Efficacy and acceptability of pharmacological interventions for tardive dyskinesia in people with schizophrenia or mood disorders: a systematic review and network meta-analysis

Tardive Dyskinesia (TD) can occur in people exposed to dopamine receptor antagonists (DRAs). Its clinical management remains challenging. We conducted a systematic review/random-effects network meta-analysis (NMA) searching PubMed/MEDLINE/PsycINFO/ClinicalTrials.gov/Cochrane Central Register (22/05/2023, pre-defined protocol https://osf.io/b52ae/), for randomized controlled trials (RCTs) of pharmacological/brain stimulation interventions for DRA-induced TD in adults with schizophrenia or mood disorders. Primary outcomes were TD symptom change (standardized mean difference/SMD) and all-cause discontinuation (acceptability-risk ratio/RR). Sensitivity analyses were conducted. Global, local inconsistencies, risk of bias (RoB-2 tool), and confidence in evidence (CINeMA) were measured. We included 46 trials (n = 2844, age = 52.89 +/- 9.94 years, males = 59.8%, schizophrenia = 84.6%, mood disorders = 15.4%), all testing pharmacological interventions versus placebo. We identified three subnetworks. In network 1, several treatments outperformed placebo on TD symptoms with large effect sizes (k = 34, n = 2269), encompassing 22 interventions versus placebo, but 18 had 1 RCTs only, and 15 had n <= 20. High heterogeneity (I-2 = 57.1%; tau(2) = 0.0797), and global inconsistency (Q = 32.64; df = 14; p = 0.0032) emerged. No significant differences emerged in acceptability. When restricting analyses to treatments with trials with n > 20 and >1 RCT, only valbenazine (k = 5, SMD = -0.69; 95% CI = -1.00, -0.37) and vitamin E (k = 7, SMD = -0.49; 95% CI = -0.87, -0.11) were superior to placebo. Deutetrabenazine outperformed placebo considering AIMS score and in low risk of bias trials only and with a moderate effect size for 24/36 mg (k = 2, SMD = -0.57/-0.60). Confidence in findings was low for deutetrabenazine and valbenazine, very low for all others. In network 2 (k = 2, n = 63), switch to molindone (k = 1, n = 9) versus switch to haloperidol worsened TD (SMD = 1.68; 95% CI = 0.61,2.76). In network 3 (k = 3, n = 194), antipsychotic wash-out+placebo (k = 1, n = 25) versus TAU+placebo (k = 1, n = 27) worsened TD (SMD = 1.30; 95% CI = 0.36,2.23). Despite large effect sizes for some treatments with very low quality/confidence, when considering higher quality evidence only valbenazine or deutetrabenazine are evidence-based first-line treatments for TD, and potentially vitamin E as second-line. Switching to molindone and antipsychotic washout should be avoided. More treatment options and higher-quality trials are needed.