This systematic review and network meta‑analysis adhered to PRISMA 2020 and PRISMA‑NMA guidance and was prospectively registered in PROSPERO (CRD420261422218). The authors searched PubMed, Europe PMC, Semantic Scholar and Crossref from inception to 29 July 2026, and supplemented electronic searches with citation chasing. The analysis used a frequentist random‑effects network meta‑analytic framework to compare adjunctive interventions for negative symptoms.
Included studies were randomized controlled trials in adults diagnosed with DSM or ICD schizophrenia or schizoaffective disorder who had prominent or persistent negative symptoms while receiving stable antipsychotic medication. Eligible interventions were adjunctive (add‑on) pharmacological or non‑pharmacological treatments. The prespecified outcomes were scores on the Positive and Negative Syndrome Scale (PANSS) negative subscale or the Scale for the Assessment of Negative Symptoms (SANS). Effect estimates were presented as standardized mean differences (SMDs) with 95% confidence intervals (CIs); negative SMD values favoured the add‑on treatment.
Multiple databases were searched to 29 July 2026, with additional citation chasing. From the search, 49 unique randomized controlled trials met the clinical and design inclusion criteria. The authors created a locked quantitative dataset for analysis containing trials meeting data completeness and connectivity requirements.
A random‑effects frequentist network meta‑analysis estimated SMDs and 95% CIs across treatments. Network P‑scores were used to rank treatments. Heterogeneity was quantified and reported (τ = 0.119; τ2 = 0.014). Because the evidence network was largely star‑shaped, inconsistency could not be evaluated formally—there were zero inconsistency degrees of freedom in the primary connected network.
Of 49 trials that met the clinical and methodological screening, 28 trials contributed to the locked quantitative dataset, comprising 2,067 randomized participants and 1,933 participants with analysable outcome data. The primary connected network included 24 trials, representing 25 treatments and 31 comparison estimates.
Against a broad control node, 14 add‑on interventions had 95% confidence intervals excluding the null. The highest P‑scores and corresponding SMDs reported in the analysis were observed for mirtazapine (SMD −2.38, 95% CI −3.55 to −1.21), granisetron (SMD −1.96, 95% CI −2.74 to −1.17), tropisetron (SMD −1.82, 95% CI −2.59 to −1.05), minocycline (SMD −1.78, 95% CI −2.55 to −1.01) and memantine (SMD −1.54, 95% CI −2.28 to −0.80). These estimates favoured the add‑on pharmacological agents over control for reduction of negative symptom scores in the included trials.
The network included non‑pharmacological interventions. The authors reported efficacy signals versus broad control for neuromodulation techniques and psychotherapeutic approaches, specifically tDCS (transcranial direct current stimulation), rTMS (repetitive transcranial magnetic stimulation) and body‑oriented psychotherapy. Exact SMDs for these non‑pharmacological interventions were reported in the full dataset; the abstract highlights that these interventions showed signals of benefit in the network analysis.
Overall heterogeneity in the primary network was low (τ = 0.119; τ2 = 0.014). However, the network geometry was predominantly star‑shaped with most comparisons coming from single, small trials feeding into a broad control node. This structure resulted in zero degrees of freedom for formal inconsistency testing, so inconsistency could not be estimated. Risk of bias assessments using Cochrane RoB 2 yielded overall judgments of low risk for eight trials, some concerns for 15 trials and high risk for five trials.
The analysis identified several pharmacological and non‑pharmacological add‑ons with potentially important efficacy signals for negative symptoms of schizophrenia, with highest P‑scores for mirtazapine, granisetron, tropisetron, minocycline and memantine. Non‑pharmacological adjuncts including tDCS, rTMS and body‑oriented psychotherapy also demonstrated efficacy signals versus a broad control node.
Key limitations noted by the authors include the sparse, star‑shaped evidence network, the reliance of many nodes on single small trials, the scarcity of direct active comparisons between interventions and common risk‑of‑bias concerns. Because inconsistency could not be evaluated and many interventions were supported by limited trial data, the authors advise that the treatment ranking should be regarded as hypothesis‑generating rather than a basis for definitive clinical recommendations.
Readers should also note that this report is a preprint and has not undergone peer review; the authors explicitly state it should not be used alone to guide clinical practice.
The network meta‑analysis suggests multiple adjunctive pharmacological and non‑pharmacological interventions may reduce negative symptoms in schizophrenia, but the evidence is limited, often based on single small trials, and at varying risk of bias. Further confirmatory, adequately powered direct comparisons are required before changing clinical practice.