An efficient distributed learning framework for integrating evidence in clinical research networks

The overarching goal of this proposal is to develop rigorous evidence synthesis methods to answer two fundamental questions from patients: (1) What are all treatment options available to me and their relative benefits and harms? and (2) What are the best treatment options for me and the associated benefits and harms? To answer the first question, we propose a meta-analytical framework to integrate evidence from published literature with studies in distributed research networks, leveraging their advantages to minimize the impacts of reporting biases in published literature. To answer the second question, we propose a novel two-stage meta-regression framework to better identify effect modifiers, which will help in characterizing heterogeneous treatment effects and informing medical decision making.

Motivated by the PCORI-funded PEDSnet consortium and the pragmatic trial COMBINE study, we will use the study of treatment options for pediatric Crohn’s disease as a working example. Stakeholders, including patients, healthcare providers and researchers, will be engaged to prioritize research questions throughout our investigation. Completion of this proposed research will directly benefit patients, families, and other stakeholders in PEDSnet. The proposed project will (1) develop a general framework for combining aggregated results from the published literature with individual patient data derived from a research network using multivariate multiple treatments meta-regression analysis, (2) propose and evaluate a powerful two-stage hypothesis testing procedure for identification of treatment-covariate interactions, and (3) evaluate the proposed methods in pragmatic trials of treatments for pediatric Crohn’s disease within the PEDSnet consortium.

The long-term objective of this research is to provide a robust statistical framework and associated software for identifying subpopulations of patients who benefit most from a treatment in the setting where multiple treatment options are available and multiple clinical outcomes (e.g., efficacy, patient outcomes, and safety) are relevant.

This project has the potential to enhance the quality of delivering (1) a full picture of relevant medical evidence on all treatment options to patients and stakeholders (aim 1), and (2) tailored evidences on the full spectrum of treatment options and outcomes for a patient in a certain subpopulation (aim 2). Our implementation of these methods within the PEDSnet consortium will provide timely information on an important condition (pediatric Crohn’s disease) for the pediatric population in the context of a multisite network.