Influence of pilot and small trials in meta-analyses of behavioral interventions: a meta-epidemiological study

Table 1 Comparison between originally reported meta-analytical effect sizes and the re-estimated effect sizes based on trial classifications

		Originally estimated summary standardized mean effect size^b			Re-estimated summary standardized mean effect size			Absolute mean difference			Percent difference in standardized mean effect size
Comparison	k	50th	(25th,	75th)	50th	(25th,	75th)	50th	(25th,	75th)	50th	(25th,	75th)
Random effects models
N > 100^a	113	0.21	(0.06,	0.40)	0.12	(0.05,	0.27)	0.05	(0.01,	0.15)	35%	(10%,	86%)
N > 370^a	84	0.16	(0.05,	0.34)	0.08	(0.02,	0.18)	0.06	(0.02,	0.18)	52%	(20%,	130%)
Excluding pilot/feasibility	83	0.21	(0.06,	0.36)	0.22	(0.06,	0.35)	0.02	(0.01,	0.05)	13%	(6%,	38%)
Excluding N ≤ 100	83	0.24	(0.07,	0.40)	0.16	(0.06,	0.32)	0.04	(0.01,	0.13)	21%	(6%,	66%)
Fixed effects models
N > 100^a	113	0.14	(0.04,	0.28)	0.10	(0.03,	0.22)	0.02	(0.01,	0.06)	24%	(6%,	54%)
N > 370^a	84	0.11	(0.03,	0.20)	0.07	(0.01,	0.15)	0.03	(0.01,	0.08)	31%	(8%,	79%)
Excluding pilot/feasibility	83	0.15	(0.04,	0.26)	0.14	(0.04,	0.26)	0.01	(0.00,	0.03)	11%	(4%,	27%)
Excluding N ≤ 100	83	0.16	(0.05,	0.29)	0.13	(0.05,	0.24)	0.02	(0.01,	0.05)	14%	(4%,	47%)

^aEstimates exclude both pilot/feasibility and N ≤ 100 studies; k = number of summary effect sizes estimated
^bDifferences in the originally estimated effect size across comparisons is because not all of the included meta-analyses contained individual studies that covered all four of the mutually exclusive categories the studies were classified into

ISSN: 2046-4053