The Assessment of Industrial Health Hazards on The Basis of Observational Data — Some Statistical Pitfalls

Because of ethical and practical difficulties, controlled experimentation is seldom possible in the field of industrial medicine. Safety standards for industrial hazards must therefore be based upon uncontrolled observational data. This paper is concerned with the effects of population selection and of the association between exposure and response upon the exposure-response relationships derived from uncontrolled studies. It is shown that serious distortions may result and that ‘safe limits’ derived from uncontrolled studies may underestimate the real hazard.