Avertir le modérateur

mardi, 15 septembre 2015



Claude Gilois 

    The first thing you have to do is to take care of your brain.  The second is to extract yourself from all    systems of indoctrination. Then, there comes a time when it becomes a reflex to list all the lies and distortions you read  and place them in your own framework. But to achieve this, you have to accept that the state, the corporations and the media will treat you as their enemy. You then have to learn how to defend yourself. If we had a proper system of education, it would give you lessons on intellectual self-defence.


Epidemiological studies generate an inaudible cacophony of results for the layman

There is not a month that passes by without some contradictory information being released by the scientific and medical community searching for links, often subtle, between diet, lifestyle, environmental factors and health. Alcohol, as the most studied subject (there are more publications on alcohol than on any other subjects in the literature) is invariably on the Olympic podium with a clockwork regularity. The pendulum swings back and forth leading to a climate of anxiety within the population. This absurd situation has led an eminent epidemiologist, Dimitrios Trichopoulos, head of Harvard Public Health School to declare: 'Epidemiologists are beginning   to be a nuisance to the population' (i). On subjects that are particularly topical or have a moralistic connotation such as alcohol, propagandists from all sides fight a war of words with the data generated by these studies, often without sufficient knowledge of their validity or even without fully understanding the results, such is the difficulty of interpreting them as they involve highly complex statistical analyses. 

Epidemiology faces its limits

Medical and scientific research is usually carried out on subjects chosen at random for the study and for the control groups. The Rolls Royce of the studies is when subjects and the controls are not known to the subjects themselves or to the researchers. These are known as double blind randomised studies, but if you want to study the effect of alcohol on the health of the population it is not possible to use such a methodology, as it would be unethical to administer a substance that could be potentially harmful to healthy volunteers.  Scientists have to resort to indirect studies such as case control (1), cohort (2) and meta-analyses (3), which are far more imprecise.

The results of these studies are usually expressed in terms of an index (a number) (4), a number equal to one means the absence of risk, the higher the number, the greater the risk. This is where the weakness of the methodology lies as the great majority of epidemiological studies report indexes that lie between one and three, far too low to overcome the inherent difficulties associated with the methodology of epidemiological studies. Many epidemiologists concede that their studies are packed with biases, uncertainties and methodological weaknesses and that they are unable to detect weak associations. When Richard Doll identified the link between tobacco and lung cancer, there was an increased risk of 3000% and there were no possible doubts, but what happens when the risk is increased by only a few percent?   ‘I have trouble imagining a system that involves human beings over a long period of time that could give you a reliable estimate of increased risks that are in the order of tens of percents’ declared Alex Walker, a Harvard epidemiologist.


wine health picture1.jpg



Can looking for a needle in a haystack give you, in the end, spectacular results ?

'We are pushing the edge of epidemiology when we do not go beyond' says Dimitrios Trichopoulos mentioned above 'and our studies generate false positive or negative results with a disturbing frequency'. 'Biases and confounding factors (5) are the plagues on the house of epidemiology' declares Philip Cole, professor of epidemiology at the university of Alabama.A ubiquitous example of a confounding factor is cigarette smoking, which can confound any study looking, for instance, at the effects of alcohol on cancer. "It just so happens," explains Trichopoulos, "that people who drink also tend to smoke" boosting their risk of cancer.

An equally ubiquitous example of a bias is assessing the exact exposure to a risk factor. "Even the sophisticated statistical techniques that have entered epidemiological research over the past 20 years — tools for teasing out subtle effects, calculating the theoretical effect of biases, correcting for possible confounders, and so on — can’t compensate for the limitations of the data”, says bio-statistician Norman Breslow of the University of Washington, Seattle.

Many eminent epidemiologists who have published erroneous results in the past state that it is very easy to be fooled.  Sir Richard Dole (the scientist who uncovered the link between tobacco and lung cancer and who uncovered an erroneous link between breast cancer and the hypertensive medication, Reserpine, suggests that no epidemiological study can be credible if the lower limit for index that measures the risk is not above 3; Dimitrios Trichopoulos suggests a lower limit of 4.  Angell from the New England Journal of Medicine declares “We need an index of at least 3 or more before we accept the paper for publication in particular if the biological mechanism is improbable or the discovery is new”.  Robert Temple from the Food and Drug Administration adds, “If the index is not above 3 or 4 then forget about the results’.


wine and health picture2.jpg



The most eminent epidemiologists therefore concede explicitly that the majority of epidemiological studies published are without value since the index is too weak to conclude anything. This begs the question as to why they are published. ‘It is not so much that the nature of epidemiology has changed’ argues Trichopoulos who adds ‘to-day there are more scientists practising the discipline, and the pressure is greater on the profession’. A scientist does not exist if he does not publish. The expression 'publish or perish’ summarised very well the need to accumulate publications sometimes at the expense of the quality of the data generated.  Conflicts of interest are rife in the scientific community. Most of the research projects are financed wholly or partly of by the pharmaceutical and chemical industries, they there have vested interests in the results. In 2006, the UK newspaper, The Guardian, reported that for years Sir Richard Dole received a salary of 1,500 $ dollars a month from Monsanto, the firm known at that time for producing highly polluting substances and now best known for its GMO. This situation most probably led him to underestimate the frequency of cancer linked to environmental causes. 

It could be argued that meta-analyses are a way to overcome the limits of individual analyses - not really - negative results tend to be under-reported in scientific and medical literature as scientists are less inclined to submit publications on negative results, and publications are generally not very keen to publish them. Furthermore, and most importantly, David Sackett, University of Oxford, declares 'If studies have the same methodological architecture and these studies have a bias, no matter if they are reproducible, a bias is always a bias even if it is multiplied by 12'.     If you add to this that the number of cases of scientific fraud has markedly increased and it is in medical and biological sciences that they flourish,it seriously undermines the credibility of some of the publications.Medical sciences come top of the league with 52% of cases of fraud where data was simply fabricated anywhere in the world. Again, the cases involving manipulation of data (a bit less serious) the medical sciences come top with 81% of frauds identified  (ii). 


Perhaps it is time to dismiss those who want to persuade us that the small risk they perceive is a big thing for us…whatever the risk is!!!  


wine and health picture3.jpg


Cave men pondering the limits of epidemiology

(1) They are retrospective studies between two groups, one presenting the disease (case) and the other without the disease (control). 

(2) These are comparative studies between a group of subjects exposed to a risk and a group not exposed to it. They tend to be more precise than case control studies. 

(3) They consist of looking though all the scientific literature on a particular topic and carrying out a global analysis on all or part of the publications using sophisticated statistical analyses 

(4) The relative risk (rr) for cohort studies or the odd ratio (or) for case control studies 

(5) Confounding factors are the hidden variables in the populations being studied, which can easily generate an association that may be real but not what the epidemiologiste think it is. bereal but is not what the epidemiologist thinks it is.  

Les commentaires sont fermés.

Toute l'info avec 20minutes.fr, l'actualité en temps réel Toute l'info avec 20minutes.fr : l'actualité en temps réel | tout le sport : analyses, résultats et matchs en direct
high-tech | arts & stars : toute l'actu people | l'actu en images | La une des lecteurs : votre blog fait l'actu