Alcohol Dehydrogenase and Alcohol Dependence: Variation in Genotype-Associated Risk between Populations

To the Editor:

Osier et al. (2002) report that haplotyping of the alcohol dehydrogenase (ADH) gene cluster at 4q21-23 showed unusually high values for F_st, an estimator of population differentiation. This was largely due to differences between populations in East Asia and those in other areas of the world. The finding was discussed in relation to the origin and maintenance of the distinct East Asian haplotype and in relation to possible association between genetic variation at this locus and the risk of alcohol dependence (MIM 103780). This letter draws attention to a potentially related difference between populations, in the magnitude of the alcohol dependence risk associated with the ADH1B (MIM 103720) Arg47His polymorphism (previously referred to as “ADH2*2”). One possible explanation for such a difference in risk is the presence of linkage disequilibrium between this marker and an undiscovered causative polymorphism, with the effect being stronger in East Asians and the relative risk associated with ADH1B Arg47His variation consequently being greater.

To update a previous meta-analysis of the effects of ADH polymorphisms (Whitfield 1997), articles reporting on ADH1B genotypes in control and alcohol-dependent subjects were identified by Medline search or from knowledge of data in conference proceedings, with elimination of articles in which subjects overlapped. Data from eight of the articles previously analyzed (all those listed in table 1 and published before 1997) and from nine new articles, were included. Information on ADH1B Arg47His genotypes in control and alcohol-dependent subjects was extracted. Data on alcoholdependent subjects with known liver disease were excluded, because of the possibility that ADH1B variation may affect the risk of liver damage in alcoholics. Odds ratios were calculated from stratified 2 × 2 tables, using StatXact 5 (Cytel Software), with tests for heterogeneity across studies and estimation of common odds ratios. Whenever possible, two 2 × 2 tables were compiled from each article: one for the ADH1B*47Arg/*47Arg versus ADH1B*47Arg/*47His genotype comparison and the second for comparison of ADH1B*47Arg/*47His against ADH1B*47His/*47His.

Table 1.

Calculated Odds Ratios and Associated 95% CI for Alcohol Dependence by ADH1B Genotype^[Note]

	Control Subjects			Alcoholics			RR vs. RH		RH vs. HH
Population, Reference,and Sources of Subjects	RR	RH	HH	RR	RH	HH	OR	95% CI	OR	95% CI
Europeans:
Gilder et al. 1993:
Englanda	77	7	0	76	6	0	1.15	.32–4.35	NA	NA
Espinos et al. 1997:
Spain	58	12	1	62	9	0	1.42	.51–4.13	NA	NA
Whitfield et al. 1998:
Australiab,c	101	18	0	36	1	0	6.37	.94–274.60	NA	NA
Borras et al. 2000:
France, Germany, Poland, Spain, Swedend	214	10	0	226	5	0	2.11	.64–7.99	NA	NA
Ogurtsov et al. 2001:
Russia (Moscow)d	15	29	6	24	12	1	3.80	1.39–10.94	2.44	.25–123.4
Frenzer et al. 2002:
Australia c,d	184	14	2	54	3	0	1.37	.36–7.70	NA	NA
Common OR (all Europeans)							2.11	1.32–3.44	NA	NA
Asians:
Thomasson et al. 1994:
China (Atayal, Taiwan)	1	10	54	3	28	63	1.07	.08–61.93	2.39	1.01–6.03
Muramatsu et al. 1995:
China (Han Chinese, Shanghai)	12	43	50	13	8	11	5.66	1.72–20.00	.85	.27–2.56
Chen et al. 1996:
China (Han Chinese, Taipei)	0	19	44	14	15	17	NA	NA	2.03	.77–5.37
Shen et al. 1997:
China (Han)	6	19	23	10	25	17	1.26	.34–5.03	1.77	.69–4.63
Korean	3	23	24	9	17	29	3.95	.82–26.11	.62	.25–1.51
Mongolian	6	14	15	11	15	5	1.69	.43–7.20	3.14	.80–14.10
Elunchun	12	22	3	13	15	3	1.58	.51–4.99	.69	.08–5.85
Osier et al. 1999:
Taipei
Han	6	56	73	40	39	49	9.42	3.52–29.86	1.04	.58–1.86
Ami	1	5	14	3	6	11	2.36	.13–156.6	1.51	.29–8.14
Atayal	0	7	13	0	6	15	NA	NA	.75	.16–3.38
Yin and Agarwal 2001:
China (Han Chinese, Taipei)e	54	242	361	152	130	137	5.22	3.54–7.79	1.42	1.05–1.91
Higuchi 1994:
Japan	31	152	247	204	224	227	4.46	2.86–7.10	1.60	1.21–2.13
Maezawa et al. 1995:
Japan	2	22	36	30	28	38	11.48	2.45–109.90	1.20	.55–2.64
Nakamura et al. 1996:
Japan	3	54	40	21	20	12	18.25	4.73–106.00	1.23	.50–3.11
Tanaka et al. 1996:
Japand	4	24	38	27	42	21	3.81	1.13–16.77	3.14	1.43–7.04
Lee et al. 2001:
Seoul, Koread	6	18	40	3	21	28	.44	.06–2.40	1.66	.70–3.98
Common OR:
Han Chinese							5.19	3.74–7.26	1.36	1.07–1.72
Japanese							5.50	3.75–8.22	1.70	1.34–2.17

Note.— RR = ADH1B*47Arg/*47Arg; RH = ADH1B*47Arg/*47His; HH = ADH1B*47His/*47His, OR = odds ratio, NA = Not applicable (odds ratio could not be calculated because of empty cells).

^aU.K. or Irish descent.

^bMen only.

^cAustralians of European descent.

^dControl subjects versus alcoholics without alcoholic cirrhosis or pancreatitis.

^eALDH2*11 and *12 subjects only.

Data from each article, exact odds ratios, and their 95% CIs are shown in table 1. For the ADH1B*47Arg/*47Arg versus ADH1B*47Arg/*47His (ADH2*1/*1 versus ADH2*1/*2) comparison, there was significant heterogeneity of odds ratios across all the studies (P<.0001). Division of studies into those from Europe (including Russia and Australia) and those from Asia, with separate analyses for the two groups, showed no evidence of within-group heterogeneity among Europeans (P=.397), and the estimated common odds ratio was 2.11 (95% CI 1.32–3.44). However, there was still significant heterogeneity (P<.0001) among Asian studies. Inspection of the data suggested that results from Japanese and from Han Chinese groups were similar, whereas the minority ethnic groups within China, as well as Koreans, had lower odds ratios. As can be seen in table 1, the Han Chinese and the Japanese groups had very similar common odds ratios associated with ADH1B*47Arg/*47Arg compared with ADH1B*47Arg/*47His, which were substantially above those for Europeans and most of the other Asian groups.

The calculated odds ratios for ADH1B*47Arg/*47His against ADH1B*47His/*47His (ADH2*1/*2 versus *2/*2) are also shown in table 1. There was no significant heterogeneity between studies (P=.405), and the estimated common odds ratio was 1.43 (95% CI 1.23–1.66). The difference in alcohol-dependence risk is therefore greater for ADH1B*47Arg/*47Arg versus ADH1B*47Arg/*47His than for ADH1B*47Arg/*47His versus ADH1B*47His/*47His, at least in the mainly East Asian populations in which the ADH1B*47His allele frequency is high enough to allow a meaningful comparison.

Two conclusions may be drawn from this summary of published results. First, the ADH1B*47His allelic effects on alcohol dependence risk are not additive. Heterozygotes are clearly more similar in risk to the ADH1B*47His/*47His homozygotes than to the ADH1B*47Arg/*47Arg homozygotes, and so the ADH1B*47His allele shows quantitative (but not complete) dominance. Proposed mechanisms for the ADH1B Arg47His effect on dependence need to account for this feature. It is worth pointing out that a study that measured hepatic ADH activity and ADH1B genotype in human livers found that activity at pH 7.5 was approximately fivefold higher in ADH1B*47Arg/*47His subjects and was only sixfold higher in ADH1B*47His/*47His subjects than in those with the ADH1B*47Arg/*47Arg genotype (Yao et al. 1997). It is not clear whether these two examples of nonadditive effects of this polymorphism are related.

Second, there was a notable difference between European and Chinese or Japanese risk estimates. At least two types of explanation for heterogeneity between populations in the relative risk conferred by, or associated with, a genetic polymorphism should be considered: genetic and social. If the polymorphism is not itself causative, then linkage disequilibrium with a causative locus will decrease with the passage of time after the original mutation event and may remain stronger in one group than in another. Alternatively, the same neutral polymorphism may have arisen independently in the two populations and may be in linkage disequilibrium with the causative polymorphism in only one. It will be seen from table 4 in the article by Osier et al. (2002) that the ADH1B*47His allele occurs on a different haplotype background in East Asians (mainly 221221) and the European/Middle Eastern/European North American groups (mainly 221211, or 212211 in some Samaritans). Although this does not demonstrate independent mutations, it does suggest that the origin of ADH1B Arg47His is not recent and that changes have occurred in the nearby sequence.

It has generally been assumed that the ADH1B Arg47His polymorphism is causative and that the effect arises from the difference in V_max for ethanol (Bosron and Li 1986) between the enzymes produced. However, there are problems in extrapolating this in vitro activity difference to alcohol metabolism in vivo, and as Osier et al. (1999, 2002) discuss, another causative polymorphism within the ADH region cannot be excluded.

On the other hand, social factors or other unlinked genetic effects may modify the ADH1B Arg47His effect in the comparatively few Europeans who have the ADH1B*47Arg/*47His or ADH1B*47His/*47His genotypes, so the genotype-associated difference in risk is smaller. There is evidence (Higuchi et al. 1994) that the size of the protective effect associated with aldehyde dehydrogenase (ALDH2) deficiency has changed during the past 20 years in Japan—a period that, although it is far too short for genetic changes, has been a time of substantial alterations in the social environment. Lee et al. (2001) also comment on the social pressures to drink in Korea. Gene-environment interaction therefore presents an alternative explanation for the heterogeneity between populations.

We cannot yet determine whether social factors or variations in linkage disequilibrium are responsible for the difference in ADH1B Arg47His effects between Europeans and two major Asian groups. The question may be resolved by haplotype data across the ADH region in alcoholics and control subjects from different countries or regions, or by studies of alcoholics and control subjects of Asian descent living in European societies.