A comparison of the statistical performance of different meta-analysis models for the synthesis of subgroup effects from randomized clinical trials

Table 1 Varying and fixed parameters of simulated datasets that were used to compare the performance measures of different models

Parameters	Value in simulation	Statistical notation in the linear predictor used for simulation of dataset^b
Varying-parameters
Number of trials	6, 10, 16, 20, 26, 30, 40, 50	–
Magnitude of the interaction effect^a	0, −0.01	beta.inter
Between-trial variance of the interaction effect^a	0, 0.005, 0.05, 0.5	V ₂
Fixed-parameters
Number of patients included in each trial	between 30 and 199 patients in small trials, and 200 and 400 patients in large trials	j
Ratio of small to large trials	1:1	–
Number of trial arms	2	treat
Ratio of randomization	1:1	–
Magnitude of the treatment effect^a	−1	beta.treat
Between-trial variance of the treatment effect^a	0.063	V ₁
Effect of age on the outcome	0.05	age
Random intercept variance	0.25	σ ²
Random slope variance	0.04	τ ²
Correlation between random intercept and random slope variances	0.8	ρ
Age of individual patients	N(mean. age, variance. age)	age
Mean age of patients within trials	N(67,20.25)	mean. age
Variance of age within trials	N(49, 1)	variance. age

ISSN: 1471-2288