Evaluation of Epidemiologic Information.
NIOSH 1985 May:12 pages
The interpretation of inconsistencies in different epidemiological studies was discussed. According to the author, results may vary and may even be contradictory in different studies due to variable factors such as the circumstances of exposure or differences in study design and methods. Different treatment of confounding variables also may cause different results. The term "pooling of inferences" has been suggested to indicate a standardized adjustment procedure for a set of common confounding variables. Such a procedure would improve the comparability of different studies. Different susceptibility to risk factors may be a property of the populations in different studies that could have a significant effect on the outcome (the so called effect modification). The time of performance of the study may play a role if the agent under study has a long latency period. Thus, studies that are carried out before the adverse effects are apparent will underestimate the risk. In addition, reactions to initial study results for popular exposures, such as coffee drinking and risks of pancreatic cancer, can result in reporting bias in further studies. The author concludes that the most difficult challenges that face the epidemiologist are not related to the statistical aspects of an investigation, but rather to the validity and reliability of the measurements of exposure and outcome variables.
Epidemiology; Biostatistics; Environmental-exposure; Leukemogenesis; Humans; Breast-cancer; Bladder-cancer; Risk-analysis;
Proceedings of a Symposium on Epidemiology and Health Risk Assessment, Columbia, Maryland, May 14-16, 1985, Centers for Disease Control/NIOSH, 12 pages, 4 references