Network meta-analysis of longitudinal data using fractional polynomials.

Area of Science:

• Biostatistics
• Clinical Epidemiology
• Health Research Methodology

Background:

• Traditional network meta-analysis (NMA) often synthesizes a single treatment effect measure per study.
• Many randomized controlled trials (RCTs) report outcomes at multiple time points, presenting analytical challenges.
• Existing methods struggle to incorporate diverse time points across studies in NMA.

Purpose of the Study:

• To present a novel network meta-analysis method for simultaneous analysis of outcomes at multiple time points.
• To model the development of outcomes over time using fractional polynomials.
• To synthesize treatment effect parameters across studies using Bayesian NMA.

Main Methods:

• Utilized fractional polynomials to model time-dependent treatment effects within RCTs.
• Employed Bayesian network meta-analysis to synthesize differences in polynomial parameters across studies.
• Applied fixed and random effects second-order fractional polynomials to a case study on knee osteoarthritis RCTs.

Main Results:

• Demonstrated the feasibility of simultaneously analyzing outcomes from multiple time points in NMA.
• Successfully synthesized treatment effects even when follow-up times varied across studies.
• The proposed fractional polynomial models proved effective for repeated measures in NMA.

Conclusions:

• The developed NMA models offer a valuable extension for synthesizing repeated measures data.
• This method accommodates studies with differing time points, enhancing data utilization.
• The approach is particularly useful for complex RCT data with multiple outcome assessments.