Model-based Meta-analysis to Support Decision-making in Clinical Drug Development

Date:  Thursday, May 16, 2013

Registration Fees:
Non-trainee ISoP members: $550
Trainee ISoP members: $150
ISoP non-members: $950

Sponsor: Metrum Research Group

Instructors/Teaching Assistants: Bill Gillespie, Jonathan French

Workshop summary:  This workshop provides an introduction to meta-analysis concepts and methods with a strong focus on model-based meta-analysis (MBMA) of summary data from clinical trials to support decision-making in clinical drug development. Participants will execute guided hands-on examples implemented in R and WinBUGS. Use of other software, e.g., NONMEM, will also be discussed. Participants will learn to write a meta-analysis plan, design a MBMA of clinical trial data to address strategic decisions in a clinical drug development program, and implement such an analysis. Participants will learn to construct a model for the relationship between an efficacy- or safety-related clinical outcome and independent variables such as dose, time and patient characteristics by analysis of summary data from multiple studies, e.g., treatment means and standard deviations, as well as execute and interpret population simulations to support decision-making in clinical drug development. 

Preliminary schedule: Note, specific times are subject to change based on alignment with concurrent ACoP workshops

Start: 0830

  • Introduction
    • Rationale and role of model-based meta-analysis in clinical drug development
      • Why do it?
      • What decisions benefit from meta-analysis and model-based meta-analysis in particular?
    • Motivating examples
  • The systematic review and planning your meta-analysis
    • Analysis plan
    • Database construction
  • Traditional meta-analysis
    • Fixed effects meta-analysis
  • Random effects meta-analysis and meta-regression
    • Measures of heterogeneity
    • What the traditional random effects model is and how it differs from fixed effects
    • Example: Measures of heterogeneity and random effects meta-analysis and forest plots

Break: 1000-1015

  •  
    • Meta-regression
  • Network meta-analysis / mixed treatment comparisons
    • Networks of evidence
    • Models for combining direct and indirect evidence
  • Model-based meta-analysis (MBMA)
    • What is it?
    • Role of MBMA: Objectives not adequately addressed by traditional meta-analysis
  • Modeling sample mean data
    • Example: Dose response model based on sample means

Lunch: 1200-1300

  • Modeling sample standard deviations: why and how
  • Population simulations
    • Simulating probable ranges of population estimands, e.g., population mean, probability of an event, etc.
    • Using simulation results to support decision-making in a competitive market environment
    • Example: Population simulations

Break: 1430-1445

  • Issues arising from analysis of summary data
    • Applying models developed to describe responses in individuals to summary data
    • Analysis of longitudinal data
    • Example: Longitudinal dose-response model based on longitudinal summary data
  • What we didn't cover
    • More detailed treatment of the mathematics and statistics
    • Extensive hands-on training with independent guidance
    • More extensive discussion, diagnosis and treatment of selection bias and missing data
    • Modeling other types of summary statistics, e.g., number or fraction of patients with a particular outcome or that experience an event. number or fraction of patients within each level of an ordinal scale, number of events per patient, or summary statistics for time-to-event measurements
    • Combining summary and individual data
  • Closing discussion

Adjourn: 1700