Antidepressants for cognitive impairment in schizophrenia--a systematic review and meta-analysis

Background: Cognitive impairment in schizophrenia is disabling, but current treatment options remain limited.

Objective: To meta-analyze the efficacy and safety of adjunctive antidepressants for cognitive impairment in schizophrenia.

Data sources and study selection: PubMed, MEDLINE, PsycINFO, and Cochrane Library databases were searched until 12/2013 for randomized controlled trials comparing antidepressant augmentation of antipsychotics with placebo regarding effects on cognitive functioning in schizophrenia.

Data extraction: Two authors independently extracted data. Standardized mean differences (SMDs) were calculated for continuous outcomes and risk ratios for categorical outcomes. SMDs of individual cognitive tests were pooled on a study level within domains (primary outcome) and across domains. When results were heterogeneous, random instead of fixed effects models were used.

Results: We meta-analyzed 11 studies (duration = 8.7 ± 3.7 weeks) including 568 patients (mean age = 39.5 ± 6.9 years, males = 67.2%, illness duration = 12.5 ± 8.0 years). Antidepressants included mirtazapine (4 studies; n = 126), citalopram (2 studies; n = 231), fluvoxamine (1 study; n = 47), duloxetine (1 study; n = 40), mianserin (1 study; n = 30), bupropion (1 study; n = 61), and reboxetine (1 study; n = 33). Statistically significant, but clinically negligible, advantages were found for pooled antidepressants compared to placebo in executive function (Hedges' g = 0.17, p = 0.02) and a composite cognition score (Hedges' g = 0.095, p = 0.012). Depression improved with serotonergic antidepressants (p = 0.0009) and selective serotonin reuptake inhibitors (p = 0.009), but not with pooled antidepressants (p = 0.39). Sedation was more common with pooled antidepressants (p = 0.04).

Conclusion: Adjunctive antidepressants do not demonstrate clinically significant effects on cognition in schizophrenia patients, however, larger studies, preferably in euthymic schizophrenia patients and using full neurocognitive batteries, are needed to confirm this finding.