Table 1.

Statistical Power to Detect Gene × Environment Interaction in 1,000 Case-Parent Trios for the Standard Test and the 2-Step Test Across a Range of Models

Model No.	Model	Standard Test Power	2-Step Test
			Optimal		Power at α1 = 1 × 10−4
			Power	α1
1	Base modela	0.25	0.87	3.7 × 10−5	0.86
	Disease susceptibility locus allele frequency (qA)
2	0.05	0.01	0.15	2.9 × 10−5	0.15
3	0.25	0.54	0.97	5.9 × 10−5	0.97
	Exposure frequency (pE)
4	0.10	0.03	0.28	7.3 × 10−5	0.28
5	0.25	0.23	0.83	5.1 × 10−5	0.83
	Main effect sizes (Rg, Re)
6	1.0, 2.0	0.23	0.66	3.5 × 10−5	0.65
7	2.0, 1.0	0.10	0.97	3.0 × 10−6	0.89
	Negative interaction effect (Rge)
8	0.45	0.08	0.35	1.1 × 10−3	0.29
9	0.35	0.37	0.77	2.0 × 10−3	0.65
	No. of single nucleotide polymorphisms (M)
10	500,000	0.29	0.89	5.4 × 10−5	0.89
11	100,000	0.40	0.93	1.4 × 10−4	0.93
12	1,500	0.74	0.99	2.0 × 10−3	0.95
	Genetic modelb
13	Dominant	0.25	0.77	1.7 × 10−4	0.77
14	Recessive	0.25	0.40	2.4 × 10−2	0.16
	Experiment-wise type I error rate (α)
15	0.01	0.17	0.82	1.5 × 10−5	0.78
16	0.10	0.29	0.89	5.4 × 10−5	0.89

^aThe base model has q_A = 0.15, p_E = 0.5, R_g = R_e = 1.0, R_ge = 2.3, M = 1 million single nucleotide polymorphisms, and experiment-wise α = 0.05. Each additional model varies the indicated parameter.

^bFor the dominant model, R_ge was increased to 2.6. For the recessive model, q_A was increased to 0.43 and R_ge to 2.6. These settings provided the same power as the base model for the standard analysis.