The aim of the study was to identify a dietary pattern predictive of subsequent annual weight change by using dietary composition information. Study subjects were 24,958 middle-aged men and women of the European Prospective Investigation into Cancer and Nutrition-Potsdam cohort. To derive dietary patterns, we used the reduced rank regression method with 3 response variables presumed to affect weight change: fat density, carbohydrate density, and fiber density. Annual weight change was computed by fitting a linear regression line to each person's body weight data (baseline, and 2- and 4-y follow-up) and determining the slope. In linear regression models, the pattern score was related to annual weight change. We identified a food pattern of high consumption of whole-grain bread, fruits, fruit juices, grain flakes/cereals, and raw vegetables, and of low consumption of processed meat, butter, high-fat cheese, margarine, and meat to be predictive of subsequent weight change. Mean annual weight gain gradually decreased with increasing pattern score (P for trend < 0.0001), i.e., subjects scoring high for the pattern maintained their weight or gained significantly less weight over time compared with subjects with an opposite pattern. However, the prediction of annual weight change by the food pattern was significant only in nonobese subjects. In this study population, we identified a food pattern characterized by high-fiber and low-fat food choices that can help to maintain body weight or at least prevent excess body weight gain.