A Single, Interpretable Vocal Biomarker for Enriching Antipsychotic Clinical Trials

Speech latency, an objective measure of verbal response time, is sensitive to cognitive, social, and motivational factors and can be measured using recordings of psychiatric interviews. Emerging evidence suggests that it can be used to enrich participants in antidepressant clinical trials, resulting in nearly double the drug-placebo effect at half the sample size. Here, we evaluated whether enrichment of participants using speech latency ratios, a variant of speech latency accounting for duration of response, improved outcomes in a clinical trial of schizophrenia.

Audio recordings from psychiatric interviews in a phase 3, 28-day, randomized, double-blind, placebo-controlled, fixed-dose, multicenter, parallel-group monotherapy study of brilaroxazine (RP5063) were examined. There were 2320 recordings for 406 participants with acute psychosis from 3 countries representing 8 languages.

Speech latency ratios showed excellent internal consistency, good temporal stability, and minimal convergence with potential confounders. Patients with (vs. without) moderate-to-severe negative symptoms showed longer speech latencies, with large effect sizes observed in nearly every country and language. A single speech latency ratio value (area under the curve = 0.71) identified 180 and 228 participants as being vocal biomarker negative (VBM neg; unremarkable speech) and VBM positive (VBM pos; relatively high ratios), respectively. Brilaroxazine showed statistically significant outcomes versus placebo from baseline to end of treatment in the VBM-pos group for nearly every outcome measure despite having fewer participants. Treatment effects for VBM-pos patients were large for nearly every outcome measure.

Speech latency ratios are a face-valid, objective biomarker derived from standard clinical assessments. As an enrichment tool, it reduced sample size needs and enhanced outcomes. This has important implications for reducing clinical trial costs and burden.