The use of complete-case and multiple imputation-based analyses in molecular epidemiology studies that assess interaction effects

Desai, Manisha; Esserman, Denise A; Gammon, Marilie D; Terry, Mary B

doi:10.1186/1742-5573-8-5

Epidemiologic Perspectives & Innovations

Table 4 Impact of Nature of Missingness With Auxiliary Relationship Based on Data from LIBSCP.

From: The use of complete-case and multiple imputation-based analyses in molecular epidemiology studies that assess interaction effects

a. Subjects Missing Data on Genotype Under Conditions 1 and 3
Scenario	Variable	Method	Mean β	Mean SE	Mean Bias	MSE	RelMSE	Coverage
J: Condition 1	Effect of Fast Metabolizing Genotype For Unexposed (True effect = 1.0)	Full	1.001	0.155	0.001	0.025	0.738	95.1 (93.8,96.4)
		CC	0.999	0.182	-0.001	0.033	1.000	94.3 (92.9,95.7)
		MI	0.989	0.182	-0.011	0.032	0.954	95.0 (93.6,96.4)
	Effect of Fast Metabolizing Genotype For Exposed (True effect = 2.5)	Full	2.508	0.157	0.008	0.025	0.664	94.3 (92.9,95.7)
		CC	2.519	0.197	0.019	0.037	1.000	96.4 (95.2,97.6)
		MI	2.090	0.193	-0.410	0.190	5.144	42.4 (39.3,45.5)
	Interaction Effect (True effect = 1.5)	Full	1.507	0.220	0.007	0.049	0.684	94.7 (93.3,96.1)
		CC	1.520	0.268	0.020	0.072	1.000	95.5 (94.2,96.8)
		MI	1.101	0.265	-0.399	0.211	2.945	70.0 (67.2,72.8)
K: Condition 3
	Effect of Fast Metabolizing Genotype For Unexposed (True effect = 1.0)	Full	1.003	0.155	0.003	0.022	0.666	96.1 (94.9,97.3)
		CC	1.000	0.185	0.0002	0.034	1.000	95.6 (94.3,96.9)
		MI	1.004	0.191	0.004	0.032	0.978	95.6 (94.3,96.9)
	Effect of Fast Metabolizing Genotype For Exposed (True effect = 2.5)	Full	2.509	0.157	0.010	0.024	0.024	95.9 (94.7,97.1)
		CC	3.463	0.233	0.963	0.983	1.000	0.2 (-0.1,0.5)
		MI	2.611	0.274	0.111	0.041	0.042	99.8 (99.5,100.1)
	Interaction Effect (True effect = 1.5)	Full	1.507	0.220	0.007	0.048	0.047	95.9 (94.7,97.1)
		CC	2.463	0.298	0.963	1.012	1.000	9.5 (7.7,11.3)
		MI	1.608	0.340	0.108	0.074	0.073	99.2 (98.6,99.8)
b. Subjects Missing Data on Alcohol Exposure Under Conditions 1 and 2
Scenario	Variable	Method	Mean β	Mean SE	Mean Bias	MSE	RelMSE	Coverage
L: Condition 1	Effect of Fast Metabolizing Genotype For Unexposed (True effect = 1.0)	Full	1.016	0.155	0.016	0.025	0.962	94.7 (93.3,96.1)
		CC	1.015	0.158	0.015	0.026	1.000	95.4 (94.1,96.7)
		MI	1.308	0.150	0.308	0.115	4.502	45.6 (42.5,48.7)
	Effect of Fast Metabolizing Genotype For Exposed (True effect = 2.5)	Full	2.504	0.156	0.004	0.022	0.737	96.2 (95.0,97.4)
		CC	2.503	0.189	0.003	0.039	1.000	94.1 (92.6,95.6)
		MI	2.449	0.183	-0.051	0.026	0.835	96.9 (95.8,98.0)
	Interaction Effect (True effect = 1.5)	Full	1.488	0.220	-0.012	0.048	0.845	95.2 (93.9,96.5)
		CC	1.487	0.247	-0.013	0.056	1.000	96.4 (95.2,97.6)
		MI	1.141	0.247	-0.359	0.172	3.055	71.1 (68.3,73.9)
M: Condition 2	Effect of Alcohol Exposure for Those without Fast Genotype (True effect = 0.5)	Full	0.499	0.075	-0.001	0.006	0.622	95.5 (94.2,96.8)
		CC	0.501	0.095	0.001	0.009	1.000	94.6 (93.2,96.0)
		MI	0.560	0.095	0.060	0.012	1.348	91.4 (89.7,93.1)
	Effect of Alcohol Exposure for Those without Fast Genotype (True effect = 2.0)	Full	2.008	0.176	0.008	0.032	0.780	94.4 (93.0,95.8)
		CC	2.006	0.202	0.006	0.041	1.000	95.1 (93.8,96.4)
		MI	1.621	0.199	-0.379	0.169	4.183	50.5 (47.4,53.6)
	Interaction Effect (True effect = 1.5)	Full	1.509	0.192	0.009	0.037	0.743	95.2 (93.9,96.5)
		CC	1.505	0.223	0.005	0.050	1.000	95.6 (94.3,96.9)
		MI	1.061	0.217	-0.439	0.219	4.392	44.5 (41.4,47.6)

Results From Fitting Full, Complete-Case, and Multiple Imputation Models to 1000 Simulated Data Sets With a Sample Size of 2058 Where the Covariate of Interest and as a Result the Interaction Term Were Missing for Approximately 20% of Subjects.

Back to article page