Table 3.

Results from Testing for an Excess of Significant Findings in Meta-Analyses Included in the AlzGene Database^a

	No. of Studies	No. of Meta-Analyses	Summary Effect Point Estimate				Sensitivity Analysis 1
			Ob	Eb	P Valuec	O:E	E	P Valuec	O:E
All meta-analyses	1,348	175	235	164.8	<10−6	1.61	90.4	<10−6	2.60
I2 ≤ 50%	770	114	85	86.7	0.91	0.98	48.8	10−6	1.74
I2 > 50%	578	61	150	78.14	<10−6	1.92	41.6	<10−6	3.61
Statistically significant associations	241	40	69	74.40	0.49	0.93	30.3	<10−6	2.28
Non-statistically significant associations	1,107	135	166	90.40	<10−6	1.84	60.1	<10−6	2.76
I2 ≤ 50% and significant associations	175	35	40	46.4	0.30	0.86	18.7	4 × 10−6	2.13
I2 ≤ 50% and nonsignificant associations	595	79	45	40.2	0.41	1.12	30.1	0.008	1.50
I2 > 50% and significant associations	66	5	29	28.0	0.80	1.04	11	10−6	2.64
I2 > 50% and nonsignificant associations	512	56	121	50.2	<10−6	2.41	30.0	<10−6	4.03

^aThe main results assume that the plausible effect in each meta-analysis is given by the summary effect (odds ratio) of the meta-analysis. Also shown are results from evaluations using half of the log odds ratio (“Sensitivity Analysis 1”) as the plausible effect.

^bE, expected number of studies with nominally statistically significant results (P = 0.05 threshold) included in the meta-analyses; O, observed number of significant studies.

^cP value for the difference between O and E (i.e., O − E).