Problems of safety levels derived from epidemiologic studies using independent variable with measurement error

点击翻译按钮获取中文摘要

In epidemiologic or experimental studies for dose-response or -effect relationship between toxic factors and biological responses or effects, parameters which express strength of relationship between independent and dependent variables are usually calculated using statistical models without errors in independent variable. When error in independent variable is not small in study based on individual data, obtained parameter using usual statistical model includes bias toward null relationship between independent and dependent variables i.e. under estimate. When the error in independent variable is not large and the magnitude of the error is known, the methods which correct the under estimate of the parameter were developed. Ecologic study based on means of independent and dependent variables in groups are sometimes conducted in order to decrease effect of error in independent variables. It is, however, pointed out that ecologic study is sometimes carried out without due consideration for confounding factors.In epidemiologic study for dose-response relationship between renal dysfunction as response and cadmium (Cd) intake calculated from Cd concentration in rice (Cd-R) as dose, error of the dose is very large. Therefore, it is impossible to correct the parameter, which express strength of relationship between independent and dependent variables, using usual statistical model without error in independent variable based on individual data. In the epidemiologic study area, ecologic studies for relationship between means of Cd-R and prevalences of renal dysfunction in Cd polluted and control areas can be carried out. The epidemiologic studies were conducted without due consideration for confounding variable which represents the other environmental factors than Cd-R. Therefore environmental safety level of Cd derived from the epidemiologic studies using Cd-R is not adequate.