or inverse associations
Study Designs: Cohort
 Strength: less susceptible to bias
 Weakness: may have low frequency of outcome,
more expensive, follow-ups
Some populations are followed-up (standard of
care for cancer patients) AND are inherently very
interested in diet, so can find a committed
population.
Study Designs: Controlled Trials

Undertaken only when there is a body of nonexperimental data that justify benefit to humans

High probability of finding lack of effect due to diets
exerting effects after long period of time; duration of
trial must be long and compliance must be high

Health conscious people tend to enroll, those at
higher risk may not be included in study

Imprecise measure of effect because if benefit is
observed, ethical considerations require stopping the
trial
Measures of Exposure
Types of measures
 Dietary habits

Food patterns, meals, foods, nutrients

Individuals, groups, populations
 Anthropometry
 Biological
 Knowledge/attitudes about food
Expression of measures
Continuous

Total cumulative dose (e.g. lifetime)

Average dose (e.g. average per day)

Percent of standard
Discrete

Eat a particular food item

Rank (quartiles)

Percent above/below standard
Measures of Outcome
Types of measures
 Dietary habits
 Anthropometry
 Biological
 Physiological
 Disease/health status


Morbidity
Mortality
Expression of measures
 Continuous
 Categorized into discrete (e.g. cut-points: BP, obese,
syndromes)
Measures for the relationship between
exposure and outcome







Average level of group outcome in discrete categories of
exposure (e.g. BP levels by thirds of fat intake)
Regression (change in outcome per unit change in exposure)
Standardized morbidity or mortality ratios
Absolute risk or absolute risk reduction
Relative risk (ratio of the rates of appearance of outcomes in
different categories of exposure)
Attributable risk percent (proportion of cases in the exposed
population that can be attributed to the risk factor)
Number needed to treat (number of subjects in whom the
exposure needs to be altered in order to save one person from
getting the outcome)
Interpretation of null results







Variation in diet is insufficient
Variation may exist but only within the no effect range
of the dose response curve
Method of measuring dietary intake is not sufficiently
precise to measure differences that truly exist
Low statistical power
Temporal relationship between the measured
exposure the occurrence of the disease did not
encompass the true latency period
Unmeasured third variable was related to exposure
and disease in opposite directions; negative
confounding
Methodological sources of bias
Issues of nutritional epidemiology
1.
2.
3.
4.
5.
Measurements of diet lack precision and specificity
Nutrient intakes are highly correlated; attribution of
causation to one nutrient considered to be acting on its
own may be misleading
Biological measurements of nutrients in tissues may not
accurately or reliably reflect dietary intake; biological
regulation of these measurements is complex, may be
influenced by other other nutrients, homeostasis
Lack of consideration of the effects of the way food is
prepared (whole, processed, extracts)
Everyone is exposed to causal/preventive factors,
i.e. everyone eats fats, fiber, vitamins over a lifetime
Issues in study design and interpretation of
results
 If null finding is observed, you should define the
conditions and limitations of the null finding
 Demonstrate that true variation exists in the diet
and that method of measuring diet is adequate
for detecting variation
 No study instrument is without some potential
source of error; use valid instrument of study
 Report confidence intervals
Issues in study design and interpretation of
results (cont.)
 Range of latency period should be discussed.
Case-control study of cancer describe over what
time period dietary intake was reported;
Prospective study may allow analysis over
different times
 Describe the dietary and non-dietary correlates of
the primary exposure what has been evaluated as
potential confounding variables
Types of Measurements
 Twenty-four-hour Recall is an attempt to define
and quantify food intake during a specific day.
 Dietary Records are detailed descriptions of types
and amounts of foods and beverages consumed,
meal by meal, over a prescribed period (3-7 days).
 Food Frequency Methods collect long-term diet
over months or years, not just a few days.
Different levels of information
concerning diet
Food Item/Food group level
red meat, fish, butter, vegetable oil, cruciferous
vegetables, coffee, beer
salted, smoked, pickled, fried, cured foods
Nutrient Level
energy, animal protein, fat, vitamin c, selenium
Other Factors
nitrates/nitrites, food additives, contaminants,
mutagens and carcinogens, antioxidants
Measurement Techniques of Food
Consumption
Food Recording Techniques
 Weighing method: (1) precise weighing (weighing
both before and after cooking); (2) weighed
inventory (weighing only after cooking)
 Recording in household measures
Interview methods
 Recall Methods (usually 24 H)
 Dietary history and food frequency methods
Food Frequency Questionnaire
based on Dietary History
 Measures the habitual diet of the subject at least
qualitatively and half-quantitatively
 Frequency and portion size of use 100-200 food
items are asked as well as the usual eating pattern.
 Refers to the previous month or year
 Seasonal difference are asked
 60-90 minutes interview
 The best method of estimating the habitual diet of
an individual over a long time period
 Idea in diet and cancer studies
New Challenge for Nutritional Epidemiology:
Cancer and Chronic Disease
 Occur with a low frequency with multiple causes
acting alone or in combination (nutrition,
smoking, drinking, occupational exposure,
physical activity, etc.)
 Occur only among exposed but also among
unexposed. (everyone eats fat, fiber, vitamins).
New Challenge for Nutritional Epidemiology:
Cancer and Chronic Disease
 Have a longer latent time (20-40 years for cancer) or
unknown.
 Not readily reversible, may result from excessive or
insufficient intake of dietary factors, which makes the
intervention difficult
Percentage of Cancer Deaths Attributed to Various
Factors (Doll R and Peto R, JNCI, 1981)
 Tobacco Use: 30% (25%-40%)
 Diet: 35% (10%-70%)
 Infection: 10%? (1%-?)
 Reproductive & Sexual Behavior: 7% (1%-13%)
 Occupation: 4% (<2%-8%)
 Alcohol: 3% (2%-4%)
 Geophysical factors (natural radiation): 3% (2-4%)
 Pollution: 2% (<1% – 5%)
 Food additives: <1% (-5% – 2%)
 Medicines/Medical Procedures: 1% (0.5%-3%)
 Industrial consumer products: <1% (<1%-2%)
 Unknown: ?
Harvard Center for Cancer Prevention
(www.hsph.harvard.edu/cancer)
 Tobacco: 30%
 Diet in adult life, including obesity: 30%
 Sedentary lifestyle: 5%
 Infectious agents: 5%
 Defects in single genes that run in family: 5-10%
Exhibit 16–1 American Cancer Society (ACS) Guidelines on Nutrition and Physical Activity for
Cancer Prevention
ACS Recommendations for Individual Choices
1. Eat a variety of healthful foods, with an emphasis on plant sources.
 Eat five or more servings of a variety of vegetables and fruits each day.
 Choose whole grains in preference to processed (refined) grains and sugars.
 Limit consumption of red meats, especially those high in fat and processed.
 Choose foods that help maintain a healthful weight.
2. Adopt a physically active lifestyle.
 Adults: engage in at least moderate activity for 30 minutes or more on five or more days
of the week; 45 minutes or more of moderate-to-vigorous activity on five or more days
per weekmay further enhance reductions in the risk of breast and colon cancer.
 Children and adolescents: engage in at least 60 minutes per day of moderate-to-vigorous
physical activity at least five days per week.
3. Maintain a healthful weight throughout life.
 Balance caloric intake with physical activity.
 Lose weight if currently overweight or obese.
4. If you drink alcoholic beverages, limit consumption.
ACS Recommendations for Community Action
Public, private, and community organizations should work to create social and physical
environments that support the adoption and maintenance of healthful nutrition and physical
activity behaviors.
 Increase access to healthful foods in schools, worksites, and communities
 Provide safe, enjoyable, and accessible environments for physical activity in schools, and
for transportation and recreation in communities.
Source: Reprinted with permission from T Byers, et al., American Cancer Society Guidelines on Nutrition
and Physical Activity for Cancer Prevention: Reducing the Risk of Cancer with Healthy Food Choices and
Physical Activity. CA: A Cancer Journal for Clinicians , Vol 52, pp. 92-119, © 2002, Lippincott Williams
& Wilkins.