Importance: Atypical antipsychotics offer modest effectiveness compared with placebo but with serious safety risks, including a boxed warning for the risk of death in the treatment of behavioral and psychological symptoms of dementia (BPSD). Their comparative effectiveness and safety are not fully known. Objective: To assess the relative benefits and safety of atypical antipsychotics in the treatment of BPSD shown in randomized clinical trials using network meta-analysis. Data Sources: PubMed/MEDLINE, Embase, PsychINFO, and Cochrane Library were searched from their inception until May 31, 2018. Key terms included dementia and atypical antipsychotics. Study Selection: Randomized clinical trials comparing any atypical antipsychotic with another atypical antipsychotic or with placebo were included in the analysis. Data Extraction and Synthesis: Two independent reviewers used a standardized data extraction and quality assessment form. Random-effects network meta-analyses were performed. Effect sizes were reported as standardized mean differences (SMDs) for continuous outcomes and odds ratios (ORs) for dichotomous outcomes with 95% CIs. In addition to ORs, the surface under the cumulative ranking curve (SUCRA) was ascertained, which represents the percentage of the effectiveness or safety for each treatment compared with a hypothetical treatment that would be ranked first without uncertainty. Main Outcomes and Measures: The primary effectiveness outcome assessed was the Neuropsychiatric Inventory (NPI); secondary effectiveness outcomes were the Brief Psychiatric Rating Scale (BPRS) and Cohen-Mansfield Agitation Inventory (CMAI). The primary safety outcomes were death and cerebrovascular adverse events (CVAEs). Secondary safety outcomes were extrapyramidal signs/symptoms; somnolence/sedation; falls, fracture, or injury; and urinary tract infection/incontinence. Results: Seventeen studies (5373 patients) were included. The mean (SD) age of all participants was 80.8 (3.1) years, and most were women (3748 [69.8%]). Compared with placebo, aripiprazole was associated with improvement in outcomes on the NPI (SMD, -0.17; 95% CI, -0.31 to -0.02), BPRS (SMD, -0.20; 95% CI, -0.35 to -0.05), and CMAI (SMD, -0.30; 95% CI, -0.55 to -0.05); quetiapine was associated with improvement in outcomes on the BPRS (SMD, -0.24; 95% CI, -0.46 to -0.01), and risperidone was associated with improvement in outcomes on the CMAI (SMD, -0.26; 95% CI, -0.37 to -0.15). Differences between atypical antipsychotics were not significant for effectiveness, death, or CVAE. Compared with placebo, risperidone (OR, 3.85; 95% CI, 1.55-9.55) and olanzapine (OR, 4.28; 95% CI, 1.26-14.56) were associated with increased risk of CVAEs. The SUCRA estimated relative ranking of treatments suggested that aripiprazole might be the most effective and safe atypical antipsychotic and that olanzapine provides the least benefit overall; however, these results should be interpreted with caution where point estimates (OR and SMD) show that there is no statistically significant difference. Conclusions and Relevance: This network meta-analysis supports the existence of a trade-off between the effectiveness and safety of atypical antipsychotics in the treatment of BPSD and confirms that a single most effective and safe treatment option does not exist. Clinicians should individualize the assessment of safety risks against expected benefits when prescribing these medications to patients with dementia.