The use of continuous data versus binary data in MTC models: a case study in rheumatoid arthritis

Area of Science:

• Health Economics
• Biostatistics
• Clinical Epidemiology

Background:

• Relative treatment efficacy estimation is vital for healthcare decisions.
• Bayesian mixed treatment comparison (MTC) models are powerful tools for estimating efficacy when direct comparative data is scarce.
• Continuous outcome measures offer advantages over binary measures in MTC models.

Purpose of the Study:

• To illustrate the advantages of using continuous outcome measures versus binary outcome measures in Bayesian MTC models.
• To compare the power of continuous and binary outcome measures in detecting treatment differences.
• To assess the impact of cut-off points on relative efficacy estimates.

Main Methods:

• A case study in rheumatoid arthritis was used to fit a Bayesian MTC model.
• The model estimated the relative efficacy of five anti-TNF agents using both continuous and binary versions of the Health Assessment Questionnaire (HAQ) improvement and the American College of Rheumatology (ACR) response measures.
• Sixteen randomized controlled trials were included in the analysis.

Main Results:

• Differences between treatments detected using binary outcome measures were subsets of those detected using underlying continuous effects for both HAQ and ACR measures.
• Continuous outcome measures demonstrated greater power in detecting differences between anti-TNF agents.

Conclusions:

• Transforming continuous data into binary measures leads to a loss of statistical power in MTC models.
• Continuous outcome measures are more sensitive to change and thus preferable for detecting treatment differences.
• The selection of cut-off points for binary measures significantly impacts relative efficacy estimates.