Part II: Methods and Approaches 1: Assessing Disease Associations and Interactions Chapter 7 Tables

Pages in this Report

Human Genome Epidemiology: A Scientific Foundation for Using Genetic Information to Improve Health and Prevent Disease

Book Cover | Preface | Acknowledgments | Contents | Contributors

Facing the Challenge of Complex Genotypes and Gene-environment Interaction: the basic epidemiologic units in case-control and case-only designs

Lorenzo D. Botto and Muin J. Khoury

Table 7-1Layout for a case-control study assessing the effect of a genotype (G) and an environmental factor (E)
G ^a	E ^a	Cases	Controls	Odds Ratio		Contrast	Main Information
+	+	a	b	ah/bg	A	A vs. D	Joint genotype and environmental factor vs. none
+	–	c	d	ch/dg	B	B vs. D	Genotype alone vs. none
–	+	e	f	eh/fg	C	C vs. D	Environmental factor alone vs. none
–	–	g	h	1	D		Common reference

aG, genotype; E, environmental factor
Other Measures	Odds Ratio	Main Information
Case only odds ratio	ag/ce	Departure from multiplicative model of interaction
Control only odds ratio	bh/df	Independence of factors in population
Multiplicative interaction	A/(B*C)	Deviation from multiplicative model of interaction
Additive interaction	A-(B+C-1)	Deviation from additive model of interaction
Stratified 1-a	ad/bc	Association with environmental factor among people with genotype
Stratified 1-b	eh/fg	Association with environmental factor among people without genotype
Stratified 2-a	af/be	Association with genotype among people exposed to environmental factor
Stratified 1-b	ch/dg	Association with genotype among people not exposed to environmental factor

[back to chapter]

Table 7-2Advantages of the 2X4 table in the study of gene-environment interactions
1)	The primary data are displayed clearly and completely.
2)	The primary measures of association-relative risk estimates for each factor alone and for the joint exposure-are readily generated. Because they use the same reference group, these estimates can be compared.
3)	Attributable fractions can be computed separately for each exposure alone and for the joint exposure.
4)	Relative risk estimates can be used to assess the relation between the joint exposure and the individual exposures. For example, the departure from additive or multiplicative models of interactions can be readily derived from the table.
5)	Risk estimates stratified by either exposure can also be calculated if needed.
6)	For case-control studies, the case-only and the control-only odds ratios can be easily computed. For adequately chosen control groups, the control-only odds ratio estimates exposure dependencies in the underlying population.

[back to chapter]

Table 7-3Analysis of oral contraceptive use, presence of Factor V Leiden mutation, and risk for venous thromboembolism
Factor V ^a	OC ^a	Cases	Controls	Odds Ratio		95% CI	AF-Exp (%) ^a	AF-Pop (%) ^a	Exposure Frequency in Controls (%)
+	+	25	2	ORge	34.7	7.83-310.0	97.1	15.7	1.2
+	–	10	4	ORg	6.9	1.83-31.80	85.6	5.5	2.4
–	+	84	63	ORe	3.7	2.18-6.32	73.0	39.6	37.3
–	–	36	100	Ref	Ref				59.2
Total		155	169

^a Factor V: +, Presence of Factor V Leiden mutation (heterozygotes and homozygotes)
– , absence of Factor V Leiden mutation
OC: +, current use of oral contraceptives
-, no current use of oral contraceptives
AF-Exp (%): Attributable Fraction (percent) among exposed cases
AF- Pop (%): Attributable Fraction (percent) among all cases in the population

Note: the departure of the observed from the expected effect of the joint exposure depends on the definition of no interaction, as shown below for simple additive and multiplicative definitions.
	Expected OR-GE		Departure from expected
Additive	(3.7 + 6.9) – 1 =	9.6	34.7 – 9.6=	25.07
Multiplicative	(3.7 * 6.9) =	25.7	34.7/25.7=	1.4

[back to chapter]

Table 7-4 Comparing the stratified and case-only with the 2×4 approacha
Comparison with stratified analysis
FactorV ^a		Factor V Present		Factor V Absent		Ratio of odds ratios
		Cases	Controls	Cases	Controls
Oral contraceptive use	+	25	2	84	63
	–	10	4	36	100
Odds Ratio (95%CI)		5.0 (0.8-31.8)		3.7 (2.2-6.1)		1.4
Case-only and control-only odds ratios
Case-only odds ratio	(2536)/(1084) =			1.1
Control-only odds ratio	(2100)/(463) =			0.8		1.4

^a The data are from Table 3.
Source: Note that ratios of odds ratios are identical to departure from multiplicative model (Table 3)

[back to chapter]

Table 7-5Advantages and disadvantages of using case-only studies to screen for complex genotypes in disease etiology
Can be used to	Considerations
—Screen for genotypes with highest potential impact on disease	Provides upper limit of attributable fraction associated with genotype
—Screen for supramultiplicative interactions between multiple loci	Tools include, for example, log-linear analyses and other methods of clustering
—Provide clues to etiologic heterogeneity and determinants of phenotypes and severity	Shows variations in genotype frequencies and interactions by different disease phenotypes, severity, or age at onset

Can improve
—Speed of study	Particularly useful in conditions for which well-designed disease registries are available
—Precision of estimates	Eliminates controls and associated variance
—Validity of findings	Assumes no population stratification
—Efficiency of subsequent studies	Contributes to efficient case-control design by highlighting notable case- and population subsets, factors with highest potential impact, and potential sample size issues

But have disadvantages
—Validity sensitive to assumptions	Results are exquisitely sensitive to independence of factors in population
—Limited information	Provides few data on marginal effects, relative and absolute risks, and non-multiplicative interactions

[back to chapter]