Background Dysfunction of N-methyl-D-aspartate receptor (NMDAR) is involved in the pathophysiology of schizophrenia. A meta-analysis of randomized controlled trials (RCTs) was conducted to examine the efficacy and safety of memantine, a non-competitive NMDAR antagonist, in the treatment of schizophrenia. Methods Standardized/weighted mean differences (SMDs/WMDs), risk ratio (RR), and their 95% confidence intervals (CIs) were calculated and analyzed. Results Included in the meta-analysis were eight RCTs (n = 452) of 11.5 ± 2.6 weeks duration, with 229 patients on memantine (20 mg/day) and 223 patients on placebo. Adjunctive memantine outperformed placebo in the measures of Positive and Negative Syndrome Scale and Brief Psychiatric Rating Scale negative symptoms [SMD: -0.63 (95% CI -1.10 to -0.16), p = 0.009, I 2 = 77%], but not in the total, positive and general symptoms [SMD: -0.46 to -0.08 (95% CI -0.93 to 0.22), p = 0.06-0.60, I 2 = 0-74%] or the Clinical Global Impression Severity Scale [WMD: 0.04 (95% CI -0.24 to 0.32), p = 0.78]. The negative symptoms remained significant after excluding one outlying RCT [SMD: -0.41 (95% CI -0.72 to -0.11), p = 0.008, I 2 = 47%]. Compared with the placebo group, adjunctive memantine was associated with significant improvement in neurocognitive function using the Mini-Mental State Examination (MMSE) [WMD: 3.09, (95% CI 1.77-4.42), p < 0.00001, I 2 = 22%]. There was no significant difference in the discontinuation rate [RR: 1.34 (95% CI 0.76-2.37), p = 0.31, I 2 = 0%] and adverse drug reactions between the two groups. Conclusions This meta-analysis showed that adjunctive memantine appears to be an efficacious and safe treatment for improving negative symptoms and neurocognitive performance in schizophrenia. Higher quality RCTs with larger samples are warranted to confirm these findings.