Genetic Epidemiology

Genetic Epidemiology

A Definition : A science that deals with aetiology, distribution, and control of disease in groups of relatives and with inherited causes of disease in populations (Morton, 1982). Note that this is double barrelled, but information about populations is inferred via correlations observed among samples of relatives.

Related Disciplines

Molecular epidemiology

Human genetics

Clinical genetics

Cancer genetics

Biochemical genetics

Molecular biology

Genomics

Bioinformatics

Statistical genetics

Evolutionary and population genetics

Some relevant human genetics

The human genome is made up of 3.1 billion nucleotide pairs arranged linearly in chromosomes. Human cells are diploid, that is, there are two complete sets of chromosomes (23 pairs of autosomes and two sex chromosomes), one originating from each parent.

There are 23000-25000 genes: these comprise directly transcribed sequence (exons and untranslated regions) 1.9% of the genome, and intronic DNA (36% of genome). Approximately 5% of the genome is segmental duplications.

Mutation

A germline mutation (error in replication of DNA) occurs every 1 Mbp per generation. Cross species comparisons suggest there are approximately 4 coding mutations per genome per generation, one third of which will be deleterious.

Mutation gives rise to 0.8 polymorphic sites per 1000 base pairs in human populations: a polymorphism is a location (locus) in the sequence where 1% or more of individuals carry a nucleotide different from the rest of the population. Alleles are the different variants at the locus.

Types of polymorphism

Satellite repeat

Minisatellite repeat

Variable Number Tandem Repeat

Simple Sequence Repeat: includes microsatellites

Indel: Insertion/deletion of short or long sequence

Transversion: Single Nucleotide Polymorphism

Build 123 of dbSNP (30/10/04) contains data for 10,079,771 SNPs

1 Mbp around the serotonin transporter gene (SLC6A4)

40 kbp around the serotonin transporter gene (SLC6A4)

100 bp within the serotonin transporter gene (SLC6A4)

Functional effects of variants

Mechanism	Number in HGMD
Nonsynonymous coding variants	40228
Affecting splicing	4668
Affecting transcription level
Affecting DNA binding	599
Distant enhancers

In three surveys of allelic-specific transcription, 20-40% of genes have been found to contain noncoding variants that alter gene expression level.

Recombination and haplotypes

Recombination occurs during meiois, and involves randomly cutting and rejoining of lengths of homologous DNA. There is always at least one recombination event per chromosome pair so that a chromosome in the haploid gamete's genome is a hybrid of the parental chromosomal pair.

A haplotype is the pattern of variants observed at a set of polymorphisms along a section of a chromosome in a gamete. It can be used to infer where a recombination event fell on the chromosome (along with knowledge of the parental haplotypes).

Haplotypes and "phylogenetic" history

If a new functional mutation occurs in an individual, and is subsequently propagated from generation to generation, then a particular ancestral haplotype of neighbouring polymorphisms will be associated with that new variant. The haplotype will become shorter with time, due to recombination, and unclear, due to mutation.

This is helpful, in that one does not have to genotype at the mutation itself to infer whether or not they carry a particular variant.

This is unhelpful, in that polymorphisms may be risk factors for a disease without being causative.

Population genetics

Population genetics (and evolutionary genetics) deal with occurence of genetic variation in groups of organisms and families, usually natural populations.

• Large ("ideal") idealised groups or populations, where models can be deterministic (though some deterministic models are accurate for even very small populations).
• Small populations, where the role of chance can play a larger role (so called genetic drift).

Genotype and allele frequencies

For a codominant trait, we can take a sample from the population, and count the different genotypes.

Race and Sanger (1975) counts for the MN blood group.

Blood Group (genotype)	M (MM)	MN (MN)	N (NN)	Total
Count (percent)	363 (28.4%)	634 (49.6%)	282 (22.0%)	1279 (100.0%)

The proportions are our best estimate of the probability that an individual will carry that genotype in the population of London, Oxford and Cambridge. The observed heterozygosity is 49.6%.

We can also make inferences about the population of gametes that gave rise to individuals tested:

Alleles	M	N	Total
Count (percent)	1360 (53.2%)	1198 (46.8%)	2558 (100.0%)

The proportions are our best estimate of the probability that a sperm or egg taken from that population will carry that particular allele.

Hardy-Weinberg Equilibrium (HWE)

Hardy-Weinberg equilibrium describes the relationship between the gametic or allele frequencies, and the resulting genotypic frequencies. It holds if the following properties are true for the given locus,

1. Random mating or panmixia: the choice of a mate is not influenced by his/her genotype at the locus.
2. The locus does not affect the chance of mating at all, either by altering fertility or decreasing survival to reproductive age.

If these properties hold, then the probability that two gametes will meet and give rise to a new genotype is simply the product of the allele frequencies (a la binomial):

Pr(MM)= Pr(M) x Pr(M)

Pr(NN)= Pr(N) x Pr(N)

Pr(MN)= 1 - Pr(MM) - Pr(NN) = 2 x Pr(M) x Pr(N).

Testing HWE

We can easily test for deviation from Hardy-Weinberg equilibrium using a chi-square. This can arise from,

1. Marital assortment: "like marrying like".
2. Inbreeding
3. Population stratification: multiple subgroups are present within the population, each of which mates only within its own group (homogamy), and the allele frequencies are different within each subgroup (Wahlund effect). Mating within each group is random.
4. Admixture: the breakdown of any of the former processes will lead to deviations until equilibrium is reached.
5. Decreased viability of a particular genotype: individuals carrying a deleterious genotype die early (or in utero).

Age of mutations

Coalescent models (branching linear birth-death process models)

For a neutral mutation in a population of constant size:

t = 2 N p

In the case of an exponentially expanding population or one where the mutant allele is undergoing selection,

t = log(2Np(r+s) + 1)/(r+s)

where r is the exponential growth rate parameter (approximately the proportional increase per generation), and s is the selection coefficient for heterozygotes.

A well-known example of such a calculation is for idiopathic torsion dystonia (locus DYT1) among the Ashkenazi Jews [Risch et al 1995]. In Eastern Europe, this population increased rapidly in size from approximately 10⁵ in 1650 to 5 x 10⁶ in 1900, giving us an estimated r of 0.40. Risch et al [1995] estimate the allele frequency in the current population at 1/6000 to 1/2000. This gives an estimate of the age of the first mutation at 16-19 generations ago (about the year 1550).

Phenotype-genotype relationships

Discrete variation within genes gives rise to altered type of gene product or altered level of gene product, and thus to a measurable human phenotype. Since both alleles of a diploid genotype in humans are usually active, a combined outcome is observed. The effects of the two alleles may interact in a nonlinear fashion (genetical dominance).

Penetrance of R305W/R419Q joint genotypes

R305W	R419Q	Blue	Green/Hazel	Brown
R/R	R/R	48.9% (517)	29.3% (310)	21.8% (230)
R/R	R/Q	26.5% (54)	45.1% (92)	28.4% (58)
R/R	Q/Q	0% (0)	57.1% (4)	42.9% (3)
R/W	R/R	47.6% (60)	17.5% (22)	34.9% (44)
R/W	R/Q	30.0% (3)	50.0% (5)	20.0% (2)

In the usual case where diploid genotypes are measured, haplotypes must be inferred via statistical modelling.

Genetic Association Analysis

Genetic association refers to the direct genotype-phenotype relationship.

The use of the term is to differentiate it from genetic linkage, which refers to within-family cosegregation of genotype with phenotype arising due to colocalization on the same chromosome.

It can be detected in conventional epidemiological designs such as a case-control or cohort design, as well as in family based designs, which are complex matched (or counter-matched).

A Genome wide association study is such a study testing for association between a phenotype and a large number of marker polymorphisms.

A candidate gene association study will test for association between a phenotype and polymorphisms within a gene that is likely to be related to that phenotype, based on biochemical or physiological evidence.

Practicalities of genotyping

Processing of samples

Technologies

Sequenom MassARRAY

Affymetrix GeneChip Mapping 100K Set

The GeneChip Mapping 100K Set is the first in a family of products for whole-genome association studies. It is comprised of a set of two arrays that enable genotyping of greater than 100,000 SNPs with a single primer.

The high throughput and ease of use of the Mapping 100K Set at a cost of one cent per SNP make whole-genome association feasible. It is usable for cancer genetics, linkage disequilibrium, case-control, family-based association studies, chromosomal copy number change analysis, and population genetics.

A 250K set is currently undergoing testing. Small sample requirement (250 ng DNA per array) High throughput Availability via the Australian Genome Research Facility (~A$1200/S)

Complex Genetic Disease

• Does not exhibit a simple Mendelian pattern of inheritance
• Likely to be oligogenic or polygenic
• Penetrance of any single gene not high
• Involves multifactorial causation (genes, environment)

What kind of groups do we study

• Population isolates eg Amish, Tristan Da Cunha, Iceland, Sardinia
• Admixed populations ie Caucasian/Japanese intermarriage in Hawaii.
• Migration of populations into different environments eg Tokelauan islanders and hypertension; American blacks and sickle cell anaemia.
• Families (nuclear or multigenerational)
• Pairs of relatives

Phenometric Studies

• "General" population samples eg degree of relatedness of pairs of cancer cases compared to that of pairs of controls eg prostate cancer in Utah (expressed as mean coefficient of relationship so clustering in terms of genetic distance); all cancers and sharing surnames in UK.
• Weinberg proband-control studies eg CASH study of breast cancer, Cooke's studies of asthma.
• Path analysis of family material eg Tambs on hypertension
• Twin studies
• Segregation analysis of family material eg Morton on total sIgE

Relatedness of disease cases

Ancestral kinship and cancer on Lastovo Island, Croatia [Rudan 2001]:

Path analysis

Tambs et al [1992] collected blood pressure measurements and familial relationships for 74994 Norwegians:

They concluded 47% of the variation in systolic BP was genetic, only 2% is due to shared family environment and 51% is due to unshared environment.

Genometric Studies

• Genetic linkage analysis
  • Parametric ("lod" score) linkage analysis
  • Combined segregation-linkage analysis
  • MOD and WROD score linkage analysis
  • Nonparametric (trait locus model free) linkage analysis
  • Variance components linkage analysis
• Genetic association analysis of unrelated individuals
  • Conventional case-control or cohort analyses
  • Genome control
  • Haplotype inference
• Genetic association analysis of families
  • Mixed-effects association analysis
  • Transmission-based tests
  • "ABAW" based approaches

Reasons to be cheerful

• The common disease-common gene hypothesis
• The density of SNPs now available
• The power of association analysis cf linkage analysis
• Haplotype blocks (fine recombination structure)
• Larger studies and metaanalyses
• Subsetting of disease eg familial breast cancer

Causes of difficulty

• Trait locus allelic heterogeneity
• Low trait locus penetrances
• High trait allele frequency: Gene x Gene interaction
• Low trait allele frequency: Between-family heterogeneity
• Between-population trait locus heterogeneity
• Between-population marker locus heterogeneity
• Gene x Environment interaction
• Gene-Environment correlation

Allelic Heterogeneity

The Human Genome Mutation Database lists 347 known mutations in BRCA1:

Locus Heterogeneity

In humans, asthma has been reported associated with variants in:

ADA ADAM33 ADRB2 ALOX5 BCL6 C3 CASP10 CFTR CYSLTR2
DCP1 DPP10 ELF5 FCER1B FCER2 Gm GPRA HLA Hp
IFNG IL4 IL4R IL9 IL10 IL12B IL13 IL18 LTC4S PAFAH
PHF11 PLA2G2D PTGDR SCYA5 SPINK5 TCRAD TGFB TLR4 TNFA UGB UGRP1

Low trait locus penetrances

The relative risks associated with particular alleles for complex diseases are seldom much above two-fold.

ID Disease/outcome	Gene (polymorphism), contrast	Sample size*	OR (F) (95% CI)	OR (R) (95% CI)	p Het		(number of studies)
1 Myocardial infarction	ACE (insertion/deletion), DD vs DI+II	18 664 (15)	1·20 (1·10­1·31)	1·28 (1·09­1·50)	0·001
2 Ischaemic heart desease	ACE (insertion/deletion), DD vs DI+II	21 876 (17)	1·16 (1·08­1·25)	1·20 (1·06­1·36)	0·004
3 ICVD	ACE (insertion/deletion), DD vs DI+II	11 394 (6)	1·18 (1·01­1·37)	1·21 (0·98­1·50)	0·187
4 Poor clozapine response	HTR2A (102T/C), CC vs CT+TT	733 (6)	1·64 (1·18­2·28)	1·65 (1·19­2·29)	0·514
5 Poor clozapine response	HTR2A (H452Y), YY vs HY+HH	676 (5)	5·55 (1·15-26·8)	3·37 (0·97­11·6)	0·941
6 Vascular disease	MTHFR (677C/T), TT vs CC	6947 (23)	1·08 (0·95­1·22)	1·12 (0·92­1·37)	0·001
7 Lung cancer	CYP2D6 (deficient oxidation),	5162 (14)	0·67 (0·53­0·84)	0·63 (0·42­0·93)	0·013
	poor metabolisers vs others
8 Dementia in Down's syndrome	APOE (2/3/4), allele 2 vs 3+4	1130 (9)	0·41 (0·22­0·78)	0·45 (0·17­1·17)	0·044
9 Schizophrenia	DRD3 (Bal1), 11+22 vs 12	5121 (25)	1·07 (0·95­1·20)	1·08 (0·95­1·23)	0·204
10 Bipolar affective disorder	MAOA (Fnu4HI), allele 1 vs 2	962 (3)	1·30 (0·96­1·75)	1·42 (0·82­2·46)	0·052
11 Bipolar affective disorder	MAOA (CA), allele 122 vs others	1932 (7)	0·95 (0·75­1·21)	0·95 (0·69­1·31)	0·115
12 Bipolar affective disorder	TH (tetranucleotide repeat), allele 1 vs others	2901 (8)	0·92 (0·78­1·08)	0·97 (0·74­1·26)	0·019
13 Unipolar affective disorder	TH (tetranucleotide repeat), allele 1 vs others	1128 (3)	1·17 (0·91­1·51)	1·17 (0·91­1·51)	0·370
14 NIDDM	KCNJ11/KIR6.2-BIR (E23K), KK vs EK+EE	888 (4)	1·94 (1·30­2·88)	1·93 (1·29­2·87)	0·777
15 Lung cancer	GSTM1 (gene deletion), null/null vs others	9724 (21)	1·17 (1·07­1·27)	1·18 (1·04­1·34)	0·007
16 Lung cancer	CYP1A1 (4889A/G), GG vs AA+AG	2392 (6)	1·45 (0·80­2·62)	2·07 (0·71­6·08)	0·106
17 Lung cancer	CYP1A1 (MspI), +/+ vs others	4263 (12)	1·21 (0·87­1·70)	1·26 (0·84­1·89)	0·294
18 Myocardial infarction	SERPINE1/PAI-1 promoter (4G/5G),	3381 (9)	1·32 (1·15­1·52)	1·55 (1·16­2·08)	0·001
	4G/4G vs 5G/5G
19 Parkinson's disease	CYP2D6 (1934G-A), allele 4 vs others	7029 (14)	1·17 (1·03­1·32)	1·18 (1·00­1·40)	0·094
20 Essential hypertension	AGT (M235T), allele T235 vs M235	4698 (6)	1·22 (1·06­1·42)	1·44 (1·04­2·00)	0·002
21 Cancer	HRAS/HRAS1 (rare alleles)	8542 (24)	1·91 (1·62­2·27)	1·84 (1·54­2·21)	0·364
	rare vs common alleles
22 Left ventricular hypertrophy	ACE (insertion/deletion), allele D vs I	8186 (12)	1·08 (0·97­1·21)	1·13 (0·95­1·33)	0·085
23 Bladder cancer	NAT2 (slow acetylation alleles),	5836 (20)	1·38 (1·23­1·55)	1·43 (1·20­1·71)	0·010
	slow/slow vs others
24 ICVD	APOE (2/3/4), allele 4 vs others	3632 (9)	1·71 (1·38­2·11)	1·69 (1·37­2·09)	0·449
25 Non­syndromic cleft lip	TGFA (TaqI), allele 2 vs 1	5272 (9)	1·50 (1·23­1·82)	1·58 (1·13­2·21)	0·012
26 Alcoholism	DRD2 (TaqIA), allele A1 vs A2	3826 (15)	1·50 (1·26­1·79)	1·60 (1·19­2·15)	0·002
27 Ischaemic stroke	ACE (insertion/deletion), DD vs DI+II	2160 (6)	1·42 (1·16­1·74)	1·58 (1·11­2·25)	0·022
28 Diabetic nephropathy	ACE (insertion/deletion), II vs ID+DD	5393 (20)	0·73 (0·63­0·83)	0·68 (0·55­0·84)	0·006
29 Neural tube defects	MTHFR (677C/T), TT vs CT+CC	3880 (13)	1·68 (1·38­2·05)	1·67 (1·26­2·23)	0·105
30 Neural tube defects	MTHFR (677C/T) mother, TT vs CT+CC	1955 (8)	1·95 (1·44­2·65)	1·98 (1·46­2·68)	0·844
31 Neural tube defects	MTHFR (677C/T) father, TT vs CT+CC	950 (5)	1·09 (0·62­1·93)	1·15 (0·65­2·05)	0·568
32 Ischaemic heart disease	APOE (2/3/4),	8 962 (9)	1·27 (1·14­1·42)	1·39 (1·11­1·73)	0·001
	4/3+4/2+4/4 vs 3/3
33 Ischaemic heart disease	LPL (D9N), ND vs DD	2 022 (3)	1·37 (0·81­2·33)	1·36 (0·80­2·32)	0·861
34 Ischaemic heart disease	LPL (N291S), SN vs NN	13 115 (4)	1·15 (0·91­1·46)	1·15 (0·91­1·46)	0·996
35 Ischaemic heart disease	LPL (S447X), XS vs SS	4067 (5)	0·84 (0·70­1·00)	0·84 (0·70­1·00)	0·969
36 Alcoholic liver disease	CYP2E/CYP2E1 (RsaI), allele c2 vs others	4178 (9)	1·54 (1·04­2·30)	1·41 (0·78­2·55)	0·119
37 Myocardial infarction	FGB/FGB promoter (455G/A), AA vs GG	1561 (3)	0·68 (0·47­0·99)	0·68 (0·47­1·00)	0·839
38 Myocardial infarction	F5 (1691G/A), AA+AG vs GG	5937 (12)	1·29 (1·03­1·61)	1·32 (1·03­1·70)	0·336
39 Myocardial infarction	F2 (20210G/A), AA+AG vs GG	5637 (7)	1·11 (0·79­1·56)	1·20 (0·80­1·80)	0·281
40 Ulcerative colitis	IL1RN (86-BP DUP), carriers of 2 vs others	2835 (8)	1·23 (1·04­1·45)	1·18 (0·88­1·57)	0·011
41 CAD	ITGB3 (L33P), A2A2 vs A1A2+A1A1	17 315 (31)	1·07 (1·00­1·14)	1·10 (0·99­1·21)	0·004
42 Fractures	COL1A1 (2046G/T), ss+Ss vs SS	3580 (13)	1·39 (1·17­1·65)	1·43 (1·13­1·81)	0·088
43 Bipolar disorder	DRD3 (Bal1), allele 1 vs allele 2	3392 (9)	1·01 (0·87­1·16)	1·01 (0·87­1·16)	0·545
44 Parkinson's disease	MAPT (allele A0), allele A0 vs others	2090 (5)	1·53 (1·23­1·91)	1·52 (1·22­1·90)	0·445
45 Bulimia	HTR2A (1438G/A), allele A vs G	1126 (3)	1·32 (1·04­1·69)	1·33 (1·04­1·69)	0·427
46 Anorexia nervosa	HTR2A (1438G/A), allele A vs G	3698 (7)	1·36 (1·19­1·57)	1·42 (1·05­1·93)	<0·001
47 Bladder cancer	GSTM1 (gene deletion), null/null vs others	4724 (15)	1·50 (1·32­1·71)	1·54 (1·27­1·86)	0·037
48 SLE nephritis	FCGR2A (R131H), RR vs RH+HH	2801 (24)	1·12 (0·93­1·35)	1·11 (0·88­1·41)	0·123
49 SLE	FCGR2A (R131H), RR vs RH+HH	4708 (21)	1·29 (1·12­1·48)	1·29 (1·10­1·52)	0·218
50 Parkinson's disease	COMT (V158M), MM+MV vs VV	964 (3)	1·31 (0·84­2·04)	1·37 (0·68­2·76)	0·097
51 Recurrent early pregnancy loss	MTHFR (677C/T), TT vs CT+CC	1097 (6)	1·37 (0·95­1·98)	1·31 (0·78­2·20)	0·146
52 Alzheimer's disease	MAPT (extended haplotypes),	3377 (7)	1·03 (0·89­1·19)	1·03 (0·89­1·20)	0·377
	H1H1 vs H1H2+H2H2
53 Alzheimer's disease	LPR/LPR exon3 (766 C/T), CC vs CT+TT	4097 (8)	1·37 (1·18­1·58)	1·37 (1·08­1·73)	0·021
54 IgA nephropathy	ACE (insertion/deletion), DD vs DI+II	1774 (7)	1·26 (1·02­1·55)	1·26 (0·99­1·61)	0·294
55 Heparin-induced	FCGR2A (R131H), RR vs RH+HH	1939 (6)	1·02 (0·81­1·29)	0·81 (0·49­1·34)	0·023
thombocytopenia

Between-population marker heterogeneity

If ethnic groups differ in frequency of polymorphisms and environmental risk factors for a condition, confounding can occur. There are perhaps three published examples of marked confounding, and these usually require obvious levels of admixture in the study sample (eg African Americans, Pima Indians), or strong association of the marker to ethnicity (eg lactase deficiency genotype and origin within Europe).

There are a number of statistical methods to deal with this:

• Traditional stratification on reported ancestry
• Imputation of ancestry based on marker genotypes
• Genomic control

Gene by Environment Interaction

ERCC2 genotype, cigarette smoking and lung cancer risk in 2575 cases and controls [Zhou et al 2002]:

• ERCC2 (19q13.2) codes for a DNA repair protein (Type D xeroderma pigmentosum).
• ERCC2*312Asp previously reported to increase risk in light but not heavy smokers.

Mendelian Randomization

Epidemiologists are interested in using mendelian randomization to assess causation between observed phenotypes, using genotype as an instrumental variable.

Minelli and coworkers (2004) note that metaanalysis of traditional epidemiological studies finds that a 3 uM (25%) decrease in serum homocysteine level is associated with an 11% decrease in coronary heart disease (CHD) risk.

The 677C>T transversion in the MTHFR gene is associated with higher homocysteine levels, therefore an increase in CHD risk would be expected if the direction of causation runs from homocysteine level to CHD, and MTHFR genotype has no other associations with CHD risk.

	Homocysteine level	CHD Odds Ratio
MTHFR C/C v T/T	2.7 uM (2.02-3.41)	1.21 (1.06-1.40)

So predicted Odds Ratio for CHD per 3 uM increase homocysteine level 1.24 (1.06-1.49).

The serotonin transporter gene (SLC6A4) promoter polymorphism

There is a common VNTR polymorphism in the promoter of the serotonin transporter gene. SLC6A4-promoter/luciferase reporter construct is expressed at higher levels if the 42 bp insertion ("long") is present [Heils et al 1994; Lesch et al 1996].

SLC6A4 promoter polymorphism and behaviour

Given the effects of drugs on the serotonin transporter, it seemed likely that there may be phenotypic effects of this polymorphism on behaviour:

Trait Anxiety			Major (unipolar) Depression
Genotype	N	Neuroticism Score (NEO)
L/L	648	48.5
S/L	1133	50.6
S/S	751	50.3

The "Short" allele weakly increases anxiety (0.2 SD) and risk of major depression.

SLC6A4 promoter polymorphism and behaviour 2

Bipolar Disorder			Violent Suicide

The "Short" allele weakly increases risk of bipolar disorder and violent suicide.

Caspi et al [2003]

Caspi and coworkers reported a study of the SLC6A4 promoter insertion-deletion in the Dunedin Multidisciplinary Health and Development Study cohort.

They argued that the weakness of the association might hide larger effects in different environments, and cite animal models.

For instance, in rhesus monkeys, the short allele predicts low CSF serotonin metabolite levels (5-HIAA), but in deprived (peer-reared) and not in maternally reared monkeys [Bennett et al 2002]. A similar interaction is seen for ACTH response to stress [Barr et al 2004].

Genotyping in Caspi et al [2003]

Genotyped 847 Caucasians followed to age 26 (265:435:147):

• 93% of DNA samples from blood, 7% buccal
• Excluded 100 participants of Maori descent
• Performed latent class analysis testing for heterogeneity
  (presumably based on additional markers)
• Found genotype frequencies comparable with other Caucasian populations
• No evidence of Hardy-Weinberg disequilibrium

Life Events in Caspi et al [2003]

Measured stressful life events from ages 21-26 (life-history calendar):

• Separate interview and interviewer
• 14 events: Employment/financial/housing/health/relationships
• Male excess of life events
• No association with genotype

Childhood maltreatment in Caspi et al [2003]

Assessed childhood maltreatment

• Observed Mother-child interactions age 3 ("rejecting")
• Harsh discipline ages 7-9 (upper decile of "smack or hit")
• Change of primary caregiver age 0-10 (2 or more changes)
• Recalled (age 26) physical child abuse
• Recalled (age 26) unwanted sexual contact
• No association with genotype

Depression in Caspi et al [2003]

Assessed depression in last 12 months (Diagnostic Interview Schedule)

• Interviews at ages 18, 21, 26
• Depressive symptoms score at age 26: mean score 5.2 (SD 10.5)
• Major depressive episode (DSM-IV): 144 Ss (58% Female)
• Suicide attempts or ideation: 25 Ss
• Supplemented by informant rating

Results from Caspi et al [2003]

Predicted values from fitted interaction models

Results from Caspi et al [2003] 2

Marginal effects of promoter genotype
on risk of recent major depression

Genotype	Odds Ratio
L/L (31%)	1
S/L (51%)	1.16 (0.88-1.53)
S/S (18%)	1.35 (0.78-2.33)

Moderation by promoter genotype of
association of depression with life events.

Genotype	Odds Ratio (per life event)
L/L (31%)	1.13 (0.83-1.56)
S/L (51%)	1.47 (1.08-2.02)
S/S (18%)	1.68 (1.22-2.30)

Replications

Kaufman et al [2004]: Depression score in children taken into care (N=57) versus controls (N=44).

Grabe et al [2004]: Distress (BL-38 scale) versus chronic disease and promoter genotype in Pomeranian women (N=674).

Manuck et al [2004]: Prolactin level following fenfluramine administration versus SES and SLC6A4 promoter genotype (N=139).

Population genetics of SLC6A4 promotor polymorphism

The frequency of the "short" allele varies across populations.

Group	N	Proportion s
Asian	1102	0.77
Askenazi Israelis	224	0.52
UK	686	0.43
European Americans	221	0.43
Italy	1104	0.41
African Americans	797	0.26