Epidemiological darknessBirger Svihus, professor of nutrition

The experiment - the gold standard of science

• Randomisation– Distributes error/other contributing factors

evenly among control groups and intervention groups

• Intervention– Gives the scientist control over the factor

studied

• Blinding– Reduces bias in data collection

Bertrand Russell (1872-1970)

• ”The word ’cause’ is so inextricably bound up with misleading associations as to make its complete extrusion from the scientific vocabulary desirable.” (Ward, Medical Health Care and Philosophy 12, 333, 2009)

John Snow, English physician (1813-1858)

• Established a link between water and cholera by epidemiological studies (Freedman, Statistical Science 3, 243, 1999)

Major observational methods used in epidemiology

• Correlation studies– A large number of data are screened to search for

associations between a response (e.g. obesity) and a factor (e.g. amount of a food)

• Cohort studies– Healthy persons are grouped according to factors of

interest, and the incidence of a response is registered over time

• Case-control studies– Persons with a health problem (cases) are studied in

regards to potential risk factors, and the odds ratio is compared with a control group without the health problem

Major problems in observational epidemiological studies

• Confounding– Other, unknown factors, can be underlying causes for

both the factor and the response, e.g. the correlation between icecream consumption and drowning incidence

• Reverse causation– A factor that is correlated to a response may not be

the cause for the response, but rather vice versa, e.g. the correlation between number of firefighters and the gravity of a fire

• Bias in data collection

Sir Ronald A. Fisher on epidemiology (Breslow, Journal of the American Statistical Association

91, 433, 1996)

• ”Statistics has gained a place of modest usefulness in medical research. It can deserve and retain this only by complete impartiality, which is not unattainable by rational minds … I do not relish the prospect of this science being now discredited by a catastrophic and conspicious howler. For it will be as clear in retrospect, as it is now in logic, that the data so far do not warrant the conclusions based on them.” (1957, on smoking and lung cancer)

Schünemann et al., Journal of Epidemiology, Community and Health 65, 392, 2011)

ProbableThese criteria are for evidence strongenough to support a judgement of a probable causal relationship.• Evidence from at least two independent cohort studies, or at least five case-control studies.• No substantial unexplainedheterogeneity between or within studytypes in the presence or absence of anassociation, or direction of effect.• Good quality studies to exclude withconfidence the possibility that theobserved association results fromrandom or systematic error, includingconfounding, measurement error, andselection bias.• Evidence for biological plausibility.

ConvincingThese criteria are for evidence strong enough to support a judgement of a convincing causal relationship.• Evidence from more than one study type.• Evidence from at least two independent cohort studies.• No substantial unexplained heterogeneity within or between study types or in different populations.• Good quality studies to exclude with confidence the possibility of random or systematic error.• Presence of a plausible biological gradient in the association. • Strong and plausible experimental evidence, either from human studies or relevant animal models.

World Cancer Fund criteria for causation from epidemiological data (http://www.dietandcancerreport.org/)

• “What is required is much more than the application of a list of criteria. Instead, one must apply thorough criticism, with the goal of obtaining a quantified evaluation of the total error that afflicts the study. This type of assessment is not one that can be done easily by someone who lacks the skills and training of a scientist familiar with the subject matter and the scientific methods that were employed. Neither can it be applied readily by judges in court, nor by scientists who either lack the requisite knowledge or who do not take the time to penetrate the work.”

Review paper by Rothman and Greenland (American Journal of Public Health s1,

s144, 2011)

A hierarcical list of criteria to use for dietary recommendations. The food

should:• 1. provide enough nutrients• 2. not provide too much energy and thus cause

obesity• 3. have a balanced content and quality of

carbohydrates and fat to hinder diabetes 2 and/or atherosclerosis

• 4. not contain too much of ingredients thought to be carcinogenic, or too little of ingredients thought to protect against cancer

Review paper by Boffetta (Critical Reviews in Food Science and Nutrition, 50:13–16, 2010)

• “In conclusion, cancer epidemiology is, to a large extent, the determination of small effects and weak associations, and poses major challenges that are easier to overcome in certain areas (e.g., genetic epidemiology) than in others (e.g., environmental or nutritional epidemiology). Identifying the causal nature of a weak association is not impossible, but requires large, well-planned, and well-conducted studies and supporting evidence from molecular and experimental studies.”

New dietary recommendations from the government

• Eat less red meat

The example of red meat

Rich in essential nutrients

The example of red meat

Rich in essential nutrients

Low in energy which protects against obesity

The example of red meat

Rich in essential nutrients

Low in energy which protects against obesity

Low cho and fat which protects against diabetes/ atherosclerosis

The example of red meat

Rich in essential nutrients

Low in energy which protects against obesity

Low cho and fat which protects against diabetes/ atherosclerosis

Associated with colon cancer

The risk for colorectal cancer due to red meat (Cross et al., PloS Medicine 4, e325, 2007)

• The risk of developing cancer was 24 % higher for persons eating 170 gram red meat per day compared with those eating 30 gram per day

• In Norway, the incidence of colorectal cancer is around 80 per 100 000. Thus, if the association is causal, cancer incidence would increase to 100 per 100 000 if meat consumption in Norway was 30 gram and increased to 170 gram (it is currently around 80 gram per day)

Review paper on diet and cancer by Key et al. (The Lancet 360, 861, 2002)

• “Despite extensive research during the last 30 years, few specific dietary determinants of cancer risk have been established, even for cancers such as colorectal cancer for which most researchers agree that diet probably has important effects.”