130. Meta-Analysis of Tardive Dyskinesia Rates in Randomized Controlled Studies Comparing First- to Second-Generation Antipsychotics

Background: Tardive dyskinesia (TD) rates with second-generation antipsychotic (SGA) treatment were estimated to be considerably lower compared with first-generation antipsychotics (FGAs). As recent data have questioned this notion, we conducted a meta-analysis on TD incidence during randomized controlled studies. Methods: We conducted an electronic database search without language restriction using (“tardive dyskinesia” OR tardive) AND (antipsychotic*) plus specific names of SGAs. Of 3438 hits (June 30, 2016), we screened 140 full-text articles for incident TD rates during randomized controlled studies comparing SGA to FGA treatment. Of these, 26 randomized studies were eligible for random effects meta-analysis and meta-regression. We extracted data on antipsychotic class-wise TD rates, exposure times, and on TD moderators (mean age and percentage of males in study population, illness duration of study cohort, study region, anticholinergic use, study sponsor, study design, FGA comparator dose, SGA subgroups: risperidone, olanzapine, and clozapine, other). Results: Data on 9157 study participants (mean age = 38.7 years, male = 65.1%, 23 of 26 studies including schizophrenia spectrum disorders; mean study duration 1.6 years) were analyzed. Treatment-emergent TD was observed in 132 of 5937 patients treated with SGAs (2.2% in 4380 person years) and in 203 of 3220 patients treated with FGAs (6.3% in 1982 person years). Annualized TD rates were lower with SGAs relative to FGAs (rate ratio (RR) = 0.37; CI = 0.28–0.48; P < .0001). The dose of the FGA comparator (below vs. above 500 mg chlorpromazine equivalent) did not significantly moderate this difference (Q = 1.13; P = .29). Similarly, the FGA–SGA TD rate ratios did not differ between SGA subgroups (risperidone k = 5; olanzapine k = 6; clozapine = 3; mixed/other: 9) and persisted independently within each subgroup (all comparisons P < .01). Moreover, the segregation of TD rates persisted regardless of the case definition used in the study (Schooler-Kane criteria vs. other TD reporting criteria) and regardless of study design (double-blind vs. open label). However, FGA-SGA TD RRs differed significantly depending on the study sponsor (industry vs. academic; Q = 8.1; P = .004). Nevertheless, RR differences independently persisted within each subgroup (academic studies, k = 9: mean RR = 0.53; CI = 0.38–0.74; P < .0001; industry sponsored studies, k = 14: mean RR = 0.27; CI = 0.20–0.37; P < .001). Subgroup analyses also showed a trend toward heterogeneity of TD rates within the SGAs (Q = 5.3; P = .15). Moderator analyses for age, sex, illness duration, study region, and anticholinergic use were nonsignificant. Acquisition of unpublished data is ongoing and may change the final results. Conclusion: This meta-analysis confirms the lower TD risk with SGAs versus FGAs. Contrasting to earlier suggestions, this advantage was not driven by studies that used high dose FGA comparators.