Table 5.

Expected relative bias and efficiency associated with non-consent, which is assumed to occur at a rate of 15% for individuals with Genotype X and 5% among others. Both Genotype X and the outcomes are assumed to have prevalence of 50%. Odds ratios are for associations between the outcome and intervention assignment.

Ratio of Gene-Specific ORs For Full Cohort	OR Between Intervention And Outcome Among Non-consenters		Results From Analysis of Consenters Only
	Genotype X	Genotype Not X	Observed Ratio of ORs	Standard Error of Observed Ratios of ORs	Relative Bias (%)	Relative Efficiency* (%)
		1.00	1.000	0.126	0.0	89.9
	1.00	2.00	1.039	0.131	3.9	89.9
1.00		4.00	1.077	0.136	7.7	89.9
(Assumed to be 2.00 in genotype X and in others)		1.00	0.892	0.112	−10.8	89.9
	2.00	2.00	0.927	0.117	−7.3	89.9
		4.00	0.960	0.121	−4.0	89.9
		1.00	0.801	0.101	−19.9	89.9
	4.00	2.00	0.832	0.105	−16.8	89.9
		4.00	0.862	0.109	−13.8	89.9
		1.00	4.638	0.594	15.9	89.4
4.00	1.00	2.00	4.872	0.625	21.8	89.2
		4.00	5.058	0.649	26.5	89.0
(Assumed to be 4.00 in genotype X and 1.0 among others)		1.00	4.160	0.531	4.0	89.9
	2.00	2.00	4.328	0.553	8.2	89.7
		4.00	4.492	0.575	12.3	89.6
		1.00	3.723	0.474	−6.9	90.3
	4.00	2.00	3.872	0.494	−3.2	90.1
		4.00	4.018	0.513	0.0	90.0

^*Square of the ratio of standard errors for the log-linear contrast defined by the 3-way interaction involving genotype, intervention, and outcome