METHODS: Measurement error with covariates can cause bias in the estimate of coefficients and variance. However, sometimes there is no “gold standard” for exposure assessment such as systolic blood pressure and serum hormones. While a single measurement is subject to random within-person error, the average of repeated measurements within a short time period could be considered a gold standard for true exposure value. Regression Calibration for reliability study was developed to assess the relation between replicates and true exposure. There are two types of reliability studies. One is internal reliability study which uses a random sample from the main study and collects their repeated measures of exposure. Another one is external reliability study which collects replicates from participants external to the main study.