Last update:


Arch Hellen Med, 22(1), January-February 2005, 36-49


Epidemiologic investigation of the etiology of chronic diseases

1Department of Hygiene and Epidemiology, Medical School, University of Athens, Athens, Greece
2Faculty of Health Professions, Athens Technological Institute (T.E.I.), Athens, Greece
3Department of Medical Epidemiology & Biostatistics, Karolinska Insitutet, Stockholm, Sweden
4Department of Epidemiology, Harvard School of Public Health, Boston, USA

Etiology in epidemiology, preventive medicine and medicine, in general, is stochastic (probabilistic) rather than strictly deterministic. In this review, the concepts of cause and causality in the context of epidemiology and preventive medicine and the epidemiologic designs which are employed for the study of the etiology of chronic diseases are considered. Randomized intervention trials (randomized trials) are optimal but in many situations are either ethically unacceptable or very difficult to implement. Non-interventional (observational) epidemiological investigations including ecologic, retrospective (case-control), prospective (cohort, follow-up) studies, are frequently used and there is also a tendency of increased use of meta-analyses and pooled analyses. Characteristics, advantages and disadvantages of prospective and retrospective epidemiological investigations are de48 Π. ΛΑΓΙΟΥ και συν scribed and the conditions that dictate the use of one or the other design are indicated. Basic concepts of epidemiology as person-time, the reference population, the main effect indicators and interaction (effect modification) are presented and explained. Since the documentation of causality presupposes minimization of the role of systematic errors, confounding influences and chance, the respective concepts are described and their consequences in epidemiologic research are considered. Problems in the interpretation of the P value which is frequently used to assess statistical significance are discussed and the inherent advantages of confidence intervals to describe results while taking into account random variation are stressed. Lastly, etiologic inference is considered in the context of a single study, a series of studies or a particular exposed individual.

Key words: Epidemiology, Εtiology, Statistical significance, Systematic errors.

© Archives of Hellenic Medicine