Power comparison of different methods to detect genetic effects and gene-environment interactions

BMC Proceedings

Table 2 Power and estimates of interaction coefficients of the four tests over 100 replicates

	Power (%)^a			Average interaction coefficients [95% CI]
Test	G+I	G	I	I_Bb	I_BB
Mean interaction test	6	8	12	-	-
Log-linear-modeling^b	87	78	53	1.33 [1.03–1.71]	1.72 [1.13–1.83]
Case-control^b	98	95	69	1.39 [0.97–1.96]	1.88 [1.08–3.10]
Case-only^b	-	-	95	1.39 [1.05–1.72]	1.86 [1.39–2.96]
Log-linear-modeling^c	33	23	20	1.35 [0.79–2.15]	1.79 [0.82–3.68]
Case-control^d	79	68	42	1.41 [0.85–2.09]	1.96 [0.99–3.37]

^a G+I, genetic effect accounting for interaction; G, genetic effect not accounting for interaction; I, G × E interaction.
^b Samples of 1500 families were used corresponding to 4500 (1500 triads), 3000 (1500 cases and 1500 controls), and 1500 genotyped individuals for the log-linear-modeling, the case-control and the case-only design, respectively.
^c Samples of 500 triads are considered corresponding to 1500 genotyped individuals.
^dSamples of 750 cases and 750 controls are considered here to limit the number of genotyped individuals to 1500.

ISSN: 1753-6561