Errors in exposure measures.

Errors in exposure measures reduce the power in statistical tests in health effect studies, and also bias the estimate of the magnitude of the effect of exposure on health. Several types of error in commonly used exposure measures are described. To study how these errors in exposure measures may obscure exposure-response relationships, an imaginary data set was generated, with (indoor) exposure to nitrogen dioxide and fine particles in relation to pulmonary function as an example. With this data set, the effect of different types of errors in the exposure measures were simulated. The simulation indicated that a serious under- or overestimation of the magnitude of the health effect can occur due to errors in the exposure measures. The bias was more pronounced for errors in strong than in weak determinants of the health effects.