TY - JOUR

T1 - Effects of exposure imprecision on estimation of the benchmark dose

AU - Budtz-Jørgensen, Esben

AU - Keiding, Niels

AU - Grandjean, Philippe

N1 - Keywords: Algorithms; Child; Confidence Intervals; Data Interpretation, Statistical; Dose-Response Relationship, Drug; Evaluation Studies as Topic; Female; Humans; Maternal Exposure; Mercury; Models, Statistical; Models, Theoretical; No-Observed-Adverse-Effect Level; Pregnancy; Prenatal Exposure Delayed Effects; Probability; Regression Analysis; Research Design; Risk; Risk Assessment; Sample Size; Sensitivity and Specificity; Software; Statistics as Topic

PY - 2004

Y1 - 2004

N2 - In regression analysis failure to adjust for imprecision in the exposure variable is likely to lead to underestimation of the exposure effect. However, the consequences of exposure error for determination of safe doses of toxic substances have so far not received much attention. The benchmark approach is one of the most widely used methods for development of exposure limits. An important advantage of this approach is that it can be applied to observational data. However, in this type of data, exposure markers are seldom measured without error. It is shown that, if the exposure error is ignored, then the benchmark approach produces results that are biased toward higher and less protective levels. It is therefore important to take exposure measurement error into account when calculating benchmark doses. Methods that allow this adjustment are described and illustrated in data from an epidemiological study on the health effects of prenatal mercury exposure.

AB - In regression analysis failure to adjust for imprecision in the exposure variable is likely to lead to underestimation of the exposure effect. However, the consequences of exposure error for determination of safe doses of toxic substances have so far not received much attention. The benchmark approach is one of the most widely used methods for development of exposure limits. An important advantage of this approach is that it can be applied to observational data. However, in this type of data, exposure markers are seldom measured without error. It is shown that, if the exposure error is ignored, then the benchmark approach produces results that are biased toward higher and less protective levels. It is therefore important to take exposure measurement error into account when calculating benchmark doses. Methods that allow this adjustment are described and illustrated in data from an epidemiological study on the health effects of prenatal mercury exposure.

U2 - 10.1111/j.0272-4332.2004.00560.x

DO - 10.1111/j.0272-4332.2004.00560.x

M3 - Journal article

C2 - 15660622

VL - 24

SP - 1689

EP - 1696

JO - Risk Analysis

JF - Risk Analysis

SN - 0272-4332

IS - 6

ER -