Accuracy of the Berger-Exner test for detecting third-order selection bias in randomised controlled trials: a simulation-based investigation

Table 2 Inflation of effect size due to 3 ^rd order selection bias

Parameter setting	Scenario 1*						Scenario 2**						ROR/OE%
Parameter setting	Group A		Group B		OR	95% CI	Group A		Group B		OR	95% CI	ROR/OE%
	n	N	n	N			n	N	n	N
TF = 10% - N = 120	19950	21000	19950	21000	1.00	0.92 – 1.09	18912	21000	20988	21000	0.04	0.03 – 0.05	0.04/96%
TF = 10% - N = 240	31500	42000	31500	42000	1.00	0.94 – 1.06	37824	42000	41976	42000	0.02	0.02 – 0.03	0.02/98%
TF = 10% - N = 480	63000	84000	63000	84000	1.00	0.96 – 1.04	75624	84000	83976	84000	0.01	0.01 – 0.01	0.01/99%
TF = 20% - N = 120	18900	21000	18900	21000	1.00	0.94 – 1.07	17117	21000	20681	21000	0.09	0.08 – 0.10	0.09/91%
TF = 20% - N = 240	37800	42000	37800	42000	1.00	0.96 – 1.07	34242	42000	41343	42000	0.08	0.07 – 0.09	0.08/92%
TF = 20% - N = 480	75600	84000	75600	84000	1.00	0.97 – 1.03	68493	84000	82717	84000	0.08	0.07 – 0.08	0.08/92%
TF = 50% - N = 120	15750	21000	15750	21000	1.00	0.96 – 1.05	13417	21000	18093	21000	0.29	0.27 – 0.30	0.29/71%
TF = 50% - N = 240	39900	42000	39900	42000	1.00	0.97 – 1.03	26858	42000	36150	42000	0.29	0.28 – 0.30	0.29/71%
TF = 50% - N = 480	79800	84000	79800	84000	1.00	0.98 – 1.02	53812	84000	72178	84000	0.29	0.28 – 0.30	0.29/71%

TF = Trait frequency; N = Subject number; OR = Pooled Odds ratio; CI = Confidence interval; n = Pooled number of subjects with failed intervention (Y = 0); N = Pooled total umber of subjects;
ROR = Ratio of Odds ratios; OE% = Percent overestimation/inflation of true effect size.
*No bias; **3^rd order selection bias.

ISSN: 1471-2288