Three phase 2 trials were conducted to assess the efficacy and long-term safety of weight-based memantine extended release (ER) treatment in children with autism spectrum disorder. MEM-MD-91, a 50-week open-label trial, identified memantine extended-release treatment responders for enrollment into MEM-MD-68, a 12-week randomized, double-blind, placebo-controlled withdrawal trial. MEM-MD-69 was an open-label extension trial in which participants from MEM-MD-68, MEM-MD-91, and open-label trial MEM-MD-67 were treated ⩽48 weeks with memantine extended release. In MEM-MD-91, 517 (59.6%) participants were confirmed Social Responsiveness Scale responders at week 12; mean Social Responsiveness Scale total raw scores improved two to three times a minimal clinically important difference of 10 points. In MEM-MD-68, there was no difference between memantine and placebo on the primary efficacy parameter, the proportion of patients with a loss of therapeutic response (defined as ⩾10-point increase from baseline in Social Responsiveness Scale total raw score). MEM-MD-69 exploratory analyses revealed mean standard deviation improvement in Social Responsiveness Scale total raw score of 32.4 (26.4) from baseline of the first lead-in study. No new safety concerns were evident. While the a priori-defined efficacy results of the double-blind trial were not achieved, the considerable improvements in mean Social Responsiveness Scale scores from baseline in the open-label trials were presumed to be clinically important.