Statistical approach to combining the results of similar experiments, with application to the hematologic effects of extremely‐low‐frequency electric field exposures

Publisher Website
Google Scholar
Abstract
A large proportion of scientific effort in investigating the possible biological effects of exposure to extremely‐low‐frequency (ELF) fields consists of laboratory studies on experimental animals. Most experiments in which hematologic properties are measured show no statistically significant effect due to exposure. However, some studies show significant effects which, in general, are not clearly reproducible. A difficult question must then be addressed: Are these relatively few indications of ELF effects statistical artifacts due to the increased risk of a type I error in multiple studies, or is there a real biological effect that is undetected in most studies due to the relatively small sample sizes commonly used? A statistical approach for examining the accumulated results of multiple experiments which results in a single test for treatment effect is presented. The technique requires very mild assumptions, and is valid for experiments that vary widely in specific characteristics such as exposure level, duration, and laboratory. The method is applied to the results of a collection of hematologic and serum chemistry experiments, and the combined results indicate the existence of experimental effects on some end points.