Analysis dataset and R code for: Marginal efficacy gain versus marginal weight cost of antipsychotic dose escalation – a systematic review and dose–response meta-analysis of fixed-dose randomised controlled trials

This record contains the analysis dataset and R code for the study "Marginal efficacy gain versus marginal weight cost of antipsychotic dose escalation: a systematic review and dose–response meta-analysis of fixed-dose randomised controlled trials" (Shiguo Liang, Jun Liu; The Third Hospital of Heilongjiang Province). The review was prospectively registered in PROSPERO (CRD420261466715). Files1. Supplementary File 2.xlsx – Data extraction workbook and analysis dataset. Sheet "Study level": one row per study record, with study characteristics, RoB 2 domain-level judgments, eligibility decisions and notes. Sheet "Arm level": one row per dose arm. Sheet "Analysis dataset": arm-level data used in the models (olanzapine-equivalent doses, outcome means, standard deviations and sample sizes). The dataset covers 48 trials (60 study records), 150 dose arms (102 active arms) and 16,932 analysed participants.2. Supplementary_File_3.R – R code that reproduces all tables of the main text (Tables 1–4) and the supplementary tables S2–S4 and S6–S9, including the bootstrap confidence intervals of the inflection dose D*. Supplementary Tables S1 (search strategies) and S5 (conversion factors) are fixed text and are not generated by the code. How to reproducePlace both files in the same folder, set that folder as the working directory, and run Supplementary_File_3.R. Supplementary File 2.xlsx is the only input. The code writes one output file, "All_paper_tables.xlsx", with one sheet per table; sheets prefixed "check_" are not for publication and are provided for checking the numbers reported in the text.Required R packages: readxl, dosresmeta (>= 2.0), rms, metafor, openxlsx.Run time: about 30–60 min with 8 cores (1,000 bootstrap replicates); about 2–4 h single-threaded. Reproducibility notesThe random seed is fixed (SEED = 20261012) and L'Ecuyer-CMRG parallel random number streams are used. Point estimates do not depend on the computing environment. Bootstrap confidence intervals may differ in the last decimal place if the number of cores (N_CORES) or the operating system differs from the original run. Original run: [R version x.x.x], [operating system], N_CORES = [n]; package versions: [dosresmeta x.x.x, metafor x.x.x, rms x.x.x].