A comparison of meta-analytic methods for synthesizing evidence from explanatory and pragmatic trials

Table 2 Comparison of meta-analytic methods for estimating overall intervention effect in trials to reduce obesity in children aged 6–12 years

Meta-analytic methods	Pooled effect size	95%CI	τ ²	BIC
Random-effects meta-analysis^a	− 0.15	[− 0.23, − 0.08]^*	0.029	8.08
Stratified meta-analysis
Explanatory trials^b	− 0.32	[− 0.88, 0.33]	0.209	− 11.97
Pragmatic trials^b	− 0.12	[− 0.19, − 0.06]^*	0.018	− 9.74
Meta-regression methods
REMR controlling for PRECIS	− 0.79	[− 1.26, − 0.29]^*	0.020	− 5.33
Mixture REMR controlling for PRECIS	− 1.01	[− 1.52, − 0.43]^*	0.010	− 10.39

τ² between-trial variance, BIC Bayesian information criterion, REMR random effects mixture regression, 95%CI 95% confidence interval
^*p < 0.05
^aUnadjusted model

ISSN: 2046-4053