Meta-analysis of binary outcomes via generalized linear mixed models: a simulation study

Table 3 Meta-analysis of simulated data

Model	Method	Optimizer	Hetero	LOR	L	U	Length	OR
			geneity				of CI
GLMM	FIM		0.3106	1.5477	1.0513	2.0442	0.9929	4.700646
GLMM	RIM		0.3021	1.5446	1.0548	2.0344	0.9796	4.686097
GLMM	NCHGN	“optim”	1169.0647	35.6833	26.9652	44.4014	17.4362	3.14∗10¹⁵
GLMM	NCHGN	“nlminb”	0.3113	1.5472	1.0502	2.0442	0.994	4.698297
GLMM	NCHGN	“bobyqa”	0.3113	1.5472	1.0502	2.0442	0.994	4.698297
GLMM	NCHGN	“newuoa”	0.3113	1.5472	1.0502	2.0442	0.994	4.698297
GLMM	NCHGN	“uobyqa”	0.3113	1.5472	1.0502	2.0442	0.994	4.698297
GLMM	ABNM		0.0160	0.6177	0.4943	0.7411	0.2468	1.854657
FEM				1.4540	1.3671	1.5409	0.1738	4.280201
REM	DL		0.3159	1.5469	1.0463	2.0474	1.0011	4.696887
REM	REML		0.3921	1.5476	0.9916	2.1035	1.1119	4.700176

Estimates and confidence intervals (CIs) for the heterogeneity parameter τ², for the overall log-odds-ratio (LOR) and for the overall odds ratios (OR); GLMM is the generalized linear mixed model, REM is the random-effects model and FEM is the fixed-effect model. L and U are the lower and upper limits of the respective 95% confidence intervals

ISSN: 1471-2288