General Information
Conflict of Interest
Request Clinical Trials
Associated Trial(s):- NCT00334126 - A Randomized, Double-blind, Placebo-controlled, Parallel Group Study to Evaluate the Efficacy and Safety of Paliperidone ER Compared to Quetiapine in Subjects With an Acute Exacerbation of Schizophrenia
- NCT00083668 - A Randomized, Double-blind, Placebo- and Active-controlled, Parallel-group, Dose-response Study to Evaluate the Efficacy and Safety of 3 Fixed Dosages of Paliperidone Extended Release (ER) Tablets and Olanzapine, With Open-label Extension, in the Treatment of Patients With Schizophrenia
- NCT00078039 - Trial Evaluating Three Fixed Dosages of Paliperidone Extended-Release (ER) Tablets and Olanzapine in the Treatment of Patients With Schizophrenia
- NCT00085748 - A Randomized, 6-Week Double-Blind, Placebo-Controlled Study With an Optional 24-Week Open-Label Extension to Evaluate the Safety and Tolerability of Flexible Doses of Paliperidone Extended Release in the Treatment of Geriatric Patients With Schizophrenia
Request Clinical Trials
Data Request Status
Status: OngoingResearch Proposal
Project Title: Bayesian multisource borrowing with baseline covariate matching in paliperidone schizophrenia trials
Scientific Abstract:
Background: Participant-level evidence synthesis across related clinical trials can improve precision, but inappropriate pooling may bias results when study populations differ. Methods that account for baseline heterogeneity before borrowing information are needed.
Objective: To develop and evaluate a Bayesian framework for selective information borrowing across related schizophrenia trials using baseline covariate matching/alignment prior to outcome modeling.
Study Design: Secondary methodological analysis of four paliperidone ER trials in adults with schizophrenia.
Participants: Adult patients with schizophrenia enrolled in the selected trials. The analysis will include participants from the 9 mg/day and placebo arms with available baseline covariates and PANSS outcome data.
Primary and Secondary Outcome Measure(s): The primary outcome will be change in PANSS total score from baseline to the primary endpoint visit.
Statistical Analysis: Baseline demographic and clinical covariates will be matched across trials using developed Bayesian approach. A participant-level statistical approach will be used to identify comparable subgroups across studies and to guide selective Bayesian borrowing of treatment information across trials. Performance will be assessed through estimation precision, robustness to between-trial heterogeneity, and comparison with models that assume full exchangeability across all studies.
Brief Project Background and Statement of Project Significance:
There is growing interest in combining participant-level data from multiple clinical trials to improve estimation of treatment effects and better understand heterogeneity across studies. In practice, however, even trials evaluating the same drug and outcome may differ in baseline symptom severity, demographic composition, prior treatment history, geographic setting, and other clinically important factors. When these differences are ignored, full pooling may produce misleading conclusions. This is especially relevant in psychiatric trials, where treatment response may vary across patient subgroups and where symptom-based outcomes such as PANSS are continuous and highly sensitive to baseline case mix.
A substantial methodological literature has developed for borrowing information across studies, including hierarchical Bayesian approaches, commensurate priors, power priors, meta-analytic predictive methods, and multisource exchangeability models. These approaches provide useful tools for evidence synthesis, but many rely primarily on outcome-level similarity and do not directly address whether participants across studies are comparable in their joint baseline covariate distributions. This creates a need for methods that use participant-level information to determine where borrowing is appropriate and where it should be limited.
The proposed research addresses this gap by developing and evaluating a Bayesian framework that first aligns participants across trials based on baseline characteristics and then selectively borrows information across comparable studies or subgroups. The requested paliperidone ER trials are well suited for this work because they are related placebo-controlled studies in adult schizophrenia with a common continuous clinical outcome and a common treatment contrast that can be harmonized to 9 mg/day versus placebo. This setting provides an opportunity to study how participant-level heterogeneity affects borrowing across trials and to assess whether covariate-informed selective borrowing yields more reliable inference than approaches that assume all studies are fully exchangeable.
The significance of this work is methodological rather than confirmatory. The goal is not simply to re-estimate a known treatment effect, but to develop and evaluate improved statistical tools for integrating evidence across related clinical trials while accounting for heterogeneity in enrolled populations. These methods may be useful more broadly in clinical trial analysis, evidence synthesis, and future trial design.
Specific Aims of the Project:
The objective of this project is to develop and evaluate participant-level Bayesian methods for selectively borrowing information across related schizophrenia clinical trials while accounting for differences in study populations.
Aim 1: Compare the proposed approach with existing approaches, with respect to robustness, precision, and sensitivity to between-trial heterogeneity through various simulation settings.
Aim 2: Use the schizophrenia clinical trials as an application example. Set a current trial and treat the other three trials as external studies. Match the populations using proposed approach. Focus on a common comparison of 9 mg/day versus placebo and a common continuous outcome based on PANSS change.
The main hypotheses are that:
1. The selected trials will exhibit meaningful but incomplete overlap in baseline covariate distributions;
2. Full pooling across all studies will not be optimal
3. A covariate-informed selective borrowing approach will provide more reliable inference than methods that ignore between-trial heterogeneity.
Study Design: Meta-analysis (analysis of multiple trials together)
What is the purpose of the analysis being proposed? Please select all that apply.: Participant-level data meta-analysis Meta-analysis using only data from the YODA Project Develop or refine statistical methods Research on clinical trial methods
Software Used: R, RStudio
Data Source and Inclusion/Exclusion Criteria to be used to define the patient sample for your study:
The proposed analysis will use participant-level data from randomized, double-blind, placebo-controlled paliperidone extended-release (ER) trials in adult patients with schizophrenia made available through the YODA Project. The primary analytic sample will be restricted to participants randomized to the paliperidone ER 9 mg/day arm or the placebo arm in each selected trial, in order to make the treatment comparison fair across studies. Participants from other active treatment arms, active comparator arms, or flexible-dose arms will be excluded from the primary analysis.
Eligible participants will include adults enrolled in the selected acute-treatment schizophrenia trials with a diagnosis of schizophrenia and available baseline covariates and PANSS outcome data. Additional inclusion criteria for the analysis will be availability of randomized treatment assignment, baseline PANSS total score, and at least one post-baseline PANSS assessment suitable for defining the primary endpoint. Participants with missing treatment assignment, missing baseline PANSS, or no evaluable post-baseline PANSS data will be excluded from the primary analysis. No data from studies outside the YODA Project will be used.
Primary and Secondary Outcome Measure(s) and how they will be categorized/defined for your study:
The primary outcome will be change in PANSS total score from baseline to the primary endpoint visit (at the end of 6 weeks of treatment phase), analyzed as a continuous outcome. No secondary outcome measures are planned for the primary analysis.
Main Predictor/Independent Variable and how it will be categorized/defined for your study:
The main independent variable will be randomized treatment assignment, coded as a binary indicator comparing paliperidone ER 9 mg/day versus placebo. For the primary analysis, treatment will be harmonized across trials by retaining only the 9 mg/day active-treatment arm and the placebo arm. Trial membership will also be included as an important study-level indicator in order to evaluate heterogeneity and selective borrowing across studies.
Other Variables of Interest that will be used in your analysis and how they will be categorized/defined for your study: Additional baseline demographic and clinical variables will be used in the participant-level matching/alignment stage of the analysis. These may include age, sex, race/ethnicity, baseline PANSS total score, illness duration, and other shared baseline characteristics available across the selected trials. These variables will be harmonized across studies and used to assess comparability of enrolled populations prior to selective borrowing
Statistical Analysis Plan:
The analysis will use participant-level data from the four selected randomized, double-blind, placebo-controlled paliperidone ER schizophrenia trials. The primary analysis population will include only participants randomized to paliperidone ER 9 mg/day or placebo. Descriptive analyses will first be used to summarize baseline demographic and clinical characteristics within each trial and to assess comparability across studies. Variables to be summarized may include age, sex, race/ethnicity, baseline PANSS total score, illness duration, and other shared baseline characteristics.
The primary outcome will be change in PANSS total score from baseline to the end of treatment phase, analyzed as a continuous outcome. Missing data patterns will be examined, and analyses will be limited to participants with evaluable baseline and endpoint data required for the primary outcome definition.
The main methodological analysis will use a Bayesian participant-level framework to evaluate selective information borrowing across randomized trials. The overall goal is to borrow information from external trials only when the enrolled participants appear sufficiently comparable to those in the target trial, while limiting borrowing when there is evidence of population or treatment-effect heterogeneity.
In the first stage, baseline covariates will be used to compare the subject populations across trials. These covariates may include demographic and clinical characteristics measured prior to treatment assignment, such as age, sex, race/ethnicity, baseline PANSS score and other available clinical variables. We will use a proposed Bayesian non-parametric model to identify latent participant subgroups shared across trials. Subjects from external studies that do not have a matching group in the primary study will be discarded from further analysis.
In the second stage, within each identified subgroup, we will evaluate whether treatment-effect information from the external trials is exchangeable with the corresponding information from the primary trial. The continuous PANSS outcome will be analyzed as the main endpoint. The model will allow trial-specific treatment effects while also permitting borrowing across trials when the data support exchangeability. This adaptive borrowing will be implemented using a multisource exchangeability model (MEM), in which each external trial may be classified as exchangeable or non-exchangeable with the primary trial. Posterior model probabilities will be used to determine the degree of borrowing from each external trial.
Finally, the group-specific treatment effects will be integrated in a weighted sum way to get the treatment effect in the primary study after borrowing. The primary outputs from the analysis will include estimates of the treatment effect, posterior credible intervals, posterior exchangeability probabilities for each external trial, subgroup-specific borrowing summaries, and measures of effective supplemental sample size.
Primary analyses will focus on estimation of the treatment effect for paliperidone ER 9 mg/day versus placebo and on evaluation of between-trial heterogeneity. All analyses will be conducted using reproducible statistical programming, and results will be reported in aggregate form.
Narrative Summary: This study will use data from several past clinical trials in people with schizophrenia to test better ways of comparing and combining study results. Different trials often include patients with different starting symptoms and health characteristics, which can make results hard to compare fairly. This project will examine whether accounting for these differences first leads to more accurate and reliable conclusions.
Project Timeline: The methodological framework has already been developed and is currently being evaluated through simulation studies. Following data access approval, the anticipated project start date for the real-data application is within two weeks of data release. Data cleaning and construction of the analytic dataset are expected to be completed within 3 weeks after access is granted. Primary analyses are expected to be completed within two weeks after access is granted. A manuscript draft is anticipated within 3 months after access is granted, with first submission to a peer-reviewed journal expected within approximately 5 to 6 months. Results and any resulting publications or presentations will be reported back to the YODA Project in accordance with YODA requirements. If additional time is needed for revision, manuscript resubmission, or follow-up analyses, an extension may be requested if appropriate.
Dissemination Plan: The anticipated primary product of this research is a peer-reviewed methodological manuscript . Findings may also be presented at academic conferences in statistics, biostatistics, or clinical trial methodology. The target audience includes statisticians, biostatisticians, clinical trial methodologists, and researchers interested in evidence synthesis and trial design. Potential target journals include Biometrics, Statistics in Medicine, and JRSS Series C (Applied Statistics), depending on the final scope and emphasis of the manuscript. Results will be reported only in aggregate form, and any dissemination will comply with the YODA data use agreement and all applicable publication and reporting requirements.
Bibliography:
Ibrahim, J. G. and Chen, M.-H. (2000). Power prior distributions for regression models. Statistical Science pages 46--60.
Hobbs, B. P., Carlin, B. P., Mandrekar, S. J., and Sargent, D. J. (2011). Hierarchical commensurate and power prior models for adaptive incorporation of historical information in clinical trials. Biometrics 67, 1047--1056.
Kaizer, A. M., Hobbs, B. P., and Koopmeiners, J. S. (2018). A multi-source adaptive platform design for testing sequential combinatorial therapeutic strategies. Biometrics 74, 1082--1094.