Applying advanced psychometrics to antipsychotic treatment trials: A protocol comparing effect estimates

Antipsychotic trials in schizophrenia quantify treatment efficacy using clinician-rated scales such as the Positive and Negative Syndrome Scale (PANSS) and Brief Psychiatric Rating Scale (BPRS). If these scales include redundant or weakly informative items, measurement noise can blur drug-comparator differences that may dilute real effect sizes. Modern psychometric methods can refine and re-weight symptom scores that may improve precision and potentially increase detectable treatment effects, although it is unclear which approaches might be best.

To 1) determine best psychometric models for the PANSS and BPRS, and 2) test whether psychometrically-informed scoring of PANSS/BPRS alters estimated antipsychotic effect sizes compared with conventional total-score analyses, and 3) to explore whether any changes differ across symptom domains and by sex.

We will conduct secondary analyses of anonymised individual participant data from randomized Phase II-IV schizophrenia trials accessed via secure data-sharing platforms (YODA Project and Vivli). Item-level PANSS/BPRS data will be harmonised at baseline and outcome. Psychometric models will be derived using confirmatory factor analysis, item response theory, and network analysis. Stability and invariance analyses will be conducted, examining covariates such as age and sex. Antipsychotic efficacy will be estimated using mixed-effects participant-level models, expressed as Cohen's d standardised mean differences (SMDs) for both conventional and psychometrically-informed outcomes. The primary outcome will be the difference in SMDs between conventional and psychometrically-informed effect sizes, summarised across trials.

This study will determine whether best practice psychometric refinement of existing measures meaningfully changes antipsychotic effect size estimates, informing outcome measurement choices and interpretation of trial results.