Chronic diseases such as heart disease, diabetes, and obesity are known to develop over many years and have been strongly linked with diet. However, the concept of time is not fully incorporated into most of the research investigating these associations. This is partially due to the lack of suitable distance measures for comparing time series corresponding to different eating patterns. This paper develops the concept of temporal dietary pattern (TDP) and presents dynamic time warping based novel distance measure, referred as Modified Dynamic Time Warping (MDTW), for comparing different eating patterns. An efficient algorithm for estimating MDTW distance is used in k-means clustering for comparing 24-hour dietary data and identifying TDPs. Efficacy of the proposed distance measure is shown by estimating TDPs for a representative sample of the adult US population (from the National Health and Nutrition Examination Survey).