We have previously provided methods for measurement error correction for cross-sectional dietary exposures (X) which allow for multiple covariates measured with error (Z) and without error (W) with possible correlation among covariates within (Z, W). However, to our knowledge, there is no previous literature on measurement error correction for change in surrogate exposures. In the Nurses’ Health Study (NHS) we have limited longitudinal validation study data, but propose an innovative indirect method for estimating correlations among Z, X and W both cross-sectionally and longitudinally under a stationary time series structure, where the longitudinal correlation depends on the time interval between repeated measures. This is made possible by the existence of longitudinal data on Z and W every 4 years among main study participants and several validation studies on X among a subset of NHS subjects. In this talk, we use this structure to perform measurement error correction based on change in exposure over time and use it to assess the association between change in alcohol intake and incident breast cancer.