[Help] [Aide] [Up]

The Cancer Journal - Volume 5, Number 5 (September-October 1992)

editorial


Epidemiology spreads disinformation



Disinformation threatens to overpower medical reasoning. From year to year medicine becomes more and more complex and splits into many sub-specialties which focus on diseases while tending to ignore the patient himself. Amidst this schism epidemiology thrives. Surrounded by an aura of genuine science it seems to provide the desired unifying principle for medicine. Yet a closer look at its basic premises reveals the opposite. Epidemiology is in a state of a profound confusion (1-4) that is still unnoticed by the medical establishment which regards epidemiology as its oracle.
Take the prestigious topic of smoking and cancer (4). Obviously in heavy smokers lung cancer is more prevalent than in non smokers; yet epidemiology lacks the means to prove that smoking causes cancer. At best it can state that lung cancer is associated or correlated with smoking. However an association between two variables has at least three interpretations: either smoking initiates cancer, or cancer triggers an urge to smoke, or the two may never interact and their observed association results from a third process that was not considered in the study. Epidemiology is incapable of distinguishing between the three. The conclusion is in the hands of medicine. In order to distinguish between them epidemiology should consider all factors that might contribute to an association between smoking an cancer, e.g., patient's history and mental state, clinical check-up, post-mortem reports etc. Instead of considering the relationship between smoking and cancer in its entire complexity, epidemiology focuses on a small number of variables, dismissing the rest as negligible, which is unfortunate since for the individual patient nothing is negligible.
The same approach underlies many epidemiological studies, particularly clinical trials, that involve simplifications and presumptions which breed false conclusions and should be regarded by the medical community with suspicion. As long as epidemiological statements do not contradict medical experience the harm to the patient may be small since they can always be corroborated by clinical observation. Yet epidemiology threatens to impose its deceptive reasoning on medical issues that are far less obvious, e.g., cancer treatment which today is directed by clinical trials.
From the viewpoint of medicine most of the basic premises of epidemiology are wrong. For instance, in order to apply parametric-multivariate models to medical phenomena, the variables have to be distributed normally and their variances equal. This requirement is seldom met. The organism is extremely complex, all its components interact, and none is distributed normally. Epidemiology assumes that the variables can be transformed so as to be distributed normally, which is generally incorrect. Since many distributions are skewed it appears as if they can be normalized by the log-normal transformation. In many cases the distribution becomes symmetric, yet this apparently harmless transformation changes the meaning of the variables. While in the normal distribution they are independent, in the log-normal they become proportional. Instead of testing the variables in their native form, epidemiology distorts them and ignores the consequences.
Thus, although epidemiological tests are mathematically consistent, they do not meet the basic requirements that are necessary for applying them to medicine. This is the essence of the confusion in epidemiology. They stun medicine with the consistency of their mathematics and hide the fact that most models are irrelevant. This is also the nature of epidemiological disinformation, since most epidemiological statements are neither true nor false. Epidemiological disinformation is a manifestation of iatrogenesis, particularly since it aims to direct patient treatment. It is therefore hazardous to the patient. Generally, statistical analyses on a small number of variables involving traditional tests, e.g., t-tests, chi square etc, are still acceptable. As a rule of thumb, it is advisable to ignore epidemiological statements based on observations involving more than five variables, particularly if contradicting medical intuition. This applies also to clinical trials and meta-analyses.
Epidemiology gained its name and glory from the study of epidemics. Now that epidemics are rare, its mission is nearly accomplished, and epidemiology should therefore vanish.


G. Zajicek
e-mail: Gershom@md2.huji.ac.il

References

1 Zajicek G. Progress against cancer. are we winning the war? Cancer J. 3:2,1990
2 Zajicek G. Cancer wars. Cancer J. 4:4-5,1991
3 Zajicek G. Meta-analysis and chaos Cancer J. 4: 152- 153, 1991
4 Zajicek G. To smoke or not to smoke? Cancer J. 5: 70, 1992

[Up]