We conducted a randomized control trial assessing the effects of two variants of word-problem intervention with third graders (n = 304) experiencing mathematics difficulty. Students were assigned to a business-as-usual condition (BaU) or one of two variants of word-problem intervention. One variant included a pre-algebraic reasoning component (relational understanding of the equal sign as well as standard and nonstandard equation solving); the other included word-problem intervention without pre-algebraic reasoning. Students in both interventions significantly outperformed students in the BaU with large effect sizes (2.66 and 2.44), but there were no significant differences between the two intervention conditions. The purpose of the present analysis was to assess maintenance of effects 6-12 months after intervention with the students (n = 229), now in fourth grade. At follow-up, only students in the word-problem intervention with pre-algebraic reasoning significantly outperformed the BaU on a measure of word-problem solving. This finding suggests an advantage for the pre-algebraic reasoning component, yet the follow-up effect between intervention conditions was not significant. The ESs of 0.43 and 0.31 and persistence rates of 16% and 13% reveal substantial forgetting for both conditions, which suggests the dose of intervention may not be adequate for many students.