This paper reports the results from a three-years follow-up study to access the long-term efficacy of a cognitive training for healthy older adults and investigates the effects of booster sessions on the cognitive performance of the participants using an innovative analytical approach from information theory. Design: semi-randomized quasi-experimental controlled design. Participants: 50 healthy older adults, (M = 73.3, SD = 7.77) were assigned into an experimental (N = 25; Mean age = 73.9; SD = 8.62) and a passive control group (N = 25; mean age = 72.9; SD = 6.97). Instruments: six subtests of WAIS and two episodic memory tasks. Procedures: the participants were assessed in four occasions: after the end of the original intervention, pre-booster sessions (three years after the original intervention), immediately after the booster sessions and three months after the booster sessions. Results: the repeated measures ANOVA showed that two gains reported in the original intervention were identified in the follow-up: Coding (F(1, 44) = 11.79, MSE = 0.77, p = .001, ηˆG2 = .084) and Picture Completion (F(1, 47) = 10.01, MSE = 0.73, p = .003, ηˆG2 = .060). After the booster sessions, all variables presented a significant interaction between group and time favorable to the experimental group (moderate to high effect sizes). To compare the level of cohesion of the cognitive variables between the groups, an entropy-based metric was used. The experimental group presented a lower level of cohesion in three of the four measurement occasions, suggesting a differential impact of the intervention with immediate and short-term effects, but without long-term effects.