To study the effects of a chemical substance or a physical agent on health, what scientific tools are available? There are two possible approaches available to us, experimental studies and epidemiological studies.
In experimental studies, two groups of cells, animals, or volunteers are compared in the laboratory. One is exposed to the substance in question, the other unexposed, is called “witness” or “control.” The samples are homogeneous, apart from the presence or absence of the substance to be tested; they are placed in the same “life” conditions and are comparable with each other (same cell strain or same breed of animals, same physical environment, same feeding conditions, same examination conditions.) Everything is done so that an observed effect is necessarily due to the only variable parameter, the substance, or the agent studied. The main drawback of these studies is the validity of the transposition of an effect observed on cells or on animals to humans.
Among the definitions of epidemiology, the most used is that of the World Health Organization (WHO). It is “the study of the distribution and determinants of health conditions and diseases in human population as well only influences which determine this distribution.” There are three branches to this discipline, descriptive epidemiology, the oldest, analytical epidemiology, which we will detail last to better focus on, and evaluative epidemiology.
Gives information on the state of health, description of the causes of diseases in a population according to sex, age. This is the case, for example, of the causes of death statistics recorded at the national level by INSERM 1 or of the cases of cancers recorded by the departmental cancer registers. These descriptive studies make it possible to know the incidence 2 of diseases and causes of death or to formulate hypotheses. For example, it was the recording of mortality statistics which in 1952 linked peaks in air pollution to an increase in cardio-respiratory mortality in people over 65 in London during an air pollution episode 3.
Makes it possible to evaluate a public health intervention. Historically include treatment of scurvy in the British navy in the middle of the XVIII the century or treatment of puerperal fever (severe infection of young birth) by hand washing described by Semmelweis 4 in 1847. Today, evaluative epidemiology studies have made it possible, for example, to confirm the merits of preventing cardiovascular diseases by taking moderate-dose aspirin. In fact, this is a new descriptive study to assess the merits - or not - of a measure taken to improve health.
There is a danger when a product is capable of altering human health. For example, arsenic is dangerous. There is a health risk when a person is exposed to an unsafe product. If a person ingests arsenic, they may be sick. If she does not ingest it, she runs no risk. Risk is, therefore, the product of danger and exposure; there are dangers without risk; there is no harmless risk.
Analytical epidemiology will make it possible to test the veracity of the hypotheses posed during previous descriptive or analytical studies. These studies are set up to research the risk factors for diseases. An epidemiological study in this context must respect a certain number of stages, which are the definition of the hypothesis to be tested, the study protocol, the data collection method, the statistical analysis, the interpretation of the results, and the search for a causal relationship.
The aim of the study is to find a relationship between an M disease and an FR risk factor. A risk factor is any factor statistically linked to a disease. It can be a risk-increasing factor (e.g., tobacco is a risk factor for lung cancer) or a risk-reducing factor (also known as a protective factor; e.g., exercise is a protective factor against cardiovascular disease).
The study is based on an initial hypothesis. For example, to find out if there is a relationship between lung cancer and tobacco, the hypothesis to be tested will be:
“Smoking increases the risk of lung cancer.” This relationship will be “measured” by the relative risk (RR), which is equal to the frequency of the disease (lung cancer) in the exposed group (smokers) and the frequency of the disease (lung cancer) in the unexposed group (non-smokers). An RR equal to 1 means that there is no relation between the disease and the factor studied. We immediately see that an important question is how to “classify” people, how many cigarettes per day are we classified as a smoker? In which group will we classify people who have quit smoking? What about people who stopped six months ago compared to those who stopped 5 or 10 years ago?
The next step will be to specify the study protocol. Among the multiple types of studies possible, we will describe two, the cohort study and the case-control study adapted to the study of cancer risk factors. Once the statistical relationship has been established, it is still necessary to look for the notion of causality, that is to say, if this risk factor is indeed involved in the disease. Indeed, statistical association does not mean cause and effect relationship. For example, the number of births fell sharply between 1965 and 1980 in Germany, as did the number of storks crossing the country during migratory periods. Although a statistically significant relationship between the two phenomena has been established, it is obviously not a cause and effect relationship.
Thus, the interpretation of an epidemiological study is difficult, even if the results seem easy to understand. This is an expert discussion because each study has its strengths and limitations. It is necessary to take into account the methodological biases inherent in any study (classification errors, confounding factors, etc.), to differentiate between statistical and causal relationships and to be familiar with previous studies on the subject.
The bottom line: a study alone does not hold the truth; it must always be placed in the context of research and integrated into previous results.