Table 5.

Impact of selective searching on results of meta-analysis from infectious diseases

		Original meta-analysis				Meta-analysis based on studies identified by selective search
Case	Summary measure	Number of trials	Effect size	Lower CI	Upper CI	Number of trials	Effect size	Lower CI	Upper CI	% of trials identified	% change in effect size
Medline + Embase (changes in 9 of 31 meta-analyses)
1a, b	MD	35	−24.76	−33.61	−15.91	30	−22.22	−31.85	−12.6	86	10.3
2a, b	RR	5	0.41	0.11	1.64	3	NE	-	-	60	-
3	RR	4	0.48	0.35	0.65	3	0.57	0.29	1.09	75	15.8
4a, b	RR	21	0.53	0.49	0.57	16	0.51	0.46	0.56	76	3.8
5a, b	MH OR	11	0.59	0.45	0.79	7	0.61	0.41	0.89	64	3.3
6a	RR	3	0.62	0.19	2.04	2	0.56	0.14	2.14	67	9.7
7a, b	MH OR	10	0.68	0.49	0.95	8	0.59	0.41	0.86	80	13.2
8a	RR	8	0.71	0.62	0.80	6	0.72	0.64	0.82	75	1.4
9a	RR	13	0.99	0.90	1.09	11	0.95	0.86	1.05	85	4.0
Medline + BIOSIS (changes in 8 of 31 meta-analyses; data below show cases where results were different from Medline + Embase results aboveb)
1	MD	35	−24.76	−33.61	−15.91	29	−23.18	−32.94	−13.42	83	6.4
2	RR	5	0.41	0.11	1.64	3	NE	-	-	60	-
4	RR	21	0.53	0.49	0.57	17	0.49	0.44	0.54	81	7.5
5	MH OR	11	0.59	0.45	0.79	8	0.64	0.48	0.87	73	7.8
7	MH OR	10	0.68	0.49	0.95	8	0.54	0.34	0.86	80	20.6

CI confidence interval, MD mean difference, MH OR Mantel Haenszel odds ratio, NE not estimable (all study groups had zero counts); RR risk ratio

^acases where results also changed for Medline + BIOSIS; ^bcases where results were different between Medline + BIOSIS and Medline + Embase analyses