No commercial breeding programs have yet started to improve feed utilization efficiency in fish, mainly because of the difficulty in accurately measuring individual feed intake in fish reared in groups. Our general goal is to propose indirect criteria to be used for breeding of commercial lines. In a previous study undertaken with rainbow trout clones, a genetic correlation was detected between residual feed intake (RFI), and body weight variation during successive periods of feed deprivation (FD) and re-feeding (RF). To assess the pertinence of such indirect criteria for future breeding programs, we set up a large experiment using sea bass (Dicentrachus labrax), a recently domesticated species with broad genetic diversity. The objectives of the present study were to analyze the relationship of FD and RF with RFI, and to assess the consistency of such indirect criteria over time and their correlation with carcass quality. Fish originating from a full factorial design combining eight dams and 41 sires, were raised in the same tank. At 306 days post fertilization 2000 fish were individually tagged and their body weight recorded over a growth period of three weeks, followed by two successive periods of three weeks of feed deprivation and three weeks of ad libitum re-feeding. Fish performances were then classified, FD-, FD+, RF- and RF+ for fish exhibiting loss (FD) or gain (RF) of weight relatively lower (-) and higher (+) than the population mean. Fish were sorted into four groups (FD-/RF-, FD+/RF+, FD-/RF+, FD+/RF). Each of the 4 groups was split between three replicated tanks of 50 fish. Body weight gain and feed intake were measured for each replicate every three weeks over a six months period to estimate residual feed intake. The fish then underwent a period of three weeks of feed deprivation followed by a period of three weeks of ad libitum re-feeding. At the end of each period, individual weight gain variations were recorded, as well as muscle fat using ultrasonic measurement. The relationship with RFI variations was high for FD (P = 0.04, n = 12), and close to significance for a criterion merging FD and RF performances (P = 0.06, n = 12). FD but not RF was negatively correlated with muscle fat (P < 0.05, n = 600). In conclusion, selecting fish losing less weight during a three week feed deprivation period should lead to improve RFI, but also to increase muscle fatness. Using a combination of FD and RF as indirect criteria would result in slower progress in RFI, but would not impact carcass quality traits. The next step for inferring the potential genetic gain that can be expected is to assess the heritability of such criteria.