Study of Time-Course Relationship Using Model-Based Network Meta-Analysis: A Linear Time-Course Model

OBJECTIVES: Network Meta-Analysis (NMA) is used to generate evidence from multiple studies assessing multiple treatments but lacks efficiency when studies report findings at different follow-up times. MBNMA allows the inclusion of studies with different follow-up times and therefore the possibility of including clinical trials from earlier research, and key information on treatment efficacy alongside testing the inconsistency between direct and indirect evidence within the network. The objective is to perform NMA on studies having results reported at multiple time points.

METHODS: The Linear time-course MBNMA has been used by considering that all studies with different treatment estimates have the same true effect and assume consistency between direct and indirect evidence. MBNMA can include different parametric time-course relationship methods such as Log-linear time-course MBNMA, Piecewise linear time-course MBNMA with a knot at some time point, and Emax time course MBNMA, Emax time and course MBNMA with two parameters, B-spline time-course MBNMA with a knot at 0.2 times the max follow-up, etc. The simulated data has been used to perform the Bayesian linear time-course MBNMA on the continuous summary outcome.

RESULTS: The presented method generated a network of comparisons across multiple treatments over different time points. The posterior median with 95% credible intervals for the change in efficacy for each treatment versus comparator of interest was calculated. The posterior median and 95% credible intervals calculated for comparative effectiveness synthesis were: (1) treatment B vs A: -0.87 (-2.32, 0.38), (2) treatment C vs A: -0.74 (-2.08, 0.49), (3) treatment D vs A: -0.64 (-1.95, 0.62), and (4) treatment E vs A: -0.59 (-1.90, 0.77).

CONCLUSIONS: MBNMA works efficiently and effectively on data sets including multiple treatments at varying follow-up time points and is statistically robust for synthesizing direct and indirect evidence to estimate relative effects- supporting its use in evidence synthesis.