ARE STRATIFICATION EFFECTS REAL? A CASE STUDY FROM THE ALCOHOL FIELD Andrew C. Heath, D.Phil. Washington University School of Medicine St. Louis, Missouri
ALCOHOL METABOLISM ALCOHOL ACETALDEHYDE ACETATE Alcohol Dehydrogenase (ADH) Aldehyde Dehydrogenase (ALDH)
Asian Studies of Japanese Alcoholics and Controls Work by Higuchi, Murumatsu and colleagues is documenting ways in which genes that influence alcohol metabolism may be associated with differences in alcohol dependence risk or alcohol consumption levels. (Higuchi et al., 1994) (Murumatsu et al., 1996)
Higuchi Data -- Japanese Alcoholics and Controls: ALDH2 locus (Higuchi, 1994) ControlsAlcoholics (N=461)(N=655) LocusGenotype(%)(%) ALDH2*1 / *15888 *1 / *23512 *2 / *270 p <.001
From Higuchi’s community data, in individuals who are also ALDH2*1/*1 homozygotes, we may estimate the penetrance of the AHD2*2/*2 genotype (the low risk genotype) as 7%, that of the high risk ADH2*1/*1 genotype as 29%. Higuchi Data -- Japanese Alcoholics and Controls: ADH2 locus (in those who are ALDH2*1/*1 homozygotes) Population ControlsAlcoholics ADH2*1 / *17.3% 30.4% *2 / *134.7% 33.7% *2 / *258.1% 35.8%
Higuchi Data -- Genetic effects on alcohol consumption levels in a community sample Average monthly alcohol consumption (ml pure alcohol) Genotype MEN ALDH2*1/* ALDH2*1/* ALDH2*2/*294.1 From other data, we can estimate that approximately ONE-THIRD of the variance in alcohol consumption levels in males is explained by this genetic locus. No significant effects of the ADH2 locus were observed. (Higuchi et al., 1996b)
Higuchi -- Temporal Changes in ALDH2*1/*2 Frequency in Japanese Alcoholics (Higuchi, 1994) (N=400)(N=400)(N=500) (%)(%)(%) ALDH2*1 / * *1 / * ***13.0** *2 / * i.e. GENOTYPE X ENVIRONMENT INTERACTION? Protective effect of a single *2 allele is being diminished in more recent cohorts.
FREQUENCIES OF GENES INFLUENCING ALCOHOL METABOLISM High RiskJapaneseEuropean LocusAlleleAncestryAncestry ALDH2ALDH2*176%100% ADH2ADH2*125%95% ADH3ADH3*26%45% NOTE: Predicted magnitude of effects is ALDH2*1 >> ADH2*1 >> ADH3*2.
CAN WE DEMONSTRATE SIGNIFICANT PROTECTIVE EFFECT OF THE ADH2*2 ALLELE IN INDIVIDUALS OF EUROPEAN ANCESTRY?
Lifetime DSM-III-R Alcohol Dependence by ADH2 Genotype at Follow-up of the Australian Alcohol Challenge Twin Study Sample Whitfield et al., 1998 ADH2*11ADH2*12 Male affected36 (26.3%)1 (5.3%) Male unaffected p<0.05 Female affected24 (14.3%)3 (30.0%) Female unaffected 144 7N.S.
QUANTITATIVE INDICES of ALCOHOL CONSUMPTION and PROBLEMS IC-1Weekly drinks in IC-2Average weekly consumption in FC-1Consumption in past week, FC-2Average weekly consumption in FC-3Weekly drinks in SC-1Average weekly consumption in FD-1Symptom count in SD-1Symptom count in (Whitfield et al., 1992)
(Whitfield et al., 1998) QUANTITATIVE INDICES of ALCOHOL CONSUMPTION and PROBLEMS
SURGEON GENERAL’S WARNING Using standard case-control comparisons to test for genetic association due to linkage disequilibrium can be dangerous! Danger of false positive findings in populations of diverse ancestry, e.g. most nations except population isolates, or possibly Finland, or Japan. Findings for the DRD2 ‘A1’ allele may be an example of this.
POPULATION STRATIFICATION Hypothetical Example Falsely infer that A1 allele is risk-factor for Roman Catholicism. OR = 2.28, 95%CI NO ASSOCIATION NORTHERN EUROPEAN ANCESTRY (N=200) SOUTHERN EUROPEAN ANCESTRY (N=200) NOT A1 allele A1 allele NOT ROMAN CATHOLIC ROMAN CATHOLIC NOT ROMAN CATHOLIC ROMAN CATHOLIC % % % % 70% 30% 90% 10% NOT ROMAN CATHOLIC ROMAN CATHOLIC NOT A1 allele A1 allele MINGLED IN AUSTRALIAN POPULATION (N=400)
SOCIOCULTURAL CORRELATES of ADH2*12 GENOTYPE in the AUSTRALIAN TWIN PANEL (I) Respondent Religious Affiliation:OR95% CI Church of England Any other Parental Religious Affiliation: Both parents COE Single parent COE 2.00 NS Neither parent COE
1981 SURVEY SOCIAL ATTITUDES ADH2*11ADH2*12OR95% CI Socialism 40.2%81.8% MEN
SOCIOCULTURAL CORRELATES OF ADH2*12 GENOTYPE (II) 1981 Survey Social Attitudes ADH2*11ADH2*12OR95% CI Working mothers 25.0%75.0% Pyjama parties 27.5%85.7% WOMEN (Does respondent agree or disagree with...)
CONCLUSIONS For phenotypes which show considerable variation in prevalence between different ancestral groups, the danger of artifactual findings is real – but, as we have discussed this week and will discuss more this session, there are also good methods for addressing these problems.