Two new methods to fit models for network meta-analysis with random inconsistency effects

Table 1 Results for the prostate cancer data using importance sampling, MCMC in OpenBUGS and REML estimation. Estimates are log odds ratios

		Parameter	Estimate	SD	MCSE	2.5 %	50 %	97.5 %
Importance sampling	B: Prostatectomy	δ ^AB	-0.25	0.19	0.0015	-0.63	-0.25	0.11
	C: CR	δ ^AC	-0.16	0.34	0.0033	-0.82	-0.15	0.50
	D: CR-HF	δ ^AD	-0.32	0.41	0.0038	-1.13	-0.32	0.49
	E: CF-LD	δ ^AE	-0.31	0.30	0.0030	-0.90	-0.31	0.27
	F: CF-HD	δ ^AF	-0.35	0.35	0.0033	-1.02	-0.35	0.33
	G: CF-HF-LD	δ ^AG	-0.66	0.87	0.0085	-2.34	-0.67	1.07
	H: Cryotherapy	δ ^AH	-0.27	0.53	0.0051	-1.32	-0.29	0.80
	Heterogeneity	\(\tau ^{2}_{\beta }\)	0.02	0.02	0.0002	0.00	0.01	0.08
	Inconsistency	\(\tau ^{2}_{\omega }\)	0.02	0.04	0.0002	0.00	0.01	0.11
MCMC	B: Prostatectomy	δ ^AB	-0.25	0.19	0.0011	-0.63	-0.24	0.12
	C: CR	δ ^AC	-0.16	0.34	0.0036	-0.83	-0.16	0.51
	D: CR-HF	δ ^AD	-0.32	0.42	0.0039	-1.14	-0.32	0.50
	E: CF-LD	δ ^AE	-0.31	0.30	0.0030	-0.90	-0.31	0.28
	F: CF-HD	δ ^AF	-0.35	0.35	0.0036	-1.03	-0.35	0.34
	G: CF-HF-LD	δ ^AG	-0.63	0.86	0.0100	-2.33	-0.64	1.06
	H: Cryotherapy	δ ^AH	-0.27	0.54	0.0048	-1.32	-0.27	0.78
	Heterogeneity	\(\tau ^{2}_{\beta }\)	0.02	0.02	0.0001	0.00	0.01	0.08
	Inconsistency	\(\tau ^{2}_{\omega }\)	0.02	0.04	0.0003	0.00	0.01	0.11
		Parameter	Estimate	SE	CI lower	CI upper	p-value
REML estimation	B: Prostatectomy	δ ^AB	-0.22	0.11	-0.43	-0.01	0.04
	C: CR-CF	δ ^AC	-0.17	0.28	-0.72	0.39	0.56
	D: CR-HF	δ ^AD	-0.33	0.32	-0.96	0.31	0.31
	E: CF-LD	δ ^AE	-0.32	0.25	-0.80	0.17	0.21
	F: CF-HD	δ ^AF	-0.35	0.28	-0.90	0.19	0.20
	G: CF-HF-LD	δ ^AG	-0.65	0.83	-2.27	0.97	0.43
	H: Cryotherapy	δ ^AH	-0.28	0.47	-1.20	0.65	0.56
	Heterogeneity	\(\tau ^{2}_{\beta }\)	0.00	–	0.00	0.07	–
	Inconsistency	\(\tau ^{2}_{\omega }\)	0.00	–	0.00	0.62	–

CR=conventional radiotherapy; HF=hypofractionated; CF=3D conformal; LD=low dose; HD=high dose

ISSN: 1471-2288